The book focusses on the grammatical feature definiteness in German, visible in the inflection of adjectives (“ein schönes Kind” vs. “das schön-e Kind”). It argues for an analysis of this effect that draws a connection to the visible categories of number and gender on nouns and related words rather than an abstract property. This conclusion rests on the conflation of the established grammatical categories into a single one, number-gender, which explains a vast body of grammatical phenomena in German and principles of language in general. ISBN 978-3-8233-8133-4 Benincasa A Phi-Syntax for Nominal Concord Marco Benincasa A Phi-Syntax for Nominal Concord Morphological Definiteness as Gender-Agreement in the German DP A Phi-Syntax for Nominal Concord Marco Benincasa A Phi-Syntax for Nominal Concord Morphological Definiteness as Gender-Agreement in the German DP Dissertation am Fachbereich für Geistes- und Kulturwissenschaften der Universität Kassel. Tag der Disputation: 10.02.2016 Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http: / / dnb.dnb. de abrufbar. © 2018 · Narr Francke Attempto Verlag GmbH + Co. KG · Dischingerweg 5 · D-72070 Tübingen Das Werk einschließlich aller seiner Teile ist urheberrechtlich geschützt. Jede Verwertung außerhalb der engen Grenzen des Urheberrechtsgesetzes ist ohne Zustimmung des Verlages unzulässig und strafbar. Das gilt insbesondere für Vervielfältigungen, Übersetzungen, Mikroverfilmungen und die Einspeicherung und Verarbeitung in elektronischen Systemen. Internet: www.narr.de E-Mail: info@narr.de Satz: pagina GmbH, Tübingen Printed in Germany ISBN 978-3-8233-9133-3 Für Luise und Franca 7 Table of Contents List of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 1 AXIOMS - Adjectival Inflection & Phrasal Categorization . . . . . . . . . . . . . . . . . . . . 13 1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 1.2 Preliminary Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 1.3 Basic Concepts - An Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 1.3.1 (Morphological) Definiteness . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 1.3.2 Adjectival Inflection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 1.3.3 Morphological Definiteness Coincides with Adjectival Inflection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German . . . . 39 1.4.1 Quantificational Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 1.4.2 Determiners and Demonstratives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 1.4.3 Lexical Possessives and Possessive Pronouns . . . . . . . . . . . . . . . . . 53 1.4.4 Pronominal Elements and (other) ‘ ein -Words’ . . . . . . . . . . . . . . . . . 69 1.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77 2 PREMISES - Phasehood & φ-Dependencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 2.2 Phasehood . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 2.2.2 Head Status and the Fine Structure of the Nominal Periphery 100 2.2.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110 2.3 φ-Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 2.3.1 Number and Gender (Categories) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112 2.3.2 Plural and Feminine (Values) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118 2.3.3 Plural and Feminine (Category) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 2.3.4 Concluding Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 2.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148 3 INFERENCE - Nominal Concord & Definiteness . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 153 3.2 Preliminary Remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155 8 Table of Contents 3.2.1 φ-Feature Dependencies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155 3.2.2 Agree and Nominal Concord . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162 3.2.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180 3.3 Splitting Nominal Concord . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 3.3.1 The Basic Pattern of SLI Inflection and Nominal Concord . . . . 187 3.3.2 Case-Assignment and Preconditions for the mixed Adjective Inflection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195 3.3.3 The Extended Pattern of SLI Concord . . . . . . . . . . . . . . . . . . . . . . . . 220 3.3.4 Adjectival Inflection in Morphology . . . . . . . . . . . . . . . . . . . . . . . . . 251 3.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 281 4 IMPLICATIONS - Optionality & Complex Formations . . . . . . . . . . . . . . . . . . . . . . . 288 4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 288 4.2 Optionality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289 4.2.1 Of Movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 290 4.2.2 Of Inflection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310 4.3 Loose Ends & Complex Formations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 319 4.3.1 Complex SOs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 320 4.3.2 Complex LIs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339 4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361 5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363 Endnotes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376 Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389 List of Short Form Citations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 406 9 List of Abbreviations masc , [m] masculine gender fem , [f] feminine gender neut , [n] neuter gender sg , [sg] singular number pl , [pl] plural number NOM , [nom] nominative case GEN , [gen] genitive case DAT , [dat] dative case ACC , [acc] accusative case strong strong adjectival inflection form/ pattern weak weak adjectival inflection form/ pattern mixed mixed adjectival inflection pattern HN head noun SLI semi-lexical item ES existential sentence PDC possessor doubling construction QST quantifier stranding SC Serbo-Croatian AI adjectival impoverishment OS object shift/ stylistic fronting/ scrambling PN partitive noun A note on technical terms Throughout the analysis, I try to maintain a lucid use of technical terms by following several simple notational conventions: For one part, I mark the introduction of a technical term by putting the first mentioning in single quotes and, moreover, flag those terms designating derivational operations and restrictions with capitalization. Established concepts furthermore follow general spelling conventions in the literature (e.g. SEM, PHON, TRANSFER in capitals, Spec,XP without a space). 10 A note on grammaticality judgments Throughout the thesis, I employ graded grammaticality judgments, utilizing the symbols ‘*’ for total ungrammaticality and ‘#’ for slight ungrammaticality/ oddness as well as for cases judged (un)grammatical by approximately 50% of my informants. The former use can furthermore be intensified by reduplication (i.e. ‘##’) while the latter will always be mentioned in the text surrounding the datum. Notes on citing data and footnotes When citing language data, I fully adopt the gloss of the source cited, including both its formatting and the abbreviations incorporated there if not indicated otherwise. When citing or referring to specific language data or footnotes from other works, I also include the number of the respective chapter in cases where the source text is organized in such a way as to restart the count with each beginning chapter. When repeatedly referring to a team of authors, it will be referenced by a short hand citation as indicated in the text. These are again compiled in a list following the bibliography. Moreover, key publications by Noam Chomsky are cited by an abbreviation of their title rather than by the year of publication - as listed in ch.1, fn. 1 - for ease of readability. 11 Introduction This book sets out to unify various agreement phenomena throughout the nominal domain in German. The patterns of agreement in this language are fairly complex, incorporating modifier-head agreement of various prenominal elements w.r.t. the so-called ‘φ-features’ number and gender of the nominal head as well as case-agreement across the complete domain. An additional agreement relation holds apart from the nominal core between various left-peripheral elements and prenominal adjectives for - what has been termed - definiteness, depicted in varying inflection on the latter. All these instances of (i) modifierhead as well as modifier-modifier (ii) definiteness-, φ- (i.e. person, number and gender) as well as case-agreement in the extended projection of the noun are standardly subsumed under the label ‘nominal concord’, a term that I will adopt here. The unifying approach to these seemingly heterogeneous instantiations of agreement builds on the narrow-syntactic relevance of φ-feature values as well as their (dependency) relations among one another, which constitutes the basis for the account labeled ‘Phi-Syntax’, as elaborated by Bejar (2003). Building on her proposals, the analysis developed here first suggests a reordering of φ-value dependencies and subsequently sketches the various paths of agreement for these configurations from the nominal core throughout diverging set-ups of its extended projection under a feature sharing version of Agree. Therein, several agreement relations reduce to one single chain of the shared feature hierarchy while, additionally, diverging categories of agreement reduce to the locus of overt realization of said structure; to wit, morphological definiteness reduces to gender-agreement in the extended nominal projection. The book is structured as follows: Chapter one will set out to justify the two core notions which are not deducible but have to be stipulated for the analysis to follow in subsequent chapters, namely the effect of the status of definiteness of left-peripheral elements on the inflectional pattern of adjectives on the one hand, as well as the structural status of various prenominal lexical items as heads of discrete and ordered phrases in the extended projection of the noun on the other hand. Although I will go through some length to legitimatize both, the refusal to acknowledge either one will render the following analysis moot. Therefore, this chapter has been labeled ‘axioms’ of my analysis. Chapter two will start out by focusing on the phasal status inside the elaborated hierarchy of projections. As I will argue predominantly with reference to Bošković (2014), phasehood is a contextual, hence variable property. As will be 12 Introduction shown, co-occurrences of left-peripheral elements in ellipsis of, and extraction from, nominal domains are accounted for therein. The second half of the chapter will be devoted to the coding of the φ-features number and gender in nominal domains. Based on a variety of morpho-syntactic as well as linearizational idiosyncrasies, a split of these decided categories in favor of intertwined values will be advocated along with the diverging structural loci for the resulting hierarchically structured bundles, mainly following proposals by Ritter (1993) and Harley & Ritter (2002a). Since the concepts elaborated therein will together form the basis for the subsequent analysis and moreover facilitate the notions in which the final argument will be stated, it has been labeled ‘premises’ of the investigation. The third chapter will set out to elaborate a modified framework of Phi-Syntax (Bejar 2003), incorporating among other things a feature sharing approach to Agree (Frampton & Gutmann 2000) and contrast the system developed therein with the default notion of Agree in Minimalist syntax. The remainder of the chapter will then employ all the above insights and modifications in the analysis of agreement in the nominal domain in German. It will be shown that morphological marking of definiteness and various patterns of φ-agreement on adjectives as well as left-peripheral lexical items are indeed two sides of the same coin, i.e. the uniform vs. disrupted chain of the shared φ-feature configuration throughout the nominal domain and beyond. Since the chapter will derive these findings in the combination of the insights from the preceding chapters, it has been labeled ‘inference’ of the current investigation. Chapter four will be concerned with various aspects, touched upon in the course of the previous analysis. The first subchapter will take a closer look at optionality in the quantified nominal domain concerning both movement from as well as inflection on quantifying prenominal lexical items and carve out the structural and featural basis for both of them. The remainder of the chapter will then tie together various loose ends from the foregone investigation by reviewing complex configurations in the nominal domain, both above and below the word-level to (i) identify mechanisms of case-assignment inside the nominal domain and (ii) to defend a decompositional approach as well as (iii) to present considerations on the semantic purpose of the syntactic workings carved out over the course of this inquiry. Since these various topics will be taken up from previous stages of the analysis, the chapter has been labeled ‘implications’ of the investigation. Chapter five provides a conclusion. 1 AXIOMS - Adjectival Inflection & Phrasal Categorization 1.1 Introduction In order to guarantee a coherent analysis, both the object of investigation as well as the notions in which it is conducted have to be carefully defined in advance. The current chapter sets out to serve this purpose: On the one hand, it is intended to set the stage for the inquiry in the chapters to follow by taking a closer look at two intertwined core phenomena, i.e. adjectival inflection and definiteness in German, and introduce a general understanding of their relation to one another. The complexity concerning the questions asked in this introductory chapter regarding the relation of the two is herein restricted to the presentation of simple patterns of co-occurrence of the elements involved, followed by a pre-theoretical categorization of them. On the other hand, the chapter is also understood to mark the outset of said analysis itself in a roundup of the structural premises that will be utilized throughout the rest of this book, i.e. the phrasal set-up of the functional domain in the nominal hierarchy of projections. Although I will do my best to motivate each of these in detail, I acknowledge that their selection, featural content and ordering to one another constitute mere stipulations that do not follow from deeper principles of the theory itself. As my analysis proceeds, however, I am positive that increasing approval will subsequently be given to the specific setup by the reader, once a broader and wildly heterogeneous selection of data has been discussed and - as I hope - satisfactorily accounted for in the analysis. Since the latter resides deeply in the premises of one framework of linguistic theorizing, viz. Generative Grammar, the establishment of a common ground is vital to the success of the analysis to follow, concerning the underlying rationale of the theory and the architecture of its model for the human capacity of speech as well as further concepts and technical terms. To this end I will prefix this chapter with a broad initial outline of the state of the theory in the following section, to be employed and modified in the course of my analysis. I will moreover include additional introductory subchapters and passages like the one below throughout the current analysis, preceding respective shifts in the focus of my discussion concerning the aforementioned model and its theoretical concepts as 14 1 AXIOMS - Adjectival Inflection & Phrasal Categorization well as elaborations of them. Therein, I hope that an accessible argumentation is provided for the reader. 1.2 Preliminary Remarks The following analysis is couched in the Minimalist Program of syntactic theory, the current framework of Generative Grammar, as outlined in recent works by Noam Chomsky (1993 et seq.). 1 Generative Grammar takes syntactic computation to be the motor of the ‘Faculty of Language’ (FL), hence poses a syntacto-centric system. Said system takes as its input a selection of ‘Lexical Items’ (LIs), copied from long term memory (the ‘Lexicon’, LEX) into active short term memory (the ‘Numeration’/ ‘Lexical Array’, LA). Its output, on the other hand, is twofold: The ‘Logical Form’, LF, constitutes appropriate information for the language-external module of the conceptual-intentional (C-I) system, transmitted from the semantic component Σ via the interface SEM, while the phonological component Φ transmits the ‘Phonological Form’, PF, to the external articulatory-perceptual (A-P)/ sensorimotor (SM) system via the interface PHON. SEM and PHON, then, correspond loosely to the meaning and sound side of language, a basic structuralist tenet. Appropriateness of the output designates legibility at the external systems; therefore, call these requirements the ‘bare output’/ ‘legibility conditions’. Call the computational path from input to output the ‘derivation’. The lexical items involved, as well as the complex structures derived thereof (call both ‘Syntactic Objects’, SOs), are taken to consist of ‘features’, basic building blocks of phonological, semantic and formal nature, the latter of which (FFs, henceforth [F]s) enter the computation of strings of (ultimately) sentential quality, the remaining ones inserted afterwards at the appropriate components (termed ‘Late Insertion’), with information at Σ/ SEM consisting exclusively of semantic, information at Φ/ PHON solely of phonological features. A feature of inappropriate category is understood as poisonous at the respective components as well as interfaces and hence taken to force the derivation to a halt (i.e. to crash it); derivations satisfying legibility conditions at both Σ and Φ, however, converge. Formal features divide into intrinsic and optional ones, the former inherently specified in LEX, the latter added at LA. Certain formal features do not survive 1 1993, henceforth MPLT; 1995a, henceforth MP; 1995b, henceforth BPS; 2000, henceforth MI; 2001, henceforth DbP; 2004, henceforth BEA; 2007, henceforth AUGB; 2008, henceforth OP; 2013, henceforth PoP. 1.2 Preliminary Remarks 15 past the syntactic core component ‘Narrow Syntax’ (NS): They delete before the derived structures are shipped off to the aforementioned components by the operation ‘TRANSFER’/ ‘Spell-Out’. Deletion applies as part of a complex process termed ‘checking’ between two compatible features consisting of (i) ‘Match’, searching for compatible items, (ii) ‘Agree’, itself a complex operation consisting of ‘Value’ and TRANSFER, as mentioned above. Compatibility is evaluated by (i) category and (ii) status of (un)interpretability: uninterpretability is unvaluedness (i.e. [ u F: ]); interpretability signals survival at Σ/ SEM (i.e. [ i F: val]), hence semantic relevance, the ‘Valuation/ Interpretability Biconditional’. The sets of semantic and formal features thus intersect. Interpretable features of a category [F], call them [ i F], are able to check uninterpretable counterparts, call them [ u F], in the appropriate structural configuration of (extended) sisterhood in a binary branching, hierarchical and two-dimensional configuration, call it ‘c-command’, via Value (Agree). Call the [ u F] initiating checking the ‘probe’, [ i F] identified by Match the ‘goal’. This is the sole motor of syntactic computation, D-NS. Features vary across another dimension, i.e. they are categorized as either strong ([ u / i F*]) or weak ([ u / i F]), hence, the system exhibits four logically possible feature set-ups. This second partition has consequences as to the locality of the checking relation, outlined above, with weak features able to check either in a non-local, i.e. extended, or local relation of sisterhood. Lexical material is (i) selected from the Numeration/ LA (in the case of terminal elements, i.e. LIs) or from a parallel workspace (in the case of complex SOs) into the computation (the current workspace) according to their undeleted (i.e. unvalued) [ u F]s, (ii) copied and (iii) combined with the highest node (the ‘Extension Condition’) by the operation ‘External Merge’ (EM). Thence, [ u F] (mutually) c-commands [ i F]. Additionally, Match is restricted to the structurally closest instance of [ i F] from [ u F] termed ‘Relativized Minimality’, the ‘Locality Condition’. Furthermore, the goal must itself bear [ u F] and hence be rendered active for further computation, the ‘Activity Condition’. Alternatively, further buildup of the hierarchical configuration in the workspace might possibly also be established without recourse to the Numeration/ LA or parallel workspaces. This is once again achieved via the operation Merge, applying workspace-internally: ‘Internal Merge’ (IM), via (i) selection of an appropriate [F] (ii) copy of the smallest SO [F] is part of (iii) addition of the copied SO at the top node of the hierarchy; EM and IM hence only differ in the source of the SOs involved (i.e. their initial steps as sketched above). 2 Thus, IM creates 2 Throughout the analysis, I represent applications of Internal Merge in language data by following traditional conventions in the employment of an indexed placeholder ‘t x ’ (trace) in the source position for ease of exposition; the reader is asked to bear in mind that said structural position is occupied by a copy of the moved SO in Narrow Syntax. 16 1 AXIOMS - Adjectival Inflection & Phrasal Categorization local sisterhood relations of formerly distant SOs (more accurately, their [F]s), the prerequisite for checking relations involving strong (interpretable or uninterpretable, [ i / u F*]) features. The combination of two elements brings about a new terminal node, a label (but cf. Chomsky PoP for a revision of the mechanism). If the selection of lexical items has been performed on the basis of their respective featural set-up, the LI checking its [ u F] is the one to be copied as the new label of the complex element, a subpart of both merging operations, introduced above, call it ‘Set-Merge’. Call the terminal elements ‘heads’, H. If no more [ u F]s are part of the featural set-up of H, which provided the label before, the next item merged with the topmost node (and bearing another [ u F]) is subsequently copied as the new label. Call the first element merged for the purpose of reducing the number of [ u F]s of H the complement of H, call the path from the lowest label provided by H to the highest one the ‘projection of H’, the ‘phrase’ HP, call all remaining elements following the complement in the projection of H and merged with a label of H, the ‘specifiers of H’, Spec,HP. Let us, for the purpose of this analysis, restrict the number of the latter to one even though this does not follow from conceptual premises (cf. Chomsky DbP: fn. 48), hence enforce a rigid phrase-structural model. Iff H corresponds to an item of an open class in LEX (arguably also prepositions), call HP a ‘lexical projection’, otherwise call it a ‘functional’ one. All heads are stored in LEX. Functional projections erect over lexical ones in a fixed order, given their heads are copied along with the lexical ones to the respective Numeration/ LA. For the course of the present analysis, the proposal of functional projections has to be backed up by overt morphological reflexes of their respective heads or Specs, a rationale captured best in Bošković’s (2014: 30) terms: “[W]hat you see is what you get.” Call the hierarchy of projections erected on top of a lexical projection HP the ‘extended projection of H’, call the extended projection of an item of category V a sentence. A hierarchy solely consisting of interpretable features is accessible for further computation at Σ and Φ and therein translatable into appropriate information to be processed at C-I and A-P/ SM respectively. However, recoverability of the semantic import of structural relations between SOs requires the retraction of applications of IM: While EM creates basic argument structure, IM creates discourse structure. Information hence splits up from D-NS as soon as all [ u F]s have been valued and hence subsequently deleted, to be further processed at Φ, accordingly. Call the subpart of D-NS prior to the split the ‘overt component’, the part of D-NS following said partition the ‘covert component’. The derivational path from Φ to PHON to A-P/ SM necessarily incorporates a module, bundling nodes in “word-like units” (Chomsky MP: 229), call it ‘Morphology’, 1.2 Preliminary Remarks 17 as well as a component translating hierarchical (i.e. two-dimensional) relations into relative linear order of said units, ‘Linearize’/ ‘Linearization’, arguably employing an algorithm based on (asymmetric) c-command relations, call this the ‘Linear Correspondence Axiom’, LCA. Transmission from NS via TRANSFER proceeds piecemeal (i) to a fixed structural extent of the elaborated hierarchy in the workspace and (ii) at a fixed derivational timing. Call these structural stages ‘phases’. Concerning (i), TRANSFER does not target phrase-levels, but the complement of the ‘phase head’ H, call it the ‘interior’ of the phase, leaving H and Spec,HP. Call this the ‘edge’ of H. Concerning (ii), TRANSFER applies once the next higher phase head is introduced into the workspace by EM, call this the weak ‘Phase-Impenetrability Condition’, PIC. Call subparts of the Numeration corresponding of phasal magnitude, ‘Lexical Subarrays’ (LAs). Erection and transmission of SOs hence both apply iteratively in successive cyclic fashion of the basic operations involved. The particular distribution of features along the dimension of strong vs. weak instances represents a language-specific idiosyncrasy; the totality of all valued [F]s, active in a specific language, hence constitutes a language-specific matrix of parameters, which becomes fixed for the particular native speaker during her/ his respective L1-acquisition. The features themselves are (mostly) taken to constitute language universal elements, part of the cognitive endowment of human beings, as is the derivational system, sketched in this section. Call it ‘Universal Grammar’, UG. L1-acquisition moreover includes the identification of phonological and morphological restrictions, arguably equally guided by tacitly known principles of the learner, as well as the aggregation of LIs in LEX. Languages do hence differ in PHON, not in SEM. They differ in the former in two ways: (i) obviously concerning the lexical items employed and (ii) structurally concerning the ordering of LIs/ SOs w.r.t. each other, dependent on the setting of parameters, as outlined above. They do not differ in the latter structurally, since reconstruction resets language-specific restrictions. Meaning is hence universal, structure is not (though, however, restricted in its variety to the factor (active [F]) 2 as well as principles of UG). Information to be further processed at Φ must hence split from the derivation at some point in the course from LEX to Σ, call the subpart of the latter path comprising NS and Σ the ‘Computational System for Human Language’, C HL . Call the design of the entire system the ‘inverted-Y model’. FL arguably poses a perfect solution to the conditions imposed by the external systems it communicates with: the legibility conditions sketched above. Call this the ‘Strong Minimalist Thesis’, SMT. The thesis itself constitutes the primary yardstick of the Minimalist Program, a guideline in the search for deeper 18 1 AXIOMS - Adjectival Inflection & Phrasal Categorization explanations in the Galilean tradition that likely will turn out too restrictive. However, it is only through scrupulous analysis but confident conclusions that we will ever know. This work hopes to contribute to this end. 1.3 Basic Concepts - An Outline It is a well-known though poorly understood phenomenon of several Germanic languages that adjectival inflection varies between two patterns, traditionally termed ‘strong’ and ‘weak’ inflection, 3 depending on what is most commonly understood to be the status of definiteness of the noun phrase that they are part of. Consider (1) in which the complex nominative noun phrase is headed by the indefinite and definite determiner respectively: (1) a. ein großes Haus a big strong house b. das große Haus the big weak house Furthermore, weak inflection is also triggered by demonstratives, certain quantifiers (to be made precise below) as well as possessive pronouns with plural head nouns (henceforth HNs), while another group of quantifiers is accompanied by adjectives bearing strong inflection. In addition, a possibly complex noun phrase, not headed by any overt determiner-like element, also exhibits strong inflection: (2) großes Haus big strong house Because of the asymmetry of (1) and (2), accounts of the strong/ weak inflectional dichotomy standardly take adjectives in the noun phrase to pose an inherently strong inflected category with weak inflection forced upon them by the subclass of left-peripheral elements, just outlined. The current subchapter will be devoted to carve out the basic notions involved in generating this surface effect, i.e. definiteness and adjectival inflection, as well as their relation to one another. This might, at first, amount to a trivial task that has already been accomplished in the preceding paragraph, but as is 3 The terms ‘strong’ and ‘weak’ ultimately go back to Jakob Grimm (1870: 718-756), as noted by Roehrs (2009). 1.3 Basic Concepts - An Outline 19 almost always the case with human language, matters are not as simple as they first seem (although, then again, much simpler on a deeper level). To this end, I will first introduce the field of work based on the assumption that it offers a fruitful account to focus on - what has been termed - ‘definiteness’ in German, solely from a syntactic viewpoint. For that purpose, I will briefly transpose data demonstrating relevant surface effects from English to German and thereafter continue by sketching a tentative categorization of lexical elements in the nominal domain based on their interaction with the inflectional properties of coordinated adjectives, before the alignment of the source of these two phenomena is reviewed. Thereby, a basic framework for the analysis in the following chapters will be established from two of the core concepts involved. 1.3.1 (Morphological) Definiteness Definiteness is a puzzling semanto-syntactic feature for language-universal linguistic theory today. This is mainly due to the manifold semantic and pragmatic concepts associated with, and expressible by, this grammatical feature comparatively as well as language-internally (e.g. specificity, situational and anaphoric familiarity, uniqueness, referentiality, associativity, deixis as well as genericity; cf. Lyons 1999: ch. 1.1, ch. 4). Several studies aimed at the unification of the aforementioned effects on meaning, associated with the grammatical feature (cf. Lyons 1999: ch. 7 and references therein); what these studies generally agree upon is that the core notions center around the availability of information in the discourse in a speaker-hearer relation (the ‘familiarity thesis’, expanded and shifted to ‘identifiability’ in Lyons’ (1999: 5f.) terms) on which all other concepts mentioned above build semantically. This line of reasoning led to a general hierarchical view of the semantic set-up of definite elements and established the approach to distribute basic semantic features to sets of definite lexical items. 4 But as Lyons (1999) extensively argues, this unifying attempt might run into conflict with two independent features of language itself: First, concerning the semantic aspect of definiteness, he observes that what is generally (cross-linguistically as well as language-internally) subsumed under semantic definiteness resists backtracking to a single such concept. 5 Second, and focusing on the syntactic side, a simple syntactic feature [± definiteness] may bear a variable relation to SEM/ LF once again on both the language-internal as well as the cross-linguistic level. The combination of these possible realms of idiosyncrasy hence leaves a 4 See e.g. Lyons (1999: ch. 8.5.3); Klinge (2008: ch. 5); Roehrs (2009: ch 5.2.2). 5 For Lyons, unifying efforts end in the identification of two decided core concepts: ‘identifiability’ and ‘identification’, of which neither is said to be reducible to (a subpart of) the other. 20 1 AXIOMS - Adjectival Inflection & Phrasal Categorization wide grey area of variably definite noun phrases. Lyons’ strategy to anticipate this idiosyncratic state of affairs in his investigation lies in the reanalysis of the relation of the two sides of the phenomenon just outlined. It is worth quoting him in some length here: 6 [T]here is a distinction to be made between grammatical definiteness and semantic/ pragmatic definiteness […]. The proposal is that definiteness stricto sensu is not a semantic or pragmatic notion as assumed by almost all writers on the subject, but rather a grammatical category on par with tense, mood, number, gender etc. But, like these, it is the grammaticalization (that is, the representation in grammar) of some category of meaning. And the crucial observation here is that the correspondence between a grammatical category and the category of meaning it is based on is never one-to-one. (Lyons 1999: 274f., emphasis in the original) In the light of Lyons’ insight, the concept of definiteness might be translated into feature theory by a split of the morpho-syntactic and semantic component of the concept. 7 This parallels Roehrs’ (in prep.) treatment of the category number in the nominal domain in German: He suggests two features, [PL morph] and [PL sem], externally merged in a number phrase, NumP, as well as present on nominals themselves. Their respective specification and configuration allow him to derive a wide spectrum of noun phrase configurations and readings together with their morphological properties in a combinatorial fashion. I will follow the rationale of Lyons’ and Roehrs’ analyses here (cf. also Etxeberria & Giannakidou 2009). Specifically, I want to separate morpho-syntactic from semantic definiteness in what follows and focus on the former. I am positive that an account which starts out by dispensing with pragmatically distorted semantic reflexes will be able to uncover principles of C HL concerning the effect under consideration, which also apply more generally apart from the domain under investigation. It is also due to this rationale that I will predominantly rely on language data in out-of-the-blue (i.e. zero) contexts throughout the book to prevent inferences based on information structure. In what follows, I will thus be concerned with elements, structural relations and operations involved in the coding of what has been termed ‘grammatical definiteness’ above by Lyons (i.e. ‘[DEF morph]’ in Roehrs’ phrasing) in German. 6 See also Kibort (2008: ch. 3) who discusses this point based in part on adjectival inflection in German. 7 A similar rationale is mentioned in Roehrs (2009: 179) discussion of intensifiers, determiners and quantifiers in the German noun phrase, building on Karnowski & Pafel (2004), who themselves refer to Lyons (1999) and Pafel (1994) in this point. His motivation therein is to capture the uniform syntactic behavior of semantically heterogeneous classes. 1.3 Basic Concepts - An Outline 21 The account to be presented herein does, however, part from the approaches just laid out in that I will refrain from the pre-analytic stipulation of a syntactic formal feature for definiteness. It will hence be left to be shown that surface effects associated with definiteness justify such a proposal or if these can be accounted for, and moreover reduced to, the workings and/ or interactions of the established featural content of nominal domains. The procedure of splitting semantic from syntactic subparts of a formal feature, outlined above, is hence to be understood as a template in spirit, not in application, for the analysis to follow. I will therefore also depart from Lyons’ term of ‘grammatical definiteness’ to forebear from the featural implication associated with it and refer to the surface effect, outlined in the introductory section, as ‘morphological definiteness’. To begin the analysis, I want to present a first set of data illustrating the syntactic relevance of what has been understood to constitute definiteness in German. This superficial and introductory run-through should moreover present an overview of the left-peripheral lexical items participating in generating said effect. Contrast (3) and (4), in which the differently inflected (bracketed) NPs are part of so-called existential constructions; the following syntactic environments are first discussed by Postal ([1966a] 1970); the existential construction is moreover the main topic of investigation in Milsark (1974, 1977): (3) a. Es gibt [ein Haus] auf dem Platz. it is [ a house ] on the square ‘There is a house on the square.’ b. Es gibt [Häuser] auf dem Platz. it is [ houses ] on the square ‘There are houses on the square.’ c. Es gibt [viele Häuser] auf dem Platz. it is [ many houses ] on the square ‘There are many houses on the square.’ (4) a. * Es gibt [das Haus] auf dem Platz. it is [ the house ] on the square ‘There is the house on the square.’ b. * Es gibt [dieses Haus] auf dem Platz. it is [ this house ] on the square ‘There is this house on the square.’ 22 1 AXIOMS - Adjectival Inflection & Phrasal Categorization c. * Es gibt [alle Häuser] auf dem Platz. it is [ all houses ] on the square ‘There are all houses on the square.’ Additionally, observe the structures incorporating a preposed adjective in (5): (5) a. * Schön wie ein Haus ist, wohnen wir gern darin. 8 beautiful as a house is live we gladly in.it ‘Beautiful as a house is, we enjoy living there.’ b. Schön wie das Haus ist, wohnen wir gern darin. beautiful as the house is live we gladly in.it ‘Beautiful as the house is, we enjoy living there.’ Bare noun phrases as well as those headed by the indefinite article ein ‘a’ and the quantifier viel ‘many’ pose no problem in the existential constructions in (3), contrary to those NPs headed by the determiner das ‘the’, the demonstrative dieses ‘this’ and the universal quantifier alle ‘all’ in (4), which all result in ungrammatical structures. The inverted pattern, however, is observed in (5): A noun phrase that results in a grammatical structure when inserted into postverbal subject position in existential contexts cannot be part of the construction in (5) a. (here: ein NP), while noun phrases headed by an element which results in ungrammaticality in such configurations (i.e. das in the example above) are perfectly fine in the context of (5) b. Both of the aforementioned authors understand the triggering factor for (un-) grammaticality of such structures to lie in the status of definiteness of the NP under consideration in English (cf. ch. 1.3.3 for a more in-depth discussion of Milsark’s original proposal). Following their insights, the data above hence depict the syntactic relevance of the partition in German. 9 8 I left out in-situ adjectives inside the noun phrases since these facilitate a generic reading of the sentences, which in turn renders the structure in (5) a. grammatical, a point of which Postal ([1966a] 1970: fn. 9) himself was also aware; all references and citations are taken from the 1970 version of his paper. 9 One reviewer suggests that a semantic focus effect could alternatively be put forth to account for the data presented in this section, which had been suggested for German as early as Grewendorf (1989: 172-175). In the same vein, he notes that the data in (4) turn grammatical when the noun phrase under consideration is focused by stress while (5) a. is rescued by its generic reading. However, note that I follow Milsark and Postal in their interpretation of parallel effects in English, who, moreover, report identical patterns of exception in their respective analyses. In the light of these parallels and in line with the general rational of the current inquiry, I feel justified to follow this research tradition in treating the effect as syntactic in nature. 1.3 Basic Concepts - An Outline 23 As is well-known, English does not display the adjectival strong/ weak bipartition sketched in the introduction to this subchapter. It is therefore to be shown that it is indeed the identical feature that is (i) responsible for the ungrammaticality of certain NPs in the syntactic contexts, laid out above, in both German and English, as well as (ii) relevant in the selection of the morphological form of adjectives in the former alone. To this end, I want to concentrate on the latter bipartition in some detail in the following section and therein deduce a first typology of left-peripheral elements concerning the adjectival inflectional pattern they co-occur with. The last subsection, then, will be concerned with the overlay of the concepts elaborated up until that point. 1.3.2 Adjectival Inflection As illustrated in in the introduction to this subchapter, (complex) noun phrases not headed by any overt determiner display strong adjectival inflection. This effect is not altered by the addition of the singular article ein , preceding it: 10 (6) a. großes Haus big strong house b. ein großes Haus a big strong house In contrast the inflection on the adjective does change if the determiner heads said structure: (7) das große Haus the big weak house It therefore suggests itself, as is almost universally done, 11 that the observed variation in featural content is introduced into the structure by the elements added, which are hence taken themselves to bear said difference. Let us (for now, but cf. chapter three below) follow this rationale. Thence, the variation 10 Apart from the nominative masculine and nominative/ accusative neuter, the article ein together with the negative element kein and possessive pronouns with singular headnouns trigger weak inflection. This constitutes the third paradigm of adjectival inflection, termed ‘mixed’ due to the property just outlined. I will come back to this phenomenon below. 11 Cf. Lyons (1999: ch. 2.1.1) for a discussion of the remaining logical possibilities of overt coding, cf. furthermore ibidem (ch. 8.3.1) for a tentative treatment of languages overtly coding indefiniteness rather than definiteness in his system. 24 1 AXIOMS - Adjectival Inflection & Phrasal Categorization in adjectival inflection allows us to identify left-peripheral lexical items in the nominal domain in German triggering weak inflection. Applying this testing environment to the various elements in the functional domain of NP in nominative case leaves us with the following partition (curly brackets indicating complementary distribution): (8) a. {das/ dieses/ jenes/ jedes} große Haus {Det/ Dem PROX / Dem DIST / every } big weak house sg b. {die/ diese/ jene/ alle/ meine/ keine} großen Häuser {Det/ Dem PROX / Dem DIST / all / Poss/ Neg} big weak house pl (9) a. {ø/ ein/ kein/ mein} großes Haus {ø/ a / Poss/ Neg} big strong house sg b. {ø/ zwei/ viele/ wenige} große Häuser {ø/ two / many / few } big strong house pl It is apparent from the data in (8) and (9) that the determiner (Det: das SG / die PL ) as well as the proximal and distal demonstrative (Dem PROX : dieses SG / diese PL , Dem DIST : jenes SG / jene PL ) always combine with weak inflected adjectives, while a complex noun phrase not headed by any of these left-peripheral items (i.e. ‘ø’ above) uniformly contains strong inflected adjectives, irrespective of the number-value of the HN. Additionally, a third group only occurs with either singular or plural HNs, which seems to be due to their semantic set-up. This is true for the article ( ein ) and the quantifier ‘every’ ( jedes ) in the singular as for all other morphologically simple quantifying elements (‘all’ alle , ‘many’ viele , ‘few’ wenige ) in the plural. Notice that these elements do not enforce uniform inflection: ‘many’ and ‘few’ accompany strong, while alle accompanies weak inflected adjectives. Interestingly though, number on HN does seem to have an impact on adjectival inflection with a last subclass of those semi-lexical items, namely the possessive pronoun (Poss: mein SG / meine PL ) and nominal negation (Neg: kein SG / keine PL ), which combine with strong inflection in singular NPs while preceding weak inflected adjectives with plural HNs. Note that I employ the term ‘semi-lexical item’ (henceforth SLI) in a pretheoretical manner as a cover-term throughout the analysis to refer to the heterogeneous set of left-peripheral elements in the German nominal domain. In doing so, I am able to stay agnostic as to the actual lexical vs. functional status of the individual categories subsumed therein. In general, there seems to exist 1.3 Basic Concepts - An Outline 25 little consensus on what comprises semi-lexicality. 12 The most prominent line of reasoning, ultimately dating back to Emonds (1985), takes semantically bleached lexical heads that have gained wider grammatical applicability to represent semi-lexical heads; Emonds originally referred to these as ‘grammatical nouns’ (e.g. one , thing , self ) and ‘grammatical verbs’ (e.g. be , have , do ) etc. This approach is further pursued by e.g. van Riemsdijk (1998) - and laid out in general in Alexiadou et al. (2007: ch. III.2) - who takes semi-lexical/ functional projections to constitute an intermediate level in the extended domain 13 of a lexical head to derive e.g. direct partitive constructions, DPCs (van Riemsdijk) as well as pseudo-DPCs (van Riemsdijk)/ pseudo-partitives (Alexiadou et al.) and N-of-N constructions (Alexiadou et al.), all of which, however, consist of concatenations of (categorically) lexical elements. 14 Another line of reasoning takes semi-lexicality to designate lexical items whose formal properties cannot be fully assigned to one of the two opposing categories ‘lexical’ vs. ‘functional’. My use of the term follows this rationale. Although it is clear that some of the left-peripheral lexical items fully fall into one of these categories (e.g. Det and Dem in (8) above, which are fully functional), the discussion in chapter 1.4 below will also present cases that do not adhere to the bipartition in such orderly fashion. This line of reasoning hence depicts a continuum with the two aforementioned categories posing as the respective endpoints. Therein, it ultimately arrives at the distinction of lexical and functional properties rather than categories. 15 Likewise, head status in the extended domain of the noun will below be attributed to, inter alia, monomorphemic closed class items (generally understood to be properties of functional elements) which nevertheless can be stranded by movement of their complement (i.e. a property associated with lexical categories). As noted above, I will not elaborate on the lexical vs. functional status of the individual categories any further. Returning to the phenomenon under investigation, the categorization on the basis of (8) and (9) above is not exhaustive, though. Once we additionally incorporate the dimension of case (especially focusing on lexical case forms) a further division emerges: 12 Cf. Littlefield (2006: ch. 3.1) for an overview of the various rationales connected to the term, moreover the contributions in Corver & van Riemsdijk (2001a) for analyses conducted in all of those. 13 Cf. van Riemsdijk (1990), moreover Grimshaw ([1991] 2005); cf. fn. 42 below. 14 Cf. Tănase-Dogaru (2011) for a summary of the criteria for semi-lexicality, following this rationale. 15 This parallels the original observations from Ross (1972, [1973] 2004) as reported by Corver & van Riemsdijk (2001b), cf. e.g. Zwarts’ (2011) analysis of the lexical and functional properties of prepositions in Dutch. 26 1 AXIOMS - Adjectival Inflection & Phrasal Categorization (10) a. großes Haus [ø big strong house sg ] NOM b. großem Haus [ø big strong house sg ] DAT (11) a. {das/ dieses/ jenes} große Haus [{Det/ Dem PROX / Dem DIST } big weak house sg ] NOM b. {dem/ diesem/ jenem} großen Haus [{Det/ Dem PROX / Dem DIST } big weak house sg ] DAT (12) a. {mein/ kein} großes Haus [{ my / no } big strong house sg ] NOM b. {meinem/ keinem} großen Haus [{ my / no } big weak house sg ] DAT The data in (10) illustrate the familiar strong inflectional forms on the adjective with nominative case (i.e. groß es ) as well as with its dative correlate ( groß em ) when no overt SLI heads the NP; (11) displays the respective pattern of weak forms ( groß e in nominative, groß en in dative respectively) with an appropriate element preceding NP. The structures in (12), lastly, parallel (10) and (11) with the elements identified as varying in adjectival inflectional patterns w.r.t. the number of the nominal head they are concatenated with: Note that we do find strong inflection on the adjective with nominative ( groß es ), while the inflection in dative case correlates with the weak inflection found in (11) b. The inflectional pattern exemplified therein hence makes use of weak as well as strong forms and is therefore labeled ‘mixed’ inflection. Variation hence cuts through both grammatical dimensions, number and case. More precisely, adjectives concatenated with singular neuter HNs (i.e. Haus in (10)-(12)), headed by the possessive pronoun and nominal negation, display strong inflection in the nominative and accusative but weak inflection in the lexical cases genitive and dative. Likewise, adjectives with singular masculine HNs display strong inflection in the nominative but weak inflection with all remaining case forms. Feminine-marked NPs, however, seem not to exhibit this scatter between inflectional paradigms, nor do plural noun phrases, since - as has been observed above - these structures always involve weak inflectional forms. Since we can hence safely concentrate on singular-marked NPs for now, a third element accompanying the mixed pattern can be identified, which has 1.3 Basic Concepts - An Outline 27 been categorized as occurring exclusively with singular HNs above, namely the article ein , *1 as can be derived from the case forms presented in (13) below. (13) a. ein großes Haus [ a big strong house sg ] NOM b. einem großen Haus [ a big weak house sg ] DAT I will discuss the paradigm exemplified therein in more detail in chapter 2.3.3 and finally get back to it at great length throughout chapter 3.3 below. For now, observe that all elements following the pattern of mixed adjectival inflection arguably share their morphological root: the article ( ein ), Poss ( m ein ) and Neg ( k ein ). Therefore, tying the idiosyncrasies of adjectival inflection accompanying these elements to the featural set-up of the root morpheme readily suggests itself. Lexical decompositional approaches along these lines are not a novel idea. 16 The identification of a root ein -, however, can as of now be nothing more than a tentative argument. I will return to this point in chapter 1.4.4 below. Turning to the remaining items categorized as exclusively occurring in specific syntactic environments, i.e. dependent on the number of HN, it is observable that this class consists exclusively of quantifying elements and hence readily combines with plural head nouns (to the exclusion of ‘every’, cf. ch. 4.3.2 a. below). There are, however, several environments to test the distribution of adjectival inflection of these elements with morphologically singular nominals. Since this analysis is conducted with the focus on morphological realizations, the working hypothesis should assume that morphologically non-distinct environments also build on structurally non-distinct configurations as well as identical featural compositions of the elements involved, regarding the phenomenon under consideration. This rationale hence covers various readings of nominal domains as well as NPs incorporating mass nouns, singularia and pluralia tantum nominals, occupational titles and proper names, all of which will surface as testing environments at one time or another in the course of the current analysis. That being said, let us turn to those LIs that are restricted to co-occurrence with plural NPs and therein accompany strong inflected adjectives, namely viel ‘many’ and wenig ‘few’. The arguably simplest of the testing environments mentioned above is the combination with mass nominals, which underlyingly bear (a kind of) plural semantics while not exhibiting the standardly accom- 16 Cf. e.g. Noyer ([1992] 1997), Harley (1994), Starke (2001), Harley & Ritter (2002a), Barbiers (2005), Bernstein (2008), Klinge (2008), Longobardi (2008), Cowper & Currie Hall (2009), Preminger (2011), moreover cf. Wiltschko (2009), Roehrs (in prep.) on German. 28 1 AXIOMS - Adjectival Inflection & Phrasal Categorization panying suffixial morphology, but are instead related to singular forms in this respect. These combine readily with the elements under consideration as can be seen in (14) using Reis ‘rice’, Milch ‘milk’ and Bier ‘beer’ as HNs bearing the gender-values masculine, feminine and neuter respectively: (14) {viel/ wenig} {Reis/ Milch/ Bier} { many / few } { rice masc / milk fem / beer neut } An adjective inserted into this environment clearly displays strong inflection across case forms where these are morphologically distinguishable (I once again take the pattern incorporating a homonymous strong/ weak form to follow the remaining unambiguously strong paradigms): *2 (15) a. {viel/ wenig} guter Reis [{ many / few } good strong rice masc ] NOM a'. {viel/ wenig} gutem Reis [{ many / few } good strong rice masc ] DAT b. {viel/ wenig} gute Milch [{ many / few } good strong/ weak milk fem ] NOM b'. {viel/ wenig} guter Milch [{ many / few } good strong milk fem ] DAT c. {viel/ wenig} gutes Bier [{ many / few } good strong beer neut ] NOM c'. {viel/ wenig} gutem Bier [{ many / few } good strong beer neut ] DAT We can hence categorize ‘many‘ and ‘few’ as exclusively accompanying strong inflected adjectives across both the number and case dimension. They are therein levelled with ‘ø’, i.e. (possibly complex) nominal domains without any overt SLI. Let us turn to the last element whose occurrence seems to be dependent on the plural of the nominal, but therein concatenates exclusively with weak inflected adjectives, as illustrated above, namely the universal quantifier all - ‘all’. This SLI likewise does not pose an exception to the test environment at hand, as is exemplified with female and neuter HNs below: (16) a. alle gute Milch [ all good strong/ weak milk fem ] NOM 1.3 Basic Concepts - An Outline 29 a'. aller guten Milch [ all good weak milk fem ] DAT b. alles gute Bier [ all good weak beer neut ] NOM b'. allem guten Bier [ all good weak beer neut ] DAT It turns out that this quantifier accompanies weak inflected adjectives across the singular/ plural distinction as well as all case forms and can hence be categorized accordingly, together with the determiner and demonstratives. Notice that once again no distinction can be drawn for the adjectival form in (16) a., since, as indicated in the gloss, the two patterns are homonymous. With dative case, however, identification of weak inflection is obvious. I hence feel justified to once again generalize over the ambiguous form due to the weak status of (16) a'. as well as the uniform pattern in (16) b. and b'., which does not show a deviance from the weak inflectional pattern, and return to this property of the feminine category in length in chapter two below. Returning to the group of quantifying items laid out above, let us concentrate on jed - ‘every’. Since this element combines exclusively with singular NPs, the combination with mass nouns parallel to (14)-(16) does not pose as a syntactic peculiarity as it does for the elements under consideration above. 17 However, we do find nominal elements whose morpho-syntactic featural make-up allows for the inversion of the test criterion employed above, namely the combination with a plural HN: Pluralia tantum nominals display said value irrespectively of their semantic number. Among these, a subpart combines with ‘every’, as in (17): (17) a. jede großen Ferien [ every big weak vacation pl ] NOM a'. jeder großen Ferien [ every big weak vacation pl ] GEN b. jede hohen Kosten [ every high weak cost pl ] NOM b'. jeder hohen Kosten [ every high weak cost pl ] GEN 17 These combinations, however, do result in a semantically specific but uniform reading of the HN in question, i.e. generic/ kind; cf. ch. 4.3.2 a below. 30 1 AXIOMS - Adjectival Inflection & Phrasal Categorization Notice that parallel to the feminine paradigms presented in the previous discussions on the combination of plural quantifiers with mass nominals above, the dative forms of the adjectives in the plural paradigm are likewise homonymous between the strong and the weak pattern. In (17) a'., b'. above, I have hence switched to the second lexical case in German, viz. genitive: There, a contrast in inflectional form is once again observable, i.e. weak: en , strong: er . Therefore, this element can be unambiguously classified to uniformly co-occur with weak inflected adjectives, in line with the determiner, demonstratives as well as all -. This leaves us with the classification subsumed in table (18): (18) strong weak mixed (dimension of case) uniform ø, ‘many’, few’ Det, Dem PROX/ DIST , ‘all’, ‘every’ pl sg varying with φ (dimension of number) Poss, Neg Poss, Neg, ‘a’ This section developed a categorization of morphologically simple left-peripheral elements in the nominal domain in German, according to the adjectival inflection accompanying them. Two grammatical dimensions have therein been identified to further divide the set of items, uniformly concatenating with strong and weak inflected adjectives, namely case and number: the former separating the mixed inflectional pattern from the uniform ones; the latter splitting elements of said pattern further, with the mixed division now solely applying in the singular, the corresponding plurals aligning with elements, categorized as weak, once - as has been done here - number-sensitive restrictions on SLI concatenation have been eliminated. Therefore, while categorizations in the uniform classes can be considered stable over variations of φ-features and case, elements corresponding to the mixed pattern observe switches in adjectival inflection w.r.t. changes in said values. Moreover, it is only in this category that we find at least some morphological uniformity in their shared root morpheme ein -. The following section, then, sets out to align the concepts and findings of the last two subchapters: Morphological definiteness is a syntactically relevant feature of German; adjectival inflection is dependent on the co-occurrence with left-peripheral items. We hence strive to derive the dependence of the latter on the status of the former. 1.3 Basic Concepts - An Outline 31 1.3.3 Morphological Definiteness Coincides with Adjectival Inflection We have seen above that the elements accompanying strong and weak adjectival inflection do so independently of the φ-featural specifications of the nominal with which they are concatenated (with the exceptions themselves a uniform class w.r.t. their morphological set-up, i.e. the root morpheme ein -). Recall from section 1.3.1 that the status of definiteness of a nominal domain relates to various syntactic restrictions, the best-known one being the exclusion of definite NPs from the postcopular logical subject position in existential sentences (henceforth ES). This fact is captured in the ‘Definiteness Restriction’ in Milsark’s (1974, 1977) influential work on the classification of noun phrases in English. His analysis therein concentrates on a language which does not display varying adjectival inflectional patterns. Still, his division of the elements under consideration, depending on their grammaticality in ES-contexts, is identical with the bisection of the categories strong and weak in (18) as carved out in the last section: 18 (19) (= Milsark 1977: (13), [reduced]) WEAK STRONG a “definites“ φ plural and mass the determiners in demonstratives nonuniversal reading possessive DET’s universals all every The division in (19) is a condensed version of Milsark’s original table. I will come back to the categories excluded for now, below. Notice that the second 18 A word of caution concerning terminology is in order here: Recall that from section 1.3.1 onwards, the inflectional forms on adjectives, not the elements preceding it, were labeled strong and weak respectively. The terminology employed by Milsark, though, does categorize these items themselves. Matters are complicated further by the fact that strong and weak in the sense of Milsark corresponds to the inversion of the elements evoking strong and weak adjective inflection. I will try to keep the separation of these different assignments of the terms as lucid as possible by referring to the respective forms of adjectival inflection as well as the full patterns resulting therefrom in small caps based on (18) while addressing Milsark’s division in lower case writing hereafter. Observe furthermore that the label ‘ mixed ’ can only refer to the second of these uses, i.e. the full paradigm, due to its composition of strong as well as weak adjectival inflectional forms. 32 1 AXIOMS - Adjectival Inflection & Phrasal Categorization instance of his class of weak elements actually designates bare plurals and mass nouns and hence parallels ‘ø’ in (18). Since the division from section 1.3.2 was drawn on the basis of inflectional properties of adjectives in the domain of NP, while (19) is based on clausal environments, ‘ø’ was categorized as strong without reservations concerning the φ-features of the respective nominal (i.e. as uniformly strong irrespective of the number of HN). Singular NPs not headed by any overt SLI are indeed also barred from occurring in argument position in German (even beyond ES-environments). 19 A question to ask here, however, is to what extent this fact can be accounted for in terms of (in-)definiteness, since the other instance of Milsark’s class of weak elements in (19), ‘a’, readily licenses argumenthood. I nevertheless feel justified to take ES-environments and adjectival inflection in German to be sensitive to the same syntactic feature, morphological definiteness, with the latter employing a slightly broader testing ground since it is not dependent on a sentential setting which prohibits singular bare nouns for reasons not immediately related to the concept under consideration. To continue the analysis, Milsark understands his classes of strong and weak noun phrases/ items to represent a more ample division than the status of definiteness (although the latter is subsumed therein). As he notes (Milsark 1977: ch. 2), the ability to transformationally derive ES-constructions from corresponding simple predicational sentences is dependent on the semantics of the predicate, with those denoting time-stable properties barring said application, those denoting short-living states allowing it (the individualvs. stage-level distinction, e.g. tall , intelligent vs. drunk , sick ). These two classes are labeled property (P) and state-descriptive (SD) predicates accordingly while the effect itself is subsumed under the term ‘Predicate Restriction’. From this, another testing ground for the status of nominal domains is uncovered: In noting that the simple predicational sentences, transformationally related to the aforementioned constructions, underlie the same constraints in the context of weak subject NPs, Milsark (1977: (33), (34)) derives the principle that P-predicates are restricted to concatenation with strong and universally quantified NPs. Therein, the Predicate Restriction reduces to the clash between said principle and the Definiteness Restriction, applying to transformationally derived ES-constructions thereof: While weak 19 Cf. Longobardi (1994) who argues on a structural basis that bare nominal projections (i.e. NPs) cannot be interpreted referentially and hence not pose as arguments, since it is the “denotational interpretation of the DP structure” (Longobardi 1994: 634) that is related to argumenthood. Hence, a “‘nominal expression’ is an argument only if it is introduced by a category D” (Longobardi 1994: 620). However, since his argument relies on narrow-syntactic proposals, grammatical cases of overtly bare nominal heads in argument-position in German, like mass generics, proper names or bare plurals, as mentioned in the main text, have to be interpreted to bear a covert DP-projection (cf. Longobardi 1994: ch. 8). 1.3 Basic Concepts - An Outline 33 NPs cannot concatenate with P-predicates, strong NPs cannot partake in the establishment of ES. Therefore, the full class of P-predicates is excluded from the establishment of ES. In what follows, I will shift Milsark’s term and refer to the aforementioned principle as well as the test environment resulting therefrom as the Predicate Restriction. As has been laid out, the test goes the other way around with respect to the dichotomy of (19) and the Definiteness Restriction. Here, weak NPs are excluded: (20) ([glosses added]) a. Everyone was drunk. (= Milsark 1977: (31) b.) strong SD b. A man was drunk outside. (= Milsark 1977: (29) b.) weak SD c. Everyone was intelligent. (= Milsark 1977: (32) b.) strong P d. * A man was intelligent. (= Milsark 1977: (30) b.) weak P Given that strong and weak are non-intersecting categories, no element should occur in both test environments. There is, however, a class of elements which at first seems to contradict this prediction. Recall that the table in (19) does not include all items from Milsark’s original partition as well as from (18). Among others I excluded ‘sm’ from his representation, an instance of the class of weak elements. This item, represented by a reduction of the caption ‘some’, is associated with one of two possible readings of the weak quantifier: 20 While the reduced written form refers to “some indeterminate but probably not large number” (Milsark 1977: 18), the articulated capitation carries an additional contrastive, partitive implication. The typographic distinction hence mimics stress assignment in speech, but as Milsark himself notices, this is not a foolproof criterion since information-structural factors can intervene. The distinction, however, is stable enough for him to base the partition of the weak quantifier upon it. The two readings properly pattern with the environments presented above: (21) a. * There are some men in the garden. a'. There are sm men in the garden. b. Some men are intelligent. *3 b'. * Sm men are intelligent. 20 As also noted by Milsark, the notational convention as well as the bipartition underlying it date back to Postal ([1966a] 1970: fn. 7). 34 1 AXIOMS - Adjectival Inflection & Phrasal Categorization The distribution of non-intersecting sets of strong and weak elements can therefore be maintained. As Milsark demonstrates with ‘many’ (and ‘mny’ respectively), a similar division can be deduced for all weak quantifiers, with a strong reading equal to the substitution of the quantifier with the more complex ‘Q of the N’ (i.e. partitive), yielding an interpretation in the sense of ‘Q (but not others)’ (i.e. contrastive). He nevertheless notes that the weak readings constitute what he calls the “more normal” (Milsark 1977: 20) interpretation of the elements in question. Incorporating the elements under consideration into the table in (19) hence leaves us with (22): 21 (22) (= Milsark 1977: (13), [modified]) WEAK STRONG a “definites“ φ plural and mass the determiners in demonstratives nonuniversal reading possessive DET’s universals all every sm / mny / fw some / many / few With this distinction brought to attention, let us turn to those quantifiers accompanying weak adjectival inflection in German. Here, I will concentrate on those two instances incorporated in the discussion from section 1.3.2 onwards, namely viel - (‘many’/ ‘mny’) and wenig - (‘few’/ ‘fw’). 22 As demonstrated therein, we do already possess some evidence to categorize them as indefinite, since these elements do precede adjectives with strong inflection (and hence pattern 21 See e.g. Musan (1999) and references therein for a different typology of nominal SLIs: Building on Milsark’s original division, she bases her analysis on semantic criteria (i.e. diverging presuppositional properties) and subsequently divides the weak reading of quantifiers (‘cardinal reading’) from their strong (‘partitive’) ones, which are grouped together with strong quantifiers, overt partitive constructions and what she calls ‘definites’, i.e. NPs headed by determiners. Therein, she sets out to capture the temporal (in-) dependence of noun phrases. For the current purpose, however, the bisection elaborated in Milsark (1974, 1977) is convenient. 22 The closest semantic equivalent to ‘some‘/ ‘sm’ in German, ein paar poses additional problems in that it is in itself a complex element whose initial subpart is frequently reduced to a nasal onset in speech, resulting in n’paar , quite analogous to the intuition of Postal ([1966a] 1970). There, stress assignment does not seem to facilitate the same contrast that Milsark reports for English quantifiers. I will, however, limit myself here to the morphologically simple quantifiers, cf. ch 4.3.2 below. 1.3 Basic Concepts - An Outline 35 with ‘ø’), a test procedure not available in English (recall the discussion following (19) above). In turn, matters of stress reduction do not seem as decisive in the domain of quantifiers in German as they are in English (cf. fn. 22 above). In compliance with the title of his article, Milsark (1977: 16) himself restricts the application of said principle to this one language. Indeed, the resulting test environment does likewise not readily carry over to German, where concatenations of weakly quantified nominal domains with P-predicates in out-of-the-blue (i.e. zero) contexts does not readily result in ungrammaticality. Nevertheless, one might approach the Predicate Restriction with the paraphrastic substitutions form Milsark (1977), mentioned above. Consider (23), in which the postcopular subject is headed by a weakly quantified noun phrase (in the sense of Milsark) without and with a partitive structure accompanying it: (23) a. Es gibt {viele/ wenige} Bäume im Garten. it is { many / few } trees in.the garden ‘There are many trees in the garden.’ b. * Es gibt [{viele/ wenige} {[der Bäume]/ [von den Bäumen]}] im Garten. it is [{ many / few } {[ the trees ] GEN / [ of the trees ] DAT }] in.the garden ‘There are {many/ few} of the trees in the garden.’ Here, the curly brackets indicate interchangeability. The genitive and dative PP-partitives in (23) b. correspond to the ‘some’ reading in English (i.e. the contrastive ‘Q (but not others)’ reading, laid out above). These are barred from existential contexts, while the simple weak quantifier results in a grammatical structure in this position. With this in mind, let us next turn to the Predicate Restriction. (24) a. {Viele/ Wenige} Männer waren betrunken. { many / few } men were drunk (SD) ‘{Many/ Few} men were drunk.’ a'. [{Viele/ Wenige} {[der Männer]/ [von den Männern]}] waren betrunken. [{ many / few } {[ the men ] GEN / [ of the men ] DAT }] were drunk (SD) ‘{Many/ Few} of the men were drunk.’ b. * {Viele/ Wenige} Männer waren intelligent. { many / few } men were intelligent (P) ‘{Many/ Few} men were intelligent.’ 36 1 AXIOMS - Adjectival Inflection & Phrasal Categorization b'. [{Viele/ Wenige} {[der Männer]/ [von den Männern]}] waren intelligent. [{ many / few } {[ the men ] GEN / [ of the men ] DAT }] were intelligent (P) ‘{Many/ Few} of the men were intelligent.’ The datum (24) b. turns grammatical under an unintended generic reading. However, if one cautiously evokes the weak reading in contrasting the string with (24) a'./ b'., the ungrammaticality becomes apparent. Nevertheless, in what follows, I will acknowledge Milsarks’s own confinement to English and dismiss the Predicate Restriction as a reliable test for the strong/ weak status of nominal SLIs in German. What is important at the current point is the simple fact that the contrast is reproducible, regardless of its strength. 23 Let us finally take a look at the third sentential environment sensitive to the distinction at hand, also initially presented in chapter 1.3.1 above, namely the preposed adjective structures discussed in Postal ([1966a] 1970). Once again, semantic reanalysis of the subject NP as generic or partitive readily rescues the otherwise ungrammatical datum in (25) a. below. (25) a. * Betrunken wie viele Männer sind, reden wir ungern mit ihnen. drunk as many men are, talk we reluctantly with them ‘Drunk as many men are, we only talk reluctantly with them.’ b. Betrunken wie [viele {[der Männer]/ [von den Männern]}] sind, reden wir ungern drunk as [ many {[ the men ] GEN / [ of the men ] DAT }] are, talk we reluctantly mit ihnen. with them ‘Drunk as many of the men are, we only talk reluctantly with them.’ As can be derived from (23)-(25), complex partitives headed by weak quantifiers in German uniformly pattern with strong SLIs in the sense of Milsark concern- 23 Still, the distinction seems to bear some weight in German, too: Contrasting the internetsearch hits of the appropriate weakly quantified strings on the basis of Milsark’s (1977: (24)) own list of P-predicates, e.g. the search string “die * waren intelligent” (‘the * were intelligent’, with the quotation marks facilitating the appropriate order of the items inside the search term and the asterisk an operator creating a slot for any lexical material) and its universally quantified derivative returned 617.000 and 479.000 hits respectively; the contrasting search term headed by the weak quantifier viel returned only 10 hits, of which 7 had to be subsequently rejected (since the open slot was filled by SOs, other than simple nominal domains, i.e. four partitives, one PP, one possessive and an adverb). Note, however, that a clear-cut contrast is not observed with all of Milsark’s P-predicates, e.g. groß (big) and leicht (light) do not exhibit such a landslide contrast in the aforementioned environment. 1.3 Basic Concepts - An Outline 37 ing their grammaticality in the respective sentential environments, while the simplex forms do so in accordance with weak items. Additionally, turning to adjectival inflection, we gain conforming evidence: (26) a. viele große Männer [ many big strong men ] NOM a'. vieler großer Männer [ many big strong men ] GEN b. viele der großen Männer many NOM [ the big weak men ] GEN b'. vieler der großen Männer [ many the big weak men ] GEN Taking the preceding discussion into account in classifying SLIs in German, one notices that the tests for (in-)definiteness from Milsark (1977) result in an identical division of elements. Two conclusions are to be drawn from this: First, the simple German strong -class quantifiers in (18), weak quantifiers viel -/ wenig in Milsark’s terminology, carry a weak default reading. Even if semantic reanalysis might be able to rescue certain structures under consideration above, this conclusion is backed-up by (i) the ungrammaticality, once the intended reading is isolated and (ii) the fact that the ambiguity ceases to exist with the corresponding partitives. Second, as was already suggested above, variation of adjectival inflection in German is sensitive to the same syntactic feature as the Definiteness Restriction and preposed adjective structures as well as (to a lesser intensity) the Predicate Restriction, i.e. definiteness. *4 The former - as Milsark argues - can be explained once we focus on the semantic set-up of the elements comprising the class of weak items: He notes that these items only ever denote a set of entities referred to by the HN they are combined with. Therefore, they can be translated into a number expression without regard to the absolute number of the set denoted by HN. The strong/ weak distinction hence turns out to divide ‘real quantifiers’ from ‘cardinality words’. On this view, ‘some’ can be translated into “a real quantity, greater than one, but not especially large” (Milsark 1977: 23), whereas strong (universal) items like ‘all’ are only understood in reference to the absolute number (cardinality) of the denotation of HN. Therein, these elements presuppose the existence of the set denoted by HN. This intuition covers the universal quantifiers (‘every’, ‘all’), partitive interpretations (‘{some/ many/ few} of the N’) and 38 1 AXIOMS - Adjectival Inflection & Phrasal Categorization the item ‘most’ in the category of strong elements. 24 If, in turn, ES-structures are understood as an expression of (semantic) existential quantification, strong items in postcopular subject position result in two instances of quantification on HN, a semantically anomalous structure. These properties, then, make up his division of strong and weak noun phrases/ items and constitute the basis for both restrictions, laid out above. Finally turning again to the split in definiteness into the semantic and syntactic component, laid out with reference to Lyons (1999) and Roehrs (in prep.) in chapter 1.3.1 above, the separation arguably also becomes visible in the test environments discussed in length above. Observe (27) below, in which a nominal domain headed by the proximal demonstrative (categorized as strong in Milsark’s sense above) occurs in an ES-configuration in out-of-the-blue, i.e. wide-scope context: (27) Es gibt (da) dieses neue Kind in meiner Klasse. it is ( there ) this new weak kid in my class ‘There is this new kid in my class.’ The datum is modelled after Etxeberria & Giannakidou (2009: (110) c., cf. also (102) b.) and judged grammatical in introducing the new kid into the discourse. Note that even though the authors refer to the equivalent DPs in their examples as “specific indefinites” (Etxeberria & Giannakidou 2009: 39), weak adjectival inflection identifies them as morphologically definite in German. Parallel to bare singular nouns at the beginning of this section, we hence have come across another case of contradictory results of the two sensitive environments, this time allowing a morphologically definite SO in ES-contexts in contrast to barring a morphologically indefinite SOs. In the light of the above disjunction, one might speculate on the sensitivity of these environments for the differing components, standardly overlapping in German but diverging in cases like the one above. Interestingly, the demonstrative moreover surfaces in another exceptional reading of the nominal hierarchy, namely in introducing kind readings: (28) Diese Menschen mit Hunden mag ich nicht. these people with dogs like I not ‘I don’t like those people with dogs.’ 24 With reference to Chomsky (1975), Milsark also re-evaluates the definite determiner as a universal (singular) quantifier, hence further unifying the elements comprising the strong class under the new notion ‘universals’. 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 39 Howsoever, the deviant readings of these complex nominal domains are moreover accompanied by a deviant stress pattern in that the SLI cannot bear stress in both cases. Thus, stress assignment does bear an import to the semantics of this SLI in German. I will not go into detail on these matters here but continue to focus foremost on adjectival inflection as the reflex of morphological definiteness in what follows, since this effect seems undisrupted by semantic variations. Combining the terminology and findings from the last three sections, then, leaves us with the simple yet powerful syntactic biconditional depicted in (29) below, coupled with the classification of elements in (18), now re-categorized as their status of definiteness. 25,*5 (29) SLI strong ↔ Adj WEAK SLI weak ↔ Adj STRONG As has already been done in this section (cf. fn. 18 above), I will use the terms ‘strong’ and ‘weak’ to specify the status of the semi-lexical items triggering weak and strong adjectival inflection respectively, throughout this book (i.e. following Milsark’s definition), if not specified otherwise. Equipped with these observations, let us next turn to the second topic of this introductory chapter (and axiom of the analysis to follow): the basic linearizational properties of German noun phrases and the composition of the phrasal configuration underlying them. 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German The current subchapter sets out to derive a phrasal cartography of the functional domain of the nominal hierarchy of projections. The aim therein is to provide the simplest set-up possible to account for all linearizational idiosyncrasies in German. I will start out with the head noun as the only identifiable, fixed category. This rationale is exemplified below: (30) [(…) NP (…)] Taking (30) as the vantage point, I will subsequently concentrate on various classes of semi-lexical items in the nominal domain over the course of the ensuing subchapters and propose the structural positions of the respective phrases 25 I again abstract away from the SLIs triggering mixed inflection here; see endnote *2. 40 1 AXIOMS - Adjectival Inflection & Phrasal Categorization in relation to the nominal core as well as to one another in what follows. The schematic representation depicted above will thereby be extended accordingly. As I devoted considerable space on the categorization of quantifying LIs in the last section, I will start out by analyzing quantifiers in German before turning to the remaining classes, namely determiners and demonstratives as well as possessives and pronouns. 1.4.1 Quantificational Elements Up to now, it has been demonstrated that quantifiers in German do not only divide into two classes w.r.t. them triggering strong vs. weak inflection but that they also cluster accordingly regarding the status of grammaticality in certain sentential environments, namely the postcopular subject position in ES (the Definiteness Restriction: weak quantified NPs are licit, while strong quantifiers and interpretations result in ungrammatical structures) and the preverbal subject position in copula-constructions with property predicates (the Predicate Restriction: strong quantifiers and interpretations can precede HN, while weak quantifiers in non-partitive interpretation are barred from heading the noun phrase). With this in mind, let us turn to another environment, namely linearizations inside the noun phrase. Consider (31), in which a complex NP composed of an adjective and HN is headed by a strong and weak quantifier respectively: (31) a. alle groß-en Häuser [ all big weak houses ] NOM b. viele groß-e Häuser [ many big strong houses ] NOM Apart from adjectival inflection (i.e. weak in a., strong in b. respectively), there is no divergence observable between the two strings on the surface. If, however, a definite determiner is added to the LA, we obtain (32) a. and b. for the linearizations in (31) respectively. (32) a. all die groß-en Häuser [ all the big weak houses ] NOM b. die viel-en groß-en Häuser [ the many big weak houses] NOM Several comments are in order here. First - and most importantly at the current point - notice that the determiner surfaces on different positions w.r.t. the 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 41 quantifier in question. 26 This is the key point I want to demonstrate here and I will elaborate on this right away. Before doing so, let me bring one additional effect to attention which goes along with the advent of the determiner. Note that the morphological shape of both quantifiers has changed from (31): While all displays a reduced form, missing the final vowel, the weak quantifier shows a different inflectional ending (n ). It is now homonymous with the weak adjectival inflection in (32), which is also observable in b., in accordance with the findings of the last subsection. I will return to this observation in chapter 3.3.3. Let us for now concentrate on the linearizational differences. Closer inspection on ‘Recursive Quantification’ 27 reveals that the diverging positions indeed pattern with the strong/ weak quantifier distinction, elaborated above: 28 (33) all die {vielen/ wenigen} Häuser [ all the { many / few } houses ] NOM Apart from inflectional and distributional divergences, we have uncovered a third, i.e. linearizational, difference between these classes of quantifying elements. In what follows, I will assume that quantifiers are heads of their respective quantifier phrases (QPs) outside the projection of the HN proper. There exist various instantiations of this rationale, ultimately dating back to Sportiche (1988), who takes these elements to be adjoined to the maximal nominal projection, deriving floated quantifier constructions basically as movement of HN which strands the Q in base position. Following Sportiche and focusing on quantifier float from VP-internal arguments in German, Giusti (1990) grants head status above DP to the elements under consideration. Further proposals in this spirit can be found in Shlonsky (1991), mainly on Hebrew, and Cardinaletti & Giusti (1992) on Italian. The focus on different kinds of quantifying elements, however, brought about diverging instantiations of said head/ phrase. Beginning with Hall (1962) and Dean (1966), as summarized in Jackendoff (1968), the dedicated position for quantifying elements (in prepositional partitive constructions) is located inside the determiner position either by simple insertion as in the latter, or by a split of said position which opens up another terminal node 26 This divergence causes Vater & Reis (1980) to locate alle in the so-called pre-determiner position. 27 The term is due to Jackendoff (1968), who - to my knowledge - was the first one to discuss the phenomenon of multiply quantified noun phrases in an accessible publication. 28 Recall from the discussion in section 1.3.2 that the second strong quantifier jed- takes morphologically singular HNs as its complement, which makes it incompatible with the combination of strong all as well as the weak quantifiers. This is not the only distortion concerning this lexical item, however. I will therefore devote some time to its peculiarities in chapter 4.3.2 a. below but leave it out of the discussion for now. 42 1 AXIOMS - Adjectival Inflection & Phrasal Categorization dubbed ‘prearticle’ or ‘predeterminer’ in the former. As far as can be deduced from Jackendoff’s discussion, quantifying elements are treated as a homogeneous class in these early accounts. Additionally, both accounts take prepositional partitives to pose as the underlying structure of all quantificational constructions and deal with simple quantifier-HN configurations by an additional rule, deleting the preposition (which shares the predeterminer position with the quantifier in the case of Hall’s analysis). Deletion rules are moreover obligatory in Dean’s account even for partitive structures, since the Det-position occupied by the quantifier primarily relates to another instance of NP (the head noun in the terminology used above), identical to the noun in the partitive PP, which is elided from the structure under identity with the head of the partitive NP. Thereby, Dean derives two distinct Det-positions surrounding the preposition (cf. ch. 4.3.1 b., ch. 4.3.2 a. below). Jackendoff himself distinguishes three groups of quantifying elements mainly on the basis of their co-occurrence with a preceding determiner and (obligatory vs. optional) (in-)definite partitive PPs. Of those, Group II and III subsume all elements under consideration here. 29 Because of the relation to the determiner mentioned above, elements of Group III are categorized as occupying N 0 while members of Group II are understood to constitute articles in Det. Jackendoff further distinguishes between - what he calls - ‘uses’ of these items: pronominal when co-occurring with a partitive PP, adjectival in simple quantifier-HN concatenations; the latter being derived from the former by application of of -deletion. 30 In his influential work on phrase structure, Jackendoff (1977, ch. 5) grants categorical, hence head status to a subset of these elements, abbreviated as Q. 31 The co-occurrence with D 0 is still a critical testing ground for his categorization: Qs are generated in the N’’ specifier and therefore may follow determiners (which occupy the N’’’ specifier) while other quantifying elements are base-generated in N’’’ (i.e. Art) themselves and are hence in complementary distribution with determiners and demonstratives. Their morpho-syntactic featural compositions only differ in a single value, viz. [± Det]. As can be deduced from the preceding discussion, the categorization of quantifying elements started out by associating these items uniformly with the cat- 29 Group II: some , each , few , which , all , both ; Group III: a few , many , one , three (relevant items underlined). 30 Jackendoff, however, notes certain divergences, which can be associated with the strong/ weak distinction. 31 N’’’ specifiers: the , this , that , these , those , which , what , a , singular some (i.e. sm in Postal’s sense, preceding a singular HN); N’’ specifiers/ Qs: each , every , any , all , no , many , few , much , little , some (relevant items underlined). Nothing has to be said about Jackendoff’s third class of numerals here. 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 43 egory of determiners (due to their general syntactic distribution) and partly with adjectives (due to their modifying semantics and to particular syntactic distributions) before postulating a discrete category/ phrase QP. Combining these approaches (Q as D 0 / A 0 , Q as the head of its own phrase) with the rationale of splitting the class of lexical items due to the syntactic and/ or semantic properties of certain subgroups, we gain an understanding of QP in more recent (i.e. post-Milsark) analyses like those mentioned at the beginning of this section. The questions (i) which elements can be assumed to occupy the head of a discrete quantifier phrase (often associated with the more general question of obligatory realization in the nominal domain), (ii) how to categorize other quantifying elements (D 0 , A 0 or even N 0 ) 32 and (iii) how such approaches might combine, have dominated generative approaches to quantification ever since. 33 I will concentrate on the division of accounts which grant quantifiers decided head status in the nominal domain and those associating them with the lexical category of adjectives in what follows: Cardinaletti & Giusti (2006, henceforth C&G) subsume these under the terms QP-Hypothesis 34 and AP-Hypothesis respectively. Their discussion arrives at the following listings of accountable idiosyncrasies of quantifiers by the two approaches: (34) a. QP-Hypothesis A diagnostics for quantifiers (= C&G: (67)) i. Q can precede definite as well as indefinite articles. ii. Q can co-occur with pronouns. iii. Q allows leftward movement of the DP to an immediately adjacent position which builds a constituent with Q and can straightforwardly be analyzed as [Spec, QP]. iv. Different classes of Q correlate with the (in)definiteness of the embedded DP. […] v. Q can occur in discontinuous position in a number of languages, contrary to other nominal modifiers and determiners. 32 An approach incorporating all three categories on the basis of Giusti’s (1991 et seq.) and Cardinaletti & Giusti’s (1992) work can be found in Giusti & Leko (2005, 1996), based on data from Slavic languages. 33 See e.g. Giusti’s (1991: ch. 1) discussion of the possible approaches sketched above, exemplified through the accounts of Abney (1987, Q as a modifier, i.e. adjectival non-head in the nominal domain, but cf. fn. 36 below), Sportiche (1988, Q as an adjunct to NP) and Belletti & Rizzi (1981, Q as a head of DP). 34 The term is due to Pesetsky (1982, ch. 2) who proposes the variable categorical status of noun phrases depending on the occurrence of quantified elements. Giusti & Leko (2005) point out the obvious shortcomings of this proposal in today’s rigid phrasal framework. 44 1 AXIOMS - Adjectival Inflection & Phrasal Categorization b. AP-Hypothesis A diagnostics for quantity adjectives (= C&G: (87)) i. Quantity adjectives are preceded by a determiner (in languages that have a definite article). ii. Quantity adjectives cannot occur with a pronoun. iii Quantity adjectives do not allow extraction of the part of the DP that does not contain it. iv. Quantity adjectives cannot appear in discontinuous position (in languages in which adjectives cannot do so). v. Quantity adjectives do not select a partitive PP. A wide variety of cross-linguistically attested properties of various quantifying elements are ordered in the bipartition above. The authors however explicitly state that the virtual overlap of the typology in (34) with the strong/ weak distinction, as elaborated in chapter 1.3, in English is merely accidental and implicate that the categorical status of quantifying elements (hence their external merging-site together with the associated syntactic characteristics) might vary between languages. I will get back to this bipartition below. Giusti (1991) discusses the syntactic position of quantifying elements with respect to the determiner mainly on the basis of English and Italian data. She focuses on different uses/ functions of quantifiers parallel to adnominal and predicative adjective phrases in the case of weak Qs and numerals. 35 Additionally, she also concentrates on selectional restrictions of lexical items to derive a phrasal set-up incorporating a single quantifier head dominating DP or NP, depending on the syntactic object at hand. Therein, Giusti contrasts the Milsarkian classes of strong and weak quantifiers, exemplified by few / many and all . Parallel to Jackendoff’s focus on the use/ function of the latter, which she calls ‘indefinite quantifiers’, the (un)grammaticality in several environments is translated into adjectival (i.e. modifier) vs. quantificational function, which itself coincides with Spec vs. head status. Her typology thus connects several proposals, elaborated above: All quantifiers are canonical heads of QP in the nominal hierarchy of projections; strong and weak quantifiers differ in the selectional properties of their complement. The latter can occur in adjectival/ modifier position, depending on use/ function, the former associates with the category D in the property of transmitting case. Her discussion moreover touches on a phrasal set-up incorporating two discrete quantifier phrases for recursive quantification, encircling DP. She however dismisses the idea (which is understood as an expansion of approaches of variable position of QPs in the nominal domain) on theory- 35 Giusti herself notes that the parallelism between weak Qs and Adjs is not perfect, though. 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 45 internal arguments. Still, in her account, the diverging behavior between (in-) definite (strong/ weak) quantifiers are partly handled by a non-uniform nominal hierarchy of projections, expressed in the selectional properties of the individual items. Furthermore, the dismissed rationale of varying syntactic positions reenters her approach in the proposal of uses/ functions of weak quantifiers coinciding with specifier (AP) vs. head (Q 0 ) status. In what follows, I will maintain the rationale of two discrete phrases for strong and weak quantifiers in the functional domain of the noun, call them Q S P and Q W P respectively. This is for now not at all an elegant approach and at least subject to the criticism laid out by Giusti (1991). Note, however, that such phrasal configurations have been put forth more recently, 36 the most prominent instantiation being Zamparelli’s ([1995] 2000) division of ‘Strong’ versus ‘Predicative Determiner Phrase’ (SDP and PDP respectively). He proposes a three-layered DP-domain mainly on the basis of semantic properties and interpretations of lexical items (based on this rationale, KiP as the host for kind-interpretations completes the tripartition as the most embedded phrase). The semantic origin is responsible for certain core assumptions of Zamparelli’s system, which are not readily derived under a syntacto-centric view: Note, first, that the classical determiner-level (DP) is split into SDP and PDP, which may house a variety of categories in their head-positions (among others determiners and demonstratives, personal pronouns as well as possessive elements), with SDP hosting all strong and referential uses/ functions (i.e. ‘interpretations’ in the 36 The origin of this approach can be found in Abney’s (1987) influential work on the hierarchy of projections in the nominal domain, if, as C&G observe, his treatment of Det and Deg[+Q] elements is to be understood as complementary to one another. More generally, as the authors note (p. 50), the account under consideration underlyingly constitutes a representative of the AP-Hypothesis despite the head status of quantifying elements: In Abney’s syntactic work (see also Bernstein 1993), prenominal adjectives might occupy functional head positions; strong quantifiers, however, are treated in parallel to determiners, following Milsark’s (1977) predominantly semantic analysis, cf. also Abney’s (1987: ch. 4.1.1 c.) treatment of strong every and each as heads of DP. A parallel structural bipartition can moreover be deduced more recently from Etxeberria & Giannakidou’s (2009, 2010; cf. also Etxeberria 2009, 2012) analysis of Q-Det interactions across various languages, to be made precise below in chapter 4.3.2 b. Therein, the authors first detach the strong quantifier (together with the focus particle nur ‘only’; cf. ch 3.3.2 below) from said lexical class (cf. Etxeberria & Giannakidou 2009: ch. 2.3.1, 2010: ch. 2.2.1, Etxeberria 2009: ch. 4.2.1, 2012: ch. 3.2 and references therein respectively, moreover cf. Brisson 1998: ch. 1.2.1, ch. 5), to be recategorized as (adverbial) modifiers of DP, before locating weak quantifiers as adjectives or cardinality predicates (with Milsark 1974; cf. also ch. 1.3.3 above) in a low number phrase NumP (Etxeberria & Giannakidou 2009: ch. 4, 2010: ch. 4, Etxeberria 2009: ch. 4.4.3, 2012: ch. 6), to be associated with an existential quantifier in Q 0 in a subset of its uses/ functions. 46 1 AXIOMS - Adjectival Inflection & Phrasal Categorization sense elaborated above) of these. 37 Thus, the system at hand dissociates lexical category from corresponding dedicated phrasal status. Following this rationale even further, these items may variably surface as SD 0 or PD 0 . Therefore, certain co-occurrences of semi-lexical elements are ruled out by a blocking effect of the items competing for, and occupying, positions associated with fixed interpretations in specific environments. To illustrate this matter, consider (35) a. below, which is ambiguous concerning - what Zamparelli calls - (in-)definiteness (since paraphrases of the possessive relation in relative clauses can be headed by weak as well as strong elements): the exhaustive/ uniqueness presupposition of the predicative possessor. Note that the first possessive is ambiguous between reference to the exhaustive, i.e. complete set of tools bearing a relation to Harrold and some subpart of said set. Therefore, the following introduction of a second unrelated set with an identical relation to the aforementioned possessor can cancel one of the available readings without resulting in ungrammaticality of the complete sentence structure. The aforementioned ambiguity of the first possessive however ceases to exist with the occurrence of the numeral in (35) b., which can only refer to the unique, i.e. exhaustive set of all tools bearing a relation to Harrold, consisting of four parts. This hence bars the addition of another such set which turns the datum ungrammatical. 38 (35) (= Zamparelli [1995] 2000: (354) a.) a. These are Harrold’s tools, and those, too, are Harrold’t [sic! ] tools. (= Zamparelli [1995] 2000: (355)) b. * These are Harrold’s four tools, and those, too, are Harrold’s four tools. In the system outlined above, this contrast receives a straightforward explanation in that the lexical possessive in (35) a. can be interpreted as hosted in either SDP or PDP, while the numeral (an exclusively weak element in Milsark’s typology) necessarily occupies the head position of PDP in (35) b., thus leaving only SDP as the structural position and locus of interpretation for the possessive. This is in itself an elegant approach, since we have likewise already encountered several simple elements with possible strong and weak interpretations (cf. (22) in 1.3.3), apart from the more complex cases just sketched. The dissociation of syntactic categories is hence accompanied by a unification in semantic composition in Zamparelli’s account. 37 As Zamparelli himself notes, the equal treatment of determiners and personal pronouns and the association of referentiality apart from HN inside the nominal domain date back to Postal’s ([1966a] 1970) work, discussed above. I will get back to the first point in chapter 1.4.4 below. 38 All citations are taken from the 1995 dissertation. 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 47 Concerning recursively quantified NPs like the structure presented in (33) above, however, the multi-layered system runs into severe problems. There, we find two quantifiers surrounding the definite determiner. 39 The latter must therefore be understood as the prototypical instance of an (exclusive) SD 0 element in a split-DP-approach with reference to the semantic criteria laid out above. Therefore, the only structural position available for the universal quantifier is the specifier of said projection. Still, as Zamparelli (referring to Giusti 1992) himself notes, this element should be granted head status in the overall spirit of the account as well as on parallel grounds with the treatment of weak quantifiers. 40 Those, in turn, necessarily represent instances of PD 0 in structures like (33) for obvious structural and semantic reasons. There is hence no structural configuration that uniformly allows these elements to occupy their default positions in complex co-occurrences. Moreover, one would expect structures like (33) to be additionally ruled out on semantic grounds in the system under consideration, since, as Zamparelli notes, these linearizations should induce a violation of ‘Redundancy’, which he holds responsible to rule out a wide variety of weak/ strong co-occurrences: *6 (36) (= Zamparelli [1995] 2000: (333)) REDUNDANCY two functional words F i , F j within the same DP give an impossible representation if the meaning of F i entails the meaning of F j or vice-versa. The principle in (36) is not without problems, however. Neither is the selection of data, presented by Zamparelli himself, which are taken to be ruled out by it. First, since entailment is said to apply bi-directionally in (36) above, the dismissal of (37) a. implicates the ungrammaticality of (37) b. (and, vice-versa, the latter datum at the same time predicts the former to also be grammatical, regarding (36)). (37) (= Zamparelli [1995] 2000: (332) c.) a. * Two sm people b. Sm two people Concerning the remaining data presented by Zamparelli in favor of (36), as to (38), C&G note that the entailment relation between the strong quantifiers on the one hand and the indefinite article on the other is not clear; additionally, as 39 A parallel criticism is found in C&G (ch. 2.3.4). 40 Zamparelli tentatively concludes that the tripartite structure can be maintained by locating said element in the specifier of SDP; C&G (p. 52), however, correctly point out that this would ignore several properties, unifying quantifiers in contrast with e.g. adjectives and numerals. 48 1 AXIOMS - Adjectival Inflection & Phrasal Categorization the authors observe with reference to Benincà & Haiman (1992), the last remaining linearization in (38) (i.e. co-occurrence of numeral and indefinite article) turns out grammatical in several (Ladin) languages. (38) (= Zamparelli [1995] 2000: (332) b.) * {One/ Every/ Each} a man An alternative approach mentioned by Zamparelli would take the position of the numerals in (37) a. and (38) to be the outcome of movement from a lower position (i.e. pre-article, complement of PDP). As the author himself notes, however, this rationale could hardly be carried over to those instances above, involving a strong quantifier, on semantic grounds. Therefore, it is left to introduce a further divergence in the syntactic treatment of semi-lexical elements in the nominal domain. But this points us to a much more pressing problem in Zamparelli’s theory. As C&G observe: Zamparelli is not clear whether the weak-strong distinction is obtained through different merging points, through movement from PD to SD, or whether the two options are both possible and, in this case, what would make either choice operative in different cases. (C&G: fn. 11) The system elaborated here takes a clear stance on the issue: Strong and weak quantifiers are merged exclusively in the head position of their respective functional projections Q S P, Q W P. The underlying rationale can hence be identified as a combination of the QPand AP-Hypothesis of C&G, outlined above. 41 Therein, I will show that syntactic behavior indeed directly depends on the fixed categorical status of an LI, without reference to varying external merging-sites. Moreover, as I will argue later in this book, deviations in interpretation of quantifying elements with accompanying linearizational and syntactic properties in complex formations will be linked to additional applications of Internal Merge in the buildup of the nominal domain (cf. ch. 4.3.1 a., ch. 4.3.2 below). As C&G mention (p. 45), “the combination of the QP and the AP hypotheses […] predicts the co-occurrence of two quantifiers in one and the same noun phrase” (which was the initial observation at the beginning of this section) but is still “sufficiently restricted as to predict the ungrammaticality of some logically possible combinations” (C&G: ibid.). 41 The proposed set-up therein parallels the rationale of Zamparelli’s account, as noted by C&G (p. 50), cf. fn. 36 above. The implications derived therefrom will surface throughout the current analysis. 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 49 A further argument in favor of these structural proposals might be found in conjoined noun phrase constructions, a test environment employed by Etxeberria (2009: ch. 4.4.2.1.1, cf. also Etxeberria & Giannakidou 2009: 22, Etxeberria 2012: ch. 4.2.1) in carving out the fine structure of nominal domains in Basque. He claims that in languages, generally allowing the coordination of multiple nominal heads under a single functional projection, it should likewise turn out grammatical to coordinate multiple instances of the functional domain under one token of said dominating head. As (33) above has demonstrated, strong and weak quantifiers do not stand in complementary distribution but might co-occur in the extended projection of the same nominal head. Transposing Etxeberria’s claims to the topic of investigation, it should hence be possible to coordinate two weakly quantified nominals under the same strong quantifier in German. This claim is borne out. In line with the structural proposals, the strong quantifier moreover takes scope over both nominal domains. (39) alle viel-en Häuser und wenig-en Gärten all many houses and few gardens Additionally, observe that the inflection on the weak quantifiers is once again homonymous with the weak adjectival inflection. I will not comment on this phenomenon here but merely highlight the parallelism of this morpho-syntactic effect (cf. chapter 3.3.3 below for a detailed analysis). In the final phrasal set-up of the nominal domain, the order and specification of the elements involved should follow from shared features/ values and further syntactic characteristics, in parallel to the proposed set-up of the CP and IP domains in finely-grained cartographic approaches to the extended verbal projection, 42 a rationale that will be taken up in chapter three below (cf. ch. 3.4). For now, updating (30) on the basis of the linearizational and inflectional data laid out above lets us arrive at the phrasal proposal in (40) below. See that due to the added phrasal categories two additional possible openings for locating further phrasal categories also found their way into the set-up. I will next turn to those elements which initially motivated the categorical and phrasal split, advocated in this section, on linearizational grounds and hence obviously can be taken to occupy the gap between the phrases Q S P and Q W P, viz. SLIs of the category D. 42 Starting with Grimshaw ([1991] 2005), the phrasal set-up of functional heads erected over lexical categories has been a major research topic. The most prominent exponents in this regard being Rizzi (1997) and Cinque (1999); cf. Shlonsky (2010), Cinque & Rizzi (2008) for a general overview. The works of Larson (1988) and Pollock (1989) can be understood as predecessors in this respect, together with Abney (1987), cited above. 50 1 AXIOMS - Adjectival Inflection & Phrasal Categorization (40) [(…) Q S (…) Q W (…) NP (…)] 1.4.2 Determiners and Demonstratives I take it that there is no need to justify the proposal of a dedicated determiner head D 0 today. More than 25 years ago, Abney (1987) was the first to elaborate such an account in detail, which can be viewed as part of the mainstream generative framework at least since Chomsky’s MP (for whom (p. 58) the first implications of this rationale date back as early as Postal 1966b, Brame 1981, 1982). 43 What is more controversial at the current point is the position of proximal and distal demonstratives ( dies and jen in German) in the set-up under consideration. I will go on to follow the (sometimes covert) mainstream rationale in locating them in the head position of DP, mainly on the simple observation of complementary distribution with the definite determiner 44 in German, a language-specific idiosyncrasy not to be taken as universal; note the Spanish example below: (41) ([Spanish]) El libro este the book this Lyons (1999: ch. 8.3) takes structures like (41) to propose an external mergingsite for demonstratives in the specifier of a projection, low in the extended hierarchy of the noun. 45 The proposal of successive, language-specific and optional or obligatory raising of the demonstrative to Spec,DP thereafter furthermore accounts for the prenominal position of demonstratives in simple Dem-N configurations across languages. This in turn is motivated by a restriction that 43 Early accounts based on a variety of different arguments are, among others, Cinque (1990, 1994), Crisma (1991, 1996), Picallo (1991), Giusti (1992), Radford (1993), Longobardi (1994) on Italian, Bernstein (1993) on Romance languages, Hellan (1986) on Norwegian, on Hungarian as early as Szabolcsi (1981 et seq.), Ritter (1988, 1991a, b) on the Construct State; on German, see Haider (1988), Löbel (1990a, b), Vater (1986) Müller (1986); of course, this claim has not gone unchallenged, cf. Payne (1993) for a discussion favoring - what he calls - the single-head hypothesis, recently Bošković (2005 et seq.), Despić (2011, 2013) for a variety of arguments against the universality of DP (‘Universal DP Hypothesis’, UDPH in Despić’s terms) and in favor of a DP/ NP-parameter. 44 Cf. Wiltschko (2009) for a proposal parallel to the one advocated here, Klinge (2008), Bernstein (2008) for morphological decompositional arguments for an equal treatment. 45 Cf. Roehrs (2009: ch. 2.4, 2.5.3.2) for a parallel account with the demonstrative externally merged in a low ArtP in the nominal domain and subsequently internally merged in DP; see references therein for predecessors of this rationale as well as parallel approaches. 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 51 determiners and demonstratives have to be linked to Spec,DP (cf. Lyons 1999: 302). In the case of the latter this happens via co-indexation when surfacing insitu; in configurations employing a raised demonstrative in Spec,DP, however, “this serves to express the presence of the DP projection, and an overt D head becomes unnecessary” (Lyons 1999: ibid.). Thence, all four attested configurations of determiner/ demonstrative linearizations, i.e. D-N, Dem-N, N-Dem and D-N-Dem are accounted for. Spec/ head status, however, cannot be determined on linearizational grounds in German since the replacement of Det and Dem is string vacuous while cooccurrence is not observed. In the development of the simplest phrasal setup of the nominal hierarchy of projections, though, I feel justified to rely on basic observations to develop the simplest working hypotheses and elaborate on them later, if/ when necessary. What is, then, most important at the current point is the equality in categorical status of the elements under consideration. Note that this rationale is even more prominent in those approaches presented above, which do not grant head status to determiners: Lyons (1999) starts out by separating the featural content of Det and Dem with only the latter bearing the additional deictic feature [± Prox], 46 before introducing a feature [± Dem] to account for non-deictic demonstratives which still have to be separated from D as well as non-demonstrative elements which also carry deictic information. The minimal divergence in featural set-up calls to mind Jackendoff’s treatment of certain quantifiers and determiners, discussed above. Lyons (1999: ch. 8) later arrives at a reanalysis of said elements uniformly located in the specifier 47 of what he calls ‘Definiteness Phrase’, the basic components of his analysis being that Det/ Dem signals the presence of a definiteness-feature in D 0 without actually encoding it and functional heads not made equal with word classes but grammatical (or semantic) categories; a rationale which I have followed with Zamparelli (cf. also Pafel 1994: 272 on [± Definite] as a constitutive feature of a lexically heterogeneous D). For now, it is not important if DP abbreviates Determiner/ Demonstrative or Definiteness Phrase; later on, I will however present evidence that at least in German it is indeed the former. Accordingly, as with all other labels discussed in this subchapter, I will henceforth refer to both the category as well as its prime instantiation, the LI Det, synonymously as D throughout the analysis. Observe that the ordering of the quantifier phrases to DP is crucial in the elaboration of a hybrid analysis of the QPand AP-Hypothesis, as presented 46 [+ proximal] = proximal: this , these ; [proximal] = distal: that , those ; cf. also Wiltschko’s (2009) feature ‘location’ in D, only encoded in demonstratives. 47 Lyons does not, however, exclude the possibility of free form D/ Dem occupying the head-position. 52 1 AXIOMS - Adjectival Inflection & Phrasal Categorization above. Only if D 0 is sandwiched between them does the rationale carry over to the current proposal. Luckily, the hierarchical position in question corresponds to the linearizational position on the surface so that no additional applications of Internal Merge have to be stipulated; a quite natural, hence welcome result: (42) (all) die (vielen) Bäume ( all ) the ( many ) trees Recall from section 1.2 above that I only take phrasal layers to project in a given structure if either head or Spec can be identified by overt lexical material. Therefore, DP might not pose as the highest phrase of the extended domain of the noun in structures either not incorporating the determiner/ demonstrative or including the strong quantifier (irrelevant of the occurrence of the former). Therein the account elaborated here can be taken with Zamparelli ([1995] 2000: ch. 4, fn. 1) to advocate the ‘split-DP hypothesis’ while maintaining the dedicated status of the phrase itself. Once again focusing on conjoined noun phrases, observe that the various instances of DP are able to both (i) head two weakly quantified nominal domains as the single highest instance as well as (ii) head two decided instances, themselves coordinated under one strong quantifier (i.e. (43) a. and b. below respectively), in line with the proposed relative positions of these phrases. Moreover, note that, concerning the former, weak quantifiers once again uniformly exhibit the form reminiscent of the weak adjectival inflection, already reported in the last section for coordinations headed by Q S . (43) a. {die/ diese} viel-en Häuser und wenig-en Gärten { the / these } many houses and few gardens b. all(-e) {die/ diese} Häuser und {die/ diese} Gärten all { the / these } houses and { the / these } gardens Updating the schematic representation accordingly, we hence arrive at (44). Obviously, the emergence of yet again two new possible slots for additional phrasal material accompanies the proposal of this dedicated phrasal head. The following section is concerned with filling two of these slots in the second instance of splitting semantically related elements in the nominal domain, namely possessives. Since the two categories to be separated are syntactically unequal (even apart from linearizational properties, e.g. concerning the structural status of the lexical material hosted therein), I will depart from the treatment of the 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 53 two phrases for quantificational elements above and discuss each of them in detail in a separate subsection, beginning with pronominal elements. (44) [Q S (…) D (…) Q W (…) NP (…)] 1.4.3 Lexical Possessives and Possessive Pronouns 1.4.3.1 Possessive Pronouns and Poss PRO P It would follow the general rationale, laid out thus far, to locate possessive pronouns in the head position of the determiner phrase, established in the last section. At least the complementary distribution of these elements with one another should pose as an argument in that direction: (45) a. das Buch the book b. mein Buch my book c. * das mein Buch the my book However, complementary distribution cannot qualify as a sufficient criterion in this case, once a wider paradigm is taken into account. Note (46) a., analogous to (45) a./ b. and (46) b. for (45) c. respectively: (46) a. Martins Buch Martin GEN book b. * das Martins Buch the Martin GEN book Lexical possessives are licensed in prenominal position analogous to possessive pronouns in German; however, they too cannot co-occur with the determiner of the HN. They can nevertheless be far more complex than the proper name in (46) above and hence undoubtedly display phrasal status. Since D 0 and a possessive phrase clearly do not compete for the same structural position and possessive pronouns and lexical possessives behave alike w.r.t. their semantic function as well as their linearizational properties, I want to tie a link between these elements in the proposal of a possessive phrase PossP in the complement 54 1 AXIOMS - Adjectival Inflection & Phrasal Categorization of D (cf. e.g. Delsing 1998). Observe that weak quantifiers surface to the right of both prenominal lexical as well as pronominal possessives: (47) a. sein-e viel-en Bücher his GEN many books b. Martins viel-e Bücher Martin GEN many books We hence arrive at the phrasal set-up in (48): (48) [Q S D Poss Q W (…) NP (…)] This is, once again, not a novel claim. The proposal of a dedicated phrase for possessive elements dates back more than 30 years. 48 The proposed phrasal configuration gains additional backup from current analyses: Despić (2011) argues for a tight connection of DP and PossP in its complement parallel to the relation of C and T in the clausal domain. His main point in this respect is that the (strong) phasal status of C/ D 49 in Chomsky’s DbP/ OP/ AUGB-framework as dependent on the specific non-phasal complement T/ Poss to which the phase head inherits the relevant u Fs (cf. also Chomsky PoP). Thereby, Despić derives various syntactic characteristics of languages which employ (definite) determiners like English and determinerless languages like Serbo-Croatian (i.e. binding phenomena, ellipsis licensing, LBE, possessive reflexives in the nominal as well as subject anaphors in the clausal domain). DP is hence not a universal category in his account but must be able to combine with Poss, when active. I will get back to his proposals in much more detail in the course of chapter two below. Let us turn to the relation of lexical and pronominal possessives. As already mentioned above, the lexical possessive can be of varying complexity (with a decline in grammaticality from a certain point onwards, marked by ‘#’ below) and is hence easily identified as a phrasal category. This is even more apparent once the distribution of case is taken into consideration, since the possessive part of the structure is recognizably marked with genitive: (49) # des großen Clans Flaggen [ the big clan ] GEN flags 48 Cf. Anderson (1983), Kayne ([1993: ch. 2.2] 2000: ch. 7.1.3), (1994: ch. 3.5), following Szabolcsi (1981, 1983, 1992). 49 See e.g. Svenonius (2004), Bošković (2005), Hiraiwa (2005). 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 55 Prenominal lexical possessives can hence be categorized in such a way that they occupy the specifier position of PossP. I take this to represent the standard view in the generative tradition today. Lexical possessors are, however, not exclusively found prenominally in German. When they surface right adjacent to HN, the decline of grammaticality with increasing complexity does not apply: (50) die Flaggen [des großen, starken und mächtigen Clans] the flags [ of . the big strong and mighty clan ] GEN ‘the flags of the big, strong, and mighty clan’ Pronominal possessives, on the other hand, are not grammatical postnominally. This is true for German, but not e.g. for Romance languages like Italian, Paduan or Spanish. As Cardinaletti (1998) observes, postnominal pronominal possessives precede the complements of HN in Italian, a fact that is also found with postnominal lexical possessives in German: (51) a. die Flaggen [des Clans] the flags [ of.the clan ] GEN b. die Flaggen [auf dem Berg] the flags [ on the mountain ] c. die Flaggen [des Clans] [auf dem Berg] 50 the flags [ of.the clan ] GEN [ on the mountain ] d. * die Flaggen [auf dem Berg] [des Clans] the flags [ on the mountain ] [ of.the clan ] GEN The data presented here hence point to the existence of an additional postnominal position for possessives in German, too. However, said position seems additionally restricted to only host lexical possessives, a point to which I will return in detail in the next subchapter. Before doing so, I want to concentrate on exclusively prenominal possessives, i.e. possessive pronouns. Cardinaletti (1998), drawing on the typology of Cardinaletti & Starke (henceforth C&S; 1996, [1994] 1999), classifies pronominal elements as either ‘strong’ or ‘deficient’ on the basis of predominantly Romance data, the latter again splitting into the categories ‘weak’ and ‘clitic’. This tripartition is proposed to parallel the structural status of the elements under consideration in C&S (1996), 50 The structure evokes semantic ambiguity as to which NP is further modified by the PP (which structurally translates to ambiguity as to where PP attaches). See, however, that the inverted order of postnominal constituents is out in every reading. 56 1 AXIOMS - Adjectival Inflection & Phrasal Categorization with strong elements exhibiting full (i.e. complete) phrasal status, deficient elements either lacking full phrasal structure or independent head status, i.e. the categories weak and clitic respectively. Cardinaletti (1998) therefore concludes that clitic possessive pronouns adjoin to D while weak possessive pronouns are located in the specifier position of a prenominal Agr-projection below DP (see ch. 1.4.3.2 below for further detail). The respective criteria for the three categories as well as their application to German possessive pronouns are given below: (52) deficient strong reduced w.r.t. forms used without overt HN  longer forms (without overt HN)  restricted to prenominal position  can occur postnominally  cannot be coordinated/ contrasted/ focalized/ modified  can be coordinated/ focalized/ modified  can have non-human referent  restricted to human referent # weak clitic can co-occur with D prenominally (and before modifiers)  allows doubling (doubling implies clitic)  can be used predicatively (causative/ epistemic/ copular)  allow N-to-D movement without cliticizing to N  licenses floating quantifiers  can be used in isolation  cannot move out of DP  displays declension  can be used with ostension (i.e. introduce a new referent)  in complementary distribution with D  Let us go through these criteria in detail. First, possessive pronouns with masculine and neuter nominals display zero inflection when concatenated with an overt HN, as opposed to elided head nouns, which in turn evoke overt suffixation (given that no other lexical element like adjectives, right-adjacent to it, can take over): 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 57 (53) a. Das ist mein (guter) Hund. that is my ( good ) dog b. Das ist meiner. that is mine masc.sg c. Das ist mein Guter. that is my good.one masc.sg d. Das ist der meinige. that is the masc.sg mine.one e. * Das ist der meinige Hund. that is the masc.sg mine.one dog f. * Das ist der meinige gute. that is the masc.sg mine.one good We see in (53) a. and c. that the suffix er encoding number and gender (along with case) has to always be present on adjectives, irrespective of the presence of HN (which itself bears intrinsic values above, not encoded suffixially). Only if no lexical material is coordinated below the pronoun are these values encoded there, which might be interpreted as an instance of declension - a criterion for clitic, i.e. deficient possessive pronouns in (52). The datum in (53) d. and its derivatives e. and f. display another morphologically complex instantiation of the pronoun, an archaic construction referred to as ‘adjectival’ in Roehrs (2013). I will not have to say much about these exceptional forms here, but merely include them for the sake of completeness. Observe that these forms neither allow Adj nor HN to be overtly present in its complement and moreover obligatorily co-occur with D (encoding number and gender, see also fn. 52 below). Returning to (53) a.-c., it seems equally possible to take the inflected/ suffixed pronoun to conform to Cardinaletti’s category strong with the stem in (53) a. an instantiation of a deficient pronoun (abstracting away from adjectival possessives) w.r.t. the first criterion. 51 Moving on to the next criterion, the locus of possessive pronouns inside the hierarchy of projections is restricted to prenominal position, as has been mentioned above. This is true of the bare as well as the inflected forms. There is, hence, no distinction in Cardinaletti’s sense in this respect. Additionally, prenominal possessive pronouns do not co-occur but are 51 Observe that this bipartition does not hold over all numberand gender-values: Feminine HNs force possessives in their hierarchy of projections to display agreement qua suffixation, irrespective of N as well as A, which makes this distinction obsolete in these cases. I will elaborate on this below. 58 1 AXIOMS - Adjectival Inflection & Phrasal Categorization generally in complementary distribution with D, which is another criterion for the subclass of clitic elements (and against the classification as a weak element) in Cardinaletti’s typology (cf. (45) c. above). 52 Ignoring the inflectional characteristics of the pronoun in overt HN vs. ellipsis structures for now, possessive pronouns can be coordinated (a.), contrasted (b.), focalized (c.) and modified (d.) in German, as illustrated below: (54) a. mein und dein Hund a'. meiner und deiner my and your ( dog ) b. mein(-er), nicht dein Hund 53 b'. meiner, nicht deiner my, not your ( dog ) c. mein Hund 54 (und nicht deiner) c'. meiner my ( dog ) ( and not yours ) d. {nur/ allein} mein Hund d'. {nur/ allein} meiner { only / exclusively } my ( dog ) Given this pattern, one might conclude that possessive pronouns are to be categorized as strong in German w.r.t. the criteria laid out in (52) above. Furthermore, as far as (non-)human reference is concerned, my informants give mixed judgments regarding the acceptability of the structures in (55) below (again indicated by ‘#’). I personally find cases like (55) a. and b. to be completely ungrammatical, however, animate non-human referents are found frequently (cf. (55) c.) and at least some cases like d. and e., making use of brand names, can be found on the internet. 52 The connection of demonstrative and simple possessive pronoun is, however, licit in German. I will have much more to say about this construction below. 53 Since inflection on the contrasted pronoun in (54) b. is optional, one might conclude that the structure at hand consists of two separate nominal hierarchies. Since none of the other structures exhibit such ambiguities, I still feel justified to categorize these pronouns as strong. 54 The small capitals indicate stress. 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 59 (55) a. (Der Deckel gehört zur Pfanne.) # Das ist ihr Deckel. ( the lid belongs to.the pan fem ) this is her fem lid b. (Die Kappe gehört auf den Stift.) # Das ist seine Kappe. ( the lid belongs on.the pen masc ) this is his masc lid c. (Die Leine gehört dem Hund.) Das ist seine Leine. ( the leash belongs to.the dog masc ) this is his masc leash d. Der große Porsche und sein kleiner Motor [ the big Porsche ] masc and his masc small engine e. Pirelli hält seine Reifen für unschuldig. Pirelli considers his masc tiers to.be innocent Based on these data, possessive pronouns might be considered to obey a nonanimate restriction at best while being completely free in reference for some speakers and hence pattern with the broader class of deficient elements. Turning to predicative, isolated and ostensive use, German possessive pronouns once again pattern with the strong category (the examples are constructed analogous to Cardinaletti 1998: (47) and (48), (14); curly brackets once again signal interchangeability): (56) a. Dieses Bild ist {meines, deines, seines, unseres}. (copular) this picture is { mine , yours , his , ours } b. Ich halte dieses Bild für {meines, deines, seines, unseres}. (epistemic) I consider this picture to.be { mine , yours , his , ours } c. Ich mache {meine, deine, seine, unsere} (Anträge). (causative) I work.out { my , your , his , our } ( applications ) (57) A: Wessen Buch ist das? (isolation) whose book is that B: {Meines, Deines, Seines, Unseres}. { mine , yours , his , ours } (58) A: Von wessen Wagen wurdest Du angefahren? (ostension) by whose car were you hit B: {Seinem, Ihrem, Ihrem}. { his, her, theirs } 60 1 AXIOMS - Adjectival Inflection & Phrasal Categorization Concerning movement out of DP, however, the elements under consideration display ambiguous behavior. While movement, e.g. into sentence-initial topic position, is illicit without additional elements (an indicator for weak elements), strong quantifiers can readily be stranded when the pronoun moves with HN (a criterion of clitic possessive pronouns): (59) a. * Seine hat er (großen) Schulden bezahlt. his has he [( large ) debt ] paid b. Seine (großen) Schulden hat er bezahlt. [ his ( large ) debt ] has he paid c. Seine (großen) Schulden hat er alle bezahlt. [ his ( large ) debt ] has he all paid Finally, Cardinaletti proposes that doubling implies clitic status of possessive pronouns. Although colloquial German does exhibit the co-occurrence of two instances of (possessive) pronouns, this construction exhibits idiosyncratic characteristics which set it apart from the doubling configurations in Romance languages, as laid out by Cardinaletti: (60) a. so pare de Toni (= Cardinaletti 1998: (22) a., [Paduan]) his father of Toni b. il libro suo di lui (= Cardinaletti 1998: (25) a., [Italian]) the book his of him c. E’ suo di lui. (= Cardinaletti 1998: (25) b., [Italian]) [ it ] is his of him d. von {ihm/ Martin} sein Vater of { him / Martin } DAT his father e. * sein Vater von {ihm/ Martin} his father of { him / Martin } DAT f. ihm sein Vater [ him DAT his GEN father ] NOM g. ihm sein-er him DAT his GEN.masc As can be observed in (60) a.-c., doubling involves one pronoun to be coordinated, arguably under a prepositional head, following the pronoun in the hierarchy of projections of HN (which might surface left or right from the nominal head). When transposed to German, the PP (with a pronominal or a lexical comple- 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 61 ment) necessarily precedes the simple pronoun. Possessives encircling HN are not grammatical either, be they pronominal or lexical PP complements. Interestingly, when occurring prenominally, none of the pronouns has to be headed by a preposition (cf. (60) f.). The two pronouns bear the exact same values of φ-features, differing only in case, with the first pronoun bearing dative and the second one bearing genitive (i.e. possessive) case. Ellipsis of HN can also apply to these structures, resulting in obligatory inflection on the pronoun that surfaced adjacent to the nominal before ((60) g.). The resulting structure parallels (51) d.-f. above and hence points to Cardinaletti’s category strong, once again. We can, then, tentatively conclude that German possessive pronouns do not fully pattern with any of the three categories in Cardinaletti’s typology. If we turn to C&S’ (1996) treatment of Germanic pronominal systems, we find that strong and weak forms are taken to be homophonous in German. Considering that possessive and genitive personal pronouns are also homophonous in German (and Slavic languages, cf. Cardinaletti 1998: ch. 2.4), the facts might carry over to the domain under consideration. We can test this by combining two of the characteristics in Cardinaletti’s typology which indicate different classes of possessive pronouns: Strong possessives can be coordinated but are restricted to human (or at least, animate) reference. Consider (54) a., a'. and (55) a., reprinted here as (61) a.-c. respectively, for ease of exposition. (61) a. mein und dein Hund (= (54) a.) b. meiner und deiner (= (54) a'.) my and your ( dog ) c. (= (55) a.) (Der Deckel gehört zur Pfanne.) # Das ist ihr Deckel. ( the lid belongs to.the pan fem ) this is her fem lid If we combine the two grammatical structures, the resulting string is ungrammatical even for speakers who readily accept non-animate referents of pronouns in German. (62) (der Deckel/ die Deckel gehör-t/ -en zur Pfanne und zum Topf.) ( the lid sg/ pl belong sg/ pl to.the pan fem and to.the pot masc ) * Das ist/ sind sein und ihr Deckel. this / these is / are his and her lid. Likewise, consider the data of pronominal doubling structures in (60) (a property of clitic pronouns): According to my informants, such configurations result 62 1 AXIOMS - Adjectival Inflection & Phrasal Categorization in ungrammaticality if used in the strong pronominal contexts, presented in (56)-(58) above (as well as in coordination, which is yet another environment of strong pronouns). (63) a. * Dieses Bild ist ihm seines. (copular) this picture is him DAT his GEN.neut b. * Ich halte dieses Bild für ihm seines. (epistemic) I consider this picture him DAT his GEN.neut c. A: Wessen Buch ist das? (isolation) whose book is that B: Ihm seines. him DAT his GEN.neut d. A: Von wessen Wagen wurdest Du angefahren? (ostension) by whose car were you hit B: Ihm seinem. him DAT his GEN.masc I therefore conclude that strong and deficient (possessive) pronouns are homophonous in German. This conclusion gains further backup from both Cardinaletti’s (1998) finding of homophonous strong, weak and even clitic possessive pronouns in Italian as well as C&S’ (1996) classification of German pronouns, reported above. I will leave open the exact classification of these homophonous elements here (strong and deficient possessive pronouns vs. strong possessive and weak genitive pronoun) and instead simply note that the data support the proposition of the possessive head status in the extended projection for the elements under consideration, since - as Cardinaletti (1998) notes - the mergingsite of the structurally most complex element suffices to also introduce all structurally poorer items. I will next turn to the second class of possessive elements with the criteria at hand, namely lexical possessives. 1.4.3.2 Lexical Possessives, Strong Possessives and Poss LEX P Returning to the criteria for strong possessives in (52) above, one linearizational property is not observable with German possessive pronouns, namely their occurrence in postnominal position. However, as can be seen from the discussion surrounding (50) as well as that on the position of pronouns following (52), this site can be occupied by other lexical material conveying possessive semantics, i.e. lexical possessives. As mentioned above, these complex SOs are located in the exact same position as strong possessive pronouns in Romance, namely the postnominal slot preceding complements: 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 63 (64) a. (= Cardinaletti 1998: (3) a., [Italian]) Questo libro qui suo di sintassi non mi convince. this book here his / her of syntax not me convinces b. das Buch Martins über Syntax 55 this book Martin GEN about syntax c. dieses Buch hier des Lehrstuhls für Linguistik über Syntax this book here [[ of.the chair ] GEN in linguistics ] about syntax As can be deduced from (64), lexical possessives in postnominal position are not in complementary distribution with determiners ((64) b.) or demonstratives ((64) c.), parallel to Romance postnominal possessive pronouns ((64) a.). Consequently, one should ask if such complex elements can be fitted into the typology above. 56 Concerning nominals and pronouns in German, C&S (1996) observe an interesting asymmetry: (65) (= C&S 1996: (40) a., b., d.) a. Ich habe  ihn gestern eingeladen. I have him yesterday invited b. Ich habe gestern * ihn eingeladen. I have yesterday him invited c. Ich habe gestern  nur ihn eingeladen. I have yesterday only him invited d.  Ich habe gestern Hans eingeladen (= C&S 1996: (43)) I have yesterday Hans invited In their earlier approach to pronominal systems, the authors introduce the following principle for the choice of an appropriate form from their typology (cf. also C&S 1996: 35): 57 55 Note that the ‘demonstrative reinforcer’ qui / hier (‘here’) is not acceptable in this position with proper names, but with lexical possessives introduced by determiners and prepositions. 56 Since most of the other properties of strong pronouns listed in (52) involve reference to the interpretation of the element under consideration, they are not testable for lexical possessives. 57 All citations are taken from the 1999 version of their paper. 64 1 AXIOMS - Adjectival Inflection & Phrasal Categorization (66) (= C&S [1994] 1999: (22)) Choice Principle Choose the most deficient possible form. The principle predicts that “clitics are chosen over weak, and weak over strong” (C&S 1996: 37) pronouns whenever two or more forms are not barred from occurring in a position. Returning to the paradigm in (65), C&S conclude that the rigid order of weak pronouns inside the clause (viz. “in a local relation with Case features” (C&S 1996: 23), i.e. Spec,Agr S/ O P in their system) prevents them from surfacing post-adverbially (cf. (65) b.). In (65) c., however, modification forces the pronoun to be strong (a reason, independent of the sentence structure of (65)) and thus the linearization involving strong ihn turns out grammatical. *7 Turning to nominal arguments, we find the exact same distribution as with modified pronouns. It is worth quoting C&S in full on this point: Since full NPs do not have several distinct variants in the languages under discussion (i.e. they are mostly strong-only, with no weak or clitic variants), whereas personal pronouns do have several variants, the Choice Principle gives different results in the two cases: with nouns, it always allows the strong form (which is trivially the most reduced form available), but with pronouns it allows the strong form only in contexts excluding deficient forms […]. Therefore, there should be some contexts in which (strong) full NPs can appear and strong pronouns cannot. (C&S 1996: 35) The restriction hence dissociates strong pronouns and lexical possessives; with respect to the latter, we can conclude that lexical possessives are always strong in German. As already mentioned in the last section, C&S go on by tying the categories in their typology to X-configurations for the languages under consideration (although they note that the properties are in principle independent from one another, cf. C&S 1996: 46). Their structural typology is given in (67): (67) (= C&S 1996: (44)) clitic elements are deficient X 0 s. weak elements are deficient XPs. strong elements are non-deficient XPs. (Non-deficient heads are the usual non-clitic heads, i.e. verbs, nouns, etc.) 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 65 Strong elements therefore are phrasal elements; deficiency translates into incompleteness: In the case of XPs, it is incompleteness of the projection, 58 hence, among other things, lacking modification (i.e. ‘c-modification’ in C&S [1994] 1999). This is in line with the phrasal status of lexical possessives reported above. Not only can coordination/ focalization/ modification apply to these structures, they can also take complements of their own, as has been shown here. Based on these observations, I want to propose the existence of a second possessive phrase for lexical possessives in the complement of the head noun, in parallel to the two surface positions for possessives, identified by Cardinaletti (1998), call it Poss LEX P (opposed to the prenominal possessive phrase for pronominal elements, call it Poss PRO P). Cardinaletti (1998), herself elaborating on the two surface positions for possessive pronouns in Italian, takes the lower position as the external merging-site of all pronominal possessives, followed by obligatory movement to prenominal position in the case of deficient elements, as opposed to strong ones. Her proposal hence parallels the rationale of the most embedded surface position of LIs as their probable external merging-site, which has been employed throughout this subchapter. 59 Observe that lexical possessives in this position are not in complementary distribution with determiners of HN, as has been deduced from (64) b.; the relevant part is reprinted and contrasted below: (68) (= (64) b., [reduced]) a. das Buch Martins [ the book ] NOM Martin GEN b. * das Martins Buch [ the Martin GEN book ] NOM Lexical possessives are hence located in the complement of HN under the view advocated here. 60,*8 This postulate of two discrete possessive phrases gains fur- 58 I.e. incompleteness of the hierarchy of functional projections qua absence of the ‘nominal CP’ in C&S’ ([1994] 1999) terms, cf. also C&S ([1994] 1999: ch. 5.2.4). 59 The same rationale is also depicted in Alexiadou et al. (2007: 554) for the location of possessives in the extended projection of the noun, cf. also fn. 60 below. 60 See Alexiadou et al. (2007: ch. IV.2.2.2.1) for cross-linguistic data in favor of the postnominal origin of lexical possessives as well as accounts incorporating this rationale. The authors, however, ultimately reject the concept on the basis of the thematic relation between the lexical head and its complement position vs. the semantic variability of possessive modifiers following Williams (1981); as is commonly known, possessives can take on a wide variety of semantic interpretations in their relation to HN. Further (syntactic) restrictions come into play if HN itself bears argument structure (cf. Roehrs 2013: ch. 4.1 on German). I will not go into detail on the interaction of possessives with agent 66 1 AXIOMS - Adjectival Inflection & Phrasal Categorization ther backup from pronominal doubling structures exemplified in (60) f. above and reprinted here as (69): (69) (= (60) f.) ihm sein Vater [ him DAT his GEN father ] NOM If we locate the genitive pronoun in the head position of the ‘classical’ (i.e. prenominal) PossP, the dative pronoun arguably occupies the specifier position of said phrase on the surface. Observe that the element preceding the genitive can be of varying complexity ((70) echoing (64) a. above): 61,*9 (70) Dem Lehrstuhl für Linguistik sein Buch über Syntax [[ the chair in linguistics ] DAT his GEN book about syntax ] NOM A hint as to the structural origin of the dative elements might be drawn from the properties of the initial pronoun in (69): Considering the criteria for strong pronouns laid out in (52), we find in (63) that they can be used in isolation and in ostensive contexts, but are illicit when used in copular and epistemic constructions. It is, however, unclear if these tests tell us anything about the first pronoun itself or the double possessive complex in toto. The same is true of referential properties, which, mimicking (55) with the relevant changes made, is out for non-animate reference even for speakers accepting it with simple possessive pronouns. There are, however, criteria which can be taken to apply to the initial pronoun only, namely contrast/ coordination/ focalization. Observe (71) in analogy to (54) above: (71) a. (contrast) Nicht ihm i , sondern ihm j seine Bücher habe ich gelesen. 62 not his DAT i , but him DAT j his books have I read and patient NPs in the hierarchy of projections of HN, however. See e.g. Valois (1991), Alexiadou (2001), Koptjevskaja-Tamm (2003a). 61 A further hint as to the correctness of this proposal might be found in the co-occurrence of complex dative-marked prenominal lexical possessives with a determiner rather than a possessive pronoun in the dialect of Lucerne German (cf. Salzmann 2011: fn. 44 and references therein), which is easily accounted for under the phrasal proposals made thus far in this chapter. See moreover Salzmann (2011: ch. 2.3, also ch. 2.5 section 2) on syntactic criteria to determine a parallel status of the complex prenominal possessive (i.e. ccommand, extraction, coordination and quantifier floating) predominantly in the Zurich dialect of Swiss German, much of which carries over to Standard German. 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 67 b. (coordination) Martin und ihm seine Bücher habe ich gelesen. Martin DAT and him DAT his books have I read c. (folcalization, stress) ihm seine Bücher habe ich gelesen (nicht deine). him his books have I read ( not yours ) d. (focalization, movement) * ihm habe ich seine Bücher gelesen. him have I his books read We can therefore tentatively conclude that pronominal ‘possessor doubling constructions’ (henceforth PDCs, cf. Roehrs 2013, Salzmann 2011 and references therein) involve a strong pronominal element, arguably originating postnominally in German on par with its lexical counterpart (if the unacceptability of extraction of the pronoun from the nominal domain in (71) d. can be accounted for by an unrelated but intervening factor, a point to which I will return in chapter 2.2 below). The system laid out here thus hints at the possibility of all strong (i.e. non-deficient phrasal) pronominal elements uniformly originating in the lower PossP. The uniform merging-site of strong (pronominal as well as lexical) possessive elements coupled with the homonymy of strong and weak possessive pronouns elaborated in the previous subsection might call to mind Zamparelli’s ([1995] 2000) account of varying merging-sites for different uses/ functions of SLIs, presented in section 1.4.1 above. The approach elaborated here, however, might be unified even further: To begin with, recall that German exhibits prenominal strong and deficient (i.e. weak) pronouns but no postnominal ones. The lowest structural position in which we find possessive pronouns in German is hence prenominal, while complex lexical possessives do also surface following the nominal head. As Cardinaletti (1998) notes, the proposed classes of elements constitute structural subsets of each other (along the ordering strong > weak > clitic). Therefore, as was already mentioned at the end of the previous subsection, the merging-site of the most complex possessive (i.e. strong pronominal/ lexical) arguably suffices to also introduce weak as well as clitic elements into the structure (but not vice versa). We might therefore unify the account in the stipulation of a single locus of External Merge for all possessive elements. For this proposal to go through, however, the parallel, obligatory raising of strong and deficient pronouns must be accounted for elsewhere, since 62 Contrasting two pronominal elements with identical sets of φ-features begs for additional extra-linguistic strategies of disambiguation, e.g. by pointing gestures. Nevertheless, the linguistic structure itself is judged grammatical by my informants. 68 1 AXIOMS - Adjectival Inflection & Phrasal Categorization it runs counter to the Choice Principle when paired with Cardinaletti’s observation that deficient possessive pronouns must raise for the purpose of licensing in a higher Spec position (since strong possessives would be expected to occur postnominally, parallel to Romance). Returning to the linearizations laid out in the last paragraph, observe also that such an account would moreover leave the properties of multiple (possessive) pronouns in the PDCs in (63) and (71) above unaccounted for. Therefore, I conclude that the higher possessive phrase must be able to introduce lexical items into the derivation. I will not elaborate further on these reductionist matters here but abstract away from possible applications of Internal Merge or default merging-sites, issues that will again be reviewed in ch. 4.3.1 a. below in the context of PDCs in the system elaborated thitherto in this book. Here, I will then focus on the most embedded surface positions of the elements under consideration by retaining the labels pro and lex for the respective phrases, introduced at the beginning of this subsection. Therein, the rationale underlying previous phrasal proposals is sustained. Furthermore, I hope to avoid terminological confusion with the concepts introduced previously in this chapter. We hence arrive at the hierarchy of projections in (72), an extended version of (48) above. Observe that the hierarchy of projections elaborated thus far displays a parallel set-up of ordered quantifier and possessive phrases, encircling the nominal core categories N and D. (72) [Q S D Poss PRO Q W NP Poss LEX ] Concentrating first on pronominal possessive instances here, observe that data from conjoined nominal domains once again verify the structural proposals arrived at above, with Q S and D dominating two decided nominal domains, each headed by an instance of the functional projection under analysis (i. e. (73) a. and b. respectively), as well as a token of the latter being able to dominate the conjunction of two weakly quantified nominals (i.e. (73) c.). Therein, properties of scope once more correspond to the structural claims. Moreover, inflection of Q W additionally once again parallels that of the weak adjectival inflectional pattern. (73) a. all(-e) mein-e Häuser und dein-e Gärten all my houses and your gardens b. diese mein-e Häuser und dein-e Gärten these my houses and your gardens 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 69 c. meine viel-en Häuser und wenig-en Gärten my many houses and few gardens Elaborating further on these data, and including lexical possessives therein, note that Koptjevskaja-Tamm (2003b: (12)), citing a personal communication with Marit Julien, demonstrates that in the south-east-Norwegian dialect of Kongsvinger, the proprial article (i.e. that co-occuring with proper names, following Delsing 1993) forms a constituent together with a proper noun possessor, since these items can likewise be conjoined with either another simple possessive pronoun or an identical structure inside a single extended projection. Roehrs (2013: 79) categorizes these constructions as instances of postnominal possessor doubling; turning to German, one can indeed find parallel patterns concerning the PDC complexes, discussed above: Observe that all instances of the possessive pronoun in (73) above can be replaced by such complex SOs. I therefore take this test to further backup the claim of parallel structural positions of possessive items in isolation and in the formation of possessor doubling constructions. In the last subsection, I want to draw the focus to the broader category of pronominal elements; this will, however, result in the discussion of a much more variable selection of semantically and syntactically diverse lexical items and therein depart from the categorizations introduced so far in this subchapter. These elements do, nevertheless, share a morphological property, namely the root ein . 1.4.4 Pronominal Elements and (other) ‘ein-Words’ In what follows I will introduce two additional notions that will become relevant in the ensuing analysis, namely the concept of ‘ ein -words’ and the ‘feature geometrical’ approach to SLIs. This is done by means of discussing the structural locus of a class of semi-lexical items, not included in the development of the hierarchy of projections above, i.e. personal pronouns. Both aforementioned notions will surface several times in the course of my analysis, the latter furthermore undergoing extensive modification therein. Therefore, their introduction at the current point can only be preliminary and the discussion will be kept superficial. In line with this, I will also refrain from further postulates concerning structural positions from this point onward and take the hierarchy in (72) as complete for the sake of my subsequent investigation. Approaching the class of pronominal elements with this in mind, observe that the system laid out thus far already provides several structural positions that could be utilized to locate these items in the extended projection of the noun. Beginning with Postal ([1966a] 1970) and structurally formalized as early as 70 1 AXIOMS - Adjectival Inflection & Phrasal Categorization Abney (1987), pronominal elements have been identified on a par with definite articles (Postal)/ determiners (Abney) and located accordingly. The main argument on which Postal justifies this ordering comes from the co-occurrence of pronouns with nominals, such as in (74); his mode of operation hence mirrors the rationale employed throughout this chapter. (74) (= Postal [1966a] 1970: (36), [reduced]) a. Did you see us guys? b. Who insulted you men? c. He didn’t like us Americans. d. He did not insult you Communists. As Postal notes, earlier analyses treating these linearizations as transformational derivatives of relative clauses fail to acknowledge the ungrammaticality of the latter e.g. in object position of interrogatives or in coordination under negation, while their proposed counterparts may occur freely in these environments, as exemplified above. Consequently, he locates these elements in a prenominal position parallel to articles, which is later identified as D 0 by Abney (1987: ch. 4.1.2.c; cf. Lyons 1999: ch. 1.2.5, 8.3.3, 8.4 and references therein for finer-grained approaches based on this proposal; I will discuss Lyons’ proposals in some detail in the context of ‘personal determiner’ constructions in ch. 2.3.3 below). Alternatively - and building on the insights of the last subsection - homophony of genitive personal pronouns and the possessive variants might be taken as a hint to locate these items together in identical structural position, namely the possessive phrases. This rationale has been extended even further: A morphosyntactic decompositional approach to SLIs in the nominal domain that combines both insights above (i.e. the categorical status as D 0 and the identity of morphological form as the basis to group SLIs) is elaborated by Roehrs for German (in prep.; cf. also Roehrs 2009, 2013), who focuses on the root morpheme ein , also part of the homonymous forms mentioned above, to derive structural parallelisms. 63 As he observes, “semantically quite diverse elements behave morphologically the same” (Roehrs in prep.: 78); a full list of the (partially) homonymous items he distinguishes as well as their further subdivision is given below: 63 An early implementation of this rationale for German can already be found in Vater (1979: 10). Considerations on the application of this decompositional rationale to other SLIs in German can be moreover found in Vater (1984: 21, especially fn. 3 on the monomophemic determiner), splitting agreement-phenomena from the semantics of the root. Additionally, a decompotitional proposal concerning the relation of the indefinite article and nominal negation in German surfaces in Pafel (1994: fn. 20, fn. 30). 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 71 (75) (= Roehrs in prep.: ch. 4, (6), [modified]) a. ein as an Article: - Indefinite Article (i.e. ( ei ) n ) - Vacuous Article: - ein as Part of a Composite: - Possessive (i.e. m ein , d ein , s ein ) - Negation (i.e. k ein ) - ein in Predicative Noun Phrases - Complex Determiner (i.e. ein jed -) b. ein as a Numeral (i.e. ein ) c. ein as an Adjective (i.e. ein ) Roehrs (in prep.) refers to the heterogeneous list in (75) a. and b. as ein-words, for obvious reasons (cf. also Roehrs 2009: 125). The basic rationale of his approach has been mentioned in this chapter before, mostly with reference to parallel approaches by Bernstein (2008) and Klinge (2008, cf. also references therein): While the former concentrates on the English root th and thereby covers a range of items spanning over the categories of determiner, demonstrative and pronouns as well as complementizers and existential there , the latter analysis implements an even wider scope, discussing the þ as well as hw -roots both across the languages English, Danish and German as well as diachronically. Discussing the relation of singular numeral and indefinite article, Klinge concludes that the latter, developed from the former, constitutes the inherently indefinite head of a quantifier phrase QP, located below D, which in turn hosts (definite) th elements as well as pronominal lexical items. Ungrammaticality of the co-occurrence of these SLIs hence stems from their incompatibility w.r.t. their status of definiteness (cf. Klinge 2008: 241) rather than competition for an identical structural position. Further distinguishing strong and weak stress patterns, he arrives at the indefinite article as a grammatical or semantic (singular) number marker (cf. Klinge 2008: 240). Returning to the cluster of ein-words in (75) above, one can identify parallels in the accounts of Roehrs and Klinge in that both approaches bundle the lexical items on the basis of their morphological form, but maintain categorical distinctions among them. Furthermore, the criteria of stressability (in the case of the numeral) and reducibility (in the case of the article) are also employed in Roehrs’ approach to further subdivide the fully homonymous instances of the root. In what follows, I also want to abstract away from the predicative use/ function (i.e. the only context of non-composite occurrences of the so-called ‘vacuous article’ in (75) a. above) and likewise concentrate on the terminal, i.e. word- 72 1 AXIOMS - Adjectival Inflection & Phrasal Categorization like instances. Initial evidence presented by Roehrs (in prep.: ch. 4) for bundling these lexical items together does not come from the partial homonymy but rather from patterns of complementary distribution, parallel to the rationale employed previously in this chapter. Note that the indefinite article and the proposed composites based on ein are not able to co-occur, while the numeral (i.e stressed ein ) can at least co-occur with possessives. As mentioned above, Roehrs subsequently discusses several grammatical environments that distinguish the three main categories of ein-words in (75) above, i.e. article, numeral and adjective, namely enclitization, stressability, semantic singularity/ morphological plurality. Building on the resulting insight and further leveling the article and numeral ein , he proposes varying merging-sites for the remaining categories, namely ArtP and Spec,AgrP, with elements of the former additionally raising into a higher functional projection (CardP or DP, depending on the reading of the ein-word) as well as at variable timings in the course of the derivation. Einwords can hence be located in different positions in the extended projection of the nominal head at Spell-Out, depending on their use/ function, a claim that Roehrs (in prep.: 80) incidentally makes with reference to Zamparelli’s ([1995] 2000) system, discussed in the context of QP in section 1.4.1 above. The position on the surface is taken to be D 0 only in strong (specific) readings. 64 The contrast in the grouping of (75) a. to b. and c. can then be found in the semantic vacuity of the root while at the same time providing a host for prefixial possession/ negation (cf. Roehrs in prep.: ch. 5.2.2, cf. also Roehrs 2009: ch. 4, fn. 27). The decompositional rationale underlying the proposal gains further support from the fact that possessive pronouns in German show simultaneous agreement with possessor as well as possessee, i.e. φ-agreement with the former via morphological form of the possessive prefix as well as φand case-agreement with the latter via suffixation (cf. Salzmann 2011: ch. 2.2.2, 2.5 section 3 and references therein on the parallel phenomenon in Zurich German as well as approaches to it), which might indicate an underlyingly two-membered SO, arguably with the root in the head position of the determiner phrase and the prefix located in its specifier position. 65 Moreover, Bernstein (2008: ch. 3) takes the parallel pattern of English first and second possessive pronouns as well as various (pronominal) 64 Cf. Roehrs (in prep.: ch. 4.4.2.3), moreover Roehrs (2009) for an account based on the variation in derivational timing of this instance of Internal Merge; particularly Roehrs (2009: ch 10.1) on the Last Resort rationale of raising ein to the DP-level in Narrow Syntax. 65 Cf. Roehrs (in prep.: ch. 4.4.2.2, 5.3), cf. also Salzmann (2011: fn. 53), moreover cf. Roehrs (2013) for the proposal of fused Poss and ein under linear rather than structural adjacency; on the operation ‘fusion’ cf. Halle & Marantz (1993), Noyer ([1992] 1997), furthermore ch. 2.3.3 below. 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 73 forms, based on the root th -, to argue for the status of initial m -, y as well as th as first, second and third person-markers respectively. Concentrating on the parallels of possessive and personal pronominal elements in the light of the system outlined here, observe that Roehrs (in prep.: ch. 2.3.3) assumes a variable position also for the latter, this time based on External Merge, resulting in categorical ambiguity between quantifier and determiner; the latter option is identified with explicit reference to Postal’s ([1966a] 1970) proposals, laid out at the beginning of this section (cf. Roehrs in prep.: ch. 6.2.3, 6.3.2). The system under consideration, then, treats personal pronouns structurally on par with possessive pronouns (as well as negation as composite forms of the root morpheme ein -) concerning the structural position on the surface in at least a subset of their uses/ functions. However, the rationale of morphological decomposition can hardly be carried over to the pronominal paradigm. Recall from the last section that possessive pronouns are homonymous with the genitive forms of the personal pronouns, which inflect for case, person and gender in singular number as well as for case and person in the plural. With three gender-values and four cases, the paradigm of personal pronouns exhibits 32 slots; the root ein -, however, is not found in any of the remaining 24 forms apart from the genitive/ possessive overlay. Concentrating on the remaining eight forms, it becomes clear that the clustering based on the shared root, moreover, is not perfectly applicable even therein: While the first and second person singular as well as the third person singular masculine and neuter constitute composites based on ein -, namely m ein -, d ein and s ein - (the latter for both third person forms), neither the form of the third person singular feminine nor of any person in the plural is based on said root (the forms are ihr and unser , euer as well as ihr respectively). 66 This point will surface again in the second part of the subsequent chapter. Even though I will refrain from further modifications to the extended projection in (72) based on this subchapter, I nevertheless believe the clustering presented here to constitute a useful perspective on SLIs in German that will be utilized in the course of my analysis in the ensuing chapters. Therein, I will restrict my use of the term ‘ein-word’ to apply only to the subset of forms, namely those in (75) a., discussed above, i.e. the indefinite article and the composite forms, if not explicitly stated otherwise. In the remainder of this subchapter, I will touch on additional accounts focusing on personal pronouns and discuss their relation to one another as well as 66 Observe that I refrain from making predictions about the morphological complexity of the remaining genitive/ possessive pronoun roots here and instead represent these forms by the most reduced strings found in their respective paradigm as well as without a hyphen; see Roehrs (2009: ch. 4.3) for a relevant discussion, cf. also ch. 3, fn. 27 below. 74 1 AXIOMS - Adjectival Inflection & Phrasal Categorization to the system elaborated here. I will restrict myself to a broader outline since the accounts under consideration will surface again in different contexts in the course of this inquiry and will be introduced therein with the relevant foci respectively. I nevertheless believe a preliminary and collective introduction to the underlying rationale to support the accessibility of the following discussion. With this in mind, and returning to the early approaches to the syntax of pronouns laid out at the beginning of this section, recall that pronouns have been uniformly taken to be located in parallel position to determiners in DP. As Déchaine & Wiltschko (2002) note, such accounts consequently have to retrace different distributions of pronominal elements to their respective internal structure and moreover account for the visibility of said internal (i.e. featural) make-up in Narrow Syntax. In opposition to this rationale, the authors propose that pronouns are not primitives and track variations in distribution back to diverging categorical status, viz. either DP-, NPor ϕP-level, the latter located between the first two, with the resulting pronominal phrases termed ‘pro-XPs’ respectively. Pronouns are hence of varying structural complexity in syntax (with more complex elements structurally incorporating all simpler ones) that mirrors their syntactic and semantic properties, 67 e.g. levelling argument vs. predicate status to DPvs. NP-level, with reference to Longobardi (1994). 68 ϕP is understood as constituting the midway between these two extremes, parallel to its structural position, and is hence characterized as variable w.r.t. most of the distinguishing properties (cf. Déchaine & Wiltschko 2002: (24)). Since the proposed divisions hold in single languages as well as cross-linguistically, the system is able to account for pronominal paradigms in a number of unrelated languages (e.g. Halkomelem pronouns, Shuswap pronominals or Japanese kare as pro-DPs, pro-ϕPs and pro-NPs respectively). In relating morpho-syntactic and semantic properties to diverging structural set-ups, the account parallels C&S’ approach to pronominal elements, laid out in the previous subchapter, as noted by Déchaine & Wiltschko themselves. Even more so, the authors propose the translatability of C&S’ three classes strong, weak and clitic pronouns into their typology (as ϕP with NP complement, ϕP without internal structure and ϕ 0 respectively, cf. Déchaine & Wiltschko 2002: ch. 6). Moreover, they expect their 67 The account is compatible with the decompositional approaches laid out previously in this subchapter, as the authors demonstrate with various English SLIs building on the root th -, which are analyzed as two adjacent but separate heads (i.e. [D-ϕ]) for - what they call - the bound D-morpheme and the clitic, cf. Déchaine & Wiltschko (2002: ch. 3.2). 68 Longobardi himself elsewhere (2008: 201) locates pronouns in DP, derived via movement universally for languages where said phrase is part of the extended projection of the noun. 1.4 A Phrasal Set-Up of the Extended Nominal Domain in German 75 approach to expand beyond pronouns and incorporate reflexives, pro, agreement inflection and even full XPs in the tripartition of the categories D/ ϕ/ N and their respective characteristics. This stands in contrast to the aim articulated in the approach to pronominal elements by Harley & Ritter, the so-called ‘Feature Geometry’, which is explicitly “designed to represent the distinctions in functional or grammatical morphemes, specifically pronouns” (Harley & Ritter 2002a: 506). 69 The theory departs from the rationale laid out above in that complexity of LIs is not coded in the extended projection. Rather, Harley & Ritter propose a complex interdependency of (φ-)feature values in - what they call - re s (referring expressions), i.e. pronouns. The authors exemplify this concept without reference to actual features by means of the sample geometry, given in (76) below: (76) (= Harley & Ritter 2002a: (5)) Adjectival Inflection & Phrasal Categorization A XIOMS 38 most of the distinguishing properties (cf. Déchaine & Wiltschko 2002: (24)). Since the proposed divisions hold in single languages as well as cross-linguistically, the system is able to account for pronominal paradigms in a number of unrelated languages (e.g. Halkomelem pronouns, Shuswap pronominals or Japanese kare as pro-DPs, pro-ϕPs and pro-NPs respectively). In relating morpho-syntactic and semantic properties to diverging structural set-ups, the account parallels C&S’ approach to pronominal elements, laid out in the previous subchapter, as noted by Déchaine & Wiltschko themselves. Even more so, the authors propose the translatability of C&S’ three classes strong, weak and clitic pronouns into their typology (as ϕP with NP complement, ϕP without internal structure and ϕ 0 respectively, cf. Déchaine & Wiltschko 2002: ch. 6). Moreover, they expect their approach to expand beyond pronouns and incorporate reflexives, pro, agreement inflection and even full XPs in the tripartition of the categories D/ ϕ/ N and their respective characteristics. This stands in contrast to the aim articulated in the approach to pronominal elements by Harley & Ritter, the so-called ‘Feature Geometry’, which is explicitly “designed to represent the distinctions in functional or grammatical morphemes, specifically pronouns” (Harley & Ritter 2002a: 506). 78 The theory departs from the rationale laid out above in that complexity of LIs is not coded in the extended projection. Rather, Harley & Ritter propose a complex interdependency of (φ-)feature values in - what they call - RE s (referring expressions), i.e. pronouns. The authors exemplify this concept without reference to actual features by means of the sample geometry, given in 0 below: (76) (= Harley & Ritter 2002a: (5)) [A] qp [B] [E] g qgp [C] [F] [H] [I] g g 3 [D] [G] [J] [K] The basic rationale behind this approach breaks up the bipartition of features and values in favor of a hierarchy of monovalent features that are only ever represented in the featural set-up of an LI (i.e. the root node [A] in 0 above) if the respective node is projected. These nodes might themselves again possess dependents of their own: Dominance hence translates to entailment between features. Additionally, markedness of values is derived therein via node-counting, with the most marked values deeply embedded in the hierarchy. Apart from these monovalent nodes, the authors also propose the existence of bare organizing nodes that do not themselves code any value. Additionally, the notion of default node/ value is employed for three nodes in the hierarchy, corresponding to the three φ-feature categories person, number and gender. The authors take the feature geometry to constitute a part of UG, hence to apply language-universally, with the paradigms of RE s across languages derived by the various patterns of co-occurrence of active nodes, a claim discussed w.r.t. pronominal systems in a corpus of 110 languages as well as data from language acquisition (cf. Harley & Ritter 2002a: ch. 3, Hanson et al. 2000). This approach will surface again in chapter 2.3 and prominently throughout chapter three below, where it is laid out in detail and discussed with the relevant foci respectively. For now, observe that the account circumvents Déchaine & Wiltschko’s criticism concerning the featural set-ups of structurally equal LIs, laid out above, since the authors neither make any claim about the categorical status of pronouns, nor elaborate on the properties identified by Déchaine & Wiltschko to separate their three classes D/ ϕ/ N (i.e. argument vs. predicate status, definiteness and binding-theoretic status). Furthermore, as Harley & Ritter state several times (cf. Harley & Ritter 2002a: 510, 518), their approach puts forward an interpretive model of morphology (cf. e.g. Halle & Marantz 1993) that interprets morphosyntactic features post-syntactically and hence challenges their narrow-syntactic relevance which is at the heart of Déchaine & Wiltschko’s criticism; a point which will become relevant in the following chapters. However, the authors also note (cf. Harley & Ritter 2002a: 518) that nothing in the theory of feature geometry hinges on the interpretive nature of the morphological component, which renders the two approaches compatible. Cowper & Curie Hall (2009), then, elaborate on a mediate account that combines the categorical destinctions from Déchaine & Wiltschko with the feature geometric ordering of monovalent φ-features/ values (and definiteness) from Harley & Ritter to account for a variety of SLIs in English (with a focus on personal pronouns): The combination applies along a split in the geometry and the division of the resulting sub-hierarchies to the functional projections of the extended nominal domain, namely #/ ϕ/ D. The former two heads represent subparts of Déchaine & Wiltschko’s mediate projection ϕ, whose variable nature is hence resolved (with all instances of arguments now either ϕ or D, all predicates either of category # or N). Cowper & Curie Hall’s distribution takes gender to be coded along with plural on the number head #, person to be located 78 Cf. also Harley & Ritter (2002b), Hanson et al. (2000), Harley (1994). The basic rationale behind this approach breaks up the bipartition of features and values in favor of a hierarchy of monovalent features that are only ever represented in the featural set-up of an LI (i.e. the root node [A] in (76) above) if the respective node is projected. These nodes might themselves again possess dependents of their own: Dominance hence translates to entailment between features. Additionally, markedness of values is derived therein via node-counting, with the most marked values deeply embedded in the hierarchy. Apart from these monovalent nodes, the authors also propose the existence of bare organizing nodes that do not themselves code any value. Additionally, the notion of default node/ value is employed for three nodes in the hierarchy, corresponding to the three φ-feature categories person, number and gender. The authors take the feature geometry to constitute a part of UG, hence to apply languageuniversally, with the paradigms of re s across languages derived by the various patterns of co-occurrence of active nodes, a claim discussed w.r.t. pronominal systems in a corpus of 110 languages as well as data from language acquisition (cf. Harley & Ritter 2002a: ch. 3, Hanson et al. 2000). This approach will surface 69 Cf. also Harley & Ritter (2002b), Hanson et al. (2000), Harley (1994). 76 1 AXIOMS - Adjectival Inflection & Phrasal Categorization again in chapter 2.3 and prominently throughout chapter three below, where it is laid out in detail and discussed with the relevant foci respectively. For now, observe that the account circumvents Déchaine & Wiltschko’s criticism concerning the featural set-ups of structurally equal LIs, laid out above, since the authors neither make any claim about the categorical status of pronouns, nor elaborate on the properties identified by Déchaine & Wiltschko to separate their three classes D/ ϕ/ N (i.e. argument vs. predicate status, definiteness and bindingtheoretic status). Furthermore, as Harley & Ritter state several times (cf. Harley & Ritter 2002a: 510, 518), their approach puts forward an interpretive model of morphology (cf. e.g. Halle & Marantz 1993) that interprets morpho-syntactic features post-syntactically and hence challenges their narrow-syntactic relevance which is at the heart of Déchaine & Wiltschko’s criticism; a point which will become relevant in the following chapters. However, the authors also note (cf. Harley & Ritter 2002a: 518) that nothing in the theory of feature geometry hinges on the interpretive nature of the morphological component, which renders the two approaches compatible. Cowper & Curie Hall (2009), then, elaborate on a mediate account that combines the categorical destinctions from Déchaine & Wiltschko with the feature geometric ordering of monovalent φ-features/ values (and definiteness) from Harley & Ritter to account for a variety of SLIs in English (with a focus on personal pronouns): The combination applies along a split in the geometry and the division of the resulting sub-hierarchies to the functional projections of the extended nominal domain, namely #/ ϕ/ D. The former two heads represent subparts of Déchaine & Wiltschko’s mediate projection ϕ, whose variable nature is hence resolved (with all instances of arguments now either ϕ or D, all predicates either of category # or N). Cowper & Curie Hall’s distribution takes gender to be coded along with plural on the number head #, person to be located in ϕ and specificity/ definiteness to be in D. The authors thereby grant narrow-syntactic relevance to the features/ values of the feature geometry: “While ϕ and D are used as syntactic category labels throughout the discussion, it is worth bearing in mind that they are, like their dependents , simply morphosyntactic features.” (Cowper & Curie Hall 2009: 103, [emphasis added, MB]). Therein, the account becomes subject to the criticism of Déchaine & Wiltschko since categorically equal elements do indeed differ w.r.t. their distribution. Focusing e.g. on the indefinite article once again, the proposed structural position is the head of DP, along with the determiner, demonstratives as well as the weak quantifier sm , which have doubtlessly been shown to differ in their distribution in German throughout this chapter and do so also in English. This instantiation of a narrow-syntactic application of the feature geometry approach hence falls short to account for the syntactic diversity that is coupled with the general structural uniformity. 1.5 Conclusion 77 However, I believe the basic rationale to be on the right track. As I will show below, the narrow-syntactic relevance of morpho-syntactic features and their ordering can be postulated while circumventing Déchaine & Wiltschko’s criticism, laid out above. This section focused on the introduction of the concept of ein-word from Roehrs (in prep., 2009) as well as the feature geometric approach to pronominal elements from Harley & Ritter (2002a). This was done by discussing a certain class of SLIs, namely personal pronouns, which are not clearly located in the hierarchy of projections of the nominal domain developed in previous subchapters. I have shown that the ordering of semi-lexical items based on a shared morphological subpart has been proposed several times in recent publications. In German, the root ein -, part of the morphological structure of various nominal SLIs, readily suggests itself for such a treatment. These accounts, however, fail to trace morphological parallels back to structural uniformity; I will present parallel syntactic behavior of the SLIs clustered in (75) a. above later in this book, namely in the discussion of various agreement patterns in chapter three. Hence, I believe the concept to be useful for the present analysis. The second concept introduced in this section will be picked up in the following chapter and will undergo modification there and more extensively throughout the course of chapter three. The relevant discussions therein will be introduced by a more in-depth presentation of the account to which the current subchapter provided a first overview. 1.5 Conclusion The current chapter set out to establish two unrelated axioms of the analysis to follow. First, following a general introduction to the theoretical framework in which the current inquiry is couched, I devoted some length to demonstrate the syntactic relevance of definiteness in German by presenting data that parallel the respective testing environments from English. I then went on to introduce the split in adjectival inflectional patterns in German, ultimately presenting two dedicated paradigms, called strong and weak , as well as a minor third one that cuts through the space of uniform inflection in the dimensions of number as well as case, called mixed . Finally, I presented evidence for a close relatedness of these phenomena as interdependent, i.e. the selection of strong or weak inflection on adjectives rests on the status of Milsarkian weak/ strong semi-lexical items, qua status of (in-)definiteness, concatenated with them (and vice versa). Second, I devoted the rest of the chapter to carve out the number and order of projections that make up the proposed functional spine, erected over nominal 78 1 AXIOMS - Adjectival Inflection & Phrasal Categorization heads in German. The hierarchy elaborated above will serve as the structural basis for the analysis to follow. Therein, every phrasal category, as well as the selection of SLIs that is considered to occupy their respective head-positions, has been cautiously justified in the review of earlier proposals, not related to the current topic of investigation. Therein, the structural foundation of the analysis to follow is shielded against allegations of ad hoc composition. The two final points are summarized in (77) a. and b., below. (77) a. (= (29)) SLI definite ↔ Adj WEAK SLI indefinite ↔ Adj STRONG b. Adjectival Inflection & Phrasal Categorization A XIOMS 40 (77) a. (= 0) SLI definite ↔ Adj WEAK SLI indefinite ↔ Adj STRONG b. Q S P 3 Q S DP 3 D Poss PRO P 3 Poss PRO Q W P 3 Q W NP 3 N Poss LEX P Note that the notions depicted above constitute the two basic postulates that the analysis to follow will rest on. Therefore, rejecting either (i) the interdependence of definiteness and adjectival inflection illustrated in 0 a. or (ii) the nominal hierarchy of projections depicted in 0 b. amounts to the simultaneous rejection of all conclusions, drawn from this point in my investigation onwards (parallel to the rejection of premises of the theory itself, laid out in section 0 above). As I hope to have demonstrated in this chapter, however, there is good evidence to follow this line of reasoning further. The following chapter, then, starts out by focusing on data from co-occurrence and ellipsis of semi-lexical items as well as patterns of extraction from nominal hierarchies to derive the phasal status of the phrase heads identified above. As will be shown therein, the hierarchy elaborated here is able to account for the complete set of linearizations under the theory of contextual phasehood from Bošković (2014). The second part of the chapter will thereafter take the linearizational and inflectional idiosyncrasies, that will have surfaced prominently in the aforementioned discussion, as a vantage point to elaborate on the coding of φ-features in the nominal domain in German and ultimately arrive at a modification of the hierarchy in 0 b. above, incorporating inter alia the approaches on the narrow-syntactic relevance and coding of nominal features laid out in the previous section. However, it will only be as late as chapter three that the insights from previous sections will be combined with those modifications just outlined to arrive at a principled account of nominal concord that subsumes the interdependency outlined in 0 a. above. Fortunately, it will moreover account for the general distribution of formal features in the nominal domain in a principled way in the end. Note that the notions depicted above constitute the two basic postulates that the analysis to follow will rest on. Therefore, rejecting either (i) the interdependence of definiteness and adjectival inflection illustrated in (77) a. or (ii) the nominal hierarchy of projections depicted in (77) b. amounts to the simultaneous rejection of all conclusions, drawn from this point in my investigation onwards (parallel to the rejection of premises of the theory itself, laid out in section 1.2 above). As I hope to have demonstrated in this chapter, however, there is good evidence to follow this line of reasoning further. The following chapter, then, starts out by focusing on data from co-occurrence and ellipsis of semi-lexical items as well as patterns of extraction from nominal hierarchies to derive the phasal status of the phrase heads identified above. As will be shown therein, the hierarchy elaborated here is able to account for the complete set of linearizations under the theory of contextual phasehood from Bošković (2014). The second part of the chapter will thereafter take the 1.5 Conclusion 79 linearizational and inflectional idiosyncrasies, that will have surfaced prominently in the aforementioned discussion, as a vantage point to elaborate on the coding of φ-features in the nominal domain in German and ultimately arrive at a modification of the hierarchy in (77) b. above, incorporating inter alia the approaches on the narrow-syntactic relevance and coding of nominal features laid out in the previous section. However, it will only be as late as chapter three that the insights from previous sections will be combined with those modifications just outlined to arrive at a principled account of nominal concord that subsumes the interdependency outlined in (77) a. above. Fortunately, it will moreover account for the general distribution of formal features in the nominal domain in a principled way in the end. 80 2 PREMISES - Phasehood & φ-Dependencies 2 PREMISES - Phasehood & φ-Dependencies 2.1 Introduction After having sketched a broad outline of the phrasal set-up in which I will conduct my analysis as well as the presentation of the empirical underpinning and theoretical predecessors, the current chapter will concentrate on the structural and featural set-up of the proposed hierarchy of projections in German. Therein, I will first take a closer look at phasehood in the nominal domain. As will be shown below, by reviewing data from ellipsis and extraction from the nominal spine, several functional heads are able to exhibit the property of acting as a phase head depending on the structural environment they are part of. An elaborate system of assignment of contextual phasal status will emerge out of this analysis. In the following subsections, I will then turn my attention to the coding of φ-features in the nominal hierarchy of projections. As I will show, several seemingly unrelated as well as unaccounted phenomena of inflection in the German NP readily fall into place once a reordering (and partly relocation) of dependencies of these features is employed. At the end of this chapter, a coherent picture of phasal and featural configurations will have emerged that readily lends itself to the detailed analysis of nominal concord and morphological definiteness therein in the ensuing chapter. Since the proposals elaborated here hence constitute the basis for my main arguments, carved out subsequently in this book, while they are at the same time based on the axioms laid out above, this section has been termed the ‘premises’ of the analysis. 2.2 Phasehood 81 2.2 Phasehood 2.2.1 Phasal Status in the Left Periphery Let me begin by concentrating first on the functional projections identified over the course of chapter 1.4 above. As mentioned therein, Despić (2011) proposes a tight connection between DP and its complement PossP in that the phasal status of the former depends on its co-occurrence with the latter. 1 This line of reasoning originates from Chomsky’s (OP) remodeling of the source of [ u F]s in a given derivation (concerning the relation of C and T) as originating on the phase head, which ‘hands down’ a subset to its non-phasal complement by socalled ‘feature inheritance’. T is hence in itself defective up to the point when C is merged: It does not trigger any operations since it does not possess any [ u F]s. The conceptual advantage of this approach lies in the timing of Agree (i.e. Value) and TRANSFER. Chomsky (DbP) notes that after the application of Agree, interpretable and uninterpretable (i.e. formerly unvalued) features are indistinguishable. TRANSFER hence has to apply “shortly after the uninterpretable features have been assigned values” (DbP: 5). As Richards (2007: 566, referring to Epstein & Seely 2002) notes, these operations, in fact, have to happen simultaneously (hence incorporating Value into TRANSFER) if the derivation is to avoid crashing at SEM (Value before TRANSFER, since [ u F]s are illegal at SEM) or at both interfaces (TRANSFER before Value, since it violates Full Interpretation). Locating [ u F]s on the next higher phase head (i.e. the next Spell-Out cycle) permits said simultaneity of operations, depicted in (1) c.-f., along the following lines (Ph a phase head, Pr its probing, non-phasal complement, G a goal for the uninterpretable feature [ u F]): 1 There is a lot of evidence that DP itself constitutes a phase in general, cf. Bošković (2014: fn. 6) and references therein. 82 2 PREMISES - Phasehood & φ-Dependencies (1) a. [Pr [[G F ] Ph 1 ]] Pr merges with the edge of the lower phase Ph 1 b. [Ph 2 [Pr [[G F ] Ph 1 ]]] Ph 2 merges c. [Ph 2 [Pr u F, EPP [[G F ] Ph 1 ]]] Ph 2 hands [ u F] (and possibly EPP 2 if Pr=T, cf. OP: 157) down to its non-phasal complement d. [Ph 2 [[G F ] Pr u F, EPP [<G F > Ph 1 ]]] Pr probes into its complement, [ u F] renders subparts of Pr and G active (cf. DbP: 4), EPP attracts G, G moves to Spec,Pr e. [Ph 2 [[G F ] Pr u F [<G F > Ph 1 ]]] Agree applies between Pr u F and G F f. [Ph 2 <[G F ] Pr u F [<G F > Ph 1 ]>] Ph 2 is transferred, leaving the edge of Ph 2 In ch. 4.2.1 below I will focus on the movement of G to Spec,Pr (depicted in (1) d.) in some length; for now, let us note that feature inheritance allows a phasal, cyclic derivation to survive at Σ/ SEM and Φ/ PHON. Richards, however, points out a conceptual flaw in the system presented above: Although it is obvious why the defective non-phase (i.e. T) needs a phasal head dominating it (i.e. C), it is not at all clear why said phase needs the non-phasal complement, i.e. why are (defective) non-phases part of FL at all if phase heads could trigger all operations on their own, 3 viz. how do non-phasal XPs conform to the Strong Minimalist Thesis? This criticism becomes even more pressing under the general rationale of the DbP/ OP-accounts to reduce derivational complexity qua memory load. Richards derives the necessity for non-phasal complements by combining (i) Value-TRANSFER simultaneity, mentioned above, with (ii) the premise of the edge of a phase (head/ label and specifier(s) of phase heads) as part of the next cycle of Spell-Out, the ‘Phase Impenetrability Condition’ (PIC): The necessity 2 ‘Extended Projection Principle’, later generalized to an optional specification for phase heads as the ‘OCC’: ‘occurrence’. Throughout the minimalist inquiry, the proposed featural content and syntactic implementation of the EPP/ OCC have undergone various radical modifications (cf. Butler 2004: ch. 1 for an overview compiling Chomsky’s own implementations). I will ultimately settle for a different relation of the EPP on the head T and its probing (φ-)features in chapter 3.2 below, but treat them as independent here for ease of exposition since it does not interfere with the point I want to demonstrate. Furthermore, I will treat the notions synonymous under the label ‘EPP/ OCC’ for the course of this study but refer to the feature on T as the EPP only. 3 In a system that allows multiple specifiers, cf. Richards (2007: 564f.) on Chomsky’s (OP) arguments for the phase-plus-non-phase set-up on the basis of the A/ Ā distinction for C-T. 2.2 Phasehood 83 for non-phasal complements follows since otherwise, no transphasal movement would be possible at all. In Richards’ (2007: 569) words: “(1) (= interpretability) and (2) (= impenetrability) together entail (3) (= inheritability).” The set-up laid out above therefore poses a natural strategy to guarantee successivity. Other than the A/ Ā-argument mentioned in fn. 3 above, the successivity-argument is applicable to all phases. Richards hence speculates on a very rigid phrasal setup on the basis of SMT/ good design, consisting solely of pairs of phases (P) and their non-phasal complements (N): (2) (= Richards 2007: (4) e.) P - N - P - N… With this much in place, let us turn to the nominal domain. Bošković (2014) argues for a contextual (in contrast to a rigid) assignment of phasehood to phrases in the extended projection of lexical categories, mainly on the basis of the contrast between English and Serbo-Croatian. He arrives at the principle in (3), due to different restrictions in languages which implement a DP layer (English) and those in which the lexical category N is arguably the highest phrase in the nominal domain (Serbo-Croatian). (3) (= Bošković 2014: (56) b.) The highest projection in a TNP [Traditional Noun Phrase, MB] is a phase. His two testing grounds are extraction and ellipsis from/ in the nominal domain. I will discuss these for German, beginning with the latter: Bošković elaborates his claim concerning ellipsis on the basis of Japanese, a determinerless language, and English nominal domains; consider the examples below: (4) ([ Japanese]) a. (= Bošković 2014: (48), [from Saito, Lin, and Murasugi 2008]) [Taroo-no taido-wa] yo-i ga, [ TNP Hanako-no [ NP taido]-wa] yoku-na-i. Taro- GEN attitude- TOP good- PRES though Hanako- GEN attitude- TOP good-not- PRES ‘Though Taro’s attitude is good, Hanako’s isn’t.’ b. (= Bošković 2014: (50), [from Watanabe 2010]) Taroo-wa yon-satsu-no hon-o kat-ta ga, sono-uti Taro- TOP four- CL - GEN book- ACC buy- PAST though that-out.of [ni-satu [hon]-o] sudeni yomi-oe-ta. two- CL book- ACC already read-finish- PAST ‘Taro bought four books, but he already finished reading two of them.’ 84 2 PREMISES - Phasehood & φ-Dependencies c. (= Bošković 2014: (59), (61), [from Takahashi 2011]) i. A sensei-wa subete-no Taroo-no tikoku-o yurusi-ta. Prof. A- TOP all- GEN Taro- GEN tardiness- ACC forgive- PAST ‘Lit. Prof. A forgave all Taro’s tardiness.’ ii. B sensei-wa hotondo Ziroo-no tikoku-o yurus-anakat-ta. Prof. B- TOP most Ziro- GEN tardiness- ACC forgive- NEG - PAST As can be deduced from (4) a., topic particles survive partial NP-ellipsis (as does the case particle in (4) b.) together with the genitive possessor remnant. Takahashi (2011) hence locates them in head position of KP (which takes NP as its complement) outside of the ellipsis-site and identifies the possessor as an adjunct of KP. The absence of Condition C violations with quantified NPs like (4) b., as opposed to non-quantified nominals, points to additional structure above KP in those configurations, 4 hence to a switch of the phasal status of KP. The head K, however, is taken to move to Q. In these structures, the numeral together with the case particle survives ellipsis. Adding a possessor to a quantified nominal structure ((4) c. ii., with i. the antecedent for ellipsis) only allows ellipsis of the head noun (HN) in combination with said possessor. Condition C violations of quantified as opposed to non-quantified nominals (the KP/ QPvariability) in combination with (3) above and the data presented in (4) thus yields (5): (5) (= Bošković 2014: (56) a.) Only phases and complements of phase heads can undergo ellipsis. I will take (5) to constitute a solid testing ground for phasehood in what follows. 5 2.2.1.1 The upper Functional Domain in German a. Ellipsis Let us turn to the functional domain elaborated in chapter one. According to (5), apart from full NP-ellipsis, the edge of the nominal domain should be able to survive this operation. With demonstratives (D), possessive pronouns (Poss PRO ) in singular and plural NPs as well as strong quantifiers (Q S ) with plural HNs, this is exactly what we find in German: 4 Bošković (2014) reports parallel effects on binding in Serbo-Croatian with certain quantifiers and numerals; cf. fn. 15 below. 5 Bošković himself (2014: fn. 22), however, notes that various independent factors might prevent ellipsis of phases/ phasal complements across languages, see Boeckx (2009), Gengel (2009), Takahashi (2011) on the claim that only phasal complements can elide. 2.2 Phasehood 85 (6) Du hast Fredericks Buch gelesen und ich habe you have F.’s book read and I have a. dieses <Buch> this < book > b. seines <Buch> gelesen. his < book > read (7) Du hast Fredericks Bücher gelesen und ich habe you have F.’s books read and I have a. diese <Bücher> these < books > b. seine <Bücher> his < books > c. alle <Bücher> gelesen. all < books > read Either the three categories display phasal status in accordance with (5) in isolation or we observe full phasal ellipsis with both singular and plural HNs above. The former would additionally suggest the possibility of full DP-deletion, an option that is generally not available in German (see section 2.2.1.2 below). What is remarkable, however, is their behavior in co-occurrence: 6 (8) Du hast Fredericks Bücher gelesen und ich habe you have F.’s books read and I have a. * diese seine <Bücher> 7 these his < books > b. alle seine <Bücher> all his < books > c. alle diese <Bücher> gelesen. all these < books > read 6 For now, I will concentrate on linearizations with plural HNs to include structures incorporating the strong quantifier and withhold the discussion of singular HNs until section 2.2.1.2 below. 7 Note that the grammaticality of the structure improves with the use of the first person singular possessive pronoun. This is surely a pragmatic effect, since the set that 86 2 PREMISES - Phasehood & φ-Dependencies Although the emphatic combination of D and Poss PRO is licit in German (recall the discussion concerning possessor doubling constructions (PDCs) in ch. 1.4.3 above), it is ruled out in ellipsis contexts like (8) a., whereas the quantifier readily precedes the possessive and demonstrative. On the assumption that (6)-(8) represent cases of full phasal ellipsis of HN in accordance with (5), (8) a. should turn out to be fine, too. Before I elaborate on this, let me present an extended paradigm including prenominal lexical possessives with D and Q S , as well as with Poss PRO in PDCs: 8 (9) Du hast Fredericks Bücher gelesen und ich habe you have F.’s books read and I have a. Martins <Bücher> Martin GEN < books > the pronoun ranges over is more easily accessible. Without a doubt, semantic as well as pragmatic constraints additionally restrict grammaticality and hence the combinatorial properties of the lexical elements under consideration. My aim, as stated in chapter one, however, is to isolate the narrow-syntactic restrictions, employed in this domain; I therefore treat such concatenations as ungrammatical here. Additionally, one reviewer suggests that a differentiation should be made in the context of this concatenation regarding the option of a so-called ‘comma intonation’, inserting a pause between the SLIs. As a narrow reply to this remark, I have to note that, while the insertion of such a (rather significant) pause is in principle possible at the indicated position, switching to a highly stylized register, the grammaticality of such configurations did not depend on a deviating intonation pattern for my informants and no such restriction had been mentioned during the course of the elicitation interviews. I personally likewise conceive the datum in parallel to e.g. the concatenation of Q S and D, devoid any of extra-syntactic constraints. A wider reply might aim at a structural contrast giving rise to the deviant intonation. Here, the remark presumably targets the uniformity of the complex nominal domain, a notion that translates structurally to the property of a shared projection line. However, I argue in chapter 3.3.3 b. below, that the possessive pronoun constitutes an intervener for probing efforts of the dominating demonstrative in concatenation with a subset of nominal φ-specifications, to be made precise below, resulting in ungrammaticality. Grammatical cases of the concatenation, however, exist with both default as well as deviant intonation; φ-values are thus in principle accessible to the higher SLI in both cases. The effect of ungrammaticality should nevertheless reduce with the deviant intonation since the structural precondition has arguably been eliminated. As far as I can tell, this is not the case and patterns of (un-)grammaticality are unaffected by intonation. I therefore conclude that deviant intonation is structurally irrelevant to the phenomenon under consideration. 8 Curly brackets in (9) b. and d. denote the varying possible surface positions w.r.t. the elements under consideration. Parentheses additionally enclose genitive case-suffixes. Note that these configurations turn out ungrammatical in every linearizational as well as case-marking option. 2.2 Phasehood 87 b. * {Martin(-s)} diese {Martin(-s)} <Bücher> { Martin DAT/ GEN } these { Martin DAT/ GEN } < books > c. Martin seine <Bücher> Martin DAT his < books > d. * {Martin(-s)} alle {Martin(-s)} <Bücher> gelesen { Martin DAT/ GEN } all { Martin DAT/ GEN } < books > read The only grammatical combination involves the two possessive elements pro and lex . We can, however, not straightforwardly conclude that the lexical possessive is located in the same hierarchical positions in a. and c. (but see endnote *10). Let me extend the data involving PDCs further by combining (9) c. with the heads Q S and D respectively. (10) Du hast Fredericks Bücher gelesen und ich habe you have F.’s books read and I have a. all Martin seine <Bücher> all Martin DAT his < books > b. * diese Martin seine <Bücher> gelesen. these Martin DAT his < books > read As can be derived from (10), the quantifier readily combines with the PDCs, whereas the combination with the demonstrative results in ungrammaticality in NP-ellipsis configurations. What conclusions can be drawn from these observations for the phasal status of the three topmost functional heads? According to (6) and (7), these heads adhere to (5) when occurring in isolation. We hence tend to conclude that they are phases in these cases, the complement of which is elided. The combination of two out of these elements in (8) yields surprising effects, however, since D as well as Poss PRO can survive ellipsis together with the strong quantifier (cf. b., c.), but not in combination with each other (cf. a.). Under the functional hierarchy elaborated here, the complement of the complement of the phase head by hypothesis (3) is elided in the former. Additionally, the latter structure would not pose a problem if we were dealing with full phasal deletion of HNs rather than phasal complements above. Turning to (9), recall that I located the prenominal lexical possessive in the specifier of Poss PRO P in PDCs. The data are hence exactly as we would expect under the highest-phrase-as-phase account of Bošković: Only with the possessive pronoun does the lexical possessive occupy the edge (i.e. specifier) of the phase and hence survives ellipsis, whereas it is ‘buried’ in the complement’s complement in both (9) b. and d. The last surprising datum, then, is (10) a., an extension of the puzzling (8) b., which 88 2 PREMISES - Phasehood & φ-Dependencies permits the (now complex) complement - the PDC - of the phase head qua highest head to survive, whereas this is not possible with D preceding the same complex structure. We can conclude that (8) b., c. and (10) a. are unaccounted for under (3) and (5). A closer look at these data reveals that it is the strong quantifier whose properties are not captured by (3) and (5). This element can co-occur with D, Poss PRO and PDCs, but as can be derived from (7) c., it can behave as a phase in isolation. Poss PRO exhibits the phasal property specified in (5) when occurring in isolation (cf. (6) b./ (7) b.) and in PDCs ((9) c.) in accordance with (3)/ (5), but ceases to constitute a phase in the presence of the demonstrative both as a head or as a part of PDCs (cf. (8) a., (10) b.). Lastly, D always adheres to (3)/ (5): It acts as a phasal head in isolation and is barred from co-occurrence with its possibly complex possessive complement ((6) a./ (7) a., (8) a./ (10) b.). 9 If we strive to maintain the phrasal set-up developed above, (3)/ (5) hence have to be modified accordingly. I want to argue that (5) holds and that the highest-phrase-as-phase rationale cannot be carried over to German; the motivation for this approach will become clear immediately. Given the data above, (5) can be retained, if phasehood in the nominal domain in German is not derived by estimation of the highest element in the extended projection, but rather selected by a non-linear hierarchy in that the presence of category α bars β from obtaining phasehood. The order of the elements under consideration in this hierarchy should be obvious by now: (11) Assignment of Phasehood in the Nominal Domain (first version) D > Poss PRO > Q S Given (11), the conspicuous properties of the strong quantifier readily fall into place with the phasal properties of its respective environment: Q S can precede D, Poss PRO and PDCs, since all of these elements/ complexes involve a terminal node which is assigned phasehood, thereby closing off the highest nominal phase. It does, however, act as a phase in the absence of demonstratives and possessive pronouns. In these cases, the lexical possessive is located (somewhere) in the complement of the phase head and can hence not co-occur with the quantifier. As can be derived from (11), Q S is the only phrase violating the generalization in (3) in its property as the least popular, but structurally highest, phase head candidate. 9 The properties of prenominal lexical possessives in PDCs should readily carry over to Spec,DP in these cases, which is not testable due to the lack of appropriate data. 2.2 Phasehood 89 b. Extraction from NP Building on (11), we can turn to the second testing ground for phasal status from Bošković (2014). Recall from the beginning of this chapter that Richards (2007) delivered substantial evidence that every phase head needs at least/ best one non-phasal complement. As has been laid out, his argument is based on the combination of TRANSFER-Value simultaneity and the PIC. Bošković, referring to Abels (2003), reports another such argument, solely couched inside Narrow Syntax, the relevant notions of which are once again the PIC paired with Anti-Locality (cf. Grohmann 2000, 2003a/ b): As Abels notes, whereas the PIC forces movement out of a phase to proceed through the edge of said phase, Anti-Locality forbids ‘too short’ movement of the complement of α to its own specifier Spec,αP. For him, Anti-Locality applies universally at every phrasal boundary. Movement of the complement of a phase head is thereby ruled out. As Bošković (2014: 29) correctly notes, “if all phrases are phases, complements are rendered completely immobile.” A second testing ground for phasehood readily presents itself: (12) (= Bošković 2014: (9)) Abels’ Generalization Complements of phase heads cannot undergo movement. Turning again to the phrasal set-up under consideration, this is exactly what we find in German concerning the hierarchy elaborated above. 10 In what follows, I have underlined the respective phase heads according to (11). (13) a. [Diese Bücher] habe ich gelesen. [ these books ] have I read a'. * Bücher i habe ich [diese t i ] gelesen. books i have I [ these t i ] read b. [Seine Bücher] habe ich gelesen. [ his books ] have I read 10 Of course, as remarked by one reviewer, deriving discontinuous nominal configurations by movement of its subpart is not the sole option and various alternatives have been proposed in the literature, ranging from e.g. base-generation of two decided nominal domains, either VP-internal (cf. Fanselow 1988, Roehrs 2009) or in VP and CP, paired with co-indexation (cf. Haider 1990) or movement of the core NP (cf. Tappe 1989), to full DP movement combined with partial elision (cf. Fanselow & Ćavar 2002). Nevertheless, movement-based approaches targeting subparts of the nominal domain constitute the standard treatment in the generative framework today. Trivially, Bošković’s (2014) testing ground, on which this section rests, likewise incorporates this premise. I thus follow the rationale without further comment. 90 2 PREMISES - Phasehood & φ-Dependencies b'. * Bücher i habe ich [seine t i ]gelesen. books i have I [ his t i ] read c. [Alle Bücher] habe ich gelesen. [ all books ] have I read c'. * Bücher i habe ich [alle t i ]gelesen. books i have I [ all t i ] read (14) a. [Diese seine Bücher] habe ich gelesen. [ these his books ] have I read a'. * [Seine Bücher] i habe ich [diese t i ] gelesen. [ his books ] i have I [ these t i ] read b. [Alle seine Bücher] habe ich gelesen. [ all his books ] have I read b'. [Seine Bücher] i habe ich [alle t i ] gelesen. [ his books ] i have I [ all t i ] read c. [Alle diese Bücher] habe ich gelesen. [ all these books ] have I read c'. [Diese Bücher] i habe ich [alle t i ] gelesen. [ these books ] i have I [ all t i ] read (15) a. [Martin seine Bücher] habe ich gelesen. [ Martin his books ] have I read a'. * [Bücher] i habe ich [Martin seine t i ] gelesen. [ books ] i have I [ Martin his t i ] read (16) a. [All Martin seine Bücher] habe ich gelesen. [ all Martin his books ] have I read a'. [Martin seine Bücher] i habe ich [alle t i ] gelesen. [ Martin his books ] i have I [ all t i ] read b. # [Diese Martin seine Bücher] habe ich gelesen. [ these Martin his books ] have I read b'. * [Martin seine Bücher] i habe ich [diese t i ]gelesen. [ Martin his books ] i have I [ these t i ] read 2.2 Phasehood 91 I will go through these data in detail. Note that (13)-(16) mirrors (7)-(10) above in the relevant details with (7)/ (13) depicting the (phase-)heads in isolation, (8)/ (14) displaying co-occurrences of two heads, (9)/ (15) incorporating PDCs and (10)/ (16) with D/ Q S preceding PDCs. Note that I therein marked the in-situ co-occurrence of D and PDC as odd since the datum is rejected by some of my informants, rather than being judged as slightly ungrammatical by all of them. Note furthermore that this variation does not occur with the equally complex configuration in (16) a. with the phasal head part of the PDC itself; I will not speculate on these matters here. As can be seen in (13) a'.-c'., then, movement of HN is always out with one of the elements from (11) in isolation. The data in (14), however, demonstrate the familiar asymmetry of immobility from the D-Poss PRO complex to the extractability of Q S -D/ Poss PRO . Also in accordance with (11), HN cannot extract out of the complement of PDCs (cf. (15) a'.), while - once again - PDCs and HNs can move from strongly quantified nominals *10 (cf. (16) a'.) but not from below the demonstrative (cf. (16) b'.). A closer look at the underlined phase heads according to (11) gives a principled explanation to the variable extraction configurations: It is with respect to (11) and the phrasal set-up elaborated in chapter one that (12) as well as (5) above hold. Only in the presence of a more prominent phase candidate (D/ Poss PRO ) can movement proceed through the non-phase Q S . The same is not true of Poss PRO w.r.t. D, since the former is the complement of the latter. Observe that the extraction configurations in (14) and (16) leave the only dominating non-phase Q S P behind. We have thus derived the precondition for ‘quantifier stranding’ (QST) in German from variable phasehood in the nominal domain. The theory elaborated above further predicts that DP will always be a phase head in the variable system, if it is present, while the status of Q S and Poss PRO can vary in this respect. I will briefly return to some of the data presented in this section in the conclusion to this chapter. 2.2.1.2 Approaching Nominal Phases from Below Let us examine the remaining phrases in the hierarchical configuration from chapter one concerning the property of phasehood. To begin, I will focus on the most deeply embedded head, the nominal core, NP. Recall from chapter 1.4.3 that the complement position of the noun is stipulated to serve as the locus of External Merge (EM) for lexical possessives, which might move into the prenominal domain in the course of the derivation. The impossibility of extraction from the complement position was one of the two criteria for phasehood in Bošković (2014), as noted in (12) above and repeated here for convenience: 92 2 PREMISES - Phasehood & φ-Dependencies (17) a. (= Bošković 2014: (56) a.) Only phases and complements of phase heads can undergo ellipsis. b. (= Bošković 2014: (9)) Complements of phase heads cannot undergo movement. Since there are no restrictions on the availability of prenominal lexical possessives with singular, plural or mass HNs and accepting the complement of HN as their locus of EM, (17) b. ex negativo derives the non-phasal status of NP. Though never a phase, concentrating on movement of HN itself reveals some interesting properties: (18) a. Ich habe [viele Freunde]. I have [ many friends ] a'. Freunde i habe ich [viele t i ]. friends i have I [ many t i ] In so-called ‘split-topicalization’ structures (cf. van Riemsdijk 1989; Roehrs 2009: ch. 4.3.2, ch. 4.10.1; Roehrs in prep.: ch. 4.2) HN might move to sentence-initial topic position without any additional lexical material. Since we have concluded that N is not a phasal head and accepted Bošković’s generalization on movement, NP constitutes either a phasal complement which moves with a covert head or the complement of the complement of a phasal head. Given Richards’ (2007) proposals concerning good (phasal) design sketched above, we should tend to interpret (18) a'. as an instance of the former. We can actually find morpho-syntactic evidence that this is the correct option, cf. (19): (19) a. Freunde i habe ich [viele t i ]. friends i have I [ many t i ] b. # Freunde i habe ich [einen t i ]. friends i have I [ a / one t i ] b'. * Ich habe [einen Freunde]. I have [ a / one friends pl ] c. * Freund i habe ich [einen t i ]. [ friend sg ] i have I [ a / one t i ] c'. Ich habe [einen Freund]. I have [ a / one friend sg ] 2.2 Phasehood 93 (19) a. equals (18) a'. in which the plural HN has moved out of the complement of the weak quantifier, stranding it in its base position Q W0 . (19) b. mirrors this structure with the plural HN moving from beneath the indefinite article/ numeral ein-word ‘a/ one’, which - as can be observed in (19) c'. - takes a singular noun complement by default. Yet, with the nominal dislocated into sentenceinitial topic position, the structure (19) b. is not entirely ruled out. The expected number-agreement between the in-situ ein-word and HN ((19) c.), in contrast, is indisputably ungrammatical. There are, however, restrictions as to the lexical material that can be left behind (abstracting away from the deictic reading, accessible with the demonstrative, cf. endnote *11; cf. also (13) a'.-c'. above and fn. 13): (20) a. * Freund(-e) habe ich diese(-n). friend ( pl ) have I this pl/ sg b. * Freund(-e) habe ich meine(-n). friend ( pl ) have I my pl/ sg Weak quantifiers hence pattern with strong ones w.r.t. property v. of the characteristics subsumed under the QP-Hypothesis of Cardinaletti & Giusti (2006), as presented in chapter one, section 1.4.1 above: the ability to occur in discontinuous position. Moreover, it correctly sets this bundle apart from other nominal modifiers and determiners, in line with Cardinaletti & Giusti’s (2006) predictions, as well as the fully functional status of the latter (cf. the discussion in chapter 1.3.2 above on the lexical vs. functional status of the individual items in the left-periphery of the nominal hierarchy of projections). One might object that an unrelated phenomenon is encountered in the ungrammaticality of split-topicalizations incorporating a raised singular bare nominal since linearizations parallel to (19) c. without a stranded residue of the nominal hierarchy are equally ungrammatical in contrast to linearizations with raised plural object nominals. As Fanselow (2013: ch. 4) notes, bare singular count nominals do indeed not suffice to fulfill the formal constraints to act as an independent NP argument (cf. also Longobardi 1994). For the paradigm above to be fully significant, it has hence to be shown that ungrammaticality is rooted in the raising and stranding of the respective subparts of the nominal hierarchy and therefore internal to the nominal domain. As further elaborated by Fanselow (2013), the restriction just outlined does not apply to predicative configurations, in which singular count nominals may in part also surface without an SLI accompanying it, cf. (21) a. Moreover, the (possibly) bare singular object nominal can be fronted from such configurations (cf. (21) b'.), particu- 94 2 PREMISES - Phasehood & φ-Dependencies larly in contexts that generally support object fronting like narrow object focus with additional pronominal subject (i.e. (21) b., cf. Fanselow et al. 2008: ch. 3.2). However, observe that even in these configurations, coupled with contexts in which fronting of a bare singular count nominal is licit, the corresponding splittopicalization constructions are ruled out, as can be deduced from (21) b''. below. (21) a. Er ist (ein) Lehrer. he is ( a / one ) teacher b. Was ist Martin von Beruf ? what is Martin of occupation b'. (Ein) Lehrer ist er. ( a / one ) teacher is he b''. * Lehrer ist er einer. teacher is he a / one We can therefore maintain the source for ungrammaticality in (19) c. above in the barred extraction from the singular nominal domain. Singular HNs can hence not undergo topicalization on their own, while plural HNs partially can (dependent on the lexical material stranded in base position). This observation fits the proposals above rather well if we assume with several authors that number/ plural morphology does not constitute a feature of the nominal head, but rather a discrete phrase (NumP) which takes N as its complement. 11 The origin of the ungrammaticality of the datum (19) c. above thereby reduces to a violation of (17) b. Turning to (17) a. concerning nominal heads, we cannot draw any conclusions regarding its phasal status since full NP-ellipsis in German only occurs together with the verbal complex dominating it ((22) mirrors (7)-(10) above): (22) Du hast Bücher gelesen und ich you have books read and I a. * habe <Bücher> gelesen. 12 have < books > read 11 Cf. Ritter (1991a, 1991b, 1992, 1993), Noyer ([1992] 1997), Lyons (1995, 1999), Roehrs (2009), in feature-geometrical terms (cf. Harley 1994) of dedicated heads, see Cowper & Curie Hall (2009); I will elaborate on the featural content and the properties of NumP in detail in chapter 2.3 below but simply stipulate it, based on the observations in the main text, for now. 12 The datum (22) a. becomes less degraded under a generic reading of the second occurrence of the verb. However, we can assume that no nominal ellipsis has taken place in that case. 2.2 Phasehood 95 b. <habe> auch <Bücher gelesen>. < have > also < books read > As can be seen, a remnant always occupies the position of the second V/ v P. A hint as to the correctness of the proposal above, however, can be drawn from topic-drop in colloquial German, if one accepts the conclusions reached on the basis of (19). Observe that sentence-initial topics can be dropped under coreference with the discourse topic. German is a V2-language, yet we find overt verb-first structures in these cases: (23) A: Es gibt [[[Aufsätze] von Chomsky] in der Bibliothek] i . 13 there are [[[ papers ] by Chomsky ] in the library ] B: e i habe ich (alle) durchgesehen. have I (all) looked.through Since NumP can extract (cf. (18) and (19), hence not a complement by virtue of (17) b.) and elide (cf. (23), hence a phase or a phasal complement by virtue of (17) a.), we can conclude that NumP is a phase in the German hierarchy of nominal projections. Observe further that the possibility of QST in (23) might imply a higher position in the Hierarchy of Assignment of Phasehood in the Nominal Domain in (11) above, a point to which I will return below. Turning to the last of the prenominal heads identified over the course of chapter 1.4 above, it follows that - at least with plural HNs where NumP projects - the next higher phrase, proposed here, (i.e. Q W ) should never constitute a phase since its complement would be frozen in-situ due to the PIC/ Anti-Locality interaction (Bošković) and/ or interpretability/ impenetrability (Richards). With respect to (17) a., this prediction seems to be borne out. A weakly quantified nominal can be elided, leaving the overt quantifier in-situ, which suggests either its phasal status or the phasal status of the NumP it dominates. 13 The postnominal PPs are necessary to render the NP specific. More generally, a complex nominal or verbal complement serves to reduce the ungrammaticality of examples like (13)/ (20), which should be attributed to memory load. I will limit myself to the analysis of simple, i.e. complementless DPs here. In the structure under consideration, however, the PPs do not pose a problem concerning the phenomena under consideration since they are themselves closed-off extended projections of P and therefore - following Bošković (2014) - exhibit phasal status inside the nominal domain. The highest phrasal category is hence still NumP. 96 2 PREMISES - Phasehood & φ-Dependencies (24) Du hast Fredericks Bücher gelesen und ich habe [viele <Bücher>] gelesen. you have F.’s books read and I have [ many < books >] read Since we already found independent evidence for the latter option above, I will assume that (24) also points to this conclusion. The generalization in (17) b., then, predicts the availability of movement of NumP from below Q W . Once again, this is correct. Weak quantifiers can be stranded with topic-movement to sentence-initial position, just like the strong quantifier in (14) b'./ c'., (16) a'. above, cf. (18) a'./ (19) a. Observe, however, that the left-dislocated nominal is severely more restricted concerning pied-piping of additional lexical material: Contrary to QST with the strong quantifier, neither demonstrative nor possessives are licit in the left periphery of the moved complex. If we turn to the phrasal set-up argued for in chapter one, the differences become clear immediately: Since the source of movement is the complement of Q W , only the lexical head and adjectival modifiers 14 are able to move, while the complement of the strong quantifier exhibits a much more articulated structure, optionally including the Dand Poss PRO -projections; these phrases readily pied-pipe with HN away from Q S . We can, however, neither grant (i) rigid phasehood to NumP, nor (ii) uniformly deny phasal status to Q W . Concerning the former (i), recall from (8) a. above that nothing accounts for the ungrammaticality of the D-Poss PRO complex if it was the outcome of full phasal ellipsis. Rather, it is accounted for under the illicit elision of the complement’s complement of a phase head (i.e. D 0 ). I therefore conclude that domination by a possible phasal head prevents Num 0 from bearing this property. Since it is not predictable upon External Merge of one of the elements subsumed in (11) if a more prominent phase head will subsequently be introduced into the derivation and thereby prevent the establishment of a structural set-up incorporating two phases dominating one another, NumP ceases to constitute a phase upon first Merger of a higher candidate for phasehood. Regarding the latter (ii), turning to mass nouns - which arguably do not project NumP but still combine with weak quantifiers - leaves us with the following observation: *11 14 Throughout the analysis, I refrained from employing adjective position as a testing ground since these elements are themselves taken to constitute phase heads in Bošković (2014); cf. Marantz (2001, 2007) on the claim of universally phasal ‘little x’ categories; cf. also fn. 13 above. 2.2 Phasehood 97 (25) a. Reis i habe ich t i gekauft. rice i have I t i bought b. ## Reis i habe ich [viel(-en) t i ] gekauft. rice i have I [ much t i ] bought It seems that Q W can act as a phase under the condition that it does not immediately dominate one, a restriction which satisfyingly also captures the properties of Q S in its respective domain, discussed above. We can hence establish a partial hierarchy of assignment of phasehood for Num > Q W . Additionally, the restrictions on variable phasehood guarantees that these elements never surface as two adjacent phases. A remaining question, then, concerns the combination of this hierarchy with the one in (11) above. To approach an answer, observe the additional ellipsis structures below. These seem - at first sight - to contradict the variable approach to phasehood advocated here: (26) Du hast Fredericks Bücher gelesen und ich habe you have F.’s books read and I have a. diese vielen <Bücher> these many < books > b. meine vielen <Bücher> my many < books > c. # diese meine vielen <Bücher> gelesen. these my many < books > read In (26), elements classified as phase heads readily occur with an overt complement from which lexical material has been elided. This straightforwardly contradicts the generalization in (17) a. above, if - as concluded before - D always constitutes a phase. Similar evidence can be put forward against (17) b.: (27) a. Ich habe diese vielen Freunde. I have these many friends b. Freunde habe ich diese vielen. friends have I these many However, equal evidence for the status of D as the highest phase in these constructions can be found as well: 98 2 PREMISES - Phasehood & φ-Dependencies (28) * [Viele Freunde] i habe ich diese t i . [ many friends ] i have I these t i Note that only with the introduction of datum (27) b. does German act contrary to the proposals of the AP-Hypothesis of Cardinaletti & Giusti (2006) in the extraction of a subpart of a higher projection (DP in the examples above) that does not contain the quantifier (i.e. property iii. in chapter 1.4.1 above), since, prior to said datum, no overt SLI dominated the structures from which topicalization took place. Furthermore, note that property iv. is never contradicted by the splittopicalization structures presented here, since Cardinaletti & Giusti themselves restrict this condition to languages in which adjectives are also disallowed in discontinuous position. As we have already seen, this is not the case in German, where adjectives may surface either raised from the extended projection of the nominal in preposed adjective structures (cf. chapter 1.3.3 above, Postal [1966a] 1970) or stranded by movement of the nominal domain, parallel to the structures discussed here. Returning to the phenomena at hand, I want to argue that the data in (26) and (27) above do not contradict the generalizations in (17), but rather show that the system advocated here obeys the economy principle put forth in Richards (2007), namely the phase-plus-non-phase-complement order (P-N-P-N). The argument goes as follows: The remnant complexes in (26) do not constitute pairs of phase heads and partly elided complements per se, they rather constitute a phasal head (viz. diese, meine in (26) a. and b. respectively) in addition to its nonphasal complement Q W . In turn, Q W ’s own complement constitutes a phase head itself, i.e. NumP, which is elided in accordance with (17) b. above. This proposal also readily captures the datum in (27) b. We hence exhibit a configuration of phases and non-phases with two separate cycles in line with Richards (2007). The ‘oddness’ of (26) c. hence reduces to a departure from phasal economy in that two non-phasal heads follow one another without phasal intervention, arguably an increase in memory load. To make this claim even stronger, note that Richards’ proposal coupled with Bošković’s observations would result in a trivial set-up of varying phasehood for all heads down to the nominal core, if no ‘buffering’ mechanisms like the relational evaluation for the elements in the lowest cycle, elaborated above, exists. Once again turning to quantified structures, we observe the same ‘least popular’ assignment of phasehood with weak and strong quantifiers in their respective phasal cycles. The relevant data are (13) c'. and (25) b., repeated here as (29) a. and b. respectively, for ease of exposition: * 12 2.2 Phasehood 99 (29) a. * Bücher i habe ich [alle t i ] gelesen. (= (13) c'.) books i have I [ all t i ] read b. ## Reis i habe ich [viel(-en) t i ] gekauft. (= (25) b.) rice i have I [ much t i ] bought In light of the conclusions drawn above, we can hence state that the respective cycles of the nominal hierarchy of projections exhibit variable assignment of phasehood and thus arrive at the set-up in (30) a. with * marking rigid and ( * ) marking variable phasal heads: (30) Assignment of Phasehood in the Nominal Domain (final version) a. Q S( * ) > D* > Poss PRO( * ) > Q W( * ) > Num ( * ) > N b. [3 1 2] [2 1] c. 1 2 Concerning the respective cycles, we moreover arrive at the hierarchy for the assignment in (30) b., dependent on the co-occurrence of two (or more) of the elements under consideration in the same structure. Finally, (30) c. displays the preference of phasal assignment regarding the two cycles. As can be observed, the only category not obeying the principle in (3) above turns out to be Q; this departure is, however, principled and furthermore identical for both instantiations, Q S and Q W . Note further that lexical possessives, externally merged post-nominally, are predicted to be able to always extract from HN’s complement. In the cases above, I concentrated on nominal possessives, hence another instance of (30) a. This is in line with the conclusions drawn above since the lexical possessive would be the sister of a non-phasal head N 0 . Following Bošković, the argument readily extends to nominal complements headed by prepositions since the highest phrase of the extended projection of every lexical core category serves as a phase (cf. fn. 13 above). Finally, another head was introduced on the basis of additional (morpho-)syntactic evidence - Num, a head carrying number-features - which projects at least with plural head nouns. One additional datum, which is not readily captured by the system elaborated above, however, has to be presented here. Returning to (14) c'., b'., reprinted here as (31) a. and b. respectively, we also find the contrasting structures in a'. and b'.: 100 2 PREMISES - Phasehood & φ-Dependencies (31) a. [Diese Bücher] i habe ich alle t i gelesen. (= (14) c'.) [ these books ] i have I all t i read a'. [Bücher] i habe ich all diese t i gelesen. [ books ] i have I all these t i read b. [Seine Bücher] i habe ich alle t i gelesen. (= (14) b'.) [ his books ] i have I all t i read b'. [Bücher] i habe ich all seine t i gelesen. [ books ] i have I all his t i read In terms of the system elaborated here, the optional stranding of the demonstrative and possessive in the bar-examples amounts to movement of the lower cycle. Since derivations proceed in a bottom-up fashion, we might therein have found a case of conflicting evaluations of phasehood assignment with Q S able to optionally inherit phasehood from D/ Poss PRO , when these are merged with Num, to ensure structural distance of phase heads. Alternatively, as I will argue in chapter 3.3.2 below, relaxing Anti-Locality from universal application at every phrase level (cf. Abels 2003) to Grohmann’s (2000, 2003a) original ban on (prolific) domain-internal movement readily accounts for the grammaticality of these data. One might, however, take it as a hint to the general correctness of the system elaborated above that these options necessarily proceed along the phasal boundaries identified in the course of the analysis. Before turning to the (featural) properties of NumP in greater detail, the remainder of this subchapter will be devoted to contrasting said system with alternative accounts to the nominal domain. Therein, I want to demonstrate that the predictive power concerning data from ellipsis and extraction is not reproducible in a structurally simpler phrasal set-up, incorporating a single head approach to the upper functional domain (i.e. the Q S -D-Poss PRO complex) and permitting multiple specifiers/ adjuncts. I will take this as an argument for the complex set-up carved out thus far in this inquiry. 2.2.2 Head Status and the Fine Structure of the Nominal Periphery Let me begin by reviewing the head status of the elements under consideration, making use of simple binding phenomena. As has been mentioned in the beginning of this chapter, Despić (2011 et seq.) as well as Bošković (2005 et seq.) argue for the absence of the DP-layer in Serbo-Croatian (henceforth SC; and other article-less languages) as opposed to e.g. English in part on the basis 2.2 Phasehood 101 of Condition B and C violations with pronominal and lexical possessors. The relevant contrast is given below: (32) a. Kusturica’s i latest movies really disappointed him i . b. His i latest movies really disappointed Kusturica i . (33) ([Serbo-Croatian]) a. * Kusturicin i najnoviji film ga i je zaista razočarao. (= Despić 2013: (12)) Kustirica’s latest film him is really disappointed ‘Kusturica i ’s latest movie really disappointed him i .’ b. * Njegov i najnoviji film je zaista razočarao Kusturicu i . (= Despić 2013: (13)) his latest film is really disappointed Kusturica ‘His i latest movie really disappointed Kusturica i .’ As can be deduced from the data above, a co-referential reading of the (prenominal and lexical) possessive in subject position with the object is illicit in SC, but not in English. Starting from there, said authors take the distinction as an argument for the DP/ NP-parameter, hence against the universality of DP (labelled the ‘Universal DP Hypothesis’, UDPH by Despić 2011, 2013). As the authors differ in their analyses; however, I will sketch a broad picture of the underlying rationale in what follows. Bošković claims that prenominal elements are adjoined to the nominal projection. As he notes, functional elements behave like adjectives in SC and are hence categorized accordingly. Making use of the insight of Kayne (1994), who, however, excludes multiple adjunction (cf. Kayne 1994: ch. 3.3, 3.4), said elements are taken to c-command out of the respective phrases they are adjoined to (diverging from the current minimalist rationale, Kayne’s system also only allows a single adjunct/ specifier per phrase, but see Chomsky: BPS for multiple adjuncts and specifiers). With the DP projection absent, possessive elements in Spec,TP are able to c-command into T’s complement, which results in Condition B and C violations for (33) a. and b. respectively. In DP languages like English, however, the elements under consideration are ‘buried too deep’ inside the nominal domain to c-command into the extended projection of the verb, hence permitting the co-referent reading with the complement of VP in (32), resembling Kayne’s own analysis following Szabolcsi (1981, 1983, 1992). Note that D hence projects in these structures, irrelevant of an overt occurrence of the determiner or demonstrative. In SC, though, D cannot rescue the grammaticality of structures like (33) a., b., since it is likewise adjoined to the nominal phrase. 102 2 PREMISES - Phasehood & φ-Dependencies Although an elegant test for the structure of the nominal left periphery, certain auxiliary hypotheses, employed in the analysis sketched above, might carry theory-internal problems. Note that locating possessives in adjunct position is not necessary (though possible, cf. Chomsky BPS: 85) for the establishment of c-command relations out of the nominal domain in a BPS-model of Narrow Syntax. As Chomsky notes (BPS: 83; see also Rezac 2003), “a category can have two heads, one a bare head that projects, the other an X max .” As far as I can tell, nothing prevents a head D/ Poss to c-command out of its phrase into T’s complement under the standard ‘Reinhartian’, as well as Kayne’s asymmetric notion of c-command, the latter is given in (34): (34) (= Kayne 1994: 16) Asymmetric C-Command X c-commands Y iff X and Y are categories and X excludes Y and every category that dominates X dominates Y. I therefore conclude that the binding phenomena presented in (32) and (33) constitute a testing ground for further phrasal structure above possessives (pronominal as well as lexical), irrespective of their phrasal status (i.e. head/ specifier/ adjunct). 15 On the basis of parallel structures in the clausal and nominal domain, Despić (2011) argues for a DP which exhibits phasal status only in co-occurrence with the non-phasal complement PossP, parallel to the C-T relation sketched in the beginning of this section. By the same token, Bošković (2012) relates the two domains by proposing the absence of T in NP-languages. Despić’s notion of phasehood is hence a semi-rigid one: DP is either a phase (if coupled with PossP) or not, but PossP never poses as a phase itself (Despić 2011: 165). By further constituting phases as (minimal) binding domains, DP/ NP-internal reflexive binding asymmetries in English and SC as well as subject orientation of objectinternal anaphors are captured by the system, if it is the edge of the next phase that is the locus for binding (i.e. v P in NP-languages). In the same vein, Despić separates lexical possessives from pronominal ones and locates them with reciprocals on the edge of the phase DP in languages which employ this phasal 15 Interestingly, demonstratives preceding structures like (32) a., b. do not enable a co-referent reading, as noted by Despić (2011), while Bošković (2014) observes an improvement of grammaticality with numerals and quantifiers, on which he bases the proposal of an independent QP heading those structures. Additionally, Bošković (2008, 2012) notes various, superficially unrelated syntactic phenomena that accompany the NP/ DP distinction (cf. Despić 2013: (2)). My aim is hence not to argue in favor of the UDPH here, but to show that adjunction is not a necessary conclusion and Condition B/ C violations can be a meaningful testing ground in the system, elaborated here. 2.2 Phasehood 103 head. 16 Binding from outside the phase is hence available for reciprocals, but not for Poss PRO in the complement of D, which can bind inside the respective DP. This grouping gains support from the observation that the clitic s in English can be found uniformly on the elements in Spec,DP (it is hence located in D 0 ) but not on possessive pronouns. The latter are taken to be located in the specifier of the complement of DP, the possessive phrase PossP, due to linearizational options w.r.t. adjectives, in line with Cinque (1994) and to the close resemblance with clausal subjects in Spec,TP. As pointed out above, phasal status of D is triggered by its co-occurrence with the non-phasal complement PossP, in line with Richards (2007) and parallel to the relation of C to T (i.e. feature inheritance) in the clausal domain. Thus, since only phasal complements can elide in Despić’s system, it is always Poss that gets deleted. 17 Additionally, binding of picture-NP anaphors can be handled without reference to PRO: (35) (= Despić 2011: ch. 3, (89)) a. John i saw [ DP [ PossP Bill j ’s picture of himself *i/ j ]]. b. John i saw [ DP the picture of himself i ]. Since PossP is not present in (35) b., the anaphora can be bound in the next higher phase and hence DP-externally at the edge of v P. (35) a. does, however, not conform to the position proposed for the respective possessive elements above, as doesn’t (36) below: (36) (= Despić 2011: ch. 3, (90) e.) * They could read their own medical records but they could not read my medical records. In (36), the possessive pronoun is said to occupy “a position within the PossP” (Despić 2011: 165), and hence would fail to ‘signal’ the existence of the relevant complement of D, though it is located in the specifier of PossP before. Despić (ibid.: ch. 3.3.3.2) in fact discusses the possibility of possessor movement to D as head movement or phrasal movement for Icelandic and Bulgarian, which exhibit post-nominal definiteness-marking with reference to Delsing (1993), but this move would yield an Anti-Locality violation under the proposed locus of External Merge of possessors in Spec,PossP (cf. fn. 16). 16 Despić (2011: ch. 3, fn. 11) notes that these elements might have moved there from within PossP, specifically from the specifier, where all possessive material is externally merged in his system. The separation of semantically related lexical elements seems unnecessary in a framework incorporating a variable phrasal set-up. 17 I argued in section 2.2.1.1 b. above, that prenominal lexical possessives suffice to signal the presence of the projection Poss PRO , cf. endnote *10. 104 2 PREMISES - Phasehood & φ-Dependencies The system sketched above hence employs a rigid type of phasehood. To account for the German data presented in (6)-(10) on extraction and (13)-(16) on ellipsis therein, some additional assumptions would be necessary. Considering the grammatical cases (7) b., c. as well as (8) b. and (9) c. again, reprinted here for ease of exposition as (37), the postulation of an additional phase head below Poss would be necessary to counter the violation of the generalization in (17) a. above, stating that only phases and phasal complements can elide. (37) Du hast Fredericks Bücher gelesen und ich habe you have F.’s books read and I have a. seine <Bücher> (=(7) b.) his < books > b. alle <Bücher> (= (7) c.) all < books > c. alle seine <Bücher> (= (8) b.) all his < books > d. Martin seine <Bücher> gelesen. (= (9) c.) Martin DAT his < books > read With the proposal of a covert D-head, the quantifier could be located in the specifier of that phrase or a separate projection, parallel to my proposals laid out in chapter one. 18 The covert D 0 would be present in the grammatical (7) b. and c, (8) b., (10) a. above, the latter reprinted below, once again. (38) (= (10) a.) Du hast Fredericks Bücher gelesen und you have F.’s books read and ich habe all Martin seine <Bücher> gelesen. I have all Martin DAT his < books > read ‘You have read Frederick’s books and I have read all of Martin’s.’ Following Despić, lexical possessives are located in specifiers of this projection, which would account for (10) a., with two specifiers or a dedicated QP, but would leave the ungrammaticality of (8) a. (co-occurrence of the overt heads D and Poss), (9) b. (overt specifier and head of DP, i.e. the lexical possessive and the demonstrative following or preceding it) and (9) d. (two specifiers with cov- 18 Bošković (2014) notes that a quantifier heading the testing grounds for Condition B/ C in SC, presented above, rescues grammaticality and hence proposes a dedicated phrase as the highest nominal head in these instances. 2.2 Phasehood 105 ert head of DP, i.e. the lexical possessive and the strong quantifier following or preceding it) unaccounted for, which are readily explained in a framework of variable phasehood. The relevant configurations are reprinted below: (39) Du hast Fredericks Bücher gelesen und ich habe you have F.’s books read and I have a. * diese seine <Bücher> (= (8) a.) these his < books > b. * {Martin(s)} diese {Martin(s)} <Bücher> (= (9) b.) { Martin DAT/ GEN } these { Martin DAT/ GEN } < books > c. * {Martin(s)} alle {Martin(s)} <Bücher> gelesen. (= (9) d.) { Martin DAT/ GEN } all { Martin DAT/ GEN } < books > read Turning to generalization (17) b., a covert DP would have to be postulated in (13) b. and (14) b'. while the employment of a dedicated QP becomes pressing with (14) c'. and (16) a'. since a phasal head otherwise would be said to move with its inner specifier, leaving the outer specifier in base position. The relevant examples are again reprinted below: (40) a. [Seine Bücher] habe ich gelesen. (= (13) b.) [ his books ] have I read b. [Seine Bücher] i habe ich [alle t i ] gelesen. (= (14) b'.) [ his books ] i have I [ all t i ] read c. [Diese Bücher] i habe ich [alle t i ] gelesen. (= (14) c'.) [ these books ] i have I [ all t i ] read d. [Martin seine Bücher] i habe ich [alle t i ] gelesen. (= (16) a'.) [ Martin his books ] i have I [ all t i ] read I conclude, then, that the system elaborated here gives a more uniform and straightforward account in the implementation of variable phasehood in the nominal domain. Let us next turn to German data concerning possessors and c-command. Above, I have concluded that said tests do not straightforwardly point to specifier/ adjunct status of lexical material in the nominal domain (when ungrammaticality arises), but simply point to additional covert structure on top of the nominal domain (when grammatical). German would naturally be classified as a DP-language in Bošković’s/ Despić’s terms. First, let us focus on simple pronominally and lexically possessivized subjects in sentence-initial position, the 106 2 PREMISES - Phasehood & φ-Dependencies German data in (41) are constructed in parallel to (32) and (33) above. Indeed, no ungrammaticality arises therein: 19 (41) a. [ Johannes Müllers i20 Filme] haben ihn i berühmt gemacht. [ J. M. i ’s movies ] have him i famous made ‘Johannes Müller’s movies have made him famous.’ b. [Seine i Filme] haben [ Johannes Müller] i berühmt gemacht. [ his i movies ] have [ J. M. ] i famous made ‘His movies have made Johannes Müller famous.’ Co-reference is possible in both structures and we can hence conclude that Condition B (in (41) a.) and Condition C in backward anaphora (in (41) b.) are not violated, since no c-command relation could be established between lexical possessive and pronoun in the former or possessive pronoun and proper name in the latter case. As should be expected by now, this does not change with the addition of the two remaining potential phase heads dominating the possessive: 21 (42) a. [Diese seine i Filme] haben [ Johannes Müller] i berühmt gemacht. [ these his i movies ] have [ J. M. ] i famous made a'. ## [Diese Johannes Müllers i Filme] haben ihn i berühmt gemacht. 22 [ these J. M. i ’s movies ] have him i famous made b. [All seine i Filme] haben [ Johannes Müller] i berühmt gemacht. [ all his i movies ] have [ J. M. ] i famous made b'. # [Alle Johannes Müllers i Filme] haben ihn i berühmt gemacht. [ all J. M. i ’s movies ] have him i famous made 19 Throughout the presented paradigm, I sought to account for a non-contrastive reading of the relevant pronominal and lexical possessives and refrained from using relational/ kinship nouns cf. Bošković (2012: fn. 24). 20 The proper name used in the following examples is an imaginary one to bar any forced interpretation as to the relation of the individual to the movies. 21 I take the mild ungrammaticality of prenominal lexical possessive headed by additional functional structure to pose an unrelated phenomenon not relevant to the effects under consideration here, see the discussion surrounding (26) c. and its relation to the datum (8) a. above; the reasoning therein naturally extends to cases like those in the main text, in which the phase head occurs with the non-phasal specifier, (42) a'., and the phasal specifier co-occurs with additional non-phasal material, (42) b'. Observe, that the former, in which Poss PRO P does not constitute a phase in my system, is judged worse than the latter, in which the specifier of Poss PRO signals the presence of the phase; see also endnote *10. 2.2 Phasehood 107 Note that it is not the quality of the co-referential reading that declines here but rather the grammaticality of the whole structure. In line with this gradual decline of grammaticality, PDCs in sentence initial topic position are unproblematic for speakers who generally accept this construction but are gradually judged worse with the addition of the two remaining possible phase heads: (43) a. [ Johannes Müller seine i Filme] haben ihn i berühmt gemacht. [ J. M. his i movies ] have him i famous made b. # [Alle Johannes Müller seine i Filme] haben ihn i berühmt gemacht. [ all J. M. his i movies ] have him i famous made c. * [Diese Johannes Müller seine i Filme] haben ihn i berühmt gemacht. [ these J. M. his i movies ] have him i famous made As a last observation concerning Condition B/ C, I want to present an interesting asymmetry concerning the structural position of the possessivized nominal domain: In (41)-(43) above, I presented sentences in which said domain occupies the subject position; if switched to the fronted internal argument position, coreference is still possible with pronouns, but illicit with lexical possessors. This is true for structurally as well as lexically case-marked objects as (44) and (45) below demonstrate. (44) a. [Seine i Filme] mag Johannes Müller i . [ his i movies ] likes J.M. i b. [ Johannes Müllers i Filme] mag er *i . [ J.M. i ’s movies ] likes he *i (45) a. [Seinen i Filmen] hat Johannes Müller i vertraut. [ his i movies ] DAT has J.M. i trusted b. [ Johannes Müllers i Filmen] hat er *i vertraut. [ J.M. GEN i ’s movies ] DAT has he *i trusted Observe that we cannot conclude on the basis of these data that binding applies before movement, since co-indexation is still possible with the head Poss PRO0 . Observe that they are, in turn, readily explained in the system of Despić (2011) 22 No variation in the acceptability from (41) a. to (42) a'. should arise under Despić’s theory of phasehood, since DP is always present in these structures, cf. also (43) c. below. 108 2 PREMISES - Phasehood & φ-Dependencies since the lexical possessive on the edge of the DP-phase would be taken to ccommand out of the nominal domain, while the pronoun in the specifier of the complement of D could not establish such a relation. This is, however, not necessarily the case since e.g. the structural position of sentence-initial internal and external arguments arguably differ, with the subject in Spec,TP and the fronted object higher above in the left periphery. I will not speculate on these matters here; cf. ch. 4.2.1 below on said divergences in movement. I proposed that Condition B/ C violations of sentence-initial possessives can be viewed as a testing ground for structural complexity of the nominal hierarchy of projections, the possessive is part of. Employing a standard view of c-command as well as a bare phrase structural approach, I moreover argued against said configurations as a straightforward argument for the structural status of specifier/ adjunct regarding the SLIs involved. These data hence point to the direction of additional phrasal structure in the left periphery of the nominal domain instead, although I will not speculate on its nature here. 23 A legitimate question, however, would be if we could in principle get away with less structure and still account for the idiosyncrasies of the left periphery of the nominal domain in German. I have shown above that the framework of Despić (2011) cannot straightforwardly account for the data at hand. Reducing structural content further, we might start with a configuration consisting of a single head, call it FP, which hosts determiners (and possibly possessive pronouns). Incorporating the proposal of Chomsky (BPS, BEA) that a single phrase can contain several specifiers, we locate demonstratives and lexical possessives in such positions, the former analogous to Bošković’s (2013) argument that these LIs exhibit adjective-like inflectional properties, the latter on simple linearizational observations. As mentioned above, possessive pronouns might be subject to the same rationale as demonstratives (as in Bošković 2013, option a1. in (46) below) and hence located in Spec,FP, or alternatively categorized as heads of said phrase, since they are in complementary distribution with determiners in German (option a2.). The strong quantifier, in turn, might be located either in the head of its own projection (see the discussion of Despić’s system above, option b1.) or the outer Spec of FP (option b2.). The phrasal set-up incorporating all the proposals laid out above would look as follows: 23 Bošković (2014: fn. 8) himself notes that the nominal domain in German might be structurally more complex than its English counterpart. I take the conclusions reached in this chapter as evidence that this is indeed so. 2.2 Phasehood 109 (46) 60 (46) QP 2 b1. Q FP 3 b2. Q F’ 3 Dem F’ 3 Poss LEX F’ 3 a1. Poss PRO F’ 3 F 0 … Det a2. Poss PRO (49) Q S P 3 3 Q S( * ) D 3 1 1 D* Poss PRO P 3 2 Poss PRO( * ) Q W P 3 2 Q W( * ) NumP 2 3 1 Num ( * ) NP 3 N Poss LEX P* (52) N 9 [m] [f] [n] (57) a. N b. N 9 9p [m] [f] [n] [m] [f] [n] sg ! ! ! sg sg sg (58) Num ! [pl] With this much in place, let us focus on the predicative power of the structural models, subsumed in (46). a. What we account for With two two-way oppositions there are effectively four set-ups represented in (46), call them a1/ 2b1/ 2 respectively, all of which capture simple linearizational properties of in-situ configurations with irrelevant subparts not overtly represented. Under a1/ 2b1, we additionally capture QST in German, which is bared under b2., since a bar-level projection would have been taken to move. 24 b. What we miss out on Incorporating the generalizations in (17) above, repeated as (47) for ease of exposition, we are unable to account for the (un-)grammaticality of a variety of linearizations in German. (47) a. (= Bošković 2014: (56) a.) Only phases and complements of phase heads can undergo ellipsis. b. (= Bošković 2014: (9)) Complements of phase heads cannot undergo movement. 24 It might, however, be possible to analyze this type of construction in successive steps, first moving Q to an intermediate position, followed by iterations of the remaining specifiers, in line with Bošković’s (2013) proposal on the contextuality of the phasal edge. This is, however, extremely uneconomical and therefore highly unlikely under current assumptions. 110 2 PREMISES - Phasehood & φ-Dependencies Starting with (47) a., and beginning with the assumption that FP constitutes a phase, we return to the data in (7)-(10) from chapter 2.2.1 above. Therein, (7) is accounted for in all instances of (46), since all elements proposed therein are taken to be located at the edge of FP. (7) b. and c. are subject to the a1/ 2 and b1/ 2 oppositions in their interpretation, however, with (7) b. a phase head under a2. and alle an additional phrase under b1. Turning to the data in (8), the grammatical b. and c. are accounted for under all instances of (46) with respect to (47) a., but nothing accounts for the ungrammaticality of (8) a., either another instance of Spec-head concatenation (under b2.), or a linearization of two specifiers of a covert F 0 (under b1.). The same is true for the ungrammatical instances of (9), i.e. b. and d., which also reduce to the latter configuration, or - for (9) d. - to the combination of XP and phasal edge under a1. Finally, (10) a. is accounted for under a1/ 2b1/ 2, while nothing bars (10) b. under b1/ 2. Focusing on the barexamples in (13)-(16), (13) b'. can only be ruled out under a2. w.r.t. (47) b., while c'. forces option b1. In (14), all examples force option a2., which is additionally also true for (16) a'. While we have seen that (47) b. accounts for the data in (13)- (16) under a1. (a modification of the axiom of variable phasehood of Q S in the theory carved out above), no option readily rules out the ungrammatical data in (8) a., (9) b., d. and (10) b. The former are repeated above as (39) in the discussion of Despić’s system, (10) b. is given again below: (48) (= (10) b.) * Du hast Fredericks Bücher gelesen und ich habe diese Martin seine <Bücher> you have F.’s books read and I have these Martin DAT his < books > gelesen. read I hence conclude that we are not able to reduce the structural complexity of the system at hand while at the same time maintain its predictive power. Only the proposed set-up is able to capture the linearizational idiosyncrasies in the nominal domain in German while straightforwardly ruling out ungrammatical configurations in accordance with - and with reference to - variable phasal status in the functional domain of the noun. 2.2.3 Conclusion I have argued in this section that an articulated structural set-up of the functional nominal domain is necessary to account for the idiosyncrasies concerning ellipsis and extraction in German. As I have demonstrated, the phrasal configuration proposed in chapter one accounts for the complete paradigm of gram- 2.2 Phasehood 111 matical and ungrammatical linearizations presented in subsection 2.2.1. With reference to Richards (2007), I identified the resulting concatenations as instances of the most economical pairing of phasal heads and non-phasal complements in the implementation of Bošković’s (2014) proposal of the contextuality of phasehood. The phasal properties hence lend further support to the grouping and the head status of the elements under consideration, as proposed in chapter one. 25 Additionally, I contrasted the set-up elaborated here with superficially simpler structural configurations and demonstrated that these do not suffice to capture the range of relevant data while at the same time barring ungrammatical configurations. All relevant conclusions of the preceding subchapter are combined in (49) below, with * marking rigid phasehood (if the respective lexical item is overtly present), ( * ) marking variable phasehood, bold numbers to the left referring to the rank in the Hierarchy of Assignment of Phasehood in the Nominal Domain, light numbers to the preference in selection of the phasal cycle. The dotted line separates said cycles. (49) 60 (46) QP 2 b1. Q FP 3 b2. Q F’ 3 Dem F’ 3 Poss LEX F’ 3 a1. Poss PRO F’ 3 F 0 … Det a2. Poss PRO (49) Q S P 3 3 Q S( * ) D 3 1 1 D* Poss PRO P 3 2 Poss PRO( * ) Q W P 3 2 Q W( * ) NumP 2 3 1 Num ( * ) NP 3 N Poss LEX P* (52) N 9 [m] [f] [n] (57) a. N b. N 9 9p [m] [f] [n] [m] [f] [n] sg ! ! ! sg sg sg (58) Num ! [pl] 25 The specific selection of non-quantifying SLIs, taken to comprise the higher phasal cycle, has been bundled in a parallel fashion in previous analyses of German nominal domains as early as Vater (1984: (Abb. 2), but cf. Fehlisch 1986: ch. 2.2.1), whose class of Determinantien (‘determiners’), contrasted with the class of quantifiers, subsumes determiners, (proximal and distal) demonstratives and possessive pronouns together with the demonstrative pronouns der -/ die -/ das jenige and der -/ die -/ das selbe , which are omitted in his subsequent discussion as well as the current analysis; note that Vater (1984: 38) moreover correctly attributes ‘limited conditions of use’ (“eingeschränkte Gebrauchsbedingungen” [MB]) to these latter items. 112 2 PREMISES - Phasehood & φ-Dependencies 2.3 φ-Features As has been shown in the preceding section on the basis of German split-topicalization structures, there is initial evidence for the proposal of a dedicated phrase for (at least) plural HNs in the phasal framework sketched above, hosting the appropriate number-feature. As has been furthermore mentioned there, a dedicated head for nominal number morphology, Num 0 , has been proposed by several authors, beginning with Ritter’s (1991a/ b et seq.) analysis of Modern Hebrew. In the current subchapter, I want to elaborate on this head, its featural content and the order and dependencies of φ-features in the nominal hierarchy of projections in general. As I will argue, this variable phase head encodes not only plural morphology but also a subset of gender-features. I will subsequently show that the resulting reordering of φ-features and their dependencies (in a feature-geometric approach) in the nominal domain allows accounting for several morpho-syntactic idiosyncrasies of German nominals. Before beginning my investigation, however, recall from chapter one that I restrict my analysis to the morpho-syntactic realization of features. In the context of definiteness, I referred to Roehrs’ (in prep.) notion of the division of numberfeatures on LIs ([αPL morph], [αPL sem]) to depict this rationale. In parallel fashion, Heycock & Zamparelli (2005) propose two distinct binary features ‘ plur ’ and ‘ latt ’, the latter corresponding to semantic singularity/ plurality, the former to “the ϕ-feature for syntactic number” (Heycock & Zamparelli 2005: 219). In the light of these approaches, I feel justified to restrict myself to the morphosyntactic reflexes of number in my analysis. This will become pressing below. 2.3.1 Number and Gender (Categories) As an Introduction to the discussion of this subchapter, let me revisit the structures that initially led to the proposal of NumP in chapter 2.2.1.2 above: (50) (= (19)) a. Freunde i habe ich [viele t i ]. friends i have I [ many t i ] b. # Freunde i habe ich [einen t i ]. friends i have I [ a / one t i ] b'. * Ich habe [einen Freunde]. I have [ a / one friends pl ] c. * Freund i habe ich [einen t i ]. [ friend sg ] i have I [ a / one t i ] 2.3 φ-Features 113 c'. Ich habe [einen Freund]. I have [ a / one friend sg ] (51) (= (20)) a. * Freund(-e) habe ich diese(-n). friend ( pl ) have I this pl/ sg b. * Freund(-e) habe ich meine(-n). friend ( pl ) have I my pl/ sg As can be deduced from (50) and (51), only plural HNs are able to move on their own from beneath weak quantifiers and numerals. In the light of Bošković’s (2014) proposal concerning movement properties of phases and in line with Richards’ (2007) notion of good phasal design, I proposed the existence of a functional head above HN (at least with plural nominals), a variable phase head carrying the morpho-syntactic number-feature, in the last section. In contrast, split-topicalization of mass HNs with weak quantifiers seems to be barred by the phasal status of Q W in the absence of Num; moreover, since Q W is unable to combine with singular nominals at all, this case is most satisfyingly handled by selectional restrictions. 26 With the focus on nominal features, let us once again turn to mass, singularia and pluralia tantum nouns as an initial step into the discussion. I follow Borer (2005: 93) in the assumption that all nominals are underlyingly mass and further featural specification allows interpretability and interaction within the functional domain in the extended projection: “[A]ll nouns, in all languages, are mass and are in need of being portioned out, in some sense, before they can interact with the ‘count’ system.” 27 Mass nouns, in turn, seem to fail to be ‘portioned out’. This is reminiscent of Bouchard’s (2002) notion of atomization, which, for him, is a strategy that serves participant identification in the event. It is worth quoting him at some length here: [M]ost languages have a second level of grammaticalization regarding the means to ‘atomize’ the set defined by a common noun. For instance, distinctions […] may be grammaticalized by means of a classifier system. That is what seems to be the case in Mandarin and Korean. However, in French and many other languages, atomization is 26 Note that said combinations often turn grammatical if a shift to mass reading of the noun is evoked, a fact to which I will return below. 27 Cf. Chierchia (1998) for an approach based on the diverging rationale, termed the ‘Inherent Plurality Hypothesis’, levelling mass nominals with plural count instances in proposing that the former are already specified for plurality in the lexicon. 114 2 PREMISES - Phasehood & φ-Dependencies done by adding features of Number, definiteness, specificity, demonstrativity, etc. and it steers the referential potential of the N to a particular possibility. (Bouchard 2002: 40) Another nominal feature that should fall into this listing is gender, which Lyons (1999, ch. 5.6) takes to be closely related to classifier systems as the ones mentioned by Bouchard above (cf. also Croft 1994, Greenberg 1978 as well as ch. 3.4 below). As noted by several authors, gender/ nominal classification differs from other φ-features in its (predominant) invariability w.r.t. particular LIs (cf. Alexiadou et al. 2007: 236, Corbett 1991: ch. 6.3.1). Additionally, as Lyons (1999: 137) further notes, gender is most prominently coded in third person across languages. I will follow the rationale of a fixed gender-value for nominals, building on Ralli (1994, 2002, 2003) with, e.g. Ritter (1993) and Lyons (1999). 28 However, I depart from the former in that I will treat gender-values as active nodes on the respective LI in a feature-geometric set-up (cf. ch. 1.4.4 above, cf. Harley 1994, Harley & Ritter 2002a). A first sketch of the dependency just outlined is depicted in (52): (52) 60 (46) QP 2 b1. Q FP 3 b2. Q F’ 3 Dem F’ 3 Poss LEX F’ 3 a1. Poss PRO F’ 3 F 0 … Det a2. Poss PRO (49) Q S P 3 3 Q S( * ) D 3 1 1 D* Poss PRO P 3 2 Poss PRO( * ) Q W P 3 2 Q W( * ) NumP 2 3 1 Num ( * ) NP 3 N Poss LEX P* (52) N 9 [m] [f] [n] (57) a. N b. N 9 9p [m] [f] [n] [m] [f] [n] sg ! ! ! sg sg sg (58) Num ! [pl] Notice that under this view a noun can surface without a gender-value, if any node fails to be activated. With Borer’s (2005) proposal and the rationale of Bouchard (2002), laid out above, this is exactly what I want to argue for in the case of mass nouns in German. Moreover, although these nouns do come with a fixed value for gender, they lose their non-specific mass connotation when externally merged bearing such a value (or for that matter, any φ-feature) in the derivation. This is a bold claim to begin with, but we can find morpho-syntactic evidence that something along these lines is actually the case in German. Recall from chapter 1.4.1 that I proposed the existence of two discrete phrases for quantificational elements as a hybrid of the QPand AP-Hypothesis of C&G. Concentrating on the adjectival characteristics of Q W apart from its linearizational properties, we can observe adjectival inflection following the strong / weak 28 Cf. also Alexiadou et al. (2007); see Picallo (1991) on Gender as a dedicated phrase GenP in the nominal domain; cf. Panagiotidis (2002), Alexiadou (2004) for critical discussions of this account. 2.3 φ-Features 115 bipartition with weak/ strong SLIs respectively. Note the complete parallelism in inflectional properties of the weak quantifier and a common adjective with a plural HN: 29 (53) strong weak nom {viel-e, groß-e} Menschen die {viel-en, groß-en} Menschen gen {viel-er, groß-er} Menschen der {viel-en, groß-en} Menschen dat {viel-en, groß-en} Menschen den {viel-en, groß-en} Menschen acc {viel-e, groß-e} Menschen die {viel-en, groß-en} Menschen { many , big } people the { many , big } people German does not display a gender distinction with plural HNs, however, adjectives do agree with the gender of singular count and mass HNs. Observe that the reference of the mass noun turns specific in the combination with the gender-agreeing LI. 30 (54) a. count masc neut nom groß-er Mann groß-es Kind big man masc big child neut ‘a big man/ child’ b. mass masc neut nom groß-er Wein groß-es Bier big wine masc big beer neut ‘a large quantity of wine/ beer’ 29 The full paradigm of possible inflectional realizations on Q W is much more heterogeneous than the tables presented in this chapter indicate. The core phenomenon laid out in the main text nevertheless suffices to introduce the basic contrast at the current point. I will come back to this matter several times in the course of chapter 3.3 and will finally discuss the complete paradigm and the derivations underlying the varying inflectional options in chapter 3.3.4. 30 An alternative approach might argue that gender-specification is always part of the nominal head and it therefore still counts as bare, since no ‘atomization’ in Bouchard’s sense takes place by said feature. Nothing hinges on this so far. 116 2 PREMISES - Phasehood & φ-Dependencies However, inflection (qua φ-agreement) is severely degraded, if not completely ungrammatical, with the weak quantifier in structural case (nom, acc): (55) mass masc neut nom viel-ø/ ##-er Wein viel-ø/ ##-es Bier many wine masc many beer neut From (53)-(55) we might tend to conclude that the weak quantifier does not probe for gender. These nominals do, however, carry gender-specification in complex Adj-N configurations, as can be seen from (54) as well as (56), coordinating Q W and Adj in one complex NP under a mass nominal head: (56) mass masc neut nom viel-ø gut-er Wein viel-ø gut-es Bier many good wine masc many good beer neut We might conclude that the quantifier probes for number rather than gender in German. I will follow Alexiadou et al. (2007: 232) in the assumption that the morphologically unmarked number-value constitutes the default value. Since the SLI is completely barred from surfacing with singular HNs in count interpretation but licit with (gender-specified) mass nominals, I will modify (52) by the addition of a singular feature/ value [sg] on the nominal, either a dependent of the respective gender-values or an additional node on the root. Both of these accounts carry obvious problems on economic grounds. For now, I will stay agnostic as to the exact dependency of gender and singular at the nominal root. (57) 60 (46) QP 2 b1. Q FP 3 b2. Q F’ 3 Dem F’ 3 Poss LEX F’ 3 a1. Poss PRO F’ 3 F 0 … Det a2. Poss PRO (49) Q S P 3 3 Q S( * ) D 3 1 1 D* Poss PRO P 3 2 Poss PRO( * ) Q W P 3 2 Q W( * ) NumP 2 3 1 Num ( * ) NP 3 N Poss LEX P* (52) N 9 [m] [f] [n] (57) a. N b. N 9 9p [m] [f] [n] [m] [f] [n] sg ! ! ! sg sg sg (58) Num ! [pl] Common nouns can hence be taken to standardly activate any gender along with its dependent or the remote singular node ((57) a., b. respectively), while mass nominals are able to carry a value for the former without simultaneously activating the latter. Since Q W cannot surface with a complement specified as [sg], we have excluded split-topicalization of singular count nouns from below 2.3 φ-Features 117 Q W from the system and gained a hint as to the bias to re-interpret them as mass in concatenation with a weak quantifier (cf. fn. 26). Split-topicalization with the latter kind, in turn, is excluded by the phasal status of the quantifier, elaborated above. Turning once again to the inflectional properties of the weak quantifier, recall from the discussion above that the gender distinction ceases to exist with plural HNs in German, the standard view of which takes plural to bear an undifferentiated gender-value. I proposed to locate plurality on a dedicated head NumP in the last section, based on linearizational divergences of singular and plural nominals in weakly quantified structures. The featural content of NumP can thus be depicted as (58), analogous to (52)/ (57). (58) 60 (46) QP 2 b1. Q FP 3 b2. Q F’ 3 Dem F’ 3 Poss LEX F’ 3 a1. Poss PRO F’ 3 F 0 … Det a2. Poss PRO (49) Q S P 3 3 Q S( * ) D 3 1 1 D* Poss PRO P 3 2 Poss PRO( * ) Q W P 3 2 Q W( * ) NumP 2 3 1 Num ( * ) NP 3 N Poss LEX P* (52) N 9 [m] [f] [n] (57) a. N b. N 9 9p [m] [f] [n] [m] [f] [n] sg ! ! ! sg sg sg (58) Num ! [pl] I additionally demonstrated that inflection on Q W is sensitive to the occurrence of plural, thus NumP, on the basis of (55) and (56) above. An important idiosyncrasy, however, can be found in parallel structures incorporating a feminine HN, Observe (59) and (60) for (55) and (56) respectively: (59) mass fem nom viel-ø/ e Milch many milk fem (60) mass fem nom viel-ø/ e gut-e Milch many good milk fem Inflection on the quantifier is not at all degraded and purely optional in these examples, irrespective of, but parallel to, the overt co-occurrence of gender on the adjective. This is surprising, given the conclusions drawn above that it is not gender that is probed by Q W . Feminine gender can hence be taken to constitute an (optional) goal for weak quantifier inflection and therefore patterns with plural HNs rather than masculine and neuter. I will come back to the underlying mechanisms of establishing agreement in detail in chapter three; cf. also fn. 29. 118 2 PREMISES - Phasehood & φ-Dependencies Below, I want to concentrate on the feature composition of, and dependencies on, Num 0 instead. 2.3.2 Plural and Feminine (Values) With the observations from the previous section in mind, let us turn to Ritter’s (1993) discussion of gender-values in Hebrew and Romance languages. There, her main argument is concerned with the divergent loci of gender-features in said languages as either N (Hebrew) or Num (Romance), to be subsequently attached to the noun by head movement of the LI to the number head in the latter case (cf. Ritter 1993: 795, 799). Her proposal is based on three diverging properties of these languages; I will go through these in succession in what follows and match the number/ gender system of German against them. a. Gender Manipulation as a Derivational Strategy As noted by Ritter, building on Bat-El (1986), gender switching by suffixation of one or more feminine markers constitutes a productive non-compositional strategy to coin semantically related neologisms in Hebrew. This stands in contrast to Romance where, as Ritter reports referring to Harris (1991), gender switching by suffixation is foremost restricted to nouns with human or animate reference to retain correspondence with the biological sex of the referent. However, Ritter reports that the connection from suffix (‘word marker’ in Harris’ terms) to gender is an indirect one since the forms themselves are not exclusively associated with this feature and moreover sometimes able to also mark non-nominal LIs. Apart from rare cases of non-animate gender-marked nominals in a parallel fashion to the Hebrew paradigm, one also finds semantically unrelated derivations of roots with different word markers. Based on the situation sketched above, Ritter (p. 798) concludes that the partition of gender from the category N is responsible for the unavailability to utilize gender switching/ marking to derive new nominals in Romance (i.e. Spanish in her analysis): “Like number, gender is a purely inflectional feature of Spanish nouns and hence is unavailable as a derivational strategy” (Ritter 1993: 798). 31 Observe that therein gender and number are treated on a par as instances of inflectional morphology. 31 Note that the discussion focusses on the grammatical feature of gender, specifically feminine gender, not to be confused or equated with the biological property of (female) sex. Although my argument will incorporate cases of overlap in that the latter is referred to by use of the former, I do emphasize the overall independence of the concepts; see Khim (2005: 483) for a parallel partition of semantic and syntactic gender (i.e. the ‘word class morpheme’ from the ‘encyclopedic feature’ masculine/ feminine), also based on Spanish data; furthermore Wiltschko (2009: 42-45) for a split of gender in the domain of 2.3 φ-Features 119 Turning to German, we find that nouns are standardly not morphologically marked for gender but carry an inherent value which surfaces on adjectives and SLIs in the extended projection of HN. We do, however, find productive gender switching with masculine animate/ human referent nominals: When referring to a biologically female referent, they are suffixed by in and thus switch to feminine gender. 32,33 (61) a. der Arzt (male) [ the doctor ] masc a'. die Ärzt-in (female) [ the doctor ] fem b. der Lehrer (male) [ the teacher ] masc b'. die Lehrer-in (female) [ the teacher ] fem In contrast, we do not find semantically unrelated neologisms formed by this strategy at all. The noun stem maintains its unambiguous reading in all cases of gender switching in German, parallel to Romance. b. The Gender of Irregular Plural Forms Turning to irregular plurals, Ritter notes that Hebrew exhibits two distinct gender forms for plural HNs, viz. masculine and feminine respectively. What first determiners as either determined syntactically (i.e. by the Agree with the nominal head, hence an inherent and purely formal feature) or pragmatically (i.e. from the discourse referent, hence a modifying feature) across languages. In her system, German constitutes the showcase language of the former type. Therein, a third instance of the disconnection of morpho-syntactic and semantic dimensions of features has been revealed; however, cf. Saab (2010: ch. 5.1) for a critical revision of Ritter’s argument, presented in the main text, as well as the Spanish data on which it is based. 32 This strategy in fact applies to a variety of animate nouns in German. There are, however, exceptions with both neuter as well as masculine nominals, cf. e.g. Opfer / * Opfer in (victim neut ) or Gast / * Gäst in (guest masc ). 33 Strategies of partly biologically motivated grammatical gender switches are numerous in German (cf. Motsch 2004: ch. 4.2.4), however, they are fairly restricted in their application (e.g. ‘masculinization’ of inherently feminine animal species-terms) or modify the semantics of the base noun (diminutivization: masc/ fem/ neut to neut; cf. Greenberg 1978: 79 for the treatment of diminutives as a (minor) gender category and its relation to the neuter gender-class across languages) parallel to what was reported for Romance in the main text, the latter of which indirectly support the final points I want to convey. 120 2 PREMISES - Phasehood & φ-Dependencies might look like counter-evidence to a genderless number phrase turns out to be a highly idiosyncratic system in that masculine nominals occasionally take - what has been called - feminine plural suffixes and vice versa. Therein, however, the original gender of the nominal stem still serves as the probe for agreement with the extended projection of HN as well as the clause. Ritter concludes that Hebrew plural suffixes are hence only specified for number. In contrast, gender switching of irregular nouns in Romanian is accompanied by a respective switch in the agreement pattern of the LIs in the nominal hierarchy of projections and the dominating verbal domain. This effect can likewise be found in German, considering that the tripartition of gender-values ceases to exist with plural HNs. The ‘plural gender’, though, triggers agreement in the upper phasal cycle of the extended projection of HN that is morphologically identical to the paradigm found with feminine nominals in the singular, which has already been informally noticed by Roehrs (2013: fn. 10). Let us, for the sake of argument, equate these patterns. We can hence fully adopt Ritter’s conclusion (p. 799) for the situation in German: “In terms of the analysis presented here, it is the plural marker itself that is feminine […]. In other words, […] plurals […] involve gender switching because Num, not the noun stem, bears the gender specification of the noun phrase.” However, in contrast to Romanian, plural formation on the basis of feminine gender constitutes a unique strategy and hence applies to all German nouns, irrespective of their gender-value in the singular. On the basis of Ritter’s insights (and in accordance with the observations on parallel probing and identical realization above), I want to propose a split in the location of gender-values to the extent that it is solely the feminine gender that is located on Num, parallel to Ritter’s claims, while the masculine and neuter gender are still taken to be coded on the lexical core projection N. This partition copes with the fact that, even if we do find non-plural semantics with feminine SLI agreement, we never find plural semantics without feminine SLI agreement in German. Feminine gender can hence be understood as a semantically numberless syntactic plural marker in German and thereby constitutes an instance of inflectional morphology initiated in Narrow Syntax, a point that will be at the heart of the argument of the next subsection. Moreover, we might have found a first hint as to why weak quantifiers optionally inflect with singular feminine HNs in (59) and (60) above: They probe for Num, which is necessary present in the extended projection of feminine nominals. In the remainder of this chapter, I will further elaborate on this proposal and put forth additional evidence in its favor. I begin by reviewing the last argument from Ritter’s paper in the light of the claim just made. 2.3 φ-Features 121 c. The Surface Position of Gender The hypothesis to (partly) locate gender on Num, rather than the nominal head, makes the prediction that languages which code said features morpho-syntactically, e.g. via suffixation, would in principle be able to display number/ gender distinct from the nominal head. Ritter only presents tentative evidence in this respect in favor of her analysis, since the proposed head movement from N to Num levels the linearizations across the Hebrew-Romance bipartition w.r.t. the effect under consideration (cf. Ritter 1993: 800, ibid.: fn. 5). Said evidence is taken from Bernstein’s (1991) analysis of Walloon, a Romance language spoken in Belgium. As Bernstein notes, Walloon nominals are not encoded for plural themselves, due to the lack of N-to-Num movement, while prenominal adjectives do show the respective φ-agreement. Therein, adjectival agreement qua suffixation applies by means of a portmanteau morpheme in the context of feminine plural nouns, arguably coding inherent feminine gender along with plural number. As Ritter (p. 801) notes, Bernstein herself analyzes said suffix as the overt realization of the number head, hence of the φ-complex apart from the nominal head. Turning to German once again, recall from the discussion above, that gender switching does not constitute a productive strategy for deriving new (i.e. semantically unrelated) LIs, but is restricted to semantically related derivations on the basis of the referent’s biological specifications of sex foremost in the context of occupational titles. A closer look at the paradigm of predicatively used instances of such nominals reveals that the features under consideration exhibit some interesting dependencies: (62) (62) a. Sie Er ist Lehrer. (s)he is teacher b. Sie Er ist der Lehrer. (s)he is the masc teacher c. Sie ist Lehrer-in. she is teacher fem d. Sie ist die Lehrer-in. she is [ the teacher ] fem ‘He/ She is the teacher.’ 122 2 PREMISES - Phasehood & φ-Dependencies e. Sie sind Lehrer. they are teachers f. Sie sind die Lehrer. they are the teachers g. Sie sind die Lehrer-in-(n)en. they are [ the teachers ] fem.pl ‘They are the teachers.’ From (62) a., we see that the predicative use of the nominal builds on the (masculine) base form, irrespective of the referent’s (i.e. the sentence-initial pronoun’s) grammatical/ biological gender. This amounts to the bare qua mass reading of the nominal in Borers theory, incorporated here. Interestingly, the φ-specific masculine DP is also licit in this construction, once again ranging over all possible co-referents ((62) b.). 34 The data depicted in (62) c. and d., then, mimic the cases above as realizations solely for female referents: the in -derivation (socalled ‘feminization’) with and without the according determiner. Turning to plurals, (62) e. and f. denote any combination of female and/ or male teachers. 35 The last datum is the most important one in the light of the proposals made above: We once again observe the nominal realization specified for the reference to female entities, but in the case of (62) g., it co-occurs with a morphologically realized plural as suffixes on the nominal base. The structure imposes a strong semantic restriction in that every individual in the set of referents has to be female, that is, the feminine suffix does not range over the plural set per se but evaluates every member individually (i.e. singularly) before the plural marker forms a set (cf. Schwarzschild 1996, also Cruse 2004: ch. 13.4.5). 36 This is reminiscent of Bouchard’s rationale of (φ-)feature coding for the purpose of atomization, laid out in the beginning of this section. In parallel, morpho-syntactic plural marking in predicatively used nominals follows the morpho-syntactic realization of the gender-value feminine, as also observed by Ritter (1993: ch. 3, cf. also ibid.: fn. 5) and furthermore in line with the generalization reported by Khim (2005: 462) that “Class exponents, both gender and noun class, are always 34 This variability in agreement/ reference has been subsumed under the term ‘hybrid noun’, cf. Corbett (1991: ch. 8). 35 Departing from the underspecified instance in (62) a. above, the predicative nominal can be taken to bear idiosyncratic zero plural morphology as can be concluded with reference to nominals that overtly mark plural in context like (62) e. Incidentally, note the identical morphological form of feminine and plural pronouns as well as the feminine and plural determiner. 36 This restriction, however, is not absolute since some of my informants accept the form in contexts referring to groups consisting predominantly rather than exclusively of female entities. 2.3 φ-Features 123 combined with Number morphemes when overt - that is, nonsingular.” 37 In contrast, note that the only equally productive strategy of grammatical gender switching in German, i.e. diminutivization, mentioned in fn. 33 above, does not exhibit a morphologically realized plural. In the light of the data presented here, I feel justified to further pursue the proposal of a close connection of the origins of morpho-syntactic realization of plural number and feminine gender in German along the lines proposed by Ritter, as modified above, and further explore the parallels and dependencies between them, to which I will turn now. 2.3.3 Plural and Feminine (Category) I argued above that the head Number encodes two φ-values, namely plural and feminine, on the basis of inflectional idiosyncrasies (ch. 2.3.1), the derivation of (gender-specific) plurals, morphological parallels as well as order and cooccurrence of suffixation (ch. 2.3.2). In this section, I want to discuss further parallels to back up this proposal. As will be shown, once the rationale of grouping said φ-values together structurally is established, further morpho-syntactic peculiarities of German nominal inflection will fall into place. However, a closer look at the exact coding of φ-features in general seems to be in order first. So far I have proposed the regrouping of φ-features in the nominal domain without being specific on the relations and dependencies of these features to one another. I want to claim with Harley & Ritter (2002a), building on Harley (1994), herself modifying Noyer ([1992] 1997), that morphosyntactic features are ordered in hierarchical fashion (cf. ch. 1.4.4 above). Noyer captures this fact in his Feature Hierarchy for categories while Harley encodes the values themselves in hierarchical order; from a feature-geometrical viewpoint, these relations are only incidental since her focus lies in the complexity and markedness of the respective values rather than structural dominance (cf. Harley 1994: 14). What unifies all these accounts is the proposed relevance of these features beyond the narrow-syntactic stage of the derivation, viz. at an early stage from the phonological component, Φ to PHON in current minimalist terms, labeled ‘Morphological Structure’ (MS) in Distributed Morphology (cf. Halle & Marantz 37 Additionally, we might propose N-to-Num movement in German in line with the analyses of Ritter/ Bernstein in the overt suffixial coding of plural (and feminine) feature values on the nominal head, cf. also the notion of m-merger in Distributed Morphology, exemplified in the beginning of the ensuing subchapter, even more so in light of Chomsky’s (DbP: 37f.) relocation of (nearly) all head-movement from Narrow Syntax to the phonological component. 124 2 PREMISES - Phasehood & φ-Dependencies 1993, also Harley 1994), ‘Morphology’ by Noyer ([1992] 1997, cf. also Chomsky MP: 229); I will follow the latter convention here. In said component, the output of NS, consisting solely of hierarchically ordered bundles of formal features (taken to generally correspond to LIs), are subject to additional but distinct derivational operations before their substitution for phonological material (‘Vocabulary insertion’) and the translation of hierarchical relations into linear order (‘Linearize’/ ‘Linearization’). Therein, “the ordering, number, feature composition and hierarchical positioning of terminal nodes may change in the derivation of MS, but only in highly constrained and fairly well understood ways” (Halle & Marantz 1993: 121). These ways are depicted as the operations ‘(m-)merger’, ‘fusion’ and ‘fission’ (as well as ‘impoverishment’, ‘enhancement’ and ‘neutralization’ in Noyer’s instantiation of the theory, to be made precise below): While merger parallels internal head-Merge in NS in that it coordinates two dedicated heads under one node, fusion further takes such structures (irrespectively of their derivational locus of creation as either NS or Morphology) as its input to return a single terminal item for Vocabulary insertion. Contrary to this, fission is the operation that splits terminal nodes into multiple ones. The outcome at PHON is therefore a disruption of the one-to-one mapping from terminal node, as handed down from NS, to morphological item. Hence, as Harley (1994) notes, Morphology is an autonomous step in the derivation. Concentrating on the processes of word formation under consideration here, observe hence that (inflectional; cf. Noyer [1992] 1997: 13) “word-formation is syntactic and postsyntactic, not lexical” (Halle & Marantz 1993: 166, cf. also ibid.: 112). Returning to the accounts to feature-coding laid out above, note that Harley & Ritter (2002a) make explicit reference to the incorporation of an interpretative morphology like the framework laid out above into their approach (cf. Harley & Ritter 2002a: 510, 518). Observe that, since feature compositions do hence not need to exhibit narrow-syntactic relevance in their system on pronouns as well as in my own on nominal core categories elaborated here, both escape the criticism of syntactic visibility of features from Déchaine & Wiltschko (2002), reported in chapter 1.4.4 above (but cf. ch. 3.2, especially fn. 4 below). Below, I give the featural specifications of N based on the proposals arrived at in the last section as well as on the basis of Harley’s (1994: 14) partial hierarchy and Harley & Ritter’s (2002a: 486) feature geometry, with the terminology of the latter in parenthesis. 38 38 Note again that Harley & Ritter (2002a: 506) explicitly state that their account is restricted to (what they call) ‘Referring Expressions’, namely pronominal paradigms and is not concerned with the featural set-up of common nouns; cf. ch. 1.4.4 above. 2.3 φ-Features 125 (63) Phasehood & φ-Dependencies P REMISES 61 (62) a. Sie Er ist Lehrer. (s)he is teacher b. Sie Er ist der Lehrer. (s)he is the masc teacher c. Sie ist Lehrer-in. she is teacher fem d. Sie ist die Lehrer-in. she is [ the teacher ] fem ‘He/ She is the teacher.’ e. Sie sind Lehrer. they are teachers f. Sie sind die Lehrer. they are the teachers g. Sie sind die Lehrer-in-(n)en. they are [ the teachers ] fem.pl ‘They are the teachers.’ (63) N 3 singular gender (minimal) (class) 3 [m] [n] (64) Num ! [f] ! [pl] Since I omitted the feminine gender from N, I also deleted the animate-node ranging over masculine and feminine in Harley & Ritter’s system. Likewise, I abstracted away from the participant-node, encoding person-values, which would be situated as the sister of the root node N. As can be deduced from (63), gender can be active independently from singular, as required for the treatment of (gendered) mass nominals, advocated above. Singular, however, constitutes the morphologically unmarked form and is hence taken to denote the default case, following Alexiadou et al. (2007: 232; see also Wiltschko 2009: 46). Additionally, as noted by Bejar (2003: 47) in her discussion of the feature geometry proposed by Harley & Ritter (2002a), observe that the relation of sisterhood bears an arbitrary character therein, which is also observable in the geometry presented in (63): The structural configuration might signal either cross-classification, complementary distribution or combination of the values subsumed under the same mother node. In the geometry above, sisterhood does indeed serve two varying purposes, namely cross-classification (of the number-value [singular] with the gender-node) in the higher instance, as well as complementary distribution (of two gender-values) in the lower case. I will get back to this in chapter 3.2.1 below. Turning to Num, we are faced with a similar question concerning the coding of plural and feminine. In the light of (63) and the analyses underlying it, a parallel set-up might suggets itself and does in fact constitute the number/ gender sub-tree from Harley (1994: (17)), mentioned above. 39 I want to argue, however, that such an ordering misses the observations made earlier for German. If, on the other hand, one understands inflectional identity to point to a single category, an alternative approach readily suggests itself. I want to propose a relation of said features that orders plural as the sole dependent of feminine (i.e. ‘uniquely licensed’ in Noyer’s terms): 39 This relational ordering ultimately dates back to Lumsden’s (1992) analysis of Romanian. 126 2 PREMISES - Phasehood & φ-Dependencies (64) Phasehood & φ-Dependencies P REMISES 61 (62) a. Sie Er ist Lehrer. (s)he is teacher b. Sie Er ist der Lehrer. (s)he is the masc teacher c. Sie ist Lehrer-in. she is teacher fem d. Sie ist die Lehrer-in. she is [ the teacher ] fem ‘He/ She is the teacher.’ e. Sie sind Lehrer. they are teachers f. Sie sind die Lehrer. they are the teachers g. Sie sind die Lehrer-in-(n)en. they are [ the teachers ] fem.pl ‘They are the teachers.’ (63) N 3 singular gender (minimal) (class) 3 [m] [n] (64) Num ! [f] ! [pl] Therein, I want to do justice to the fact, mentioned in the discussion of the gender of irregular plurals in the last subsection, that while we do find feminine/ plural agreement on semi-lexical items with singular HNs (i.e. feminine), we do not find plural HNs without said agreement pattern. *13 Returning to the Walloon data discussed in the last subsection in the light of this proposal, if the structure carries over to said language, its Num-heads could be taken to be always maximally specified, when present, since both these values are always simultaneously coded overtly therein. A last refinement, however, is in order concerning the proposals of chapter 2.2 above. Recall that I proposed the variable phasehood of NumP as solely dependent on (the phasal status of) the element dominating it in the nominal hierarchy of projections. Thereby, the system was able to capture split-topicalizations with plural HNs while at the same time excluding movement from below determiners, demonstratives, possessive pronouns and strong quantifiers (i.e. the upper phasal cycle). In this section, however, I (tentatively) adopted the proposal of N-to-Num movement in plural as well as singular feminine nominal domains, in line with Ritter (1993: 795, 799; but cf. fn. 37 above). However, feminine nominals are barred from split-topicalization along with their remaining singular counterparts. Hence, the phasal status of Num 0 must be further restricted to plurally marked instances. Turning to the featural set-up argued for above, only maximally specified Num-heads exhibit phasal status. *14 Returning to the featural set-up proposed above, observe that the combinatorial options of active nodes from (63) and (64) account for all the classes of nominal elements introduced so far. Hence, we find common singular nouns with masculine and neuter as well as feminine gender, mass nouns in unspecified and gender-specific instances as well as plural nouns. These are listed below with the active features specified in the third column: 2.3 φ-Features 127 (65) a. proper singular nouns der Mann • masc/ fem/ neut • sg die Frau das Kind the masc/ fem/ net man/ woman/ child b. φ-specific mass nouns (ch. 2.3.2) der Wein • masc/ fem/ neut die Milch das Bier the masc/ fem/ net wine/ milk/ beer c. plural nouns die Lehrer • (masc/ fem/ neut) • pl the pl teachers However, once we turn to the strategy of feminization from masculine base forms, the system seems to run into severe problems. See (62) d. and g., reprinted here as (66): (66) a. Sie ist die Lehrer-in. (= (62) d.) she is [ the teacher ] fem b. Sie sind die Lehrer-in-(n)en. (= (62) g.) they are [ the teachers ] fem.pl ‘They are the teachers.’ While the datum in (66) a. can be accounted for in the system elaborated here, in that arguably two different gender-values are part of the derivation, introduced by two different heads (viz. N 0 , Num 0 ) in line with Ritter’s claims, the plural derivative of the feminized form in (66) b. seems problematic under the featural set-up laid out in (64), once it is contrasted with (65) c., i.e. simple plural forms. The reason lies in the parallel suffixation of feminine and plural morphology: Since plural is the dependent of feminine in the model laid out above, and the latter possesses a morphological realization in the environment depicted in (66), namely masculine base forms, we expect said realization to surface in every morphological string that displays plural morphology. In the framework of Distributed Morphology as laid out in Noyer [1992] 1997, a filter of the type *[f pl] would impoverish the complex ‘morpho-syntactic constituent’ M 0 (the output of fusion; cf. Noyer [1992] 1997: 13) in the first phase of Morphology, exemplified above, by ‘Delinking’ of the most deeply embedded feature from 128 2 PREMISES - Phasehood & φ-Dependencies the illicit combination. 40,41 Noyer captures this fact in his rule of ‘systematic neutralization’: 42 (67) (= Noyer [1992] 1997: ch. 0, (55)) Systematic Neutralization Systematic Neutralizations are the result of Feature Cooccurrence Restrictions (filters). Filters combined with feature hierarchies effect Impoverishment. With respect to (64), however, this would leave us with feminine on Num (but cf. Harley 1994: ch. 3.3.1). Additionally, structures like (66) b. would be derived by ‘Panini’s Principle’ (the ‘Elsewhere Principle’) depicted in (68) a., which, w.r.t. (64), results in the ill-formed (68) b.: (68) a. Panini’s Principle (= Noyer ([1992] 1997): ch. 1, (56.1)) If one rule’s structural description is contained in the other’s, the rule with the more specific structural description applies first. b. *N-[pl]-[f] However, Noyer’s system is able to account for the problematic string in his notion of ‘enhancement’: He proposes that logically predictable and therefore redundant features may be absent due to a filter, active in the lexicon (and hence throughout the derivation). This second strategy of neutralization therefore applies to the highest feature in an illicit configuration. As Noyer notes (p. 56), due to the redundancy of the omitted feature no semantic effect results from the underspecification. I want to adopt Noyer’s notion into the system under development here (but see Harley 1994: ch. 3.1.3 for a critique under feature-geometrical considerations). Since feminine is logically predictable in plural nominals in German, it is standardly omitted morphologically. However, since the overt occurrence in the context of feminization of male base forms is derivationally regular, I have to locate the filter in the derivation rather than the lexicon (presumably at Morphology together with Impoverishment by Delinking): Feminine gender surfaces (‘discharges’ in Noyer’s terms) only in those plural-marked M 0 s where it yields a semantic contribution, i.e. feminization. Featurally speaking, 40 The notions ‘Impoverishment’ and ‘Delinking’ are due to Bonet i Alsina’s (1991) treatment of pronominal clitics in Romance; for a previous approach combining these notions in the framework of Noyer ([1992] 1997) with an earlier version of Harley & Ritter’s feature-geometric account see Vinka’s (2001) analysis of agreement in Sámi. 41 Since M 0 s are the input of filters as well as the output of fusion in Noyer’s framework we can propose that the complex N-Num has undergone fusion and hence that either IM or m-merger has taken place beforehand. 42 All citations are taken from the 1992 dissertation. 2.3 φ-Features 129 it is fully specified N 0 s (bearing singular as well as the gender-value masculine) that trigger the surfacing of the form in (66) b. when N-to-Num movement applies at NS or Morphology (i.e. IM or m-merger respectively) to a probe specified as plural. With Noyer, this fact can be paraphrased as (69). Observe that the affixial realization is moreover in line with Chomsky’s (DbP: 38) proposals of post-syntactic complex head-formation (cf. also fn. 37 above). (69) f (m,sg) in In Noyer’s terms, -in is the ‘principle exponent’ of feminine and the ‘secondary exponent’ of masculine singular (cf. Noyer [1992] 1997: ch. 1.4), translatable as ‘feminine in the environment of masculine singular’. Likewise, the dominant plural allomorph for feminine as well as feminized nominals can be captured by (70). 43 (70) pl (f) en We might hence trace the order of semantic decoding of groups of solely female individuals to the derivational application of multiple suffixation (discharge) and further to the featural content of the complex LI. Additionally, the order of feminine and plural suffixes mirrors the levels of feature specification on Num. 44 This is also in line with Ritter’s (1993: ch. 3, ibid.: fn. 5) findings. Lastly, it follows from the feature-geometrical rationale that these forms constitute marked occurrences for a very specific interpretation which can optionally also be expressed by the more general and hence less marked simple plural that, in turn, does not bear any semantic reference to single entities (as also predicted by the system elaborated here). Below I give all discussed nominal forms and the respective featural specifications for N and Num. Parenthesis signal complementary distribution, angled brackets indicate enhancement: 43 Abstracting away from the linking element -nin written form. 44 Observe that this argument follows the rationale of Baker’s (1985) ‘Mirror Principle’, when transposed to intra-LI feature geometries. 130 2 PREMISES - Phasehood & φ-Dependencies (71) a. proper singular nouns N: Num: (m)/ (n), sg (f) b. φ-specific mass nouns (ch. 2.3.2) N: Num: (m)/ (n) (f) c. plural nouns N: Num: (m)/ (n) <f>, pl d. feminized nominals (ch. 2.3.2) N: Num: m, sg f e. feminized plural nouns (ch. 2.3.2) N: Num: m, sg f, pl In what follows, I want to present further evidence that the regrouping of φ-features along the lines presented above accounts for several unrelated linearizational and morpho-syntactic idiosyncrasies of German. a. Umlaut First tentative evidence can be found in the morphological derivation of plurals and feminized forms. So far I have only discussed masculine nominals which form their plural by means of zero-morphology, however, nouns belonging to the so-called strong declensional class predominantly form their respective plural by umlaut of the stem vowel, often also accompanied by plural suffixation. 45 Note (72) in which the plural of the masculine noun Arzt (‘doctor’) incorporates a change of the initial vowel: [aː] is shifted to [ɛː]. (72) a. Er/ Sie ist Arzt. (s)he is doctor b. Sie sind Ärzte. they are doctors pl ‘They are doctors.’ However, turning to Noyer’s work once again, I have to refrain from incorporating umlaut into the strategy of plural formation. Noyer ([1992] 1997: ch. 0.1.1.5) distinguishes between ‘structure-changing’ and ‘structure-building’ effects, associated with morphological rules with and without affixation respectively. The derivational differences are captured in his ‘rule-affix hypothesis’: 45 Only two of the six inflectional patterns of the strong declension do not form their plurals by suffixation. Those nominals either seem to only mark it by umlaut or have homonymous singular and plural forms. I will come back to these below. 2.3 φ-Features 131 (73) (= Noyer [1992] 1997: ch. 0, (25)) The Rule-Affix Hypothesis (1) A morphological rule introducing an affix is a structural-building rule. Structure-building rules discharge features and positions-of-exponence. The affix so introduced is the principal exponent of the feature discharged. (2) A structure-changing (or feature-changing) rule does not discharge any features or positions-of-exponence. In the following discussion, Noyer makes explicit reference to German pluralmarking. Additionally, his theory copes with zero-morphology plural-marking by means of an affix, -ø (cf. e.g. my examples in section 2.3.2). Since suffixation and vowel change are “freely cross-classifying” (Noyer [1992] 1997: 29), we expect to find cases in which plural is overtly coded by the latter alone as well as by none of the two strategies described in (73) above (i.e. the instances described in fn. 45). Concerning the standard case depicted in (72), then, it is only the suffix e that discharges the morpho-syntactic feature: “On my analysis, ‘plural’ is not in such cases a bipartite morpheme; rather, the ablaut rule accompanies plural as a secondary exponent” (Noyer [1992] 1997: 30). With this in mind, let me extend the paradigm of (72) by the respective feminized plural form: (74) Sie sind Ärztinnen. they are doctors fem.pl As we would expect under a bi-morphemic analysis of German plurals, we observe the same vowel change as with simple plurals, shifting the initial [aː] to [ɛː]. However, the singular equivalent to (74) under this view leaves us with a puzzling observation: (75) Sie ist Ärztin. she is doctor fem Again, umlaut is observable in (75), however, this form is obviously singular. Returning to the proposal in (64) above, it can be argued that vowel change is associated with (movement to/ fusion with) NumP rather than the φ-value plural. Therein, we do find a first unification of otherwise separate and idiosyncratic morphological phenomena in German. b. Genitive Suffixation Another observation comes from the genitive case forms of German nominals. Recall from the end of the last section that I concluded that N-to-Num move- 132 2 PREMISES - Phasehood & φ-Dependencies ment/ merger and subsequent fusion must apply in German due to the suffixation of feminine and/ or plural morphemes on the nominal, in line with Ritter (1993) and Bernstein (1991). If this is the case, the nominal stem might be expected to display further instances of suffixation that are triggered from outside of the nominal domain, such as case-assignment from a dominating predicate, to apply after said movement and hence the respective suffixes to be ordered accordingly (i.e. N-[gender]-[case]). This is, however, not what we find in German. Instead, feminine and plural nominals uniformly do not exhibit genitive case suffixation at all: (76) masc fem neut pl nom Mann Frau Kind Menschen gen Mann-es Frau Kind-es Menschen Observe that this is also true of feminized nominals and their plural forms: (77) fem pl nom Lehrer-in Lehrer-in-nen gen Lehrer-in Lehrer-in-nen I will refrain from speculations on the mechanisms underlying it but simply take the repeatedly parallel pattering of feminine and plural nominals as a hint as to the correctness of the proposals made above. 46 c. Personal Determiners The discussion of pronominal elements in chapter 1.4.4 touched on the research tradition of locating these SLIs in a prenominal position, parallel to the deter- 46 Observe that the effect cannot readily be attributed to string vacuous N-to-Num-movement, bleeding suffixation, since in Minimalism all Agree relations are established under c-command, which is not altered by the proposed movement inside the nominal hierarchy of projections (given that the case-assigner dominates the complete extended projection of the noun). Moreover, note that given the refinements to the contextual phasal status of Num 0 , made previously in this section, the effect exemplified in (76) and (77) cannot be traced to the application/ timing of TRANSFER/ Spell-Out either. Likewise, application of fusion, taken to preserve the featural content of both input nodes (hence the case-feature of the nominal), seems equally unfitting to account for the phenomenon at hand in a straightforward manner; however, the latter might be taken to deprive the resulting structure of the locus of overt exponence. I will not go into these matters further. 2.3 φ-Features 133 miner. As reported there, said treatment can be traced back to Postal ([1966a] 1970), one of his main arguments being the concatenability of pronouns with nominal heads, as depicted in the example in (78) below (which is, however, not a universal property, cf. Lyons 1999: 142): (78) we linguists As has been also been laid out above, the structural position has been identified as the functional category D as early as Abney (1987); thence, Lyons (1999: ch. 3.4) refers to the prenominal forms in such configurations as ‘personal determiners’. The proposal is still implemented in a wide variety of works on nominal domains but underwent various changes. Lyons (1999) himself links the structural location of pronouns to their morpho-syntactic complexity by means of syntactic properties which loosely resemble those elaborated by Cardinaletti (1998) and discussed in chapter 1.4.3 above on possessive pronouns in German (e.g. stressability and reducibility, cf. Lyons 1999: ch. 3.4.2): Like Cardinaletti, he arrives at a tripartition of strong, weak and clitic pronouns. In his system, only the clitic variant constitutes an instance of the head position D 0 , while weak and strong free-form pronouns are located in the specifier position. Lyons differentiates between them by their determinervs. demonstrative-like content (cf. Lyons 1999: 304), parallel to the use of those latter SLIs in weak and strong pronominal contexts like the one mentioned at the beginning of this section (cf. Lyons 1999: ch. 3.4.3, 3.4.4). All three instances moreover have in common that they do not (necessarily) exhibit concatenation with an NP-complement. A related refinement to the DP-proposal for pronouns takes these SLIs to be externally merged in the lexical domain and subsequently moved to the DPlevel in the course of the derivation. Longobardi (2008), elaborating on the relation of argumenthood and the determiner phrase, takes this route, which is explicitly rejected by Lyons (1999: 303) on the basis of the parallel morphological form of pronouns and determiners as well as demonstratives across languages. For Longobardi, parallel linearizational properties to nominals w.r.t. adjectives in non-argumental (hence, non-DP) interjection configurations as well as the impossibility of any DP-element surfacing in these constructions, suffice to propose an analogous external merging-site: (79) (= Longobardi 2008: (39) a., b.) a. Poverto cane! ([Italian]) Poor dog! b. ? * Poverto il cane! ([Italian]) * Poor the dog! 134 2 PREMISES - Phasehood & φ-Dependencies The obligatory movement to DP - if present - is justified by the identification of said projection with the person-feature as well as the association of the latter with pronouns as well as determiners, to the exclusion of nominal heads. The rationale of a linkage between the φ-feature and these SLIs can also be found in Lyons’ (1999: ch. 3.4.3) discussion of personal determiners, laid out above. There, he notes that the grammaticality of such constructions varies w.r.t. the form in the person-number-paradigm (with the emergence of ungrammatical instances sometimes circumvented by the substitution of the pronoun for the determiner or demonstrative) as well as the argumental vs. exclamatory use across languages. A cross-linguistic tendency, identified therein, is the parallel patterning of first and second person apart from third person forms. This fact is taken up again by Dechaine & Wlitschko (2002, cf. Bernstein 2008: ch. 4.2, Cowper & Curie Hall 2009: ch. 6; cf. also ch. 1.4.4 above) in their analysis of the English pronoun paradigm. Combining data from personal determiners with a decompositional approach to the respective forms, the authors arrive at a mapping of person-values to varying structural positions and complexity, with we / us D-heads, parallel to the bound morpheme th - (cf. ch. 1, fn. 67). Thereby the account accommodates for both the parallel structural behavior of determiners and pronouns as well as the parallel morphological form of the former with a subpart of the pronominal paradigm. Turning to German, we find personal pronouns in prenominal position in both exclamatory as well as argument use in the first as well as the second person singular and plural. Moreover, in both functions, the determiner or demonstrative can be substituted to derive a reading, parallel to the third person pronoun as reported by Lyons (1999: ch. 3.4.3) for various languages. Roehrs (2009: ch. 4.10.3.1) consequently locates these elements in DP on the surface with reference to Postal. However, in line with Longobardi, those SLIs are taken to move to said position, parallel to the class of ein-words, discussed in ch. 1.4.4 above. Evidence for this proposal once again comes from adjectival inflection: As Roehrs argues, “personal pronouns are similar to ein -words such that singular structurally case-marked pronouns take a strong adjective and oblique and/ or plural pronouns a weak one” (Roehrs: ibid., emphasis in the original). This state of affairs is exemplified in (80) below: (80) (= Roehrs 2009: ch. 4, (78)) a. {kein/ ich/ du} armer Idiot {no/ I/ you} poor( strong ) idiot b. {keine/ wir/ ihr} armen Idioten {no/ we/ you} poor( weak ) idiots 2.3 φ-Features 135 This observation is, however, premature in two respects. First, observe that pronouns diverge from the parallel inflection with ein-words in lexical case, as is exemplified below (glosses are modelled after Roehrs 2009): *15 (81) a. keinem armen Idioten [no poor( weak ) idiot] DAT b. dir armem Idioten [you poor( strong ) idiot] DAT Hence, personal pronouns in the singular seem to uniformly co-occur with strong adjectival inflection in contrast to ein-words. With plural nominals, Roehrs (ibid.) himself only denotes degraded grammaticality for strong inflection when the adjective is preceded by a pronominal SLI, which is, however, judged as a clear-cut case in favor of weak inflection for my informants. This generalization, however, is still not complete: Taking the general topic of this subchapter into account, observe that the nominal head in the examples above bears masculine gender. It seems therefore promising to focus on nominal hierarchies built on feminine HNs, once again. However, linearizations paralleling (80) above only return ambiguous data, since strong and weak adjectival inflection are homonymous there as well as in the other structural case, a point to which I will turn in greater detail in the next section. With the genitive furthermore inaccessible on independent grounds (cf. endnote *15), linearizations constructed in parallel to (81) above constitute the only valid testing ground. Following the discussion above, adjectives in the syntactic environment just outlined are expected to surface with strong inflection. (82) a. Ich schreibe [dir schön-er Frau] einen Brief. I write [ you beautiful strong woman ] DAT a letter b. Ich schreibe [dir schön-en Frau] einen Brief. I write [ you beautiful weak woman ] DAT a letter ‘I write a letter to you beautiful woman.’ In (82) above I refrained from marking grammaticality since the intuitions concerning the opposition therein evenly splits the group of my informants. Most interestingly, irrelevant of the individual choice, the informants each experience a distinct preference, fully rejecting the grammaticality of the contrasting option. A quick google-search, only involving the particular combination of adjective, noun and pronoun from (82) above, however, returns roughly three times 136 2 PREMISES - Phasehood & φ-Dependencies more results for the linearization employing weak inflection (63 vs. 226 hits). 47 Returning to the structural proposals laid out above, pronouns indeed have an effect on adjectival inflection, parallel to determiners, with plural and feminine head nouns. 48 Parallel to the distinct coding of strong adjectival inflection in masculine and neuter nominal hierarchies, exemplified in structural and lexical case in (80) a. and (81) b. respectively, we can generalize from the single unambiguous and grammatical form in the feminine - i.e. the dative as illustrated in (82) - to the full paradigm: Plural and feminine personal determiners pattern alike in their concatenability with weak inflected adjectives. I will not elaborate on the implications for the structural locus of pronouns here but rather concentrate on the relationship between the adjectival inflection patterns in the light of current proposals in the next section. d. mixed Adjectival Inflection Recall from chapter one, section 1.3.2 that German exhibits a third inflectional pattern on adjectives, apart from the strong and weak dichotomy. As reported in chapter 1.4.4 above, all elements evoking this so-called mixed pattern seem to share the same morphological base ein and hence make up the group of einwords, viz. the indefinite article ( ein ), nominal negation ( k ein ) as well as Poss PRO ( m ein ), but see chapter 3.3.1 below. Focusing on the resulting concatenations in complex nominal domains, however, note that these SLIs only induce a deviation in the inflectional pattern of the adjective with singular HNs, while plural nominal domains exhibit uniform weak inflection. Since I argue for the regrouping of the φ-feature values of the categories number and gender here, this observation should call our attention. Below I give the full paradigm of mixed adjectival inflection: (83) mixed masc fem neut pl nom gut-er Mann gut-e Frau gut-es Kind gut-en Menschen gen gut-en Mannes gut-en Frau gut-en Kindes gut-en Menschen dat gut-en Mann gut-en Frau gut-en Kind gut-en Menschen acc gut-en Mann gut-e Frau gut-es Kind gut-en Menschen 47 Cf. that the respective strong and weak forms are only homonymous in the other lexical case (i.e. genitive), which has been argued to be barred in these configurations on independent grounds (cf. endnote *15). Hence, the numbers can be taken as undisrupted. 48 As one informant, preferring option (82) b. (i.e. weak adjectival inflection) notes, the contrasting datum turns grammatical in appositional reading which can be understood as separating the pronoun from the extended nominal projection, hence erasing the effect of the former on adjectival inflection. 2.3 φ-Features 137 As can be deduced from (83) above, the pattern of adjectival inflection seems to switch with the number-value of HN. This variation is observed with both morphologically complex ein-words, i.e. Poss PRO as well as the nominal negation k ein , while combining plural nominal projections with the third instance from chapter 1.4.4 above, viz. the morphologically simple indefinite article ein , in contrast, is barred by selection altogether. To approach this apparent switch in inflectional paradigm, let me first turn to the particular morphological realizations therein. Recall from chapter one above that the mixed pattern is composed of strong as well as weak inflectional forms, a subset of which is, however, homonymous. The respective categorizations are represented in (84) with homonymy additionally marked in greyscale: (84) mixed Adjectival Inflection: Homonymous Forms masc fem neut pl nom STRONG HOMONYM STRONG WEAK gen HOMONYM WEAK HOMONYM WEAK dat WEAK WEAK WEAK HOMONYM acc HOMONYM HOMONYM STRONG WEAK As can be derived from the table above, there is no fully unambiguous number/ gender-paradigm of mixed inflection (i.e. a paradigm containing no homonymous forms). I believe, however, that the distribution of unequivocal forms suffices to elaborate on the status of the ambiguous ones and hence of the inflectional paradigms in general. Concentrating on the singular for now, it might be fruitful to start from the unambiguous forms and generalize over all identical case forms. Hence, the strong form in nominative masculine and neuter could be generalized to the ambiguous nominative feminine, rendering all nominative singular forms strong . Likewise, the only unambiguous accusative form, namely accusative neuter could be utilized to conclude that both remaining accusative forms are also instances of the strong inflectional pattern. The same abstraction from one unequivocal to two ambiguous forms would then be found in the case of genitive, where the feminine weak occurrence might suffice to conclude that the masculine and the neuter forms are part of this paradigm as well. This would leave us 138 2 PREMISES - Phasehood & φ-Dependencies with the partition in (85), which uniformly divides structural from lexical cases into strong and weak adjectival inflection across all genders. 49 (85) mixed Adjectival Inflection: 1 st Disambiguation masc fem neut nom STRONG STRONG STRONG gen WEAK WEAK WEAK dat WEAK WEAK WEAK acc STRONG STRONG STRONG However, once we take the insights of the preceding sections into account again, we should tend to focus on the columns of (84), i.e. the gender-values, rather than rows, i.e. case forms. By this switch in perspective, it comes to attention that the feminine pattern is the only singular one, not displaying unambiguous strong forms, with the structural case forms homonymous between the two patterns. However, both the masculine as well as the neuter paradigm display at least one unequivocal strong and weak form each (which additionally occupy identical slots in the case-paradigm, i.e. nominative and dative respectively). If we elaborate on the findings from the last subsections, the reordering of numberand gender-values might be a promising approach to the disambiguation of the mixed adjectival inflection: (86) a. mixed masc neut nom STRONG STRONG gen HOMONYM HOMONYM dat WEAK WEAK acc HOMONYM STRONG b. mixed fem pl nom HOMONYM WEAK gen WEAK WEAK dat WEAK HOMONYM acc HOMONYM WEAK 49 I will come back to the overlap of the partitions of strong and weak inflection with structural and lexical case in masculine and neuter gender in the course of my final argument on nominal concord and definiteness in chapter three below. 2.3 φ-Features 139 Starting from the two pairings in (86) a. and b., and applying the rationale of generalizing over case forms from there on leaves us with a different ordering: Observe first, that the masculine and neuter paradigms in (86) a. straightforwardly show two similar values for nominative ( strong ) and dative ( weak ). By the observation made above that structural cases pattern uniformly, I feel justified (for now) to disambiguate the accusative masculine form to strong , in what follows; the genitive, however, might only be unambiguous with recourse to the patterns in (86) b., which is too weak an argument and therefore should be postponed for now (cf. chapter 3.3.4 below). Turning to (86) b. itself, however, we do not find any case, in which both forms are ambiguous. Moreover, we do not find strong inflection in either structural or lexical cases. Taking the results arrived at so far in this chapter seriously, I hence feel safe to classify all homonymous forms in (86) b. as weak and therefore completely exclude feminine and plural from the pattern of mixed adjectival inflection. Recall that every other conclusion drawn on this parallel would also challenge the existence of a uniform weak inflectional pattern for feminine and plural HNs in general. Thence, the mixed pattern cuts through another grammatical dimension, i.e. gender, in addition to number and case, identified in chapter 1.3.2 above; all elements classified as varying between (i) weak in the plural and (ii) mixed in the singular (i.e. the ein-words Neg and Poss PRO ) therein must hence be recategorized accordingly as weak in the feminine as well as plural and mixed in the masculine and neuter paradigms. e. Target, Controller and Subgender I have, then, presented several morpho-syntactic idiosyncrasies in the German nominal domain to strengthen my claim of a reordering of the φ-feature values feminine and plural, above. The rationale was first introduced from Ritter (1993) in the discussion of gender of irregular plural forms in section 2.3.2. Recall that I stated therein that we never find plural semantics without feminine SLI agreement; hence, German displays only a single gender in the plural and therein corresponds to the well-known observation that the numbers with fewer gender distinctions are always non-singular (cf. Greenberg 1963: 95, universal 37; cf. also Bernstein 2008: 220 on the cross-linguistic tendency of gender neutralization in plural SLIs). This point needs further elaboration. Before doing so, however, it will be necessary to introduce a structuralized approach to genderagreement in what follows. Corbett (1991) employs a division between simple agreement paradigms w.r.t. other φ-features and their respective grouping into complex patterns, termed ‘target’ and ‘controller genders’ respectively. As has been elaborated above and also emphasized by Corbett (1991: 154), gender itself is inherent to the noun and 140 2 PREMISES - Phasehood & φ-Dependencies (predominantly) invariable. However, target genders “may be found in combination with other categories and so they may vary according to the other categories involved” (Corbett 1991: 154). The prime category in this respect is number. Controller gender classifications are hence inherent to the noun, while target genders are marked on agreeing elements (cf. Corbett 1991: 151f.). 50 A first merit of this approach lies in the accountability of - what he calls - ‘convergent’ and ‘crossed gender systems’ (with reference to Heine 1982) in opposition to simple ‘parallel systems’, i.e. the specific linkage of numberdependent paradigms (“sets of agreement-forms” Corbett 1991: 152, i.e. target genders) into various complex patterns (“sets into which nouns are divided” Corbett 1991: 152, i.e. controller genders). Each of these systems is exemplified in (87): (87) Phasehood & φ-Dependencies P REMISES 62 (87) a. parallel gender system of French (= Corbett 1991: (Figure 6.2)) singular plural masculine masculine masculine feminine feminine feminine b. convergent gender system of Tamil (= Corbett 1991: (Figure 6.8)) singular plural masculine I rational feminine II III neuter neuter c. crossed gender system of Rumanian (= Corbett 1991: (Figure 6.1)) singular plural ø I i III ă II e (88) (= Corbett 1991: (Figure 6.7), [modified]) singular plural masculine feminine feminine neuter (93) (= Corbett 1991: (Table 6.1), [reduced]) ovaj ‘this student student’ zakon ‘law’ singular nominative ovaj student ovaj zakon accusative ovog studenta ovaj zakon genitive ovog studenta ovog zakona dative ovom studentu ovom zakonu instrumental ovim studentom ovim zakonom plural nominative ovi studenti ovi zakoni accusative ove studente ove zakone genitive ovih studentātā ovih zakonā dative ovim studentima ovim zakonima instrumental ovim studentima ovim zaonima (94) (= Corbett 1991: (Figure 6.13), [reduced]) singular plural animate masculine masculine inanimate 50 Cf. Giurgea (2014) for a narrow-syntactic, featural implementation of this bipartition, additionally employing a different approach to the relocation of gender to NumP: While controller genders (his ‘nominal agreement classes’) are specified in the lexical entries of nominals, target genders are values of the number head which selects for varying nominal classes (hence establishing the crosscutting interdependencies exemplified in the main text, cf. Giurgea 2014: 52f.). Therefore, the Num-N complex ultimately determines the gender of the nominal domain. 2.3 φ-Features 141 Hence, parallel systems exhibit a one-to-one relation of target genders in one number-value to those in the other one, while convergent systems display a many-to-one relation in this respect (focusing on the common rational plural target gender in (87) b.). Crossed systems, finally, display a many-to-many relation on the axis, laid out above. Gender in either number, thus, determines gender in the other and vice versa in parallel systems (e.g. masc sg ↔masc pl in (87) a.), gender in one number determines gender in the other but not vice versa in convergent systems (e.g. fem sg →rational pl but rational pl ↛fem sg in (87) b.). Lastly, gender in neither number determines gender in the other in crossed systems (cf. the reference to irregular plurals in Romanian in the context of Ritter’s 1993 proposals in section 2.3.2 above). Turning to German, Corbett himself classifies the agreement paradigm as a prime instance of convergent gender systems; howsoever, he refrains from labeling the target gender in the plural in his graphical representation. I believe we gain a deeper insight by incorporating the finding of the current chapter into Corbett’s representation of the German gender system: (88) (= Corbett 1991: (Figure 6.7), [modified]) Phasehood & φ-Dependencies P REMISES 62 (87) a. parallel gender system of French (= Corbett 1991: (Figure 6.2)) singular plural masculine masculine masculine feminine feminine feminine b. convergent gender system of Tamil (= Corbett 1991: (Figure 6.8)) singular plural masculine I rational feminine II III neuter neuter c. crossed gender system of Rumanian (= Corbett 1991: (Figure 6.1)) singular plural ø I i III ă II e (88) (= Corbett 1991: (Figure 6.7), [modified]) singular plural masculine feminine feminine neuter (93) (= Corbett 1991: (Table 6.1), [reduced]) ovaj ‘this student student’ zakon ‘law’ singular nominative ovaj student ovaj zakon accusative ovog studenta ovaj zakon genitive ovog studenta ovog zakona dative ovom studentu ovom zakonu instrumental ovim studentom ovim zakonom plural nominative ovi studenti ovi zakoni accusative ove studente ove zakone genitive ovih studentātā ovih zakonā dative ovim studentima ovim zakonima instrumental ovim studentima ovim zaonima (94) (= Corbett 1991: (Figure 6.13), [reduced]) singular plural animate masculine masculine inanimate This point, however, needs to be elaborated further, but it was not until now that I believe to have presented sufficient unrelated evidence that the basic claim of this subchapter will remain unaffected thereby. To start out, let me present the full paradigm of inflection of SLIs under consideration in this analysis, namely D/ Dem, Poss PRO and Q S , the elements making up the upper phasal cycle in the German nominal domain. From this point onwards, I will maintain the ordering of inflectional paradigms from (86) above, namely masc/ neut and fem/ pl. 51 51 Note that the hyphens in (89) are implemented for expository purpose only, viz. levelling of the data, and therefore do not contain any prediction about the internal structure of the determiner. 142 2 PREMISES - Phasehood & φ-Dependencies (89) a. D masc neut nom d-er d-as gen d-es d-es dat d-em d-em acc d-en d-as b. D fem pl nom d-ie d-ie gen d-er d-er dat d-er d-en acc d-ie d-ie (90) a. Dem masc neut nom dies-er dies-es gen dies-es dies-es dat dies-em dies-es acc dies-er dies-es b. Dem fem pl nom dies-e dies-e gen dies-er dies-er dat dies-er dies-en acc dies-e dies-e 2.3 φ-Features 143 (91) a. Poss PRO masc neut nom mein mein gen mein-es mein-es dat mein-em mein-em acc mein-en mein b. Poss PRO fem pl nom mein-e mein-e gen mein-er mein-er dat mein-er mein-en acc mein-e mein-e (92) a. Q S masc neut nom all-er all-es gen all-em all-em dat all-em all-em acc all-en all-es b. Q S fem pl nom all-e all-e gen all-er all-er dat all-er all-en acc all-e all-e As can be derived from the tables in (89) b.-(92) b., the homonymy of feminine and plural forms is not perfect, though regular (with the diverging forms in the 144 2 PREMISES - Phasehood & φ-Dependencies fem/ pl paradigms marked in greyscale): Divergence is limited to dative suffixes (er in feminine, en in plural respectively), while in masculine and neuter, only lexically case-marked forms overlap. 52 Corbett (1994, see also Corbett 1991: ch. 6.2), referring to Zaliznjak (1964), takes nominals to be part of the same agreement class (e.g. same gender) iff “they take the same agreements under all conditions” (Corbett 1994: 1348). Therefore, strictly speaking, feminine and plural cannot be conflated into a single inflectional category (observe, however, that this conclusion does not make any predictions as to the structural proximity, proposed above). However, the variation between the patterns is both too minimal and regular as to ignore it and fully separate the inflectional categories, even more so in contrast to the remaining gender-values masculine and neuter. Turning to Corbett’s (1991) proposals once again, we do find a way out of this dilemma in the notion of ‘subgender’: Subgenders are agreement classes which control minimally different sets of agreement, that is, agreements differing for at most a small proportion of the morphosyntactic forms of any of the agreement targets. (Corbett 1991: 163) Corbett argues on the basis of SC demonstrative agreement that minimally diverging paradigms should be grouped together, therein relaxing Zaliznjak’s original claim to include those paradigms which only differ minimally from one another w.r.t. the “general level of difference between other agreement classes” (Corbett 1991: 164). Focusing on the respective data from the SC paradigm, namely the masculine animate and inanimate categories, Corbett is able to further elaborate on this concept. The appropriate data are given below: 52 This is also true of the parallel strong inflectional paradigm on adjectives. In the weak paradigm with the adjective the highest element of the structure, however, fem/ pl is homonymous only in the non-structural cases genitive and dative, while masc/ neut additionally coincide in the morphological form of the nominative. 2.3 φ-Features 145 (93) (= Corbett 1991: (Table 6.1), [reduced]) Phasehood & φ-Dependencies P REMISES 62 (87) a. parallel gender system of French (= Corbett 1991: (Figure 6.2)) singular plural masculine masculine masculine feminine feminine feminine b. convergent gender system of Tamil (= Corbett 1991: (Figure 6.8)) singular plural masculine I rational feminine II III neuter neuter c. crossed gender system of Rumanian (= Corbett 1991: (Figure 6.1)) singular plural ø I i III ă II e (88) (= Corbett 1991: (Figure 6.7), [modified]) singular plural masculine feminine feminine neuter (93) (= Corbett 1991: (Table 6.1), [reduced]) ovaj ‘this student student’ zakon ‘law’ singular nominative ovaj student ovaj zakon accusative ovog studenta ovaj zakon genitive ovog studenta ovog zakona dative ovom studentu ovom zakonu instrumental ovim studentom ovim zakonom plural nominative ovi studenti ovi zakoni accusative ove studente ove zakone genitive ovih studentātā ovih zakonā dative ovim studentima ovim zakonima instrumental ovim studentima ovim zaonima (94) (= Corbett 1991: (Figure 6.13), [reduced]) singular plural animate masculine masculine inanimate The only diverging forms are found in the accusative singular, printed in italics above. As Corbett (1991: 164) notes, some rules can refer to the feature [masculine] without any reference to [animate]. His formal representation of the animate and inanimate masculine gender system is given below: (94) (= Corbett 1991: (Figure 6.13), [reduced]) Phasehood & φ-Dependencies P REMISES 62 (87) a. parallel gender system of French (= Corbett 1991: (Figure 6.2)) singular plural masculine masculine masculine feminine feminine feminine b. convergent gender system of Tamil (= Corbett 1991: (Figure 6.8)) singular plural masculine I rational feminine II III neuter neuter c. crossed gender system of Rumanian (= Corbett 1991: (Figure 6.1)) singular plural ø I i III ă II e (88) (= Corbett 1991: (Figure 6.7), [modified]) singular plural masculine feminine feminine neuter (93) (= Corbett 1991: (Table 6.1), [reduced]) ovaj ‘this student student’ zakon ‘law’ singular nominative ovaj student ovaj zakon accusative ovog studenta ovaj zakon genitive ovog studenta ovog zakona dative ovom studentu ovom zakonu instrumental ovim studentom ovim zakonom plural nominative ovi studenti ovi zakoni accusative ove studente ove zakone genitive ovih studentātā ovih zakonā dative ovim studentima ovim zakonima instrumental ovim studentima ovim zaonima (94) (= Corbett 1991: (Figure 6.13), [reduced]) singular plural animate masculine masculine inanimate Elaborating further, he notes that none of the diverging morphological forms is exclusive to the diverging slot in the case-paradigm (indicated by the brackets in (93) above), with the accusative in the animate paradigm parallel to the morphological realization of the genitive in both paradigms, and the respective inanimate form identical to the nominative. Thereupon, Corbett establishes the notion of dependent target gender: A dependent target gender is a target gender consisting of a set of morphological realizations which mark agreement with members of a given agreement class by an opposition involving only syncretism (and no independent form). (Corbett 1991: 164) 146 2 PREMISES - Phasehood & φ-Dependencies Turning to the German paradigm with these notions at hand, feminine and plural obviously differ in the most minimal opposition (w.r.t. the level of difference between paradigms found elsewhere, but cf. fn. 52). Moreover, the morphological form of the dative feminine coincides with the form of the genitive in both number/ gender-values. Yet, the contrasting form in the plural is not found elsewhere in the pattern of SLI agreement. However, if we additionally take the other lexical category agreeing with HNs in the complex number/ gender-value into account, namely adjectives, the solely independent form can also be reinterpreted as a syncretism: (95) Phas ehood & φ- (95) nom geni dativ accu (97 -Dependenc ) fe SL minative / e itive / e ve / e usative / e ) Auf M Michael hat d cies eminine LI / STRONG e/ er/ er/ e/ Michaels Dachb diese Krimis ge plural SLI / STRONG / e/ / er/ / en/ / e/ boden findet An elesen. fem WE / e/ / en / en / e/ ndrea eine Kiste Krimis hat M minine EAK plur WEA / en/ n/ / en/ n/ / en/ / en/ e mit Krimis. Michael diese ge ral A K elesen. P REMISES 633 Thence, the paradigms under consideration might be understood to pose instances of dependent target genders under the incorporation of the strong / weak inflectional bipartition. In any case, given Corbett’s restrictions on minimal divergence of morphological form, the feminine and plural target paradigms doubtlessly constitute subgenders of one common target paradigm. However, in contrast to the two SC subgenders presented above, these are divided on the basis of the same φ-feature that is standardly taken to categorize target gender-paradigms from one another, namely number. Therefore, the system of German gender-agreement cannot be formalized parallel to (94) above, since the category on which the formalization rests is at the same time the feature that separates the subgenders. We are, then, left with the observation that German indeed exhibits three target genders (masc, fem, neut), one of which splits into two subgenders (i.e. fem sg and fem pl respectively). Below, I will continue to employ the standard labels ‘feminine’ and ‘plural’ to counter any confusion in the ensuing analysis. A last point to be mentioned in this context, then, is the homonymy of weak adjectival inflectional forms in the masculine and neuter paradigms, as reported in fn. 52, which is likewise robust across the structural/ non-structural case partition and equally covers three out of four forms. I will refrain from transferring the notion of subgender (and its implications for the ordering of features in the case at hand) to these remaining φ-values on the nominal head since they overlap in a broader and, above that, case-sensitive way in SLI agreement under con- 2.3 φ-Features 147 sideration here, i.e. exclusively in the lexical cases. This observation will become crucial in my analysis in the following chapter, where I will ultimately propose a deduction of weak adjectival inflectional forms from their strong counterparts. 2.3.4 Concluding Remarks In this subchapter I have presented extensive evidence for the reordering of φ-feature values of the categories gender and number in the nominal domain in German. I started out by reviewing inflectional idiosyncrasies of weakly quantified noun phrases with plural and mass HNs, which uncovered parallel restrictions between the former and the subclass of feminine mass nouns. In the subsequent discussion of Ritter’s (1993) proposal for the placement of the category gender on the number head Num 0 , I showed that findings presented therein parallel phenomena found in German for the subclass of feminine nominals. In the analysis of gender-switching on nominals with animate referents, I then carved out the dependencies between the two morpho-syntactic values on the aforementioned head in terms of Noyer’s ([1992] 1997) model of Distributed Morphology and categorized various nominal occurrences as configurations of N and Num heads, accordingly. Lastly, I presented further, arguably unrelated, morpho-syntactic phenomena in the nominal domain in German as well as their unification under the proposals made above: [f]/ [pl] are values of Num 0 in German, while [m]/ [n] and [sg] are coded on the head noun N 0 . The conflation of (parts of) number with (parts of) gender is not without problems, however, as has been mentioned e.g. by Corbett (1991), since we seem to be dealing with two different categories of feature-coding here (gender a fixed, lexical feature, number a variable, formal one). However, as I believe to have shown, gender, at least in the context of the value feminine and the strategy of feminization, is not a fixed property but on a par with number as instances of variable inflectional morphology. Further backup for the special status of feminine gender might then be found in this exact property with Di Domenico’s (1997) analysis of variable gender on Num and invariable gender on N (interpretable and uninterpretable respectively) in Italian, as reported by Alexiadou et al. (2007: ch. II.3.2). 53 Incidentally the partition therein coincides with the one proposed above for German. 53 The difference in lexical specification of number and gender additionally constitutes Giurgea’s (2014: 52) primary argument against the conflation of these categories. Apart from the arguments in the main text, observe that my analysis circumvents this critique in that the reanalysis targets the value feminine, rather than the category gender. However, I will demonstrate below that the findings of this chapter do indeed favor a complex feature (notion), comprising number and gender in chapter three. 148 2 PREMISES - Phasehood & φ-Dependencies Furthermore, the phasal status of this functional head was revised in the reordering of φ-features, based on diverging properties of nominal domains incorporating the two varying values w.r.t. split-topicalization from chapter 2.2 above (with plural nominals able to raise into the verbal domain but feminine nominals unable to undergo said movement): It is only the fully specified (i.e. plural) instance of Num 0 that exhibits phasal status. This proposal is in itself not problematic once we take into account standard assumptions about the phrasal set-up of the verbal domain. The lowest phase head therein is v P, however, this is likewise not a rigid phase head for Chomsky (DbP: 12); said property is rather ‘propositional’ and relies on the ‘completeness’ qua realization of an external argument. In passive and unaccusative structures, v therefore fails to act as a phase head and is hence understood to be defective (i.e. v def as opposed to v *) in this respect. Though widely criticized in its motivation, the completeness-argument readily applies to the featural specification of Num 0 , argued for above, even more so in the light of cartographic approaches (starting out with Grimshaw [1991] 2005), seeking parallels in the hierarchies of projections erected above the core lexical categories, 54 as well as Distributed Morphology (Halle & Marantz 1993), proposing category-imposing projections (f-morphemes) to combine with category neutral roots (l-morphemes); for an analysis incorporating iterative applications of this rationale cf. Hiraiwa (2005). Interestingly for the present purpose, Alexiadou (2001: ch. 2.3.4) hints at the conclusion that NumP can function in this respect and hence parallels v 0 , while Khim (2005) argues for a parallel point from the opposite direction in locating Class (viz. gender and noun class) on n . Howsoever, as noted above, the weak/ strong phase-approach has not gone unchallenged and for it to be fully implemented into the present analysis, the upper phasal cycle should bear a parallel resemblance to CP, whose weak/ strong status depends on the presence of “force indicators” (DbP: 12, i.e. sentential status), a point which we will come across again in the ensuing chapter. 2.4 Conclusion In this chapter I have reviewed the phasal status and featural set-up of those heads in the nominal hierarchy of projections carved out in chapter one. Concerning the former, I have argued on the basis of data from ellipsis as well as extraction out of the nominal domain in German that phasehood is a variable 54 Cf. e.g. Aboh (2004) who parallels the categories Num and D to Fin and Force (following Rizzi 1997) in the clausal domain respectively. 2.4 Conclusion 149 property, contextually evaluated w.r.t. the co-occurrence of SLIs in a given structure. Evaluation therein applies in accordance with a fixed hierarchy that turned out to be stable across derivations featuring multiple SLIs in that phasal status is unambiguously resolved upon EM of a second candidate, in line with the weak version of the Phase Impenetrability Condition, PIC (DbP: (11)). Furthermore, two dedicated cycles of phasehood assignment were identified in the analysis of complex extraction patterns from weakly quantified nominal domains. The system therein conforms to Richards’ (2007) considerations on good phasal design, i.e. the most minimal configuration of phase-plus-non-phase-complement (P- N-P-N). Moreover, both cycles diverge from the rationale of the highest-phraseas-phase from Bošković (2014: (56) b.) in identical manner: It is the quantifier phrase (strong and weak respectively), heading each cycle, that simultaneously poses the least popular candidate. From thereon, (strong) QST and split-topicalization were understood as structurally equal phenomena, applying at different phasal boundaries. Additionally, the data presented above forced the abandonment of the rationale of solely lexically filled, i.e. ‘visible’ phrases from Bošković (2014), introduced in chapter one, with the establishment of a dedicated number phrase, NumP. However, parallel to the remaining phrases in the nominal spine discussed in chapter one, much unrelated evidence as well as previous analyses were put forward for the justification and implementation of this projection into the proposed nominal hierarchy of projections. The remainder of this chapter, then, elaborated on the featural content of said head, Num 0 . Therein I advocated a split in the category of gender and relocated the feminine value as part of the featural set-up of Num 0 , dominating [pl], mainly on the basis of morphological identity of the respective paradigms. Therefore, [f] was reanalyzed as a ‘numberless plural marker’ in German. A number of seemingly unrelated inflectional and linearizational phenomena in German seem to be unified under these proposals, once [f]/ [pl] are grouped together and contrasted with [masc]/ [neut], located on the core nominal projection N 0 . One additional topic, then, demands to be addressed in the context of this chapter. Two reviewers have expressed doubts concerning a few of the grammaticality judgements put forth predominantly in the current chapter, based - as both of them note - on their own intuitions. I understand it is my obligation to address this criticism in some detail here. Rather interestingly, however, there seems to be no consensus between the reviewers themselves concerning the exhaustive set of critical data. I can therefore only concentrate on the two occurrences that have unanimously been judged contrary to the categorizations presented by me, i.e. the ungrammatical (13) a'. and (15) a'., repeated here as (96) a. and b. respectively for ease of exposition. 150 2 PREMISES - Phasehood & φ-Dependencies (96) a. * Bücher i habe ich [diese t i ] gelesen. (= 13 a'.) books i have I [ these t i ] read b. * [Bücher] i habe ich [Martin seine t i ] gelesen. (= 15 a'.) [ books ] i have I [ Martin his t i ] read In what follows, I want to put forth additional judgements by my original core group of informants to counter this critique. They serve to reinforce my claim that the data compiled in (96) and their in-situ counterparts are not fully synonymous and that it is at least the most unmarked contexts that splits them apart concerning their status of grammaticality. Here, I rely heavily on Schütze’s (2016) work on methods of data acquisition in linguistics. Thus, in order to reproduce (and strengthen) my original findings, I switched both the method of elicitation (from the original form of interviews to a questionnaire study) as well as the set of stimuli (to a larger paradigm 55 ), both designed according to the guidelines laid out by the aforementioned author. 56 Relevant instructions have been provided 57 while rating was conducted on a three-point scale analogous to that employed by Quirk & Svartvik (1966), as laid out in Schütze (2016: ch. 3.2), with the category denominations of the former translated into German. 58 Each datum was affixed with said scale (including its category denominations) and the contrasted in-situ and floating configurations displayed in juxtaposition (again in randomized order). A sample item is given in (97) below. 55 All data had consistently been designed according to the structural template provided by those in (96) above as well as their in-situ counterparts, i.e. transitive main clauses with and without movement of a bare noun from internal argument position with the respective remnant of the nominal domain stranded VP-internally. 56 Cf. Schütze (2016: ch. 6.3.1) on randomized order, varied (but thoroughly common) lexical content of test data with identical structure, sole usage of common words, supply of supporting contexts, facilitation for side-by-side comparison of contrasted constructions, as well as cf. Schütze (2016: ch. 6.3.2) on control for the so-called ‘maturation effect’ by restricting the number of critical items and the incorporation of consistently grammatical fillers serving as ‘anchors’ (in Schütze’s terms) for the subject’s capability to judge grammaticality correctly over the course of a full study. 57 Subject were asked to rate the following sentences by intuition in the context provided, regarding their status as normal sentences of German that might be expectable in spoken language. They were furthermore told that no correct answers exist and asked to reach their rating decisions quickly. Finally, it was specified that only closed off, complete sentences are employed to exclude an anaphoric interpretation of the demonstrative that turns grammatical when co-referring with a succeeding list of the relevant items. 58 To wit: ‘wholly natural and normal’ into völlig natürlich und normal , ‘wholly unnatural and abnormal’ into völlig unnatürlich und abnormal as well as ‘marginal or dubious’ (sandwiched in between the former) into grenzwertig oder zweifelhaft . 2.4 Conclusion 151 (97) Auf Michaels Dachboden findet Andrea eine Kiste mit Krimis. Michael hat diese Krimis gelesen. Krimis hat Michael diese gelesen. The contexts preceding the stimuli divided along the criteria (i) ± contrasted with additional subsets of a common superset and (ii) proper name vs. pronominal form of the subject argument crossed by movement in the case of (96) a., i.e. floating of a nominal from the demonstrative, with five items per configuration and six filler items. In the case of (96) b., i.e. floating from PDCs, contexts split along the lines of (i) ± contrasted possessor and (ii) ± contrasted possessee, again with five items per configuration and eight filler items. In both cases, the nominal under consideration was consistently mentioned in the preceding contexts. Turning to the results of the studies, beginning with floating from the category D, judgements turned out consistently clear cut across all context types in the unanimous grammatical judgement of the in-situ variant in contrast to the thorough rejection of its floating counterpart. One subject only receded from this division with stimuli incorporating pronominal subjects in rating two of the five floating alternatives as marginal or dubious, rather than outright ungrammatical, but still retained a contrast with the in-situ variant, consistently marked as wholly natural and normal. Results gathered on the possessor doubling construction turned out more diverse, with far greater rejection of the construction in general, as has been hinted at several times throughout chapter one. Thus, in-situ PDC-configurations have been increasingly judged as marginal or dubious by my informants. Howsoever, the outlined contrast was still maintained in the consistent classification of the floated alternatives as wholly unnatural and abnormal. While this distinction remained clear cut across most cases, some informants ruled out both DPC-configurations in single instances. This was again observed across all contexts. For one subject, in-situ as well as floating DPC-configurations simultaneously increased in acceptability with contexts contrasting the possessee. Nevertheless, the contrast between in-situ and floating construction remained stable both across contexts as well as subject. These outcomes are further, though only informally, backed up by the results of a small scale corpus study, as also suggested by the reviewers. Modelled after (96) a., a rudimentary search in the annotated Tagged-C corpus of Cosmas 2 for 152 2 PREMISES - Phasehood & φ-Dependencies sentence final nominative strings of type [noun plural -verb-pronoundiese ] with and without an additional verb succeeding it returned 20 and 5 hits respectively. In all of these, however, the demonstrative had been employed in anaphoric reference to a noun phrase located outside the sentence frame of the demonstrative (either a new clause or a coor superordinate one). Therefore, the data should be ignored since they do not correspond to the configuration under analysis. Moreover, search strings modelled after (96) b., i.e. sentence final [noun plural verb-pronoun-proper nounseine / ihre (-verb)] consequently returned no results at all in either of the four configurations. In contrast, QST-constructions based on analogous configurations with and without sentence final verb returned relevant occurrences both with the strong (16 vs. 10 hits) as well as the weak quantifier (4 vs. 153 hits). These have likewise been judged as grammatical by my informants (cf. (14) c'., (18) a.). Since Tagged-C exclusively consists of formal written sources (viz. predominantly newspaper articles) rather than spoken language data, a low count of colloquial configurations is expected and the reported figures can only pose as a supplemental argument. Nevertheless, even though the figures are vanishingly low, the contrast is clear cut and conforms to the judgements reported by my informants. I thus feel justified to retain my original classification of the data. Returning to the general objects of investigation, namely nominal concord and morphological definiteness in German, the system established in chapter one and elaborated above will be put to work in an approach to the inflectional idiosyncrasies in the nominal domain in the following chapter. Therein, I will draw attention to the featural content of inflectional morphology in a narrow-syntactic implementation of the feature-geometric modifications, elaborated above, as well as to the underlying operations deriving agreement. Thereby, morphological definiteness will ultimately reduce to the distribution of φ-features on the semi-lexical heads identified above. As Harley & Ritter (2002a: 518f.) note, one of the major research questions arising from their approach concerns the characterization of the relation between the geometry and Narrow Syntax, especially w.r.t. agreement. I trust it that the current investigation contributes to converge towards an answer to this question. 3.1 Introduction 153 3 INFERENCE - Nominal Concord & Definiteness 3.1 Introduction Chapter 2.2 took a close look at the (variable) phasal status of semi-lexical items in the nominal domain in German, following Bošković’s (2014) claim of the structural contextuality of phasehood. I concluded that the nominal hierarchy of projections constitutes at most two independent phasal cycles on the basis of data from extraction and elision. Therein, I identified another functional projection dominating the core NP, NumP (following Ritter 1993), which acts as a phase in linearizations with plural head nouns. The phasal set-up elaborated therein was concluded to constitute an instance of the most minimal phase-plus-non-phase configuration, as proposed by Richards (2007). In chapter 2.3, I took a closer look at the featural set-up of said functional head Num 0 . It was concluded on the basis of morpho-syntactic and phonological phenomena that Num is not only the host of the syntactic plural marker but that said plural marker is itself a dependent of another feature located on Num, namely feminine gender, which reduces to a ‘numberless plural marker’ in German. The final findings of these subchapters are reprinted here as (1) a. and b. respectively: (1) 120 3 I NFERENCE Nominal Concord & Definiteness 3.1 Introduction Chapter 2.2 took a close look at the (variable) phasal status of semi-lexical items in the nominal domain in German, following Bošković’s (2014) claim of the structural contextuality of phasehood. I concluded that the nominal hierarchy of projections constitutes at most two independent phasal cycles on the basis of data from extraction and elision. Therein, I identified another functional projection dominating the core NP, NumP (following Ritter 1993), which acts as a phase in linearizations with plural head nouns. The phasal set-up elaborated therein was concluded to constitute an instance of the most minimal phase-plus-non-phase configuration, as proposed by Richards (2007). In chapter 2.3, I took a closer look at the featural set-up of said functional head Num 0 . It was concluded on the basis of morpho-syntactic and phonological phenomena that Num is not only the host of the syntactic plural marker but that said plural marker is itself a dependent of another feature located on Num, namely feminine gender, which reduces to a ‘numberless plural marker’ in German. The final findings of these subchapters are reprinted here as 0 a. and b. respectively: (1) a. (= chapter 2, (30), [modified]) Q S( * ) > D* > Poss PRO( * ) > Q W( * ) > Num ( * ) > N [3 1 2] [2 1] 1 2 b. (= chapter 2, (64)) Num ! [f] ! [pl] Equipped with these notions, we are now ready to turn to nominal concord and morphological definiteness in German. In what follows, I want to argue that the combination of these insights suffices to capture the diverging paradigms of nominal inflections and their relation to the status of definiteness. No recourse to a dedicated feature [± def] (proposed for German as early as Haider 1992) or phrase for definiteness will be necessary. In fact, I will argue for 0: (2) Equipped with these notions, we are now ready to turn to nominal concord and morphological definiteness in German. In what follows, I want to argue that the combination of these insights suffices to capture the diverging paradigms of 154 3 INFERENCE - Nominal Concord & Definiteness nominal inflections and their relation to the status of definiteness. No recourse to a dedicated feature [± def] (proposed for German as early as Haider 1992) or phrase for definiteness will be necessary. In fact, I will argue for (2): (2) Morphological Definiteness in German The highest phase head must bear a gender-value. 1 As will be demonstrated below, the definiteness effect can be fully traced back to (2) in the terms of the preceding chapters. Every idiosyncrasy of definiteness in German follows from the selectional restrictions of the respective SLIs and their relations in a cyclic derivational process. Some comments on the subject of investigation and the modus operandi of the following chapter are in order here. Recall from chapter 1.3 that I went through some length to equate the source of varying adjectival inflection with the status of definiteness in the nominal domain. Furthermore, I have shown that dependencies inside the nominal hierarchy of projections, viewed in isolation, constitute a most rigid testing ground for the phenomena under consideration, since no unrelated factors introduced by sentential derivations may disrupt the picture. I will therefore start the investigation in the nominal domain and focus on the phenomenon of nominal concord before turning to the relation of (in-)definite nominal hierarchies with the extended projection of the verb. In doing so, I am positive to deduce the full paradigm of SLI inflectional idiosyncrasies in German. To achieve this, however, it is necessary to first gain insight into the dependencies and interactions of features in the nominal domain, to which I will turn now. 1 Cross-linguistically, definiteness has been associated with another φ-feature, namely person, several times in the literature. The most direct account being Lyons’ (1999: ch. 8.5.2) complete conflation of the two, which he both associates with the category D; later approaches, in contrast, rather focus on an indirect connection either via a linkage of the feature to the aforementioned structural position and hence to other features taken to be hosted there (cf. Longobardi 2008: 200f.) or by addition to the featural set-up of lexical material base-generated in said position; cf. Bernstein 2008: ch. 3 for a decompositional approach to English th- along these lines, to wit, her account ties definiteness (itself not taken to be inherent to the SLI) only informally to the person-marker th-, and hence to said φ-feature in their “derivational association” (Bernstein 2008: 221). 3.2 Preliminary Remarks 155 3.2 Preliminary Remarks The current subsection sets out to review the basic notions for the analysis to follow, namely the exact coding of φ-features (ch. 3.1.1) and the interaction of these due to the operation Agree (ch. 3.1.2). As it will turn out for both cases, standard assumptions (Chomsky: MPLT, MP et seq.) naturally have not gone unchallenged but alternative accounts have moreover gained wide acceptance in the last decade. This section will discuss alternatives for each of the notions under consideration and finally end on the implementation of independently justified concepts into the current analysis. 3.2.1 φ-Feature Dependencies Formal Features are understood as the ‘atoms’ of syntactic computation, the basic building blocks that drive the derivation in Narrow Syntax. Lexical Items (LIs) are viewed as composed of phonological, semantic and formal bundles of such atoms, the latter of which enter the syntactic derivation, while the former ones are inserted into the structure post-syntactically (referred to as ‘Late Insertion’ in Distributed Morphology, cf. Halle & Marantz 1993) at Morphology (cf. Chomsky MP: 229). Formal features split along two relevant dimensions: A feature is either valued or unvalued, meaning that it either possesses or needs/ seeks a specification for the (syntactic) category it represents, as well as either interpretable or uninterpretable with respect to the semantic outcome. Chomsky (DbP: 5) couples these seemingly unrelated properties by - what Pesetsky & Torrego (2007: 266, henceforth P&T) call - the ‘Valuation/ Interpretability Biconditional’: (3) (= P&T: (3)) Valuation/ Interpretability Biconditional A feature F is uninterpretable iff F is unvalued Uninterpretable features are ill-formed at the semantic interface SEM (due to Full Interpretation) and must hence be eliminated before the (subpart of the) derivation reaches the interfaces. This is done by checking via the operation Agree qua Value, the precondition for which is the operation Match. The latter searches the c-command domain of the unvalued (and uninterpretable) feature, i.e. the LI bearing said feature, for the structurally closest compatible (‘identical’ in earlier (Chomsky: MI), ‘non-distinct’ in Chomsky’s (DbP, OP) later terms) and active counterpart (cf. Chomsky MI: 122), i.e. an LI which has not yet received case (the ‘Activation Condition’, ‘Active Goal Hypothesis’ in Rezac 2003, which trivially also applies to the LI in need of a value), bearing a value and hence 156 3 INFERENCE - Nominal Concord & Definiteness being either interpretable or valued uninterpretable, which reduces to a trivial distinction by (3). An LI bearing the valued instance of the feature [F] moreover has to be φ-complete to enter into an Agree relation (DbP: (3)). Call the LI/ feature initiating Match the ‘probe’ and the LI bearing a matching instance the ‘goal’; call the process ‘probing’. Upon successful identification of a valued instance of [F], Value copies the diverging structure between [F] on goal and [F] on probe onto the latter and marks the newly valued instance for deletion at the following phase-level. If the probe additionally bears an EPP/ OCC-feature, the goal is subsequently attracted and internally merged (IM) at the edge of the phase, i.e. the specifier of the probe. In OP, Chomsky considers all steps following Match to apply at TRANSFER (Spell-Out), the Value-TRANSFER simultaneity from chapter 2.2, whereby lexical material is ‘shipped’ to the interfaces, since uninterpretable features might have a phonological reflex at PHON/ PF even though they are poisonous at SEM/ LF. The steps preceding Value, in contrast, are taken to apply with EM of each phase head, which are taken to be the sole bearer of (uninterpretable) features, to be ‘handed down’ to non-phasal complements via feature inheritance, as described in chapter 2.2, which formed the basis for Richards’ claim of the phaseplus-non-phase configuration as the most minimal phasal set-up. I will postpone commenting on the operation Agree for now since it will be the topic of the following section. Let us instead concentrate on the elements driving the operation in what follows: Features are taken to be primitive. They consist of two parts, a category and a value, e.g. a number-feature with a plural value, [ u num: pl]. However, φ-features are standardly taken to constitute a set or bundle. Above we saw the requirement of φ-completeness for goals to enter into an Agree relation with a probe, moreover, Chomsky’s (MI, DbP) well-known treatment of (i) raising/ ECMand (ii) ES-constructions rests on this very rationale. In the latter (ii), an expletive (Expl) enters the derivation with a defective set of φ-features (expletives are taken to only possess an unvalued person-feature), (φ-complete) T’s EPP is satisfied by IM of Expl into T’s specifier, Match of Expl and T succeeds, but Value only applies to Expl’s person-feature after establishment of (long-distance) Agree relation of T with the DO. Intervention of Expl is therein barred by (4): (4) (= Chomsky DbP: (14)) Maximization Principle Maximize matching effects. 3.2 Preliminary Remarks 157 The Maximization Principle hence forces φ-complete goals; φ-Agree on T is taken to apply to all uninterpretable features of T simultaneously with all φ-features taken “as a unit” (Chomsky MI: 124). In the former (i), φ-incomplete and hence defective (raising) T, whose only φ-feature is person (MI: 124), 2 is completely barred from initiating case/ agreement relations, which only leaves its EPP-feature to be checked by EM of Expl or IM of Obj (although Chomsky speculates that movement might proceed in “one fell swoop” (DbP: 9) to the non-defective matrix T comp , which would even render the EPP on T def moot). We have thus seen that the notion of φ-completeness on probe as well as on goal is made responsible for the successful establishment of Agree relations. In the light of these analyses and the treatment of φ-features as a set/ bundle therein, the primitive nature of number, gender and person becomes even more dubious. Recall from chapter 2.3.3 that the split of nominal features on Num 0 and N 0 was conducted on the basis of Harley’s (1994) and Harley & Ritter’s (2002a, henceforth H&R) 3 feature geometries, expressing universal relations and dependencies between φ-feature categories and values in pronominal elements, called re , ‘referring expression’, in H&R. Their complete geometry is given in (5) below: (5) (= H&R: (6)) (5) (= H&R: (6)) RE qp PARTICIPANT INDIVIDUATION 3 qgp Speaker Addressee Group Minimal CLASS g gp Augmented Animate Inanimate/ Neuter 3 Feminine Masc… Some comments are in order here. H&R motivate the hierarchy in parallel to phonological feature geometries, the goals of which are adopted from Noyer ([1992] 1997): (6) (= H&R: (4) a, c, [from Noyer [1992] 1997]) a. Subtrees define natural classes of features for phonological rules. b. The dependency relation encodes contrastiveness. Small caps in 0 represent organizing nodes while their dependents are monovalent features, which only ever enter into the structure when active. The underlined features represent defaults which may not be represented but rather inserted by default rules; hence, reduced organizing nodes may themselves surface as unmarked values as in H&R’s treatment of Pirah- pronouns, inflecting only for person. I will get back to this below. (7) Some comments are in order here. H&R motivate the hierarchy in parallel to phonological feature geometries, the goals of which are adopted from Noyer ([1992] 1997): 2 The person-feature on T def is necessary for Chomsky to motivate Match for any further IM relation of T with a nominal category to apply in a framework dispensing with categorical features. Likewise, the motivation to locate person on Expl can be found in the establishment of its relation to T, so that IM by EPP can proceed (cf. Chomsky DbP: 7) as well as to obey the Activation Condition. 3 Cf. also Harley & Ritter 2002b, Hanson et al. 2000; cf. also chapter 1.4.4 above. 158 3 INFERENCE - Nominal Concord & Definiteness (6) (= H&R: (4) a, c, [from Noyer [1992] 1997]) a. Subtrees define natural classes of features for phonological rules. b. The dependency relation encodes contrastiveness. Small caps in (5) represent organizing nodes while their dependents are monovalent features, which only ever enter into the structure when active. The underlined features represent defaults which may not be represented but rather inserted by default rules; hence, reduced organizing nodes may themselves surface as unmarked values as in H&R’s treatment of Pirah- pronouns, inflecting only for person. I will get back to this below. (7) Some comments are in order here. H&R motivate the hierarchy in parallel to phonological feature geometries, the goals of which are adopted from Noyer ([1992] 1997): (6) (= H&R: (4) a, c, [from Noyer [1992] 1997]) a. Subtrees define natural classes of features for phonological rules. b. The dependency relation encodes contrastiveness. Small caps in 0 represent organizing nodes while their dependents are monovalent features, which only ever enter into the structure when active. The underlined features represent defaults which may not be represented but rather inserted by default rules; hence, reduced organizing nodes may themselves surface as unmarked values as in H&R’s treatment of Pirah- pronouns, inflecting only for person. I will get back to this below. (7) a. (= H&R: (Table 13), [from Thomason & Everett 2001]) 1 st ti 2 nd gi/ gia 3 rd hi b. (= H&R: (25)) 1 st person 2 nd person 3 rd person RE RE RE g g PART PART g Addressee ti gi/ gia hi H&R understand 0 as constituting a part of UG. Their discussion of pronominal systems therefore includes data from systematic co-occurrences in a variety of languages as well as acquisitional considerations and gaps in their paradigm of 110 languages. On the basis of H&R’s proposals, I advocated a split in syntactic locus of H&R understand (5) as constituting a part of UG. Their discussion of pronominal systems therefore includes data from systematic co-occurrences in a variety of languages as well as acquisitional considerations and gaps in their paradigm of 110 languages. On the basis of H&R’s proposals, I advocated a split in syn- 3.2 Preliminary Remarks 159 tactic locus of φ-features between N and Num in chapter 2.3.3. The relevant structures are reprinted here: (8) Nominal Concord & Definiteness I NFERENCE 125 (8) a. (= chapter 2, (63)) N 3 singular gender (minimal) (class) 3 [m] [n] b. (= chapter 2, (64)) Num ! [f] ! [pl] Recall that 0 was taken as a statement on the status of the feature/ value [f] (the only ‘relocated’ element w.r.t. H&R’s original ordering) in German rather than a statement about the ordering of features/ values in UG: [f] is a numberless plural marker in German. However, this proposal interweaves the two formerly discrete φ-feature categories into one complex category for which the primitive feature-value bipartitions [unum: val] and [ugen: val] do not seem fit. Yet, 0 readily suggests an alternative account: If φ-features are taken to be structured in terms of hierarchical relations (dependency, sisterhood) on heads bearing them, the most natural consequence would be to apply these relations to their uninterpretable counterparts on heads probing for them. Observe that by doing so, the φ-features/ values on Num 0 and N 0 regain narrow-syntactic relevance that was only attributed to the heads bearing them in chapter 2.3 (cf. P&T: fn. 21). 4 The implementation of ordered features and the utilization of their dependencies in narrow-syntactic computation is not a novel claim. Starke (2001) implements these in a ‘feature-tree’, introducing hierarchical relations of φ-, case, and several Ā-features to unify various island effects. The feature geometry of H&R on which my proposal of the split of φ-features was based in chapter 2.3, furthermore, has also undergone this treatment recently by Preminger (2011) and most prominently by Bejar (2000, 2003, Bejar & Rezac 2009), both of whom underline the narrow-syntactic applicability of H&R’s original proposal (cf. Preminger 2011: 35; Bejar 2003: 30). Bejar (2003) argues on the basis of Georgian split personand number-agreement that (relative and complete) underspecification of (nominal) goals for φ-features influences the verb’s strategies in probing for said features in a second cycle, applying to an extended search space, viz. the specifier of the probing head (following Rezac 2003). I will leave the exact 4 Note that by granting these features narrow-syntactic relevance, the account elaborated here finally becomes subject to the criticism of Déchaine & Wiltschko (2002). As has been laid out in chapter 1.4.4 above, approaches tracing diverging distributions back to varying featural structure have to account for the visibility of the latter in the syntax. However, I believe to satisfy this requirement in the adequate modification of Match and Value in the following subchapter. Recall that (8) was taken as a statement on the status of the feature/ value [f] (the only ‘relocated’ element w.r.t. H&R’s original ordering) in German rather than a statement about the ordering of features/ values in UG: [f] is a numberless plural marker in German. However, this proposal interweaves the two formerly discrete φ-feature categories into one complex category for which the primitive feature-value bipartitions [ u num: val] and [ u gen: val] do not seem fit. Yet, (8) readily suggests an alternative account: If φ-features are taken to be structured in terms of hierarchical relations (dependency, sisterhood) on heads bearing them, the most natural consequence would be to apply these relations to their uninterpretable counterparts on heads probing for them. Observe that by doing so, the φ-features/ values on Num 0 and N 0 regain narrow-syntactic relevance that was only attributed to the heads bearing them in chapter 2.3 (cf. P&T: fn. 21). 4 The implementation of ordered features and the utilization of their dependencies in narrow-syntactic computation is not a novel claim. Starke (2001) implements these in a ‘feature-tree’, introducing hierarchical relations of φ-, case, 4 Note that by granting these features narrow-syntactic relevance, the account elaborated here finally becomes subject to the criticism of Déchaine & Wiltschko (2002). As has been laid out in chapter 1.4.4 above, approaches tracing diverging distributions back to varying featural structure have to account for the visibility of the latter in the syntax. However, I believe to satisfy this requirement in the adequate modification of Match and Value in the following subchapter. 160 3 INFERENCE - Nominal Concord & Definiteness and several Ā-features to unify various island effects. The feature geometry of H&R on which my proposal of the split of φ-features was based in chapter 2.3, furthermore, has also undergone this treatment recently by Preminger (2011) and most prominently by Bejar (2000, 2003, Bejar & Rezac 2009), both of whom underline the narrow-syntactic applicability of H&R’s original proposal (cf. Preminger 2011: 35; Bejar 2003: 30). Bejar (2003) argues on the basis of Georgian split personand number-agreement that (relative and complete) underspecification of (nominal) goals for φ-features influences the verb’s strategies in probing for said features in a second cycle, applying to an extended search space, viz. the specifier of the probing head (following Rezac 2003). I will leave the exact mechanisms of what Bejar calls ‘context sensitive agreement’ for the next subchapter and concentrate on the formal representation of features in what follows. Bejar’s (2003) treatment of the feature geometry from H&R dismisses the distinction of organizing nodes and monovalent features, presented above, and takes each node to constitute a possible value on pronominal and nominal elements (cf. Bejar 2003: 45, but cf. Bejar 2000, where the distinction is still maintained). The system therein derives varying values of a single dependent node by ‘entailment’ and ‘underspecification’ (Bejar’s (2003: 38) “two desiderata for a theory of features”). As Bejar notes, the latter is a systematic device, employed to signal contrast. She proposes two kinds of underspecification: total vs. relative. Total underspecification designates the absence of any feature of a specific category, while relative underspecification captures any intermediary nodes in the system of which no dependent, however, is active in a given structure under consideration. Total underspecification hence parallels H&R’s treatment of bare organizing nodes, discussed above. These contrasts are illustrated in (9) with Bejar’s interpretation of the active nodes for the first, second and third personfeature; R corresponds to H&R’s root node. (9) (= Bejar 2003: ch. 2, (24), [reduced]) Nominal Concord & Definiteness I NFERENCE 126 mechanisms of what Bejar calls ‘context sensitive agreement’ for the next subchapter and concentrate on the formal representation of features in what follows. Bejar’s (2003) treatment of the feature geometry from H&R dismisses the distinction of organizing nodes and monovalent features, presented above, and takes each node to constitute a possible value on pronominal and nominal elements (cf. Bejar 2003: 45, but cf. Bejar 2000, where the distinction is still maintained). The system therein derives varying values of a single dependent node by ‘entailment’ and ‘underspecification’ (Bejar’s (2003: 38) “two desiderata for a theory of features”). As Bejar notes, the latter is a systematic device, employed to signal contrast. She proposes two kinds of underspecification: total vs. relative. Total underspecification designates the absence of any feature of a specific category, while relative underspecification captures any intermediary nodes in the system of which no dependent, however, is active in a given structure under consideration. Total underspecification hence parallels H&R’s treatment of bare organizing nodes, discussed above. These contrasts are illustrated in 0 with Bejar’s interpretation of the active nodes for the first, second and third person-feature; R corresponds to H&R’s root node. (9) (=Bejar 2003: ch. 2, (24), [reduced]) 3 rd person 2 nd person 1 st person R R R ! ! PART PART ! SPEAKER total relative underspecification Bejar however continues her investigation with the proposal of an additional node, a direct dependent of R, [π], which forms the basis for all person-values. Therein, all total underspecification of the category person is shifted to relative underspecification. This stipulation is necessary for her to capture the intervention effects of maximally underspecified elements in her analysis; this insight will play a crucial role in my analysis to be elaborated below. In what follows, I will adopt Bejar’s treatment of φ-features. Specifically, I will take the dependencies of features argued for in chapter 2.3 and repeated in 0 and unify these along the geometry of H&R in 0. We hence arrive at 0: (10) INDIVIDUATION 9 [f] [sg] class ! 3 [pl] [m] [n] Naturally, several comments are in order here. First, note that I continued to omit the animate-node from the geometry, as advocated in chapter 2.3 above, since with the relocation of feminine gender, masculine would be the sole dependent. Second, observe that the 3.2 Preliminary Remarks 161 Bejar however continues her investigation with the proposal of an additional node, a direct dependent of R, [π], which forms the basis for all person-values. Therein, all total underspecification of the category person is shifted to relative underspecification. This stipulation is necessary for her to capture the intervention effects of maximally underspecified elements in her analysis; this insight will play a crucial role in my analysis to be elaborated below. In what follows, I will adopt Bejar’s treatment of φ-features. Specifically, I will take the dependencies of features argued for in chapter 2.3 and repeated in (8) and unify these along the geometry of H&R in (5). We hence arrive at (10): (10) Nominal Concord & Definiteness I NFERENCE 126 mechanisms of what Bejar calls ‘context sensitive agreement’ for the next subchapter and concentrate on the formal representation of features in what follows. Bejar’s (2003) treatment of the feature geometry from H&R dismisses the distinction of organizing nodes and monovalent features, presented above, and takes each node to constitute a possible value on pronominal and nominal elements (cf. Bejar 2003: 45, but cf. Bejar 2000, where the distinction is still maintained). The system therein derives varying values of a single dependent node by ‘entailment’ and ‘underspecification’ (Bejar’s (2003: 38) “two desiderata for a theory of features”). As Bejar notes, the latter is a systematic device, employed to signal contrast. She proposes two kinds of underspecification: total vs. relative. Total underspecification designates the absence of any feature of a specific category, while relative underspecification captures any intermediary nodes in the system of which no dependent, however, is active in a given structure under consideration. Total underspecification hence parallels H&R’s treatment of bare organizing nodes, discussed above. These contrasts are illustrated in 0 with Bejar’s interpretation of the active nodes for the first, second and third person-feature; R corresponds to H&R’s root node. (9) (=Bejar 2003: ch. 2, (24), [reduced]) 3 rd person 2 nd person 1 st person R R R ! ! PART PART ! SPEAKER total relative underspecification Bejar however continues her investigation with the proposal of an additional node, a direct dependent of R, [π], which forms the basis for all person-values. Therein, all total underspecification of the category person is shifted to relative underspecification. This stipulation is necessary for her to capture the intervention effects of maximally underspecified elements in her analysis; this insight will play a crucial role in my analysis to be elaborated below. In what follows, I will adopt Bejar’s treatment of φ-features. Specifically, I will take the dependencies of features argued for in chapter 2.3 and repeated in 0 and unify these along the geometry of H&R in 0. We hence arrive at 0: (10) INDIVIDUATION 9 [f] [sg] class ! 3 [pl] [m] [n] Naturally, several comments are in order here. First, note that I continued to omit the animate-node from the geometry, as advocated in chapter 2.3 above, since with the relocation of feminine gender, masculine would be the sole dependent. Second, observe that the Naturally, several comments are in order here. First, note that I continued to omit the animate-node from the geometry, as advocated in chapter 2.3 above, since with the relocation of feminine gender, masculine would be the sole dependent. Second, observe that the sub-geometry in (10) employs a bipartition as well as the only tripartition of features in H&R’s hierarchy. As also observed by Bejar (2003: 47), sisterhood can have (one of) several consequences in the theory of H&R, namely cross-classification, complementary distribution and/ or combination (the latter again split between conjunction and intersection, cf. H&R: fn. 10). Bejar (p. 45f.) understands the only bipartition in the sub-geometry under investigation, i.e. the person-hierarchy, as an instance of the latter, so no opposition is expressed by branching in her system. For the bipartition of masculine and neuter gender, complementary distribution seems a straightforward choice (but see chapter 2.3.2, fn. 33 on diminutivization for a possible counter-argument). Concerning the tripartition of feminine, singular and class, I will refrain from formulating the exact interpretative content of the structural relation, which would have to take into account every combinatorial possibility of the categories and their respective dependents. However, the discussion in chapter 2.3.3 touched upon nominal structures in which features of all three branches are arguably simultaneously active (contra complementary distribution); moreover, I began the discussion in this section with the observation that the findings of chapter 2.3 interweave two formerly differentiated categories of φ-features into one complex category (contra cross-classification). This suggests 162 3 INFERENCE - Nominal Concord & Definiteness that the tripartite relation in (10) simply constitutes an instance of combinatorial possibility. Furthermore, in adopting Bejar’s treatment of features for the current analysis, a twofold transposition is in order: As reported above, Bejar’s investigation is predominantly concerned with person-feature agreement in the verbal domain, while the current inquiry focusses on numberand gender-agreement in nominal concord. As observed by several authors 5 person-agreement is at best rare in nominal concord cross-linguistically (see however Baker 2008 for exceptions) and doubtlessly non-existent in German. Since person-agreement therefore falls out of the scope of the current investigation, I have omitted all structure above the individuation node in (10). This does, of course, not imply that person-features are not part of the φ-feature bundle; however, as Danon (2011) presumes, the feature person is likely to be introduced into the derivation as late as at the DP-level cross-linguistically. In what follows, I will henceforth refer to individuation as the root node [φ]. Observe that underspecification w.r.t. φ-features is hence always relative underspecification. I will argue that this gives the correct predictions for intervention effects in nominal concord in German. The diverging focus will become crucial in the exact implementation of Match as well as Value qua Agree, to which I will turn now. 3.2.2 Agree and Nominal Concord We have seen in the introduction to this subchapter that the complex operation Agree is cyclic in that it successively applies to each set of uninterpretable formal features newly introduced into the derivation, in line with the Earliness Principle (Pesetsky 1989, Pesetsky & Torrego 2001: 400). It has been argued numerous times that it is this core characteristic of Agree that renders it unsuitable for the analysis of nominal concord. Recently, Norris (2014) argued that verbal agreement and concord cannot be analyzed in the same terms; he lists a number of characteristics that split these phenomena (and hence, the operations underlying them) apart: 6 5 Cf. Lehmann (1988: 56-58), cf. Baker (2008: ch. 2 and references therein), also Norris (2012: 2, 2014: 162). 6 Observe that Norris’ main concern lies with extended nominal projections in Estonian, traditionally classified as an articleless language and moreover also lacking the adjectival inflectional partition, exemplified above. Nominal concord hence reduces to various instances of modifier-head agreement in his analysis (cf. Norris 2014: ch. 3.1), while the approach elaborated below additionally strives to include agreement between left-peripheral LIs, i.e. modifier-modifier agreement. 3.2 Preliminary Remarks 163 (11) (= Norris 2014: (Table 3.2)) Subject-verb agreement Concord Number of loci of expression (in Estonian) one many Structural position of agreeing elements head head, specifier, adjunct Feature origin external internal Case-dependence yes no Let me go through these for the language under consideration. The first property doubtlessly also applies to German. Concerning the second property, recall that I have shown in chapter 1.4 that, although there is no general consensus as to the categorical status of the elements under consideration here, sufficient evidence has been put forward from previous analyses for all the SLIs in the nominal domain (in German) to be analyzed as heads of their respective projections. The third property is evaluated with respect to the extended projection of an LI, with nominal features transmitted into the clausal spine in verbal agreement, while they remain in the same nominal hierarchy (at various SLIs) in concord. Finally, the last property depicts the phenomenon of verbal agreement that is dependent on the case-value of the subject. As Norris notes, such patterns of agreement are derivationally captured if T is taken to interact only with a subset of all case-valued nominal domains. Under standard assumptions, however, nominative case is associated with T in such a way that it poses a reflex, rather than a precondition, for agreement. Moreover, observe that the proposal also runs counter to the Activation Condition as has been laid out above. Nevertheless, I will argue in chapter 3.2.2 below that patterns of agreement in concord are influenced by (mechanisms of) case-assignment. Therein, the list of diverging properties reduces to number one and number three, which, as I will show below, are tightly interrelated: It is the domain-internal origin of features which facilitates multiple loci of exponence in a highly derivational approach to nominal concord on the basis of Agree. Norris continues to list several recent modifications of the operation Agree (cf. Norris 2014: ch. 3.2.2, 3.2.3 and references therein) and finally dismisses a full narrow-syntactic analysis of nominal concord. Instead, he identifies two components of the grammar to be at play in the derivation of the phenomenon, to wit: Syntax and Morphology. Therein, the syntactic side mimics the highly derivational mechanisms of Agree, sketched above, in that features might percolate up the nominal spine until they are blocked by a differently specified 164 3 INFERENCE - Nominal Concord & Definiteness instance of the same feature. Norris takes this to apply language-universally. However, a phonological reflex of said features is determined later in the morphological component of the grammar by the insertion of Agr-heads as sisters to the respective SLIs in the nominal domain. Said Agr-nodes vary w.r.t. the feature categories they incorporate and are valued by the closest dominating head via copying. His partition of the phenomenon is given below: (12) (= Norris 2014: (Table 3.8)) Concord Syntax Morphology feature percolation Agr 0 node insertion case concord feature copying from closest dominating source Norris arrives at a system that is able to capture ‘disrupted’ patterns of concord that exceed the standard notion of agreement with the head noun. However, concentrating on the exact implementation of the components of grammar taken to participate in concord, the split advocated by Norris in fact splits the phenomenon into two different kinds of Agree relations: i.e. upward and downward agreement. While feature percolation in Syntax follows the standard rationale of Agree in that a hierarchically higher element receives a value from a lower source, feature copying in Morphology takes the exact opposite way. This proposal hence makes Agree in the nominal domain bi-directional, which has been independently argued for by Baker (2008). He does so by relaxing the c-command restriction between probe and goal to be a transitive one and hence renders either configuration P > G, G > P (‘>’ indicating c-command) legal. His account rests on the assumption that features have to be unified across the nominal domain at its highest head (and beyond that point w.r.t. case). The rationale of unification/ collection of feature values has also been employed in Norris’ (2012) treatment of concord by means of - what might be termed - a ‘mediator head’ on top of the nominal domain (viz. KP, Kase Phrase, cf. Lamontagne & Travis 1987) which receives values for φ-features from the heads it dominates and a case-value from the clausal domain. Therein, some form of Agree is again part of the mechanism deriving nominal concord but the morphological insertion of Agr-nodes to the elements participating in concord assures the uniform pattern inside the nominal hierarchy of projections. The second modification from Baker (2008) concerning the directionality Agree has been explored by numerous authors. Recently, Zeijlstra (2012) has argued for the complete reversion of agreement relations on the basis of several phenomena, problematic for the standard version of Agree, one of them being clausal concord phenomena (i.e. negative concord and sequence of tense). In his 3.2 Preliminary Remarks 165 system, goals bearing an interpretable instance of a formal feature c-command their uninterpretable counterparts with which Agree relations are established (but see Preminger 2013). In his revision, Zeijlstra contrasts the standard approach as well as his own proposals with P&T’s feature sharing version of Agree as well as Bošković’s (2007) “Activation Condition/ Phase Impenetrability Condition-free conception of Agree” (Bošković 2007: 634), which are uniformly downward directional. I will review the former in detail later in this chapter. As has been laid out, (nominal) concord is taken to pose a problem for the default notion of Agree in the minimalist framework, mainly on the basis of the multiple loci of exponence of the features involved. Alternatively, analyses revert to strategies of collecting and copying of features to derive the uniform pattern of case-/ φ-agreement across elements in the nominal hierarchy. However, I want to show that nominal concord in German constitutes a highly derivational phenomenon to be analyzed exclusively in terms of Agree, parallel to argument-predicate agreement (‘A-P agreement’, Norris 2012). Hence, all modifications to Chomsky’s original formulation of Agree introduced hereafter have been applied in the analysis of clausal agreement phenomena. The only restriction that cannot be carried over is the requirement of φ-completeness of goals for the successful establishment of a probe-goal relation. I take this to follow naturally from the object of investigation (but see the comments on feature chains below for possible reinstantiations). However, as we will see below, certain idiosyncrasies of nominal concord in German are arguably not syntactic in nature and hence a split of the complex phenomenon to be handled by different domains of the grammar seems inevitable. The split to be elaborated below will nevertheless be categorically determined and hence not interfere with derivational dependencies. I will postpone this issue until a clear picture of the narrow-syntactic derivation is established. With these initial concepts in place, I begin by presenting Bejar’s (2003) versions of Match and Value as Agree, which themselves were worked out on the basis of Georgian A-P agreement, applying to complex feature notions like those elaborated in the last section. I will contrast them with Chomsky’s own proposals and finally discuss an unrelated instance, namely feature sharing approaches to Agree, 7 which have also been successfully applied to A-P agreement, and show with Danon (2011) how a coherent derivational system of nominal concord phenomena in German can be developed in merging key ideas from both approaches in the subchapters to follow. As reported above, Bejar dismisses the [category: value] notion of features and instead takes every node in the feature geometry to constitute a possible 7 Cf. Frampton & Gutmann (2000), P&T, Bošković (2007). 166 3 INFERENCE - Nominal Concord & Definiteness value for [F], which ambiguously seem to constitute a value from the viewpoint of nodes dominating [F] (e.g. [φ[F]]) and at the same time appear to depict a category from the viewpoint of further dependents of [F] (e.g. […[F[G]]]). As she elaborates, Match, as operating on the basis of feature-identity (Chomsky MI: 122), is not applicable to these notions: “[S]ince probe is by definition unvalued, the value of the goal cannot enter into the evaluation of match” (Bejar 2003: 51). Hence, preconditions for Match (and Value) must take into account the specification of the goal relative to the specification of the probe. The notion Bejar implements is ‘entailment’ (‘implication’ in Bejar 2000). Her definition of Match is given in (14); the locus of application is the c-command domain of the probe (i.e. its ‘domain’), following Chomsky (MI). His formulation is given in (13): (13) (= Chomsky MI: (40) b., c.) a. D(P) is the sister of P. b. Locality reduces to ‘closest c-command.’ (14) (= Bejar 2003: ch. 2, (35)-(38)) a. Entailment Condition Match is defined by entailment. b. Probe (F) and Goal (F’) match if Goal (F’) entails Probe (F). c. Match is evaluated at the root. where d. Probe (F) and Goal (F’) match if Goal (F’) entails root Probe (F). Additionally, Bejar has to posit a distinction between uninterpretable and interpretable features in that only the latter are automatically specified for every feature between [F] and the root node (i.e. above [F] in H&R’s geometry), since otherwise Match would apply in every c-commanding relationship between two items that bear φ-specifications, because they share the same root. We can ignore this additional constraint, since - as I will argue below - all elements in the nominal domain are at least relatively underspecified w.r.t. the root node φ. Match as formulated in (14) hence applies in those cases in which the goal is at least exactly specified like the probe, i.e. entailment applied at the root includes identity: (15) probe goal match a. [φ] [φ]  b. [φ] [φ[F]]  3.2 Preliminary Remarks 167 Observe that (14) also makes the prediction that Match is successful when the goal is more highly specified than the probe. Apart from this default case, depicted in (15) b., we have two configurations in which Match may apply: “Where the probe and the goal have equally specified features [i.e. (15) a., MB], structure-building is vacuous (but we will assume, successful) […] where the goal is less specified than the probe [i.e. (16), MB], match may succeed, but value fails” (Bejar 2003: 60): (16) probe goal match [φ[F]] [φ]  It has not yet been noted what consequences follow if no goal can be identified at all. In this case, trivially, no Match relation can be established, irrelevant of the complexity of the feature structure of the probe: (17) probe goal match a. [φ] ø  b. [φ[F]] ø  Let us next turn to Agree, i.e. Value. As Bejar notes, conditions on Value have to be more restrictive than conditions on Match (cf. Chomsky MI: 122), since it is possible for the former to succeed but the latter to fail. Therefore, conditions on Value pose as subset of the conditions on Match, just reviewed. Bejar proposes the following: (18) (= Bejar 2003: ch. 2, (58)) Condition on Value G(oal) values P(robe) iff f ’(G) entails f(P) The condition in (18) states that the features of the goal must be a superset of the features of the probe. The difference between Match and Value boils down to their locus of evaluation, with Match evaluated at the root node and Value evaluated w.r.t. the complete featural structure. Her predictions concerning the establishment of Value for the structural configurations given in (15) and (16) are therefore: 168 3 INFERENCE - Nominal Concord & Definiteness (19) probe goal value a. [φ] [φ]  b. [φ] [φ[F]]  c. [φ[F]] [φ]  The only configuration, in which Value is predicted to fail, hence, occurs when the probe is more highly specified than the goal. Failure to Value is thus a consequence of relative underspecification, while failure to Match results from total underspecification, a case which I already excluded above for the nominal domain in German. Not discussed in Bejar’s treatment of person-features is the case of (equally) complex, but diverging, specifications of probe and goal (cf. that Bejars’s modification of H&R’s feature geometry takes participant to be the sole dependent of [π] with the only binary branching node encoding combination and hence no contrast, while the geometry proposed in chapter 3.1.1 shows two instances of arguably contrasting branching nodes, cf. the discussion at the end of said subchapter). It follows from the conditions above that Match may succeed in these cases, but Value fails: (20) probe goal match value [φ[F]] [φ[G]]   We therefore expect intervention effects arising from configurations of the form (19) c. and (20), in which Match succeeds but Value fails (including the default case of inactive goals due to case-assignment, cf. Chomsky MI: 123; what Rezac 2003 refers to as the ‘Active Goal Hypothesis’). What is the consequence of unsuccessful applications of Match or Value in the system at hand, that is, what happens when either configuration in (17), (19) c. or (20) occurs? In the standard version of the Minimalist Program, the unchecked feature would be ‘buried’ in the derivation after additional applications of EM and thereupon cause the derivation to crash at SEM due to a violation of Full Interpretation (FI). Bejar, however, sketches an interesting alternative on the basis of default number and person A-P agreement in Georgian. Observe (21) and (22): 3.2 Preliminary Remarks 169 (21) (= Bejar 2003: ch. 2, (77), [Georgian]) V-xedav-(ø) 1-see-( number ) ‘I see him.’ (22) (= Bejar 2003: ch. 2, (78), [from Sigurðsson 1996, Icelandic]) Strákunum leiddist/ *leiddust The.boys. dat . pl bored.3 sg / *3. pl ‘The boys were bored.’ She concludes that “the success of match and value are (at least in some instances) irrelevant to convergence” (Bejar 2003: 77) and the cases in (21) and (22) to display “default agreement […] made available by the computation to rescue/ delete unvalued uninterpretable features” (Bejar 2003: 77-78.). Bejar proposes that default agreement applies in exactly those cases listed above and summed up in (23): “[D]efault agreement is triggered upon the failure of value. The idea is that this occurs either when a probe is halted by an intervener [i.e. (23) c.,d., MB] or when […] a probe exhausts its search domain without finding a match [i.e. (23) a.,b., MB]” (Bejar 2003: 78). (23) probe goal match value a. [φ] ø   b. [φ[F]] ø   c. [φ[F]] [φ]   d. [φ[F]] [φ[G]]   How, then, is default agreement captured in featural terms? Bejar proposes that the failure to value in (23) results in deletion of all featural structure below the root node ([φ] in the system under consideration), signaling that an attempt was made to value the feature. She calls the resulting [φ] on the probe ‘partial default agreement’, which coincides with relative underspecification: “[T]he feature is impoverished, but remains active” (Bejar 2003: ibidem). This stands in contrast to ‘total default agreement’, which deactivates a certain feature completely. Partial default agreement, Bejar presumes, has to turn into total default agreement at some point in the derivation (although she later states that partial default agreement obviates a crash of the derivation). When exactly the deactivation of a default feature applies remains an open question. Bejar herself takes it to be the next application of EM and states that nothing hinges on this. However, 170 3 INFERENCE - Nominal Concord & Definiteness I will later show that deactivation has to apply later in the derivation, if at all. See that for a partial default feature to obviate a crash it has to be non-distinct from a valued uninterpretable/ interpretable feature. Deactivation would hence be gratuitous. I will come back to this below. Note that the above instances of underspecification and default agreement are taken to derive from Narrow Syntax in the system at hand, contrary to the claims from Distributed Morphology that impoverished/ reduced outputs are the result of morphological processes as elaborated in chapter 2.3.3, based on Noyer ([1992] 1997; cf. Bejar 2003: ch. 2, fn. 8). I believe, however, that these dedicated processes do not run counter to one another since the featural impoverishment under consideration here has further syntactic import, to which I will turn now. As stated above, partial default agreement reduces the feature structure of the probe to the bare root node. Hence, in Bejar’s system, an unsuccessful application of Match/ Value does have an impact on the featural content of the probe. This is a novel claim in itself but the exact implementation bears additional consequences for the proceeding derivation. Observe that, whatever the selectional restriction of the probe was before the unsuccessful application of Match/ Value, it is trivially specified for any value of the root node after transformation into partial default agreement if Agree would happen to apply again thereafter. Bejar (2000) actually proposes this derivational variant for the analysis of Georgian person A-P agreement in (24), varying glosses from Bejar (2000) and (2003) are indicated by parentheses: (24) (= Bejar 2000: (12) c., e.; Bejar 2003: ch. 2, (66), [Georgian]) a. m-xedav-s 1-seetns ‘(s)he sees me.’ b. g-xedav(-s) 2-seetns ‘I see you.’ c. v-xedav (= Bejar 2000: (13) a.; Bejar 2003: ch. 2, (67), [Georgian]) 1-see ‘I see him.’ The object acts as an intervener for subject A-P agreement concerning person in Georgian: Subject agreement only surfaces when the object bears 3 rd person (i.e. the bare root [π] in Bejar 2003). She concludes that there are actually two dedicated cycles of Agree involved in (24): The verb in Georgian probes for [π[ part ]] from its base position and - in structures involving first and second person objects - Agree successfully applies between the predicate and the ob- 3.2 Preliminary Remarks 171 ject. However, since the third person-value is relatively underspecified, Agree (qua Value) between predicate and object cannot apply: (25) probe v goal obj value [π[ part ]] [π]  By v -to-T movement, more accurately, IM of v in T, a second application of probing is triggered, with the subject DP now the highest argument in the domain D(P), i.e. D(T). Hence, an external argument, at least specified for [π[ part ]], can now match and value v ’s person-feature. The second application of Agree hence applies to an extended search space due to movement of the probe. A different implementation of this rationale can be found in Rezac (2003, see also Bejar & Rezac 2009). He takes the insights of Chomsky’s bare phrase structure seriously in that the projection of a head actually constitutes the exact same structure as the head itself. As Chomsky (MI: 133, with Rezac’s (2003: (7)) addition in angled brackets) notes: (26) Label(α) = α, for α an LI [or its modification under Agree]. From there, Rezac derives the specifier of a probe as the extended domain for the means of probing. The argument goes as follows: If the label is identical with the head from which it projects (by means of its featural content) and probing takes place on the basis of c-command (cf. (13) a. above), then further Merger to the head α extends its checking domain D(α). This expansion happens cyclically, i.e. the first attempt to find a goal targets the traditional domain of the probe. Rezac refers to this as the ‘dynamic approach to search space’: (27) (= Rezac 2003: (3)) {α, complement} → { α, {specifier, { α, complement}}} He takes the Earliness Principle (Pesetsky & Torrego 2001: 400) as the economy condition preventing unchecked uninterpretable features below the root node (which would induce a crash at Spell-Out due to violation of FI) to account for this effect. Hence, e.g. v does not initiate Agree with the external argument in its specifier if/ since it finds a goal for φ-feature agreement in its domain (in run-ofthe-mill sentences). However, Chomsky’s very own formulation of ‘cyclicity’ in MI (p. 132) makes a different prediction, not addressed by Rezac (2003) or Bejar (2003). I give both principles in (27): 172 3 INFERENCE - Nominal Concord & Definiteness (28) a. Earliness Principle (= Pesetsky & Torrego 2001: (91)) An uninterpretable feature must be marked for deletion as early in the derivation as possible. b. Cyclicity (= Chomsky MI: (53)) Properties of the probe/ selector α must be satisfied before new elements of the lexical subarray are accessed to drive further operations. I follow Bejar in treating “marked for deletion” and “satisfied” in (28) a., b. synonymously. Cyclicity (but not the Earliness Principle) hence states that the extension of the domain for probing of α to Spec,αP can only happen by means of Internal Merge. Returning to Bejar’s (2003) own proposals, this is exactly what the deactivation of partial default agreement at the next introduction of LIs from the Numeration predicts. Second cycle Agree is hence solely agreement with an element, already introduced into the derivation in Bejar (2003). However, returning to partial default values, the probe’s selectional restrictions are relaxed in the 2 nd cycle: They can be trivially met by any element sharing the root with the probe (by Match as well as Value). To recap, 2 nd cycle Agree is bound to movement in all accounts under consideration so far: It is restricted to probes moving to extend their domain in Bejar (2000), while only goals already part of the derivation at the stage where probing applies can move to the extended domain (i.e. the specifier position) in Bejar (2003). In Rezac (2003), we saw that this position might additionally be available for External Merge of suitable goals, once the first probing fails to identify a proper feature in the traditional domain of the head (i.e. its sister). We have seen, then, that 2 nd cycles of Agree are always triggered. Bejar (2003: 71), reviewing her earlier account, explicitly states that it is the movement of the probing head that forces all uninterpretable features to probe again; her later system (2003: 80) likewise takes re-Merge (by projection and possibly also head-movement) to force the second probing of the head’s uninterpretable features. The changes in the structural configurations are themselves triggered by unsuccessful applications of Agree (cf. Bejar 2003: 68). For Rezac, 2 nd cycles are much more regular in that they obey the c-command restriction by reinterpreting the properties of Label(α). Still, it is the failure to find a goal in the traditional domain D(α) that triggers the head to ‘look elsewhere’ once. All accounts hence take the failure of the head’s probing to motivate a second cycle in which different hierarchical configurations allow for the search in an altered domain. Bošković (2011), elaborating on findings from Bošković (2009), implements yet another variant of 2 nd cycle Agree on the basis of conjunct A-P agreement 3.2 Preliminary Remarks 173 in SC. 8 Observe the obligatory gender-agreement of the participle with the first (neuter) conjunct in (29) a. and with the second (feminine) conjunct in the configuration with a fronted argument (incorporation of the adverb in (29) a. is obligatory since enclitic su cannot surface sentence-initially). (29) (= Bošković 2011: (6), [Serbo-Croatian]) a. Juče su uništene [ &P sve varošice I sva sela]. yesterday are destroyed.pl.fem all towns.fem and all villages.neut ‘All villages and all towns were destroyed yesterday.’ b. [ &P Sve varošice I sva sela] su uništena. All towns.fem and all villages.neut are destroyed.pl.neut The analysis of the varying gender-agreement on the predicate rests on the assumption that the syntactic properties of grammatical versus semantically contentful gender-features vary even language-internally. The argument goes as follows: Assuming that the highest head of the conjunct (&P) bears a plural specification for number but none for gender, the predicate probes for φ-features and finds [pl] on &P and [f] on the first NP of the conjunct in (29) a. This is not critical. With a dislocated conjunct in (29) b., however, the predicate bears an EPP-feature for which the two goals induce a problem: Both &P matching for number as well as the first conjunct matching for gender are able to check the EPP on the probe, since first conjunct extraction is licit in SC. Bošković (following McGinnis’s 1998) presumes that this ambiguity cancels the valuation. The important step in the derivation follows this cancellation: Since the gender of the first conjunct is solely grammatical, (unsuccessful) probing by the predicate suffices to delete it from the derivation. The predicate thereafter probes for a second time, now able to value its gender-feature since the former intervener is no longer part of the derivation and the search space of the predicate has hence extended to the second conjunct. Since the goal for number (&P) is able to raise, but the goal for gender is not, the ambiguity is circumvented in the second cycle and the complex subject raises to sentence-initial position. The 2 nd cycle proposed by Bošković differs from the implementations reviewed above in a number of key respects. First, it is not strictly the failure of Match or Value that enables a second cycle, but the ambiguity arising from two successful applications thereof. More importantly, it differs in that no structural change is initiated for either the probe or a newly accessible goal, but it is the featural set-up in the domain of the probe that changes and extends the search 8 Cf. Corbett (1991: chapter 9) for a detailed overview of conjunction agreement across languages. 174 3 INFERENCE - Nominal Concord & Definiteness space. We have hence identified an instance of in-situ 2 nd cycle Agree. Bošković (2009) understands Chomsky’s (MI) analysis of expletive constructions, discussed in chapter 3.1.1 above, to pose another instance of this operation: Matrix T probes for φ and attracts the expletive due to its EPP-feature into its specifier before subsequently agreeing with the lower argument. However, Bošković (as well as Bejar 2003: 62) takes the attraction of the expletive to follow from Value of the person-feature of T by the expletive, not from simple Match (as in Chomsky: MI). As I tried to lay out above, Expl never posed a suitable goal for T’s φ-features since it is not φ-complete and is hence overlooked in the search for a valuator (cf. Frampton et al. 2000: 2). On another plane, Match (as the precondition for movement) applies to Expl because of its categorical status as well as its active person-feature, resulting in the movement to Spec,TP. I therefore conclude that there are no 2 nd cycle effects in Chomsky’s (MI) treatment of expletive constructions. An important point, shared by all accounts examined above, stays dubious in Bošković’s proposal, to wit, what triggers the probe to instantiate Agree a second time? In Bejar (2000), the failure to find a suitable goal forces movement of the probe, while it forces the probe to extend its domain in a cyclic fashion in Bejar (2003) and Rezac (2003). This is not true for the analysis advocated by Bošković, in which the exact same domain is probed a second time for the same feature. If the purely grammatical gender-value of the first conjunct deletes “as soon as [it is] targeted by a probing operation” (Bošković 2009: 457), the question is: How can the probe know that Match/ Value will now have a successful - or actually, different - outcome? I want to adopt Bošković’s proposal of in-situ 2 nd cycle Agree but modify it accordingly: A probe is able to initiate a second search for a suitable goal in the same domain; however, this operation must be triggered. Only if the probe is itself forced to bear a value for some feature, will it instantiate a second attempt to gain it from the domain accessible at that moment. This turns the probe of the second cycle to simultaneously act as a goal (for the trigger) and a probe (for the second cycle) and stands in stark contrast to the Earliness Principle. However, we have seen that partial default features behave as saturated (valued uninterpretable/ interpretable) features in the ongoing derivation. A second cycle is hence not necessary in that it circumvents the crash of the derivation, but motivated/ triggered. Further refinements to this proposal are doubtlessly needed, but I will postpone these until the discussion of the paradigm of nominal concord in German further below. Returning to Bošković’s (2011) proposals, his concerns lie elsewhere, namely the establishment of valued uninterpretable features (i.e. [ u F: val]) as legal elements in Narrow Syntax. Bošković’s first objects of investigation, as has been 3.2 Preliminary Remarks 175 shown above, are gender-features. His split between semantically motivated (i.e. interpretable) and purely syntactic (i.e. valued uninterpretable) gender-features is reminiscent of Ritter’s (1993) split, reported in chapter 2.3.2 a., between languages which employ gender switching as a derivational strategy and those in which the effect is restricted or absent. While Ritter arrives at a language-specific location for gender, Bošković (also referring to Despić 2010) understands the variation to occur inside a single language. This is made possible by the shift in notions held responsible; while it is the featural set-up of particular heads in Ritter’s account, it is the set-up of the feature itself in Bošković’s analysis: Valued uninterpretable features are said to enter the derivation with the value already implemented. Therefore, they can act as goals or interveners. Still, as we have seen above, these elements need not partake in a successful Agree relation to delete. As Bošković argues, probing of the element is sufficient to delete them from the derivation. The kind of feature just described violates Chomsky’s (DbP: 5) implementation of the Valuation/ Interpretability Biconditional as described by P&T, mentioned in chapter 3.1.1 above. They were first proposed by said authors as one of the results of rejecting the biconditional, which - for them - is a questionable coupling of unrelated properties of lexical items in the lexicon (cf. P&T: 266f.). The result of said rejection is an expansion of the possible feature types from two to four logical compositions: (30) (= P&T: (9)) Types of features (boldface = disallowed in MI/ DbP) u F val uninterpretable, valued i F val interpretable, valued u F [ ] uninterpretable, unvalued i F [ ] interpretable, unvalued P&T note that thereby two probing feature types emerge, namely uninterpretable and interpretable unvalued features. Syntax is hence not sensitive to the property of interpretability, but valuation, the consequence being that the applicability of interpretability should be located outside the syntax proper, since it is now of exclusively semantic relevance. P&T follow Brody’s (1997) ‘Radical Interpretability’ in this point: (31) (= P&T: (15)) Thesis of Radical Interpretability Each feature must receive a semantic interpretation in some syntactic location. Thereby, the need to value uninterpretable features, which is unmotivated and stipulated in MI/ DbP, is derived, since syntax has no knowledge which ‘part’ of the feature will be interpreted (probe or goal) after syntax: “[R]adical Inter- 176 3 INFERENCE - Nominal Concord & Definiteness pretability requires all syntactic elements to be semantically interpretable, but not necessarily actually interpreted in a given structure” (Brody 1997: 143-144). Those ‘parts’, then, have to be revised to ensure that they meet the condition to be potentially interpreted later in the derivation. P&T refer to Brody (1997), Frampton & Gutmann (2000) as well as works conducted in HPSG (cf. Pollard & Sag 1994) to propose an alternative version of Agree as feature sharing: (32) (= P&T: (5)) Agree (Feature sharing version) i. An unvalued feature F (a probe ) on a head H at syntactic location α (F α ) scans its c-command domain for another instance of F (a goal ) at location β (F β ) with which to agree. ii. Replace F α with F β , so that the same feature is present in both locations. Call [F], part of the shared feature, an ‘instance of [F]’; [F], not yet undergone Agree, an ‘occurrence of [F]’. P&T point out that all (valued and unvalued) occurrences of [F] must turn into instances of [F] by means of Agree. These, then, can act as goals for further probing for [F] thereafter. Agree creates a consistent link between two instances, rather than a one-time checking relation as the standard version does (cf. Danon 2011: 10). Derivationally, even two unvalued features are able to Agree if a later application with a valued instance turns all three occurrences of [F] into instances, a case taken to be “either vacuous or impossible” (P&T: 269) in the standard approach, following Frampton et al. (2000, cf. also P&T: fn. 9). With these notions in place, observe that preconditions for Agree are far more relaxed in that uninterpretable as well as interpretable unvalued features can act as probes (i.e. the lower row in (30) above). This suffices to exclude what Bošković (2011) refers to as ‘reflex feature checking’: the need to posit distinct uninterpretable features on probe and goal to achieve mutual valuing on both elements in two unidirectional (downward) applications of Agree; cf. (33): (33) α > β [ u F] [ i F] [ i G] [ u G] [EPP] Even though a value for [G] on β exists in the derivation, α[ i G] is not in the domain of β, D(β), and hence unable to value β[ u G]. Probing of α[ u F] and β[ i F] as well as subsequent attraction by the EPP on α has to be posited to create the structural configuration, in which [ i G] is located in D(β), its sister. This is the ra- 3.2 Preliminary Remarks 177 tionale of structural case-assignment in MP (though, (structural) case is strictly not a feature assigned by Value and arguably not a feature of the head T at all, cf. Frampton et al. 2000: 3, Bošković 2011). P&T’s system, however, understands case to be unvalued uninterpretable u T[ ] on DP, as well as unvalued interpretable i T[ ] on T 0 (‘Tns’ in their terms). In simple run-of-the-mill configurations, the higher unvalued (but interpretable) occurrence is able to probe for, and agree with, the lower instance of T[ ]. Two instances of T are thereby created, followed by Agree between T 0 and v , which bears another uninterpretable but valued instance, resulting in the configuration (34): 9 (34) Nominal Concord & Definiteness I NFERENCE 133 With these notions in place, observe that preconditions for Agree are far more relaxed in that uninterpretable as well as interpretable unvalued features can act as probes (i.e. the lower row in 0 above). This suffices to exclude what Bošković (2011) refers to as ‘reflex feature checking’: the need to posit distinct uninterpretable features on probe and goal to achieve mutual valuing on both elements in two unidirectional (downward) applications of Agree; cf. 0: (33) α > β [ u F] [ i F] [ i G] [ u G] [EPP] Even though a value for [G] on β exists in the derivation, α[ i G] is not in the domain of β, D(β), and hence unable to value β[ u G]. Probing of α[ u F] and β[ i F] as well as subsequent attraction by the EPP on α has to be posited to create the structural configuration, in which [ i G] is located in D(β), its sister. This is the rationale of structural case-assignment in MP (though, (structural) case is strictly not a feature assigned by Value and arguably not a feature of the head T at all, cf. Frampton et al. 2000: 3, Bošković 2011). P&T’s system, however, understands case to be unvalued uninterpretable u T[ ] on DP, as well as unvalued interpretable i T[ ] on T 0 (‘Tns’ in their terms). In simple run-of-the-mill configurations, the higher unvalued (but interpretable) occurrence is able to probe for, and agree with, the lower instance of T[ ]. Two instances of T are thereby created, followed by Agree between T 0 and v , which bears another uninterpretable but valued instance, resulting in the configuration 0: 9 (34) T 0 > DP > v i T[* val ] u T[* val ] u T[ val ] Structural case on the subject is hence not dependent on φ-agreement between T 0 and DP. 10 In what follows, I will reject P&T’s approach to structural case for reasons that will become clear in the discussion of the paradigm of nominal concord in German and assume with Chomsky that φ-agreement has to apply between T 0 and the external argument. As for the notion of feature sharing, I will follow P&T’s claims, though not without some modifications, to which I will turn forthwith. The combination of P&T’s version of Agree and Bejar’s notion of φ-features seems evidently problematic since with the dismissal of the Valuation/ Interpretability Biconditional in 0 and the banishment of the interpretable/ uninterpretable distinction from the syntax proper due to Radical Interpretability in 0, the system at hand, which disposed of the categoryvalue bipartition in feature structure, is in danger to predict a vast over-application of Agree. This seems to be a real bug in the system; cf. the derivational snapshot in 0 when α is externally merged: (35) α > β > γ [φ[F[G]] [φ[*F]] [φ[F]] Following P&T in the implementation of Brody’s Radical Interpretability, I will henceforth not distinguish interpretable and uninterpretable instances of features by the respective diacritics [ i F] and [ u F]. With the Valuation/ Interpretability Biconditional in place and the rejection of Radical Interpretability, α would find a match for [φ] in β, but Value could not apply since the structure of [φ] on the latter does not entail the structure on α (an instance of the intervention effect 0 c. 9 Here, I employ Frampton & Gutmann’s use of the asterisk to indicate formerly unvalued instances of feature occurrences, the greater-than sign illustrating c-command relations as well as the lines to indicate a shared feature/ value. 10 However, observe that T’s EPP-feature is not part of the approach to structural case in P&T’s account and is hence again an isolated phenomenon (i.e. the subject still moves to T 0 in several languages), which the authors leave as an open matter (cf. P&T: fn. 22). Structural case on the subject is hence not dependent on φ-agreement between T 0 and DP. 10 In what follows, I will reject P&T’s approach to structural case for reasons that will become clear in the discussion of the paradigm of nominal concord in German and assume with Chomsky that φ-agreement has to apply between T 0 and the external argument. As for the notion of feature sharing, I will follow P&T’s claims, though not without some modifications, to which I will turn forthwith. The combination of P&T’s version of Agree and Bejar’s notion of φ-features seems evidently problematic since with the dismissal of the Valuation/ Interpretability Biconditional in (3) and the banishment of the interpretable/ uninterpretable distinction from the syntax proper due to Radical Interpretability in (31), the system at hand, which disposed of the category-value bipartition in feature structure, is in danger to predict a vast over-application of Agree. This seems to be a real bug in the system; cf. the derivational snapshot in (35) when α is externally merged: (35) Nominal Concord & Definiteness I NFERENCE 133 With these notions in place, observe that preconditions for Agree are far more relaxed in that uninterpretable as well as interpretable unvalued features can act as probes (i.e. the lower row in 0 above). This suffices to exclude what Bošković (2011) refers to as ‘reflex feature checking’: the need to posit distinct uninterpretable features on probe and goal to achieve mutual valuing on both elements in two unidirectional (downward) applications of Agree; cf. 0: (33) α > β [ u F] [ i F] [ i G] [ u G] [EPP] Even though a value for [G] on β exists in the derivation, α[ i G] is not in the domain of β, D(β), and hence unable to value β[ u G]. Probing of α[ u F] and β[ i F] as well as subsequent attraction by the EPP on α has to be posited to create the structural configuration, in which [ i G] is located in D(β), its sister. This is the rationale of structural case-assignment in MP (though, (structural) case is strictly not a feature assigned by Value and arguably not a feature of the head T at all, cf. Frampton et al. 2000: 3, Bošković 2011). P&T’s system, however, understands case to be unvalued uninterpretable u T[ ] on DP, as well as unvalued interpretable i T[ ] on T 0 (‘Tns’ in their terms). In simple run-of-the-mill configurations, the higher unvalued (but interpretable) occurrence is able to probe for, and agree with, the lower instance of T[ ]. Two instances of T are thereby created, followed by Agree between T 0 and v , which bears another uninterpretable but valued instance, resulting in the configuration 0: 9 (34) T 0 > DP > v i T[* val ] u T[* val ] u T[ val ] Structural case on the subject is hence not dependent on φ-agreement between T 0 and DP. 10 In what follows, I will reject P&T’s approach to structural case for reasons that will become clear in the discussion of the paradigm of nominal concord in German and assume with Chomsky that φ-agreement has to apply between T 0 and the external argument. As for the notion of feature sharing, I will follow P&T’s claims, though not without some modifications, to which I will turn forthwith. The combination of P&T’s version of Agree and Bejar’s notion of φ-features seems evidently problematic since with the dismissal of the Valuation/ Interpretability Biconditional in 0 and the banishment of the interpretable/ uninterpretable distinction from the syntax proper due to Radical Interpretability in 0, the system at hand, which disposed of the categoryvalue bipartition in feature structure, is in danger to predict a vast over-application of Agree. This seems to be a real bug in the system; cf. the derivational snapshot in 0 when α is externally merged: (35) α > β > γ [φ[F[G]] [φ[*F]] [φ[F]] Following P&T in the implementation of Brody’s Radical Interpretability, I will henceforth not distinguish interpretable and uninterpretable instances of features by the respective diacritics [ i F] and [ u F]. With the Valuation/ Interpretability Biconditional in place and the rejection of Radical Interpretability, α would find a match for [φ] in β, but Value could not apply since the structure of [φ] on the latter does not entail the structure on α (an instance of the intervention effect 0 c. 9 Here, I employ Frampton & Gutmann’s use of the asterisk to indicate formerly unvalued instances of feature occurrences, the greater-than sign illustrating c-command relations as well as the lines to indicate a shared feature/ value. 10 However, observe that T’s EPP-feature is not part of the approach to structural case in P&T’s account and is hence again an isolated phenomenon (i.e. the subject still moves to T 0 in several languages), which the authors leave as an open matter (cf. P&T: fn. 22). ] 9 Here, I employ Frampton & Gutmann’s use of the asterisk to indicate formerly unvalued instances of feature occurrences, the greater-than sign illustrating c-command relations as well as the lines to indicate a shared feature/ value. 10 However, observe that T’s EPP-feature is not part of the approach to structural case in P&T’s account and is hence again an isolated phenomenon (i.e. the subject still moves to T 0 in several languages), which the authors leave as an open matter (cf. P&T: fn. 22). 178 3 INFERENCE - Nominal Concord & Definiteness Following P&T in the implementation of Brody’s Radical Interpretability, I will henceforth not distinguish interpretable and uninterpretable instances of features by the respective diacritics [ i F] and [ u F]. With the Valuation/ Interpretability Biconditional in place and the rejection of Radical Interpretability, α would find a match for [φ] in β, but Value could not apply since the structure of [φ] on the latter does not entail the structure on α (an instance of the intervention effect (23) c. above). Subsequently, the feature on α would reduce to partial default. With the implementation of the insights from P&T, however, (36) seems to turn out a legal, and moreover forced, outcome of the derivation (see also P&T: fn. 9 on the case of unvalued, here: less valued features and directionality): (36) Nominal Concord & Definiteness I NFERENCE 134 above). Subsequently, the feature on α would reduce to partial default. With the implementation of the insights from P&T, however, 0 seems to turn out a legal, and moreover forced, outcome of the derivation (see also P&T: fn. 9 on the case of unvalued, here: less valued features and directionality): (36) α > β > γ [φ[F[G]] [φ[*F[*G]] [φ[F[*G]] Since [φ] on α is more highly specified, it can value [φ] on β (and subsequently also γ). This amounts to a system that parallels what Danon (2010) calls ‘Relativized φ-completeness’. 11 Agree therein turns out a bidirectional operation, an unwelcome result. P&T (fn. 9) acknowledge the problem concerning Agree between two unvalued instances, which carries over to the directionality in unvalued-valued interactions of a feature. In the framework at hand, the problem applies to every pair of differently complex (but non-distinct) specifications. The authors attribute directionality to recoverability considerations. However, I will take the directionality of Agree to follow neither from recoverability nor the feature configuration of the system. Consider again parts of Bejar’s version of Agree (0, referring to Chomsky (MI: (40) b.) and 0 as well as P&T’s notion in 0): (37) (= Chomsky MI: (40) b., c.) a. D(P) is the sister of P. b. Locality reduces to ‘closest c-command.’ (38) (= Bejar 2003: ch. 2, (58)) Condition on Value G(oal) values P(robe) iff f’(G) entails f(P) (39) (= P&T: (5), [reduced]) Agree (Feature sharing version) i. An unvalued feature F (a probe ) on a head H at syntactic location α (F α ) scans its c-command domain for another instance of F (a goal ) at location β (F β ) with which to agree. The structural preconditions in both accounts take the domain of the probe as the locus of application. Hence, a feature can gain structure through Agree from its sister, while being barred from transmitting it there. The structure of features is hence irrelevant to the directionality of Agree - a welcome result. How, then, can the insights from P&T be formally incorporated into Bejar’s system? That is, when do occurrences of [F] turn into instances of [F]? In a system that strongly utilizes the cut between Match and Value, this is not a trivial question. Turning to P&T’s feature sharing version of Agree, (i.) might be understood as the former, while (ii.), reprinted in 0, constitutes the latter: (40) (= P&T: (5), [reduced]) Agree (Feature sharing version) ii. Replace F α with F β , so that the same feature is present in both locations. Feature sharing would hence constitute a property of Value. However, P&T argue that two unvalued instances of a feature can share one - arguably non-existent - configuration, to be valued later in the course of the derivation: “Agree between two unvalued occurrences of F (F α [ ] and F β [ ]) is far from vacuous, since its output will be a structure that contains only one occurrence of F with two instances” (P&T: 269). In the system under elaboration, this characterization fits the operation Match. There is nothing to copy in unvalued features; still, following P&T, a link between them can be established. In a framework without the category-value partition of features, I propose that Value is the only possible locus of application for 0; however, I want to exclude Agree of two bare root nodes from those configurations resulting in a shared feature (i.e. 0 a. above) since no replacement - or comparison - of any structure (not even identical ones as in configurations of the form [φ[F]] > [φ[F]]) arguably takes place at that level. 12 The establishment of a link between two 11 Danon (2010: 15), however, maintains the valued/ unvalued distinction and hence the outcome of his proposal is different from the one sketched in the main text. 12 Alternatively, it might be possible to modify Match to such an extent that the Entailment Condition applies to the complete goal, therein partly recovering the constraint of φ-completeness for goals (see the beginning of this section) and thereby allowing feature sharing at that point. However, intervention effects would no longer be ] ] ] Since [φ] on α is more highly specified, it can value [φ] on β (and subsequently also γ). This amounts to a system that parallels what Danon (2010) calls ‘Relativized φ-completeness’. 11 Agree therein turns out a bidirectional operation, an unwelcome result. P&T (fn. 9) acknowledge the problem concerning Agree between two unvalued instances, which carries over to the directionality in unvalued-valued interactions of a feature. In the framework at hand, the problem applies to every pair of differently complex (but non-distinct) specifications. The authors attribute directionality to recoverability considerations. However, I will take the directionality of Agree to follow neither from recoverability nor the feature configuration of the system. Consider again parts of Bejar’s version of Agree ((37), referring to Chomsky (MI: (40) b.) and (38) as well as P&T’s notion in (39)): (37) (= Chomsky MI: (40) b., c.) a. D(P) is the sister of P. b. Locality reduces to ‘closest c-command.’ (38) (= Bejar 2003: ch. 2, (58)) Condition on Value G(oal) values P(robe) iff f ’(G) entails f(P) 11 Danon (2010: 15), however, maintains the valued/ unvalued distinction and hence the outcome of his proposal is different from the one sketched in the main text. 3.2 Preliminary Remarks 179 (39) (= P&T: (5), [reduced]) Agree (Feature sharing version) i. An unvalued feature F (a probe ) on a head H at syntactic location α (F α ) scans its c-command domain for another instance of F (a goal ) at location β (F β ) with which to agree. The structural preconditions in both accounts take the domain of the probe as the locus of application. Hence, a feature can gain structure through Agree from its sister, while being barred from transmitting it there. The structure of features is hence irrelevant to the directionality of Agree - a welcome result. How, then, can the insights from P&T be formally incorporated into Bejar’s system? That is, when do occurrences of [F] turn into instances of [F]? In a system that strongly utilizes the cut between Match and Value, this is not a trivial question. Turning to P&T’s feature sharing version of Agree, (i.) might be understood as the former, while (ii.), reprinted in (40), constitutes the latter: (40) (= P&T: (5), [reduced]) Agree (Feature sharing version) ii. Replace F α with F β , so that the same feature is present in both locations. Feature sharing would hence constitute a property of Value. However, P&T argue that two unvalued instances of a feature can share one - arguably non-existent - configuration, to be valued later in the course of the derivation: “Agree between two unvalued occurrences of F (F α [ ] and F β [ ]) is far from vacuous, since its output will be a structure that contains only one occurrence of F with two instances” (P&T: 269). In the system under elaboration, this characterization fits the operation Match. There is nothing to copy in unvalued features; still, following P&T, a link between them can be established. In a framework without the category-value partition of features, I propose that Value is the only possible locus of application for (40); however, I want to exclude Agree of two bare root nodes from those configurations resulting in a shared feature (i.e. (19) a. above) since no replacement - or comparison - of any structure (not even identical ones as in configurations of the form [φ[F]] > [φ[F]]) arguably takes place at that level. 12 The establishment of a link between two feature occurrences hence 12 Alternatively, it might be possible to modify Match to such an extent that the Entailment Condition applies to the complete goal, therein partly recovering the constraint of φ-completeness for goals (see the beginning of this section) and thereby allowing feature sharing at that point. However, intervention effects would no longer be captured in the system just outlined. I will therefore not pursue this line of reasoning further and assume 180 3 INFERENCE - Nominal Concord & Definiteness reduces to a reflex of Value. The system elaborated in this section can be summarized with the conditions on Match and Value given in (41)-(44) below: (41) (= Chomsky MI: (40) b., c.) a. D(P) is the sister of P. b. Locality reduces to ‘closest c-command.’ (42) (= Bejar 2003: ch. 2, (35)-(38)) a. Entailment Condition Match is defined by entailment. b. Probe (F) and Goal (F’) match if Goal (F’) entails Probe (F). c. Match is evaluated at the root. where d. Probe (F) and Goal (F’) match if Goal (F’) entails root Probe (F). (43) (= Bejar 2003: ch. 2, (58)) Condition on Value G(oal) values P(robe) iff f ’(G) entails f(P) (44) Agree (Feature sharing version) Replace f(P) with f ’(G), so that the same feature is present in both locations. 3.2.3 Conclusion Formal Features come in hierarchically structured bundles that make up the syntactically relevant part of an LI (cf. H&R), possibly introduced into the derivation by several functional and lexical heads across languages. 13 Features are hence not primitive entities consisting of a link between a category and a value, rather, categories of (φ-)features are interwoven, but constitute natural classes (cf. H&R: (4) a.); these result from the combinatorial properties of monovalent nodes (Bejar 2003). Therefore, features might act as a unit for the purpose of some syntactic phenomena (cf. the condition of φ-complete goals (i.e. nominal domains) in sentential derivation) while also being able to trigger syntactic reflexes in isolation (e.g. in split-φ-agreement languages). Features are hence only that the exclusion of bare root Agree and the retention of Bejar’s original notion of Match constitutes the convincing alternative option. 13 Cf. Starke (2001), Bejar (2003), Preminger (2011). 3.2 Preliminary Remarks 181 ever relatively specified: (i) w.r.t. the nodes (values) dominating/ depending on them and (ii) w.r.t. the features they (try to) interact with; the latter being what Bejar (2003: 50) refers to as “reducing relativized minimality […] to conditions on match.” The system elaborated in this subchapter takes serious the partition of Match and Value. Syntax has to be able to (i) identify compatible feature structures (Match) and (ii) enable interactions between them (Agree qua Value) to constitute a derivational system. Conditions on the former are less strict than on the latter, As Chomsky (MI: 122) puts it: “Not every matching pair induces Agree.” If the conditions would be equal, the partition would be unmotivated and redundant, if they were reversed, these operations would be completely separated: Match could not pose a precondition for Value. Furthermore, Match without Value is intervention. This is due to Relativized Minimality, barring further search of the probe in its domain once Match applies successfully. In these cases, as well as the exhaustion of the search space of the probe without the identification of a suitable goal, the probing feature reduces to partial default (i.e. root node), but remains active. It is hence able to probe anew, now trivially specified for any specification of said root (as well as the root node itself). This 2 nd cycle Agree may apply to a modified domain: the specifier of the probe, since in bare phrase structure the label of the probe constitutes a complete copy of the head (cf. Bejar 2003, Rezac 2003, including the reduced root node feature) which c-commands its sister. Said position can be filled by IM of a goal, formerly inaccessible due to intervention (Bejar 2003), or by EM of a suitable goal (Rezac 2003). For Bejar, partial default features deactivate after the next application of EM. There exists evidence, however, that deactivation has to occur later, namely in-situ 2 nd cycle Agree (Bošković 2011). The subsequent probing of an identical domain must - as I have argued - be triggered from outside, i.e. the probe of the second cycle must be ‘motivated’ to undertake yet another search for the feature under consideration into the identical search space. Having found a matching goal with which to establish an Agree relation, the features of both elements thereafter constitute a single, shared feature structure, present on two locations (Frampton & Gutmann 2000, P&T), with the respective instances still being able to act as goals for further applications of Agree that turn additional occurrences of the feature into instances of a single feature structure (cf. P&T: 268). Subsequent changes to one of these instances carry over to the remaining ones. Feature sharing is hence (i) the process of copying structure from goal to probe (i.e. Value) and (ii) the establishment of a link between them. With (i) a precondition for (ii), no feature sharing applies with Agree of root nodes. 182 3 INFERENCE - Nominal Concord & Definiteness Interpretability is evaluated in a post-syntactic stage of the derivation (Brody 1997, P&T). Valuation is relativized to feature specification w.r.t. a probing head (Bejar 2003). Hence, there are neither unvalued nor uninterpretable feature structures per se in the syntax proper. The result is a system in which all features attempt to probe at least once, the domain for probing being the sister of the probe: the c-command domain. Unsuccessful probing does not (necessarily) result in a crash of the derivation but (possibly) in reduction of the feature structure to partial default. We therefore arrive at a highly cyclic derivational system, suitable to account for all narrow-syntactic dependencies comprising nominal concord (and definiteness therein). I will turn to these now. Before doing so, I believe that one additional analysis has to be mentioned in the context of the current subchapter. Schoorlemmer’s (2009, cf. ch. 5 on German) approach to agreement patterns of attributive adjectives in various Germanic and Romance languages superficially seems to share various key aspects with the one elaborated here, most obviously regarding the incorporation of a feature sharing approach to Agree as well as the dissociate treatment of the gender values masculine/ neuter from feminine. However, note that even the implementation of these properties already differs significantly from my proposals: On the one hand, Schoorlemmer maintains the category-value bipartition of morpho-syntactic features and hence of the dedicated categories number and gender. Therein, he treats feminine gender on the surface as a reflex of underspecification, i.e. unvaluedness of said category (on a par with the singular value of the number-feature) in Narrow Syntax. Schoorlemmer (2009: 196) is hence able to parallel the feminine and plural paradigms based on their underspecification for gender. Turning to his implementation of Agree, maintaining feature-value pairs has the consequence that it has to apply independently to the various φ-categories as well as to case. Therefore, the existence of a feature α, shared between LIs A and B, does not imply simultaneous sharing of the remaining featural specification of either element. His account to adjectival inflection in German is, then, based on the ability of the adjective to successfully establish a probe-goal relation with strong inflection the result of uniform agreement between the adjective and the nominal head. The introduction of a lexically filled D-head, to which adjectives are taken to adjoin, however, disrupts the relation: Since the determiner merges first, it probes for φ-features and deactivates (for the respective feature only) in case a value is received. Thereby, the adjective, successively merged to the maximal projection, remains unvalued for a subpart of its features. Through the incorporation of various Vocabulary insertion rules in post-syntactic Morphology, Schoorlemmer derives the morphological forms of the weak adjectival inflectional pattern from this impoverished featural content. 3.3 Splitting Nominal Concord 183 Additionally, he locates both the determiner and the indefinite article, utilized to evoke mixed adjectival inflection, in the head of DP (at least in German) and hence does also not include the various categories of semi-lexical items employed here, their featural set-up or possible co-occurrences discussed at lengh below. Even more so, he also desists from the analysis of constructions beyond the nominal domain. I will thus refrain from a thorough discussion of the approach in what follows but acknowledge the independent arrival of both accounts at the special status of feminine gender as a hint to the correctness of the underlying rationale. 3.3 Splitting Nominal Concord As demonstrated in the last subchapter, nominal concord has been taken to pose problems for a derivation comprised of successive applications of Agree. In this subchapter I want to counter that claim and show that inflectional idiosyncrasies in nominal concord in German follow from the successive cyclic character of the derivation, i.e. the selectional restrictions of the SLIs involved and the way in which they combine and interact in Narrow Syntax. However, as has already been mentioned above, the analysis to follow will have to acknowledge the insights of former studies, stating that certain parts of the phenomenon, to be made precise below, cannot be handled by the syntax proper without modifications to the system that are otherwise (i.e. outside of the scope of the investigation) unmotivated and hence unwelcome. I do not take the proposal of such a joint effect to pose a weakening of the analysis, though. The syntactic system to be elaborated in this chapter will still be both uniform and independently motivated, as will be those domains of the analysis that are ‘sourced out’ to other components of the grammar. This section will motivate said splits in the phenomenon and take a closer look at those effects taken to emerge from parts of FL outside of the syntax proper. The rest of chapter 3.2 will then be devoted to nominal concord phenomena that are taken to be part of NS, i.e. the derivation of inflectional patterns on SLIs. I will review the basic paradigms of the participating elements and extract their selectional properties therefrom (chapter 3.2.1) before I will take the interactions between the nominal and the clausal domain into account (i.e. the interaction of φ and case, chapter 3.2.2) and show that my analysis is able to span the gap that Danon (2010) referred to as ‘encapsulation’ between research on the DP and the CP domain. Thenceforward, I will extend the paradigm to account for inflectional interactions between multiple SLIs (chapter 3.2.3). Further refinements will follow in the subsequent sections, to wit, I will demonstrate how the system is able to account for the 184 3 INFERENCE - Nominal Concord & Definiteness inflectional and linearizational properties of floating constructions before turning to those parts of nominal concord, split from syntax hereafter. I believe that, therein, a holistic and coherent grasp of the phenomena of nominal concord, adjectival inflection as well as morphological definiteness in German will have been worked out by the end of the current chapter. a. Narrow Syntax In chapter 3.1 I elaborated on the operations that enable interactions of LIs in Narrow Syntax (chapter 3.1.2) as well as the elements driving them (chapter 3.1.1). As I will show, all inflectional idiosyncrasies of semi-lexical items in the nominal domain in German will follow from the interplay of these refined notions: Overt inflections are reflexes of the manipulation of their featural set-up in the course of the derivation. We do not have to assume an unmotivated split between members of certain categories, or worse, a single category: Overt inflection on SLIs (i.e. Q 0 s, Poss PRO0 , D 0 ) is born in Narrow Syntax. One additional comment, however, is in order before reviewing the full set of SLI inflectional paradigms below. Recall from chapter 2.2 that I began my analysis on the phasal set-up of the nominal domain in German with Richards’ proposal of the phase-plus-non-phase-complement (P-N-P-N) configuration constituting the most minimal phasal set-up. As elaborated above, the derivational life of valued uninterpretable features has to be maximally short to ensure a licit derivation at Σ/ SEM; Value-TRANSFER simultaneity follows from this. However, the phase impenetrability condition (PIC) renders the phase head (and its specifier, i.e. the edge) a part of the next higher phasal cycle; uninterpretable valued features of the phase head hence survive and should crash the derivation. In OP, the introduction of formal features happens at EM of each strong phase head. These were subsequently ‘handed down’ to the complement by feature inheritance. Richards proposes that this inheritability of features from phases to their non-phasal complements is the only way to guarantee that those features do not survive and carry over to the next cycle, in his (Richards 2007: 569) words: “(1) (= interpretability) and (2) (= impenetrability) together entail (3) (= inheritability).” Richards’ first premise (interpretability), however, was negated in chapter 3.1.2 with the implementation of Brody’s (1997) Radical Interpretability. Hence, it is worth reviewing the status of feature inheritance in the system at hand, which does no longer follow necessarily. In a framework incorporating contextual phasal status following Bošković (2014), it seems curious that a head α might transfer features to its complement β when carrying phasal status, while β comes equipped with those exact same features in the absence of α. I therefore exclude feature inheritance from the 3.3 Splitting Nominal Concord 185 system at hand, since, as I have shown above, nothing hinges on this: (S)LIs enter the derivation equipped with their formal features, independently of their phasal status. Observe that the analysis in chapter 2.2 is not weakened by this decision since Richards’ observations were utilized to introduce the most minimal phasal set-up, later strengthened by Bošković’s claims on different grounds. b. Morphology Apart from the core of the phenomenon, there are good reasons to assume that one specific part of nominal concord is derived on a different plane, namely in post-syntactic Morphology (cf. Chomsky MP: 229, cf. also ch. 2.3.3 above). Fanselow (2013) argues on the basis of paradigms of discontinuous nominal domains like those depicted in (45) that morphological form can be taken to be evaluated after movement: (45) (= Fanselow 2013: (5)) a. Rotem polnischen Wein vertrauen wir nicht. red.sg.dat.strong Polish.sg.dat.weak wine trust we not ‘We do not trust red Polish wine.’ b. Polnischem Wein vertrauen wir nur rotem. Polish.dat.sg.strong wine trust we only red.dat.sg.strong ‘We only trust red Polish wine.’ Concentrating on the morphological form of the second adjective in (45) a., part of the dislocated nominal structure in (45) b., we observe a change in adjectival inflection, dependent on its concatenation with another adjective dominating it. Let me extend the paradigm with (46), dispensing with topical movement of the object NP altogether (the gloss is modelled after Fanselow 2013): (46) Wir vertrauen polnischem Wein. we trust Polish.sg.dat.strong wine ‘We trust Polish wine.’ Here, the adjective again bears strong inflection. It is questionable if the changes in morphological realization can be traced back to processes of a cyclic derivation. In what follows, I want to show why. Let us - for the sake of argument - posit a formal feature [F] on adjectives, solely responsible for the strong / weak inflectional partition. Let [F] take Boolean values, i.e. true/ false, 1/ 0, +/ -, hence be bivalent; with the value [- F] resulting in weak adjectival inflection, [+ F] in its strong counterpart. Deriving (45) b. in a cyclic manner from an underlyingly 186 3 INFERENCE - Nominal Concord & Definiteness continuous nominal domain would necessarily incorporate the derivational stages illustrated in (47) at some point (abstracting away from case): (47) a. polnischem Wein (= (46)) Polish .[+F] wine b. rotem polnischen Wein (= (45) a.) red .[+F] Polish .[-F] wine c. Polnischem Wein vertrauen wir nur rotem. (= (45) b.) Polish .[+F] wine trust we only red .[+F] The value for [F] on the embedded adjective would hence have to change twice to accommodate for the variation. Additionally, it is an open question as to what syntactic processes might be able to trigger the featural change in the nominal domain ([+ F]→[- F]) as well as in the clausal domain ([- F]→[+ F]). The narrow-syntactic approach seems even more dubious under the current treatment of φ-features as complex bundles; observe the reiteration of derivational stages from (47) below, with [F] the feature structure resulting in weak , [F[G]] resulting in strong inflection: (48) a. Adj1.[F[G]] N (= (46)) b. Adj2.[F[G]] Adj1.[F] N (= (45) a.) c. Adj1.[F[G]] N … Adj2.[F[G]] (= (45) b.) Adj1’s featural content reduces under c-command with Adj2, then extends with subsequent movement into the clausal domain. The Agree relation developed above clearly cannot account for the interaction in (48) b. Observe that the same oddity also carries over to the reversion of the levels of complexity for strong and weak adjective inflection (i.e. strong = [F], weak = [F[G]]): The derivational steps leading to the snapshots in (48) b. and c. would therefore remain completely unexplained. Fanselow concludes that morphological form is derived in a post-IM stage of the derivation. I will propose below (chapter 3.2.4) that this process is not part of Narrow Syntax at all, but derived in Morphology by Impoverishment and Late Insertion. I will henceforth refer to the phenomenon of strong / weak adjectival inflectional dichotomy as ‘Adjectival Impoverishment’ (AI), to be made precise below, and take its application to be a property of (classes of) LIs. That section 3.3 Splitting Nominal Concord 187 will rest heavily on the analysis as well as the underlying rationale of Sauerland (1996; see also Roehrs 2009). c. Morpho-Phonology The split of the complex phenomenon of nominal concord to be handled in separate domains of FL was advocated on the basis of the lexical category of the elements involved above: Nominal concord on all SLIs is derived in Narrow Syntax with Adjectival Impoverishment applying later in post-syntactic Morphology, hence retaining a well-regulated, ordered system. However, when focusing on complex concatenations of the elements involved (ch. 3.2.3), it will become obvious that at least one inflectional idiosyncrasy can easily be analyzed in morpho-phonological terms, while it posits an unexplained gap in the inflectional paradigm in the arguably highly ordered domain of Narrow Syntax, as stated above. The phenomenon will still be highly regular, once this diverging viewpoint is implemented. I will postpone a detailed presentation of the analysis until the relevant structural configuration is encountered in chapter 3.2.3 and proceed by introducing and categorizing the basic pattern of inflectional properties of SLIs in the nominal domain in German in the following section. 3.3.1 The Basic Pattern of SLI Inflection and Nominal Concord I will commence my analysis by focusing on nominal SLI inflection in isolation, i.e. ignoring case as well as the co-occurrence of multiple SLIs for now. Recall that I hinted at the weak quantifier viel probing for NumP in German in chapter 2.3. This conclusion was of course premature; such observations will, however, build the basis of the analysis in the current subchapter. The phenomenon at hand is much more complex but, luckily, regular. In fact, I will argue that there are only two probing φ-feature configurations found on the various SLIs in the nominal domain in German, namely [φ] and [φ[f]]. What might at first glance appear to constitute a simple paraphrase of the aforementioned claim from chapter 2.3 actually entails several key properties of the derivational interactions found in the nominal domain in German. Observe the table in (49), a transposition of Bejar’s (2003: 91f.) own categorizations of personand numberfeature configurations and their respective effects across various languages. Recall from chapter 3.1.1 that all elements in the nominal domain are taken to be at least relatively underspecified (i.e. minimally bear the root node [φ]). The interactions proposed by the system are therefore: 188 3 INFERENCE - Nominal Concord & Definiteness (49) A (Underspecification of interpretable features) B (Specification of probes) Relative underspecification only [φ] • [m], [n], [f] and [pl] are interveners • [m], [n], [f] and [pl] value on 1 st cycle • there are no 2 nd cycle effects [φ[f]] • [m], [n], [f] and [pl] are interveners • [f] and [pl] value on first cycle • [m], [n] trigger 2 nd cycle (i.e. partial default here) • [m], [n], [f] and [pl] value on 2 nd cycle With (49) in place, let us again turn to the nominative paradigm in German. Due to the Observation from chapter 2.3 mentioned above, I will begin my analysis in focusing on linearizations of Q W with masculine, feminine and neuter singular, as well as plural HNs. (50) a. viel(*-er) Wein much wine masc b. viel(*-es) Bier much beer neut c. viel(-e) Milch much milk fem d. viel-e Menschen many people pl Inflection on the quantifier is out with masculine and neuter head nouns (while adjectival suffixes in concord with these nominals do exist), optional with feminine nominals and obligatory when Q W is concatenated with plural HNs. For now, I will postpone an analysis on the source of optionality of inflection on Q W and simply presume that if SLIs are able to display a realization of the feature category [F] on the surface, they have acquired it in the course of the derivation. Incorporating the premise of probing featural content of Q W from the beginning of this section, let me sketch the derivational snapshots when Agree takes place between these items in (50). What I want to propose is that the weak quantifier bears, and probes for, the more highly specified φ-structure from (49) above, viz. [φ[f]]. I will - once more - apply Frampton & Gutmann’s notation, established at the end of ch. 3.1.2 (cf. fn. 9): 3.3 Splitting Nominal Concord 189 (51) Nominal Concord & Definiteness I NFERENCE 140 (51) a. viel > Wein Q W N i. [φ[f]]  [φ[m]] ii. [φ] [φ[m]] Match:  Value:  b. viel > Bier Q W N i. [φ[f]]  [φ[n]] ii. [φ] [φ[n]] Match:  Value:  c. viel(-e) > Milch Q W Num [φ[f]] [φ[f]] Match:  Value:  d. viel-e > Menschen Q W Num [φ[f[*pl]]] [φ[f[pl]]] Match:  Value:  As predicted in 0, masculine and neuter HNs are compatible with Q W by Match (i.e. they share the same root), but not by Value. Therefore, no link is established between probe and goal. Subsequently, the feature structure on the probing element reduces to partial default and hence results in a surface structure without inflection. 14 In Narrow Syntax, however, the reduced feature stays active. With feminine HNs, though, Value succeeds since the two φ-bundles are identical (see the exact formulation of the Entailment Condition from Bejar 2003 and the discussion on identical values in chapter 0) and the two instances of the feature thus turn into two occurrences. With plural nominals in 0 d., the probe furthermore gains additional structure by Value, being specified for [f[pl]] after copying from the goal (hence the asterisk on the newly acquired value, cf. fn. 9 above). Before moving on to the restrictions of the remaining SLIs, I want to make one additional point concerning the formulation of Match from Bejar (2003). The notions employed so far rest on the assumption that Match and Value operate on the basis of an identical principle (i.e. the Entailment Condition), but apply to diverging structural configurations, namely Match on the root node and Value to the full feature structure of the goal w.r.t. that of the probe. In the standard minimalist framework of Chomsky’s DbP and OP, maintaining the category-value partition on features, in contrast, Match is taken to apply to ‘non-distinct’ feature configurations. Taking non-distinctness to be evaluated w.r.t. the full structure of the goal (and hence against the full structure of the probe), the configurations resulting in successful applications of Match but unsuccessful instances of Value in 0 above (i.e. those incorporating masculine and neuter HNs in a. and b.) would arrive at different outcomes of these operations under ‘Chomskyan’ Match. I illustrate this with a masculine nominal, parallel to 0 a. above: (52) viel > Wein Q W N i. [φ[f]]  [φ[m]] ø ii. [φ] [φ[m]] Match:  Value:  In 0 i. viel probes for [φ[f]], but [φ[m]] does not match and is hence overlooked. In what follows, no non-distinct instance of φ-bundles can be found in the domain of the probe. Observe, however, that the result of this unsuccessful probing is the 14 Bejar (2003: ch. 2, fn. 10) actually refrains from equating zero-morphology with underspecified/ default realizations of LIs. She notes that a blocking pattern also has to be observable. I will show below that those intervention effects are exactly what we find in the nominal domain in German, once multiple probing SLIs are introduced in chapter 0, and that zero-morphology and default values hence coincide in the phenomenon under investigation here. As predicted in (49), masculine and neuter HNs are compatible with Q W by Match (i.e. they share the same root), but not by Value. Therefore, no link is established between probe and goal. Subsequently, the feature structure on the probing element reduces to partial default and hence results in a surface structure without inflection. 14 In Narrow Syntax, however, the reduced feature stays active. With feminine HNs, though, Value succeeds since the two φ-bundles are identical (see the exact formulation of the Entailment Condition from Bejar 2003 and the discussion on identical values in chapter 3.1.2) and the two instances of the feature thus turn into two occurrences. With plural nominals in (51) d., the probe furthermore gains additional structure by Value, being specified for [f[pl]] after copying from the goal (hence the asterisk on the newly acquired value, cf. fn. 9 above). Before moving on to the restrictions of the remaining SLIs, I want to make one additional point concerning the formulation of Match from Bejar (2003). 14 Bejar (2003: ch. 2, fn. 10) actually refrains from equating zero-morphology with underspecified/ default realizations of LIs. She notes that a blocking pattern also has to be observable. I will show below that those intervention effects are exactly what we find in the nominal domain in German, once multiple probing SLIs are introduced in chapter 3.2.3, and that zero-morphology and default values hence coincide in the phenomenon under investigation here. 190 3 INFERENCE - Nominal Concord & Definiteness The notions employed so far rest on the assumption that Match and Value operate on the basis of an identical principle (i.e. the Entailment Condition), but apply to diverging structural configurations, namely Match on the root node and Value to the full feature structure of the goal w.r.t. that of the probe. In the standard minimalist framework of Chomsky’s DbP and OP, maintaining the category-value partition on features, in contrast, Match is taken to apply to ‘non-distinct’ feature configurations. Taking non-distinctness to be evaluated w.r.t. the full structure of the goal (and hence against the full structure of the probe), the configurations resulting in successful applications of Match but unsuccessful instances of Value in (51) above (i.e. those incorporating masculine and neuter HNs in a. and b.) would arrive at different outcomes of these operations under ‘Chomskyan’ Match. I illustrate this with a masculine nominal, parallel to (51) a. above: (52) Nominal Concord & Definiteness I NFERENCE 140 (51) a. viel > Wein Q W N i. [φ[f]]  [φ[m]] ii. [φ] [φ[m]] Match:  Value:  b. viel > Bier Q W N i. [φ[f]]  [φ[n]] ii. [φ] [φ[n]] Match:  Value:  c. viel(-e) > Milch Q W Num [φ[f]] [φ[f]] Match:  Value:  d. viel-e > Menschen Q W Num [φ[f[*pl]]] [φ[f[pl]]] Match:  Value:  As predicted in 0, masculine and neuter HNs are compatible with Q W by Match (i.e. they share the same root), but not by Value. Therefore, no link is established between probe and goal. Subsequently, the feature structure on the probing element reduces to partial default and hence results in a surface structure without inflection. 14 In Narrow Syntax, however, the reduced feature stays active. With feminine HNs, though, Value succeeds since the two φ-bundles are identical (see the exact formulation of the Entailment Condition from Bejar 2003 and the discussion on identical values in chapter 0) and the two instances of the feature thus turn into two occurrences. With plural nominals in 0 d., the probe furthermore gains additional structure by Value, being specified for [f[pl]] after copying from the goal (hence the asterisk on the newly acquired value, cf. fn. 9 above). Before moving on to the restrictions of the remaining SLIs, I want to make one additional point concerning the formulation of Match from Bejar (2003). The notions employed so far rest on the assumption that Match and Value operate on the basis of an identical principle (i.e. the Entailment Condition), but apply to diverging structural configurations, namely Match on the root node and Value to the full feature structure of the goal w.r.t. that of the probe. In the standard minimalist framework of Chomsky’s DbP and OP, maintaining the category-value partition on features, in contrast, Match is taken to apply to ‘non-distinct’ feature configurations. Taking non-distinctness to be evaluated w.r.t. the full structure of the goal (and hence against the full structure of the probe), the configurations resulting in successful applications of Match but unsuccessful instances of Value in 0 above (i.e. those incorporating masculine and neuter HNs in a. and b.) would arrive at different outcomes of these operations under ‘Chomskyan’ Match. I illustrate this with a masculine nominal, parallel to 0 a. above: (52) viel > Wein Q W N i. [φ[f]]  [φ[m]] ø ii. [φ] [φ[m]] Match:  Value:  In 0 i. viel probes for [φ[f]], but [φ[m]] does not match and is hence overlooked. In what follows, no non-distinct instance of φ-bundles can be found in the domain of the probe. Observe, however, that the result of this unsuccessful probing is the 14 Bejar (2003: ch. 2, fn. 10) actually refrains from equating zero-morphology with underspecified/ default realizations of LIs. She notes that a blocking pattern also has to be observable. I will show below that those intervention effects are exactly what we find in the nominal domain in German, once multiple probing SLIs are introduced in chapter 0, and that zero-morphology and default values hence coincide in the phenomenon under investigation here. In (52) i. viel probes for [φ[f]], but [φ[m]] does not match and is hence overlooked. In what follows, no non-distinct instance of φ-bundles can be found in the domain of the probe. Observe, however, that the result of this unsuccessful probing is the subsequent reduction of Q W ’s φ-bundle to partial default in ii.; as Bejar (2003: 86, [emphasis added, MB]) puts it: “Failure to match […] may result in match with a lower NP, or last resort default agreement , or cyclic expansion of the search domain […]. These three outcomes are also potential consequences of match without value […].” For the purpose of the analysis in this section, the two notions are hence identical w.r.t. their outcome even though they differ in the steps leading there. Observe that it is only the first Merger of an SLI bearing the more complex of the two featural bundles, proposed above (i.e. [φ[f]] vs. [φ]), with an incompatible φ-bundle on the goal HN in which diverging outcomes are predicted by the two versions of Match. Every subsequent Merger of items bearing either [φ[f]] or [φ] will result in successful Match under both notational variants. I will have to say more about Match and the various co-occurrence configurations in the subchapter on complex nominal domains below. For now, it suffices to note that the version of Match from Bejar (2003) and Chomsky (DbP, OP) are identical w.r.t. the outcome. There is, however, a third notion to be mentioned in this context, namely Anti-Identity as employed by Starke (2001) in his analysis of island effects, 3.3 Splitting Nominal Concord 191 mentioned in the discussion of feature geometries in 3.1.1 above. His principle is given in (53): (53) (= Starke 2001: (12)) Anti-Identity * α . . . γ . . . α if class(α) = class(γ) and γ intervenes between the two occurrences of α As discussed above, Starke implements a feature tree vastly diverging (in underlying rationale as well as in category of the elements involved) from H&R’s original geometry. Bejar (p. 55f.) nevertheless discusses the application of the Anti-Identity Principle to Match in her version of the person-geometry. Just as I translated Chomsky’s non-distinctness to apply to the full featural structure of the goal, Bejar does so for the principle in (53), which therefore allows Match only to apply to goals categorically equally, but structurally more highly, specified than the respective probe. Again, Bejar’s system does not require the incorporation of configurations with categorically distinct probes and goals (see the discussion in ch. 3.1.2 above). Hence, the configurations under discussion in her account reduce to (54): (54) (= Bejar 2003: ch. 2, (42)) probe goal match a. [π[ part ]] [π]  b. [π[ part ]] [π[ part ]]  c. [π[ part ]] [π[ part [ adressee ]]  Taking the additional case of two differently, but equally complex, specified bundles on probe and goal, Anti-Identity evaluated w.r.t. the complete goal obviously results in an unsuccessful application of Match: (55) probe goal match [φ[F]] [φ[G]]  The outcomes of all three implementations of Match are hence identical for the simple cases of concatenation of SLI and HN under consideration in this chapter so far, with the underlying applications of Match in Starke’s and Chomsky’s account patterning together against Bejar’s proposals: Match of the probe Q W with the differently complex, but categorically equally, specified goal HN is taken to fail in the former, but succeed in the latter. It seems, then, that these accounts to Match only differ with respect to their locus of application (i.e. the 192 3 INFERENCE - Nominal Concord & Definiteness root vs. full geometry of the goal). I will for now follow Bejar without comment but note that the topic has to be addressed again once the full paradigm of cooccurrences of SLIs in the nominal domain and their respective interactions is presented below. Let us, then, continue to review the basic selectional restrictions of the remaining SLIs. I will move up the nominal spine from Q W , discussed above, to the next higher head, the first of the context sensitive phase heads of the highest cycle: Poss PRO0 . Applying the same test as before, I will concatenate the possessive pronoun with three mass nominals and a plural HN: (56) a. mein(*-er) Wein my wine masc b. mein(*-es) Bier my beer neut c. mein-e Milch my milk fem d. mein-e Menschen my people pl Not much has to be added to the description of Poss PRO that does not readily carry over from the weak quantifier. Observe that the optionality of inflection in the concatenation with feminine HNs is not observed in (56), making the possessive pronoun a prime candidate for the selectional feature configuration [φ[f]]. I present once again the derivational snapshots at the time of φ-probing of Poss PRO0 . (57) (57) a. mein > Wein Poss PRO N i. [φ[f]]  [φ[m]] ii. [φ] [φ[m]] Match:  Value:  b. mein > Bier Poss PRO N i. [φ[f]]  [φ[n]] ii. [φ] [φ[n]] Match:  Value:  3.3 Splitting Nominal Concord 193 c. mein-e > Milch Poss PRO Num [φ[f]] [φ[f]] Match:  Value:  d. mein-e > Menschen Poss PRO Num [φ[f[*pl]]] [φ[f[pl]]] Match:  Value:  Climbing the nominal spine further, we arrive at D 0 , the invariable phase head of the highest phasal-cycle. I will propose without comment that D is the first proponent of the second, less restrictive selectional feature configuration [φ]. The reason will become clear immediately once the appropriate data as well as the proposed underlying derivational snapshots are presented in parallel to those laid out above: (58) Nominal Concord & Definiteness I NFERENCE 142 my beer neut c. mein-e Milch my milk fem d. mein-e Menschen my people pl Not much has to be added to the description of Poss PRO that does not readily carry over from the weak quantifier. Observe that the optionality of inflection in the concatenation with feminine HNs is not observed in 0, making the possessive pronoun a prime candidate for the selectional feature configuration [φ[f]]. I present once again the derivational snapshots at the time of φ-probing of Poss PRO0 . (57) a. mein > Wein Poss PRO N i. [φ[f]]  [φ[m]] ii. [φ] [φ[m]] Match:  Value:  b. mein > Bier Poss PRO N i. [φ[f]]  [φ[n]] ii. [φ] [φ[n]] Match:  Value:  c. mein-e > Milch Poss PRO Num [φ[f]] [φ[f]] Match:  Value:  d. mein-e > Menschen Poss PRO Num [φ[f[*pl]]] [φ[f[pl]]] Match:  Value:  Climbing the nominal spine further, we arrive at D 0 , the invariable phase head of the highest phasal-cycle. I will propose without comment that D is the first proponent of the second, less restrictive selectional feature configuration [φ]. The reason will become clear immediately once the appropriate data as well as the proposed underlying derivational snapshots are presented in parallel to those laid out above: (58) a. der/ dies-er > Wein D N [φ[*m]] [φ[m]] Match:  Value:  b. das/ dies-es > Bier D N [φ[*n]] [φ[n]] Match:  Value:  c. die/ dies-e > Milch D Num [φ[*f]] [φ[f]] Match:  Value:  d. die/ dies-e > Menschen D Num [φ[*f[pl]]] [φ[f[pl]]] Match:  Value:  With the determiner/ demonstrative, then, no recourse to default forms is necessary since D is trivially specified to Agree with every instantiation of φ in the nominal domain in German. With the determiner/ demonstrative, then, no recourse to default forms is necessary since D is trivially specified to Agree with every instantiation of φ in the nominal domain in German. The last element under consideration, viz. the highest SLI in the nominal hierarchy of projections and least likely candidate for phasehood in the higher cycle, is the strong quantifier Q S . It is the only SLI of which a categorical counterpart 194 3 INFERENCE - Nominal Concord & Definiteness has been identified in the left periphery of the nominal domain. Recall that the weak quantifier was analyzed as an instance of the selectional pattern [φ[f]] above. The strong quantifier, however, seems not to pattern with its categorical equivalent in this respect but with the adjacent category of D: 15 (59) Nominal Concord & Definiteness I NFERENCE 143 The last element under consideration, viz. the highest SLI in the nominal hierarchy of projections and least likely candidate for phasehood in the higher cycle, is the strong quantifier Q S . It is the only SLI of which a categorical counterpart has been identified in the left periphery of the nominal domain. Recall that the weak quantifier was analyzed as an instance of the selectional pattern [φ[f]] above. The strong quantifier, however, seems not to pattern with its categorical equivalent in this respect but with the adjacent category of D: 15 (59) a. all-er > Wein Q S N [φ[*m]] [φ[m]] Match:  Value:  b. all-es > Bier Q S N [φ[*n]] [φ[n]] Match:  Value:  c. all-e > Milch Q S Num [φ[*f]] [φ[f]] Match:  Value:  d. all-e > Menschen Q S Num [φ[*f[pl]]] [φ[f[pl]]] Match:  Value:  Even though the SLI in 0 constitutes an instance of the category of Q, it overtly agrees with all number/ genderspecifications, which amounts to its φ-featural specification as probing for bare [φ]. Moreover, contrary to its categorical counterpart, no optional instances of inflection can be observed in this pattern of simple Q S P-HN concatenations. I will, then, have to stay agnostic as to the existence of default (reduced) forms on the strong quantifier as well as the category D since their selectional restrictions are trivially met in all linearizations in 0 above. The analysis of simple SLI-N concatenations in German returned a coherent bipartition of the selectional φ-probes in the left periphery of the noun. The lower categories Q W and Poss PRO pattern together in their selectional configuration [φ[f]], satisfiable by either feminine or plural values on Num, while the higher phrase heads D and Q S observe the loose, trivial specification [φ] which is - as I have argued above - met by all instantiations of nominal heads as well as number phrases in German. The former categories, however, reduce to partial default and surface without inflection accordingly, once Match succeeds but Value fails in the course of the derivation (i.e. with masculine and neuter head nouns). Moreover, a further split between the semi-lexical items Q W and Q S was identified in their selectional, in addition to their linearizational, properties. The patterns of probing φ-bundles on SLIs are compiled in 0 below. Additionally, the ability to surface with partial default agreement has also been included. (60) Q S D Poss PRO Q W Probing φ [φ] [φ] [φ[f]] [φ[f]] Default Agreement ? ?   The aim of a theory of the derivation of morphological form in the nominal domain, however, is hardly accomplished by the collocation of simple statements in 0 which, in addition, could have been easily stated in the terms of standard Minimalism without the derivational apparatus elaborated at length in chapter 0 above. The findings of this subsection therefore can only constitute a starting point (though a promising one) for the analysis of the derivation(s) of morphological form in the nominal domain. As the proof of the pudding is in the eating, the benchmark of the theory lies in its accountability for the full paradigm and the particular variation of the forms incorporated therein. For an analysis of nominal concord, this foremost applies to the varying adjectival inflection, the basis of which I will turn to next, to wit, the 15 It will be shown, however, that the etymological origin of Q S still leaves its footprints with respect to AI, as elaborated below in the context of more complex nominal configurations and adjectival forms. Even though the SLI in (59) constitutes an instance of the category of Q, it overtly agrees with all number/ gender-specifications, which amounts to its φ-featural specification as probing for bare [φ]. Moreover, contrary to its categorical counterpart, no optional instances of inflection can be observed in this pattern of simple Q S P-HN concatenations. I will, then, have to stay agnostic as to the existence of default (reduced) forms on the strong quantifier as well as the category D since their selectional restrictions are trivially met in all linearizations in (59) above. The analysis of simple SLI-N concatenations in German returned a coherent bipartition of the selectional φ-probes in the left periphery of the noun. The lower categories Q W and Poss PRO pattern together in their selectional configuration [φ[f]], satisfiable by either feminine or plural values on Num, while the higher phrase heads D and Q S observe the loose, trivial specification [φ] which is - as I have argued above - met by all instantiations of nominal heads as well 15 It will be shown, however, that the etymological origin of Q S still leaves its footprints with respect to AI, as elaborated below in the context of more complex nominal configurations and adjectival forms. 3.3 Splitting Nominal Concord 195 as number phrases in German. The former categories, however, reduce to partial default and surface without inflection accordingly, once Match succeeds but Value fails in the course of the derivation (i.e. with masculine and neuter head nouns). Moreover, a further split between the semi-lexical items Q W and Q S was identified in their selectional, in addition to their linearizational, properties. The patterns of probing φ-bundles on SLIs are compiled in (60) below. Additionally, the ability to surface with partial default agreement has also been included. (60) Q S D Poss PRO Q W Probing φ [φ] [φ] [φ[f]] [φ[f]] Default Agreement ? ?   The aim of a theory of the derivation of morphological form in the nominal domain, however, is hardly accomplished by the collocation of simple statements in (60) which, in addition, could have been easily stated in the terms of standard Minimalism without the derivational apparatus elaborated at length in chapter 3.1 above. The findings of this subsection therefore can only constitute a starting point (though a promising one) for the analysis of the derivation(s) of morphological form in the nominal domain. As the proof of the pudding is in the eating, the benchmark of the theory lies in its accountability for the full paradigm and the particular variation of the forms incorporated therein. For an analysis of nominal concord, this foremost applies to the varying adjectival inflection, the basis of which I will turn to next, to wit, the interaction of the nominal and the clausal domain in the reciprocity of φand case-features. While the actual derivation of adjectival forms will be postponed until after the properties of SLIs have been laid out and examined in detail, I am positive that the processes in Morphology, taking part in the derivation of AI, readily fall into place once the preconditions have been dissected in full in the two subchapters to follow. 3.3.2 Case-Assignment and Preconditions for the MIXEd Adjective Inflection Danon (2011), reviewing the state of research on DP-internal derivations, identifies a gap between the generative research traditions on nominal and clausal syntax in such a way that each of these objects of investigation is standardly analyzed without close attention to the structural and featural composition of the other. Danon refers to Chomsky’s (DbP: fn. 8) treatment of NP as a simple yet φ-complete nominal domain in his analysis of clausal derivations as a prime example. 196 3 INFERENCE - Nominal Concord & Definiteness We have seen above that N 0 cannot always be assumed to bear a full set of φ-features, an objection that becomes even more pressing in the light of the discussion on the location of person-features in chapter 3.1.1 but not under investigation in this analysis. Additionally, as has been exemplified in chapter 2.3.1 and likewise included in Danon’s critique, dedicated functional heads of vast featural divergence have been proposed to partake in the composition of the nominal spine, with the NumP from chapter 2.3 just one of the many options proposed in the literature (cf. Alexiadou et al. 2007: Part II, chapter 3 for an extensive overview). For Danon, what follows from this rationale under a standard minimalist feature-theory constitutes a bug in the system, a questionable assumption tacitly shared by all approaches to sentential derivation proceeding along the lines just sketched, i.e. the ‘DP Encapsulation Assumption’; paraphrased into a derivational conundrum termed ‘Danon’s Puzzle’ in Norris (2014). I give both notions below: (61) (= Danon 2011: 3) The DP Encapsulation Assumption In every non-expletive noun-phrase, the highest head bearing φ-features is φ-complete. (62) (= Norris 2014: (241)) Danon’s Puzzle v 0 [i.e. heads in the clausal domain, MB] looks for φ-features on DP, but φ-features originate elsewhere. Under standard minimalist assumptions about the nature of Agree, (61) is a necessary, though highly questionable, notion. 16 Danon demonstrates this in the simple summary of the respective loci of the discrete φ-features in the nominal domain, taking HN to bear interpretable genderand number-features, D to bear interpretable person as well as uninterpretable, thus deleted, gender and number when the nominal domain is inserted into the clausal spine. He subsequently pursues several options to circumvent the assumption stated in (61), setteling on a feature sharing approach to φ-features, parallel to the one incorporated into the theory at hand above, as the favorable choice. Recall from chapter 3.1.2 that valued features that are marked for deletion can still act as goals for P&T (p. 268), however, these authors ultimately also refer to feature deletion applying phase-final in their system (p. 284, in line with stand- 16 As for (62), however, the paraphrase shifts the focus of the matter from φ-completeness to the locus of φ-features; I believe that Danon’s Puzzle will resolve once a more finegrained approach to featural CP-DP interdependencies is elaborated, to which this section is devoted. 3.3 Splitting Nominal Concord 197 ard minimalist assumptions) to all uninterpretable instances of a shared feature (in line with Brody’s Radical Interpretability). P&T elaborate on this deletion operation in the light of Brody’s insights: Since Narrow Syntax does not have access to the status of interpretability of a feature, contrary to the semantic interface SEM, it is problematic to attribute identification as well as subsequent deletion of uninterpretable instances to any component of C HL applying previous to the latter. P&T (p. 290), referring to Fox (1995, 1999), hence speculate on a bidirectional relation between NS and Σ/ SEM to ‘source out’ the identification and deletion of uninterpretable instances and subsequently hand the structure back into syntax in a cyclic manner applying at each phase-level. I will not go into detail on the exact locus of deletion here but focus on its timing instead. Danon’s argument for the preference of feature sharing accounts in the treatment of CP-DP interactions rests on the accessibility of φ-features in the nominal domain for probes beyond the DP-level, i.e. heads of the clausal domain. Recall, however, from chapter 2.2 above that the former domain is taken to constitute a rigid set of context sensitive phase-plus-non-phase configurations, parallel to the sentential domain. Coupled with P&T’s claims, we hence arrive at a system whose properties concerning the accessibility of φ-features are parallel to the standard model. No head in a phasal nominal domain can be taken to be φ-complete at Merge with the clausal domain, at least under the strict version of the PIC, given below: (63) (= Chomsky MI: (21)) Phase-Impenetrability Condition (PIC, strict version) In phase α with head H, the domain of H is not accessible to operations outside α, only H and its edge are accessible to such operations. The weaker version of the PIC from Chomsky (DbP et seq.), unfortunately, does not serve the purpose at hand any better if one assumes that the evaluation of the next phase head carries over from the nominal to the verbal domain, i.e. the trigger for inaccessibility (i.e. Merger of a phase head) is a domain-independent notion: (64) (= Chomsky DbP: (11)) Phase-Impenetrability Condition (PIC, weak version) The domain of H is not accessible to operations at ZP; only H and its edge are accessible to such operations. Taking inaccessibility in both versions of the PIC to refer to the application of TRANSFER, hence, deletion of uninterpretable instances of formal features, let us see why both notions of the PIC are problematic for the present analysis. 198 3 INFERENCE - Nominal Concord & Definiteness Assuming the strict PIC in (63), D 0 (or Poss PRO0 , Q S0 ; viz. the highest phase head in the nominal domain) enters the derivation via EM, completing the isolated derivation of the nominal spine at a parallel workspace. TRANSFER (including deletion) then applies to the domain of the phase head, at most returning a featurally deprived syntactic object (under the bidirectional syntax-semantics interface envisioned in P&T) solely bearing interpretable instances of the formerly shared features, parallel to the outcome of a standard minimalist derivation. Thus, the DP Encapsulation Assumption is not met by the output: N bears interpretable numberand (with masculine or neuter nominals) gender-features, D solely bears interpretable person (following standard assumptions, cf. Danon 2011: 7 and references therein). In contrast, under the weak PIC in (64), the domain of D is accessible until EM of the next phasal head (cf. Chomsky DbP: fn. 28). At this point the scope of evaluation of the next phase becomes crucial: Taking a simple transitive v P, the internal argument can be taken to be φ-complete until v 0 enters the derivation. How, though, is TRANSFER of the external argument, merged in Spec, v P, to be evaluated? If evaluation spans the gap between nominal and clausal domain, does the phase merged as the sister of another phase head’s copy (i.e. its barprojection) be immediately transferred due to structural dependence on the latter? Does it trigger TRANSFER of its sister’s (i.e. v 0 ) complement, domain( v ), of whom it occupies the edge due to derivational timing of EM? Do they survive until the EM of CP as the next phasal head? This is not a trivial question. All DPv P interactions predicted by the weak PIC are highly problematic and virtually impossible to uphold; further problems for the third option above, though, arise with the (absolute) ban on Improper Movement (viz. Ā-to-A movement, cf. Chomsky 1986) brought into question several times in minimalist research on the analysis of raising/ ECM constructions. 17 Here, formal feature-driven IM becomes dubious in the domain of the raising predicate if it arguably applies to a feature structure differently specified from the one available to the lower CPphase (i.e. deprived by Radical Interpretability at TRANSFER); take e.g. movement to the specifier of matrix TP, initiated, as elaborated in the last subchapter, by φ-completeness of DP. Again, the DP Encapsulation Assumption is not met. We see that both versions of the PIC are too strong to allow any formal feature-driven interaction between nominal and verbal domain in the framework incorporating nominal phase heads: The strict version never allowing uninterpretable instances of features to survive until EM with the verbal hierarchy of projections, the weak version (incorporating the assumption that the next phase head evaluation is a mechanism that is independent of the notion of domain) 17 Cf. Bošković (1997), Hornstein (2001), Saito (2001). 3.3 Splitting Nominal Concord 199 predicting either instant TRANSFER of the external argument, of the domain of v or the featural deprivation of the nominal domain in derivational mid-stream (i.e. at the next CP-phase). Luckily, these problems can easily be circumvented if one assumes TRANS- FER/ deletion of uninterpretable instances to be postponed for the highest nominal phasal cycle to apply along with the clausal domain it is merged with. Call this the ‘DP Inclusion Assumption’: (65) DP Inclusion Assumption The highest nominal phase is part of the clausal spine for the means of TRANS- FER/ Spell-Out. With (65) implemented into the current theory, Danon’s DP Encapsulation Assumption is met under P&T’s approach to feature sharing, yet the economic benefits of phasal derivations of DP are maintained. 18 As we will see, this is a welcome state of the theory. We are now in a position to focus on the role of case in nominal concord. Case-features remain suspicious in the generative tradition concerning peculiar properties of (i) the locus of the feature (cf. Bošković 2009), (ii) the featural setup (cf. Bošković 2009, 2011; P&T), as well as (iii) the means of assignment (cf. Frampton et al.: 3, Witkoś 2004) under standard assumptions, all of which are exemplified in the quote from Chomsky below: [i] Structural case is not a feature of the probes (T, v ), but it is assigned a value under agreement, then removed by Spell-Out from the narrow syntax. [ii] The value assigned depends on the probe: nominative for T, accusative for v […]. [iii] Case itself is not matched, but deletes under matching of φ-features. (Chomsky DbP: 6) Case-agreement hence departs from the mechanisms of Agree in its most basic premises. 19 Observe that this treatment of (structural) case not only disposes 18 It must be noted that the testing grounds for phasal status that rest on the notion of the Phase Impenetrability Condition in chapter 2.2 are not weakened by the Inclusion Assumption. The ban on movement skipping the edge (specifier) position of a phase is by no means dependent on the timing of Spell-Out of the latter, as can be derived from the two versions of the PIC above. Movement through the edge is an independent property of phases (cf. also Abels 2003: ch. 2), therefore the extraction tests for the highest phasal cycle of the nominal hierarchy remain viable. 19 P&T’s framework, however, offers a different treatment: Case is the showcase uninterpretable feature in Minimalism. The postulation of an interpretable instance of case (on probes) for the purpose of the establishment of Match/ Value relations is hence proble- 200 3 INFERENCE - Nominal Concord & Definiteness with the category-value bipartition, it furthermore lacks any structure to be transmitted from probe to goal. In fact, there is no transmission at all: Structural case is an undifferentiated feature whose ‘value’ is solely determined by the structural position of the probe. Case-assignment therein reduces to a reflex of φ-agreement between the DP and the functional head in the verbal domain with no dedicated establishment of a Match/ Value relation; as Danon (2011: 2) puts it, case is “parasitic” on φ-agreement. Interestingly, H&R (p. 507f.) do hint at the incorporation of case into their feature geometry. They further conjecture that these features do not only demand the implementation of another organizing node but that a basic split of said nodes will have to be proposed along the partition of structural and oblique cases (cf. H&R: fn. 32). I will follow the standard rational of φ-parasitic case-assignment here and show that nominal concord in German poses an argument for Chomsky’s view in that case-assignment on SLIs is indeed dependent on the availability of φ. a. Structural Case At the end of chapter 3.2.1, I arrived at the listing of distributions of selectional feature structures and the availability of default forms on SLIs in German, reprinted below: (66) (= (60)) Q S D Poss PRO Q W Probing φ [φ] [φ] [φ[f]] [φ[f]] Default Agreement ? ?   Taking (66) as a starting point, I will exemplify the assignment of structural case by focusing on the respective interactions of the elements described therein with the functional head T. All relevant operations readily carry over to the second structural case-assigner v 0 . To begin, recall that T is assumed to bear [φ] and EPP, the former triggering the Match relation established between T and the SLI, the latter triggering the structural configuration in which case is assigned. I will diverge from the procedure of chapter 3.2.1 above in commencing my analysis with the determiner/ demonstrative D since this element provides the matic. Under the abandonment of the Valuation/ Interpretability Biconditional, however, case-agreement can be at least partly leveled with standard Agree in that two uninterpretable instances may be proposed: unvalued on the nominal but valued on the probe (see Bošković 2011). As mentioned above, P&T take a different route in reinterpreting case as uninterpretable unvalued u T[ ] on DP and interpretable unvalued i T[ ] on Tns (i.e. T 0 ), to be subsequently valued by Agree of T with v . I will, however, not follow this move, for reasons to become explicit below. 3.3 Splitting Nominal Concord 201 most straightforward derivation: Observe the derivational moment exemplified in (67), in which the corresponding nominal domain with a masculine HN has merged as the external argument into the verbal spine after v 0 has established Agree with the object NP in the specifier of the lexical verb (irrelevant details omitted): (67) Nominal Concord & Definiteness I NFERENCE 147 which the corresponding nominal domain with a masculine HN has merged as the external argument into the verbal spine after v 0 has established Agree with the object NP in the specifier of the lexical verb (irrelevant details omitted): (67) [D > N] > v 0 > NP > V 0 [φ[*m]] [φ[m]] [φ] [φ] D’s φ-feature has already probed prior to the conflation of the parallel workspaces and successfully agreed with N’s complex feature structure [φ[m]]; therefore, D and N share said feature (i.e. they bear instances of φ). Next, T 0 is externally merged, probes for [φ] and Agree successfully applies between T and the external argument: (68) T 0 > [D > N] [φ[*m]] [φ[*m]] [φ[m]] The complex φ-feature is now shared by three elements. Subsequently, T 0 ’s EPP-feature attracts the nominal complex into its specifier. Case-assignment occurs on the basis of φ-agreement; the external argument surfaces with nominative case. 20 (69) [D > N] > T 0 [φ[*m]] [φ[m]] [φ[*m]] [*nom] [*nom] Nothing more has to be said about derivations incorporating SLIs which are categorized as probing for [φ] (i.e. D and Q S ) since they are trivially specified to successfully agree with any feature structure of φ on HN. Turning to the SLIs bearing the more restrictive analogue [φ[f]], namely Poss PRO and Q W , observe that those derivations match the one in 0-0 above if the nominal is either specified for feminine or plural, i.e. if NumP is present in the derivation: (70) a. (= 0, [modified]) [SLI > Num > N] > v 0 > NP > V 0 [φ[f[*pl]]] [φ[f[pl]]] [φ] [φ] b. (= 0, [modified]) T 0 > [SLI > Num] [φ[*f[pl]]] [φ[f[*pl]]] [φ[f[pl]]] c. (= 0, [modified]) [SLI > Num] > T 0 [φ[f[*pl]]] [φ[f[pl]]] [φ[*f[pl]]] [*nom] [*nom] The SLI agrees with Num, irrespective of the φ-feature structure of N, prior to insertion into the verbal domain. T merges and probes for φ, agreeing with the φ-complete SLI/ Num, which subsequently raises to the specifier of the probe and thereby receives case. An important question at the current point is this: What happens when the φ-features of the SLI do not match those of the lexical domain, viz. HN? We have seen in the last subsection that those elements which bear the more restrictive φprobe are also those able to reduce to partial default in the case of failed applications of Value. Here I want to argue that this initial mismatch has further consequences for the derivation in the sentential domain. Observe 0, a derivational sketch along the line of 0 a. above with SLI[φ[f]] concatenated with a masculine head noun, N[φ[m]], parallel to 0: (71) (= 0, [modified]) [SLI > N] > v 0 > NP > V 0 20 I diverge from Frampton & Gutmann (2000) in the overt notion of case on elements in the nominal domain for reasons that will become clear immediately below. Note that for those authors the establishment of a shared φ-feature from the nominal to the case-‘assigner’ suffices to signal successful case-agreement (cf. Frampton & Gutmann 2000: 5). D’s φ-feature has already probed prior to the conflation of the parallel workspaces and successfully agreed with N’s complex feature structure [φ[m]]; therefore, D and N share said feature (i.e. they bear instances of φ). Next, T 0 is externally merged, probes for [φ] and Agree successfully applies between T and the external argument: (68) Nominal Concord & Definiteness I NFERENCE 147 which the corresponding nominal domain with a masculine HN has merged as the external argument into the verbal spine after v 0 has established Agree with the object NP in the specifier of the lexical verb (irrelevant details omitted): (67) [D > N] > v 0 > NP > V 0 [φ[*m]] [φ[m]] [φ] [φ] D’s φ-feature has already probed prior to the conflation of the parallel workspaces and successfully agreed with N’s complex feature structure [φ[m]]; therefore, D and N share said feature (i.e. they bear instances of φ). Next, T 0 is externally merged, probes for [φ] and Agree successfully applies between T and the external argument: (68) T 0 > [D > N] [φ[*m]] [φ[*m]] [φ[m]] The complex φ-feature is now shared by three elements. Subsequently, T 0 ’s EPP-feature attracts the nominal complex into its specifier. Case-assignment occurs on the basis of φ-agreement; the external argument surfaces with nominative case. 20 (69) [D > N] > T 0 [φ[*m]] [φ[m]] [φ[*m]] [*nom] [*nom] Nothing more has to be said about derivations incorporating SLIs which are categorized as probing for [φ] (i.e. D and Q S ) since they are trivially specified to successfully agree with any feature structure of φ on HN. Turning to the SLIs bearing the more restrictive analogue [φ[f]], namely Poss PRO and Q W , observe that those derivations match the one in 0-0 above if the nominal is either specified for feminine or plural, i.e. if NumP is present in the derivation: (70) a. (= 0, [modified]) [SLI > Num > N] > v 0 > NP > V 0 [φ[f[*pl]]] [φ[f[pl]]] [φ] [φ] b. (= 0, [modified]) T 0 > [SLI > Num] [φ[*f[pl]]] [φ[f[*pl]]] [φ[f[pl]]] c. (= 0, [modified]) [SLI > Num] > T 0 [φ[f[*pl]]] [φ[f[pl]]] [φ[*f[pl]]] [*nom] [*nom] The SLI agrees with Num, irrespective of the φ-feature structure of N, prior to insertion into the verbal domain. T merges and probes for φ, agreeing with the φ-complete SLI/ Num, which subsequently raises to the specifier of the probe and thereby receives case. An important question at the current point is this: What happens when the φ-features of the SLI do not match those of the lexical domain, viz. HN? We have seen in the last subsection that those elements which bear the more restrictive φprobe are also those able to reduce to partial default in the case of failed applications of Value. Here I want to argue that this initial mismatch has further consequences for the derivation in the sentential domain. Observe 0, a derivational sketch along the line of 0 a. above with SLI[φ[f]] concatenated with a masculine head noun, N[φ[m]], parallel to 0: (71) (= 0, [modified]) [SLI > N] > v 0 > NP > V 0 20 I diverge from Frampton & Gutmann (2000) in the overt notion of case on elements in the nominal domain for reasons that will become clear immediately below. Note that for those authors the establishment of a shared φ-feature from the nominal to the case-‘assigner’ suffices to signal successful case-agreement (cf. Frampton & Gutmann 2000: 5). The complex φ-feature is now shared by three elements. Subsequently, T 0 ’s EPPfeature attracts the nominal complex into its specifier. Case-assignment occurs on the basis of φ-agreement; the external argument surfaces with nominative case. 20 (69) Nominal Concord & Definiteness I NFERENCE 147 which the corresponding nominal domain with a masculine HN has merged as the external argument into the verbal spine after v 0 has established Agree with the object NP in the specifier of the lexical verb (irrelevant details omitted): (67) [D > N] > v 0 > NP > V 0 [φ[*m]] [φ[m]] [φ] [φ] D’s φ-feature has already probed prior to the conflation of the parallel workspaces and successfully agreed with N’s complex feature structure [φ[m]]; therefore, D and N share said feature (i.e. they bear instances of φ). Next, T 0 is externally merged, probes for [φ] and Agree successfully applies between T and the external argument: (68) T 0 > [D > N] [φ[*m]] [φ[*m]] [φ[m]] The complex φ-feature is now shared by three elements. Subsequently, T 0 ’s EPP-feature attracts the nominal complex into its specifier. Case-assignment occurs on the basis of φ-agreement; the external argument surfaces with nominative case. 20 (69) [D > N] > T 0 [φ[*m]] [φ[m]] [φ[*m]] [*nom] [*nom] Nothing more has to be said about derivations incorporating SLIs which are categorized as probing for [φ] (i.e. D and Q S ) since they are trivially specified to successfully agree with any feature structure of φ on HN. Turning to the SLIs bearing the more restrictive analogue [φ[f]], namely Poss PRO and Q W , observe that those derivations match the one in 0-0 above if the nominal is either specified for feminine or plural, i.e. if NumP is present in the derivation: (70) a. (= 0, [modified]) [SLI > Num > N] > v 0 > NP > V 0 [φ[f[*pl]]] [φ[f[pl]]] [φ] [φ] b. (= 0, [modified]) T 0 > [SLI > Num] [φ[*f[pl]]] [φ[f[*pl]]] [φ[f[pl]]] c. (= 0, [modified]) [SLI > Num] > T 0 [φ[f[*pl]]] [φ[f[pl]]] [φ[*f[pl]]] [*nom] [*nom] The SLI agrees with Num, irrespective of the φ-feature structure of N, prior to insertion into the verbal domain. T merges and probes for φ, agreeing with the φ-complete SLI/ Num, which subsequently raises to the specifier of the probe and thereby receives case. An important question at the current point is this: What happens when the φ-features of the SLI do not match those of the lexical domain, viz. HN? We have seen in the last subsection that those elements which bear the more restrictive φprobe are also those able to reduce to partial default in the case of failed applications of Value. Here I want to argue that this initial mismatch has further consequences for the derivation in the sentential domain. Observe 0, a derivational sketch along the line of 0 a. above with SLI[φ[f]] concatenated with a masculine head noun, N[φ[m]], parallel to 0: (71) (= 0, [modified]) [SLI > N] > v 0 > NP > V 0 20 I diverge from Frampton & Gutmann (2000) in the overt notion of case on elements in the nominal domain for reasons that will become clear immediately below. Note that for those authors the establishment of a shared φ-feature from the nominal to the case-‘assigner’ suffices to signal successful case-agreement (cf. Frampton & Gutmann 2000: 5). Nothing more has to be said about derivations incorporating SLIs which are categorized as probing for [φ] (i.e. D and Q S ) since they are trivially specified to successfully agree with any feature structure of φ on HN. Turning to the SLIs bearing the more restrictive analogue [φ[f]], namely Poss PRO and Q W , observe that those derivations match the one in (67)-(69) above if the nominal is either specified for feminine or plural, i.e. if NumP is present in the derivation: 20 I diverge from Frampton & Gutmann (2000) in the overt notion of case on elements in the nominal domain for reasons that will become clear immediately below. Note that for those authors the establishment of a shared φ-feature from the nominal to the case-‘assigner’ suffices to signal successful case-agreement (cf. Frampton & Gutmann 2000: 5). 202 3 INFERENCE - Nominal Concord & Definiteness (70) Nominal Concord & Definiteness I NFERENCE 147 which the corresponding nominal domain with a masculine HN has merged as the external argument into the verbal spine after v 0 has established Agree with the object NP in the specifier of the lexical verb (irrelevant details omitted): (67) [D > N] > v 0 > NP > V 0 [φ[*m]] [φ[m]] [φ] [φ] D’s φ-feature has already probed prior to the conflation of the parallel workspaces and successfully agreed with N’s complex feature structure [φ[m]]; therefore, D and N share said feature (i.e. they bear instances of φ). Next, T 0 is externally merged, probes for [φ] and Agree successfully applies between T and the external argument: (68) T 0 > [D > N] [φ[*m]] [φ[*m]] [φ[m]] The complex φ-feature is now shared by three elements. Subsequently, T 0 ’s EPP-feature attracts the nominal complex into its specifier. Case-assignment occurs on the basis of φ-agreement; the external argument surfaces with nominative case. 20 (69) [D > N] > T 0 [φ[*m]] [φ[m]] [φ[*m]] [*nom] [*nom] Nothing more has to be said about derivations incorporating SLIs which are categorized as probing for [φ] (i.e. D and Q S ) since they are trivially specified to successfully agree with any feature structure of φ on HN. Turning to the SLIs bearing the more restrictive analogue [φ[f]], namely Poss PRO and Q W , observe that those derivations match the one in 0-0 above if the nominal is either specified for feminine or plural, i.e. if NumP is present in the derivation: (70) a. (= (67), [modified]) [SLI > Num > N] > v 0 > NP > V 0 [φ[f[*pl]]] [φ[f[pl]]] [φ] [φ] b. (= (68), [modified]) T 0 > [SLI > Num] [φ[*f[pl]]] [φ[f[*pl]]] [φ[f[pl]]] c. (= (69), [modified]) [SLI > Num] > T 0 [φ[f[*pl]]] [φ[f[pl]]] [φ[*f[pl]]] [*nom] [*nom] The SLI agrees with Num, irrespective of the φ-feature structure of N, prior to insertion into the verbal domain. T merges and probes for φ, agreeing with the φ-complete SLI/ Num, which subsequently raises to the specifier of the probe and thereby receives case. An important question at the current point is this: What happens when the φ-features of the SLI do not match those of the lexical domain, viz. HN? We have seen in the last subsection that those elements which bear the more restrictive φprobe are also those able to reduce to partial default in the case of failed applications of Value. Here I want to argue that this initial mismatch has further consequences for the derivation in the sentential domain. Observe 0, a derivational sketch along the line of 0 a. above with SLI[φ[f]] concatenated with a masculine head noun, N[φ[m]], parallel to 0: (71) (= (67), [modified]) [SLI > N] > v 0 > NP > V 0 [φ] [φ[m]] [φ] [φ] 20 I diverge from Frampton & Gutmann (2000) in the overt notion of case on elements in the nominal domain for reasons that will become clear immediately below. Note that for those authors the establishment of a shared φ-feature from the nominal to the case-‘assigner’ suffices to signal successful case-agreement (cf. Frampton & Gutmann 2000: 5). The SLI agrees with Num, irrespective of the φ-feature structure of N, prior to insertion into the verbal domain. T merges and probes for φ, agreeing with the φ-complete SLI/ Num, which subsequently raises to the specifier of the probe and thereby receives case. An important question at the current point is this: What happens when the φ-features of the SLI do not match those of the lexical domain, viz. HN? We have seen in the last subsection that those elements which bear the more restrictive φ-probe are also those able to reduce to partial default in the case of failed applications of Value. Here I want to argue that this initial mismatch has further consequences for the derivation in the sentential domain. Observe (71), a derivational sketch along the line of (70) a. above with SLI[φ[f]] concatenated with a masculine head noun, N[φ[m]], parallel to (67): (71) (= (67), [modified]) Nominal Concord & Definiteness I NFERENCE 147 which the corresponding nominal domain with a masculine HN has merged as the external argument into the verbal spine after v 0 has established Agree with the object NP in the specifier of the lexical verb (irrelevant details omitted): (67) [D > N] > v 0 > NP > V 0 [φ[*m]] [φ[m]] [φ] [φ] D’s φ-feature has already probed prior to the conflation of the parallel workspaces and successfully agreed with N’s complex feature structure [φ[m]]; therefore, D and N share said feature (i.e. they bear instances of φ). Next, T 0 is externally merged, probes for [φ] and Agree successfully applies between T and the external argument: (68) T 0 > [D > N] [φ[*m]] [φ[*m]] [φ[m]] The complex φ-feature is now shared by three elements. Subsequently, T 0 ’s EPP-feature attracts the nominal complex into its specifier. Case-assignment occurs on the basis of φ-agreement; the external argument surfaces with nominative case. 20 (69) [D > N] > T 0 [φ[*m]] [φ[m]] [φ[*m]] [*nom] [*nom] Nothing more has to be said about derivations incorporating SLIs which are categorized as probing for [φ] (i.e. D and Q S ) since they are trivially specified to successfully agree with any feature structure of φ on HN. Turning to the SLIs bearing the more restrictive analogue [φ[f]], namely Poss PRO and Q W , observe that those derivations match the one in 0-0 above if the nominal is either specified for feminine or plural, i.e. if NumP is present in the derivation: (70) a. (= (67), [modified]) [SLI > Num > N] > v 0 > NP > V 0 [φ[f[*pl]]] [φ[f[pl]]] [φ] [φ] b. (= (68), [modified]) T 0 > [SLI > Num] [φ[*f[pl]]] [φ[f[*pl]]] [φ[f[pl]]] c. (= (69), [modified]) [SLI > Num] > T 0 [φ[f[*pl]]] [φ[f[pl]]] [φ[*f[pl]]] [*nom] [*nom] The SLI agrees with Num, irrespective of the φ-feature structure of N, prior to insertion into the verbal domain. T merges and probes for φ, agreeing with the φ-complete SLI/ Num, which subsequently raises to the specifier of the probe and thereby receives case. An important question at the current point is this: What happens when the φ-features of the SLI do not match those of the lexical domain, viz. HN? We have seen in the last subsection that those elements which bear the more restrictive φprobe are also those able to reduce to partial default in the case of failed applications of Value. Here I want to argue that this initial mismatch has further consequences for the derivation in the sentential domain. Observe 0, a derivational sketch along the line of 0 a. above with SLI[φ[f]] concatenated with a masculine head noun, N[φ[m]], parallel to 0: (71) (= (67), [modified]) [SLI > N] > v 0 > NP > V 0 [φ] [φ[m]] [φ] [φ] 20 I diverge from Frampton & Gutmann (2000) in the overt notion of case on elements in the nominal domain for reasons that will become clear immediately below. Note that for those authors the establishment of a shared φ-feature from the nominal to the case-‘assigner’ suffices to signal successful case-agreement (cf. Frampton & Gutmann 2000: 5). The complex nominal domain, bearing two separate φ-configurations, merges into Spec, v P. 21 Next, T merges and probes for φ. However, the highest compatible goal in domain D(T) does not exhibit a uniform feature configuration. Recall from the introduction to chapter 3.1.1 that T skips Expl in ES in this exact application of probing, since the expletive bears only a person-value and is hence not φ-complete. With Expl constituting a simple LI, T then probes further down into its c-command domain and finds a suitable goal for φ in the complement of the verb. My proposal here is that this exact same mechanism is also available 21 Observe that unlike Bejar (2003) I take the feature on the SLI to constitute an instance of partial default, i.e. an active feature, even this late in the derivation. 3.3 Splitting Nominal Concord 203 for the probing into the complex nominal domain, still part of D(T). The SLI bearing a partially reduced feature structure is overlooked by T which can hence continue to probe and find a goal in the φ-complete head noun: (72) (= (68), [modified]) Nominal Concord & Definiteness I NFERENCE 148 (71) (= (67), [modified]) [SLI > N] > v 0 > NP > V 0 [φ] [φ[m]] [φ] [φ] The complex nominal domain, bearing two separate φ-configurations, merges into Spec, v P. 21 Next, T merges and probes for φ. However, the highest compatible goal in domain D(T) does not exhibit a uniform feature configuration. Recall from the introduction to chapter 0 that T skips Expl in ES in this exact application of probing, since the expletive bears only a person-value and is hence not φ-complete. With Expl constituting a simple LI, T then probes further down into its ccommand domain and finds a suitable goal for φ in the complement of the verb. My proposal here is that this exact same mechanism is also available for the probing into the complex nominal domain, still part of D(T). The SLI bearing a partially reduced feature structure is overlooked by T which can hence continue to probe and find a goal in the φ-complete head noun: (72) (= (68), [modified]) T 0 > [SLI > N] [φ[*m]] [φ] [φ[m]] Subsequently, N, matching in category and in φ-features, is attracted to the specifier of TP. Recall from chapter 2.2, however, that the SLI is a phase head, in fact the most minimal phase of the nominal domain. IM therefore forces piedpiping of the complete nominal domain. Since the SLI is not part of the φ-chain, the basis of case-assignment, the element surfaces with default morphology for φ as well as case, which is zero-morphology on German SLIs, as has been elaborated above. (73) (=,(69), [modified]) [SLI > N] > T 0 [φ] [φ[m]] [φ[*m] [*nom] I will have to stay agnostic as to the possibility of abstract case on SLIs in linearizations like 0 for now. As we will see later in the analysis of inflectional impoverishment, there are indeed good reasons to assume that the SLI is omitted in the assignment of case and surfaces with default inflection (i.e. zero-morphology) for both φ as well as case. Turning to H&R’s speculations on a potential integration of case into the feature geometry, one might initially tend to code said feature on their root node RE , i.e. [ RE [φ[number/ gender]][case]]. Alternatively, and therein following implications from Greenberg’s universal 39, cited by H&R (p. 507f.), stating that number-morphology is always located closer to the noun-stem than case-morphology, the latter might be argued to be a dependent of the φ-node, i.e. [φ[number/ gender][case]], or even of this node’s respective specification, i.e. [φ[number/ gender[case]]]. Note that the second option, however, would always require a number/ gender-specification to constitute the basis for case-specifications and would thus not allow overt case marking without φ-agreement. Additionally, obvious problems emerge concerning the varying locus of the dependence of case (i.e. [φ[m[case]]] with masculine nominals, [φ[n[case]]] with neuter HNs etc.), which is why this latter approach is readily dismissed in what follows. I will get back to this observation and the various structural implementations laid out above in the discussion of non-structural case below. Turning to overt realizations, note that φ/ case-morphology is realized as a portmanteau-morpheme in German. In what follows, I will nevertheless continue to mark case distinct from the φ-specification of lexical items for ease of exposition and to do justice to the tentative character of the observation above as well as the uncertainty of H&R themselves on this point, but cf. chapter 0 below. The theory at hand readily distinguishes semi-lexical items that partake in the sharing of φ- (and case-)features (i.e. D, Q S ) from those whose selectional restrictions prevent them from doing so with a subset of such specifications (i.e. Poss PRO , Q W ). Since φ-completeness is the prerequisite for agreement with the structural case-‘bearing’ head T 0 (as well as v 0 ), a mismatch in the nominal domain has further consequences once the complex is merged into the verbal spine: SLIs bearing [φ[f]] reduce in concatenation with masculine and neuter nominals during the derivation of the nominal domain and are hence overlooked in T’s probing for a φ-complete goal, parallel to expletives. The complement of the SLI, still part of 21 Observe that unlike Bejar (2003) I take the feature on the SLI to constitute an instance of partial default, i.e. an active feature, even this late in the derivation. Subsequently, N, matching in category and in φ-features, is attracted to the specifier of TP. Recall from chapter 2.2, however, that the SLI is a phase head, in fact the most minimal phase of the nominal domain. IM therefore forces piedpiping of the complete nominal domain. Since the SLI is not part of the φ-chain, the basis of case-assignment, the element surfaces with default morphology for φ as well as case, which is zero-morphology on German SLIs, as has been elaborated above. (73) (= (69), [modified]) Nominal Concord & Definiteness I NFERENCE 148 (71) (= (67), [modified]) [SLI > N] > v 0 > NP > V 0 [φ] [φ[m]] [φ] [φ] The complex nominal domain, bearing two separate φ-configurations, merges into Spec, v P. 21 Next, T merges and probes for φ. However, the highest compatible goal in domain D(T) does not exhibit a uniform feature configuration. Recall from the introduction to chapter 0 that T skips Expl in ES in this exact application of probing, since the expletive bears only a person-value and is hence not φ-complete. With Expl constituting a simple LI, T then probes further down into its ccommand domain and finds a suitable goal for φ in the complement of the verb. My proposal here is that this exact same mechanism is also available for the probing into the complex nominal domain, still part of D(T). The SLI bearing a partially reduced feature structure is overlooked by T which can hence continue to probe and find a goal in the φ-complete head noun: (72) (= (68), [modified]) T 0 > [SLI > N] [φ[*m]] [φ] [φ[m]] Subsequently, N, matching in category and in φ-features, is attracted to the specifier of TP. Recall from chapter 2.2, however, that the SLI is a phase head, in fact the most minimal phase of the nominal domain. IM therefore forces piedpiping of the complete nominal domain. Since the SLI is not part of the φ-chain, the basis of case-assignment, the element surfaces with default morphology for φ as well as case, which is zero-morphology on German SLIs, as has been elaborated above. (73) (= (69), [modified]) [SLI > N] > T 0 [φ] [φ[m]] [φ[*m] [*nom] I will have to stay agnostic as to the possibility of abstract case on SLIs in linearizations like 0 for now. As we will see later in the analysis of inflectional impoverishment, there are indeed good reasons to assume that the SLI is omitted in the assignment of case and surfaces with default inflection (i.e. zero-morphology) for both φ as well as case. Turning to H&R’s speculations on a potential integration of case into the feature geometry, one might initially tend to code said feature on their root node RE , i.e. [ RE [φ[number/ gender]][case]]. Alternatively, and therein following implications from Greenberg’s universal 39, cited by H&R (p. 507f.), stating that number-morphology is always located closer to the noun-stem than case-morphology, the latter might be argued to be a dependent of the φ-node, i.e. [φ[number/ gender][case]], or even of this node’s respective specification, i.e. [φ[number/ gender[case]]]. Note that the second option, however, would always require a number/ gender-specification to constitute the basis for case-specifications and would thus not allow overt case marking without φ-agreement. Additionally, obvious problems emerge concerning the varying locus of the dependence of case (i.e. [φ[m[case]]] with masculine nominals, [φ[n[case]]] with neuter HNs etc.), which is why this latter approach is readily dismissed in what follows. I will get back to this observation and the various structural implementations laid out above in the discussion of non-structural case below. Turning to overt realizations, note that φ/ case-morphology is realized as a portmanteau-morpheme in German. In what follows, I will nevertheless continue to mark case distinct from the φ-specification of lexical items for ease of exposition and to do justice to the tentative character of the observation above as well as the uncertainty of H&R themselves on this point, but cf. chapter 0 below. The theory at hand readily distinguishes semi-lexical items that partake in the sharing of φ- (and case-)features (i.e. D, Q S ) from those whose selectional restrictions prevent them from doing so with a subset of such specifications (i.e. Poss PRO , Q W ). Since φ-completeness is the prerequisite for agreement with the structural case-‘bearing’ head T 0 (as well as v 0 ), a mismatch in the nominal domain has further consequences once the complex is merged into the verbal spine: SLIs bearing [φ[f]] reduce in concatenation with masculine and neuter nominals during the derivation of the nominal domain and are hence overlooked in T’s probing for a φ-complete goal, parallel to expletives. The complement of the SLI, still part of 21 Observe that unlike Bejar (2003) I take the feature on the SLI to constitute an instance of partial default, i.e. an active feature, even this late in the derivation. I will have to stay agnostic as to the possibility of abstract case on SLIs in linearizations like (73) for now. As we will see later in the analysis of inflectional impoverishment, there are indeed good reasons to assume that the SLI is omitted in the assignment of case and surfaces with default inflection (i.e. zeromorphology) for both φ as well as case. Turning to H&R’s speculations on a potential integration of case into the feature geometry, one might initially tend to code said feature on their root node re , i.e. [ re [φ[number/ gender]][case]]. Alternatively, and therein following implications from Greenberg’s universal 39, cited by H&R (p. 507f.), stating that number-morphology is always located closer to the noun-stem than case-morphology, the latter might be argued to be a dependent of the φ-node, i.e. [φ[number/ gender][case]], or even of this node’s respective specification, i.e. [φ[number/ gender[case]]]. Note that the second option, however, would always require a number/ gender-specification to constitute the basis for case-specifications and would thus not allow overt case marking without φ-agreement. Additionally, obvious problems emerge concerning the varying locus of the dependence of case (i.e. [φ[m[case]]] with masculine nominals, [φ[n[case]]] with neuter HNs etc.), which is why this latter approach is readily dismissed in what follows. I will get back to this observation and the various structural implementations laid out above in the discussion of non-structural case below. Turning to overt reali- 204 3 INFERENCE - Nominal Concord & Definiteness zations, note that φ/ case-morphology is realized as a portmanteau-morpheme in German. In what follows, I will nevertheless continue to mark case distinct from the φ-specification of lexical items for ease of exposition and to do justice to the tentative character of the observation above as well as the uncertainty of H&R themselves on this point, but cf. chapter 3.2.4 below. The theory at hand readily distinguishes semi-lexical items that partake in the sharing of φ- (and case-)features (i.e. D, Q S ) from those whose selectional restrictions prevent them from doing so with a subset of such specifications (i.e. Poss PRO , Q W ). Since φ-completeness is the prerequisite for agreement with the structural case-‘bearing’ head T 0 (as well as v 0 ), a mismatch in the nominal domain has further consequences once the complex is merged into the verbal spine: SLIs bearing [φ[f]] reduce in concatenation with masculine and neuter nominals during the derivation of the nominal domain and are hence overlooked in T’s probing for a φ-complete goal, parallel to expletives. The complement of the SLI, still part of domain(T), is accessible for T’s probing analogous to the verbal domain beyond the expletive in the derivation of existential sentences. Hence, φ-Agree is established between T and the highest φ-complete element of the nominal domain but IM is forced to pied-pipe the full nominal domain due to the phasal status of the SLI. Since case is assigned on the basis of φ-Match, the (S)LIs bearing an instance of the shared, complete φ-structure are assigned case, while those bearing the partially reduced root node surface with default, viz. zero-morphology. I have made extensive use of varying labels in the derivations above for ease of exposition. For explanatory purposes, however, I will end this section by backing up the claims made above with exemplary surface structures of the respective SLIs in concatenation with mass as well as plural head nouns together with their featural specifications before turning to the second class of case forms, i.e. non-structural cases and the means of their assignment. 3.3 Splitting Nominal Concord 205 (74) Nominal Concord & Definiteness I NFERENCE 149 domain(T), is accessible for T’s probing analogous to the verbal domain beyond the expletive in the derivation of existential sentences. Hence, φ-Agree is established between T and the highest φ-complete element of the nominal domain but IM is forced to pied-pipe the full nominal domain due to the phasal status of the SLI. Since case is assigned on the basis of φ-Match, the (S)LIs bearing an instance of the shared, complete φ-structure are assigned case, while those bearing the partially reduced root node surface with default, viz. zero-morphology. I have made extensive use of varying labels in the derivations above for ease of exposition. For explanatory purposes, however, I will end this section by backing up the claims made above with exemplary surface structures of the respective SLIs in concatenation with mass as well as plural head nouns together with their featural specifications before turning to the second class of case forms, i.e. non-structural cases and the means of their assignment. (74) a. D/ Q S - N[φ[m]] Dies-er/ All-er Wein schmeckt gut. this / all wine tastes good D/ Q S > N > T 0 > v 0 … [φ[*m]] [φ[m]] [φ[*m]] [*nom] [*nom] b. D/ Q S - N[φ[n]] Dies-es/ All-es Bier schmeckt gut. this / all beer tastes good D/ Q S > N > T 0 > v 0 … [φ[*n]] [φ[n]] [φ[*n]] [*nom] [*nom] c. D/ Q S - N[φ[f]] Dies-e/ All-e Milch schmeckt gut. this / all milk tastes good D/ Q S > Num > T 0 > v 0 … [φ[*f]] [φ[f]] [φ[*f]] [*nom] [*nom] d. D/ Q S - N[φ[f[pl]]] Dies-e/ All-e Kekse schmecken gut. this / all cookies taste good D/ Q S > Num > T 0 > v 0 … [φ[*f[pl]]] [φ[f[pl]]] [φ[*f[pl]]] [*nom] [*nom] (75) a. Poss PRO / Q W - N[φ[m]] Mein-ø/ Viel-ø Wein schmeckt gut. my / much wine tastes good Poss PRO / Q W > N > T 0 > v 0 … [φ] [φ[m]] [φ[*m]] [*nom] b. Poss PRO / Q W - N[φ[n]] Mein-ø/ Viel-ø Bier schmeckt gut. my / much beer tastes good Poss PRO / Q W > N > T 0 > v 0 … [φ] [φ[n]] [φ[*n]] [*nom] c. Poss PRO / Q W - N[φ[f]] Mein-e/ Viel-e Milch schmeckt gut. my / much milk tastes good Poss PRO / Q W > Num > T 0 > v 0 … [φ[f]] [φ[f]] [φ[*f]] [*nom] [*nom] d. Poss PRO / Q W - N[φ[f]] Mein-e/ Viel-e Kekse schmecken gut. my / much cookies taste good Poss PRO / Q W > Num > T 0 > v 0 … [φ[f[*pl]]] [φ[f[pl]]] [φ[*f[pl]]] [*nom] [*nom] 206 3 INFERENCE - Nominal Concord & Definiteness (75) Nominal Concord & Definiteness I NFERENCE 149 domain(T), is accessible for T’s probing analogous to the verbal domain beyond the expletive in the derivation of existential sentences. Hence, φ-Agree is established between T and the highest φ-complete element of the nominal domain but IM is forced to pied-pipe the full nominal domain due to the phasal status of the SLI. Since case is assigned on the basis of φ-Match, the (S)LIs bearing an instance of the shared, complete φ-structure are assigned case, while those bearing the partially reduced root node surface with default, viz. zero-morphology. I have made extensive use of varying labels in the derivations above for ease of exposition. For explanatory purposes, however, I will end this section by backing up the claims made above with exemplary surface structures of the respective SLIs in concatenation with mass as well as plural head nouns together with their featural specifications before turning to the second class of case forms, i.e. non-structural cases and the means of their assignment. (74) a. D/ Q S - N[φ[m]] Dies-er/ All-er Wein schmeckt gut. this / all wine tastes good D/ Q S > N > T 0 > v 0 … [φ[*m]] [φ[m]] [φ[*m]] [*nom] [*nom] b. D/ Q S - N[φ[n]] Dies-es/ All-es Bier schmeckt gut. this / all beer tastes good D/ Q S > N > T 0 > v 0 … [φ[*n]] [φ[n]] [φ[*n]] [*nom] [*nom] c. D/ Q S - N[φ[f]] Dies-e/ All-e Milch schmeckt gut. this / all milk tastes good D/ Q S > Num > T 0 > v 0 … [φ[*f]] [φ[f]] [φ[*f]] [*nom] [*nom] d. D/ Q S - N[φ[f[pl]]] Dies-e/ All-e Kekse schmecken gut. this / all cookies taste good D/ Q S > Num > T 0 > v 0 … [φ[*f[pl]]] [φ[f[pl]]] [φ[*f[pl]]] [*nom] [*nom] (75) a. Poss PRO / Q W - N[φ[m]] Mein-ø/ Viel-ø Wein schmeckt gut. my / much wine tastes good Poss PRO / Q W > N > T 0 > v 0 … [φ] [φ[m]] [φ[*m]] [*nom] b. Poss PRO / Q W - N[φ[n]] Mein-ø/ Viel-ø Bier schmeckt gut. my / much beer tastes good Poss PRO / Q W > N > T 0 > v 0 … [φ] [φ[n]] [φ[*n]] [*nom] c. Poss PRO / Q W - N[φ[f]] Mein-e/ Viel-e Milch schmeckt gut. my / much milk tastes good Poss PRO / Q W > Num > T 0 > v 0 … [φ[f]] [φ[f]] [φ[*f]] [*nom] [*nom] d. Poss PRO / Q W - N[φ[f [pl] ]] Mein-e/ Viel-e Kekse schmecken gut. my / much cookies taste good Poss PRO / Q W > Num > T 0 > v 0 … [φ[f[*pl]]] [φ[f[pl]]] [φ[*f[pl]]] [*nom] [*nom] b. Non-Structural Case As H&R (fn. 32) note, the basic distinction of a hypothetical case-node in their feature geometry would have to be made on the basis of structural vs. local or oblique cases. Woolford (2006) proposes that despite the apparent difference e.g. in predictability/ regularity (inherent case is predictable, while lexical case is idiosyncratic), inherent and lexical cases (i.e. genitive and dative in German) should be bundled as non-structural case forms in opposition to the structural ones under analysis in the last section (i.e. nominative and accusative). She reviews a number of tests for structural case-hood and concludes that the reliable ones (e.g. preservation under A-movement in passive and raising, preservation on subjects in tensed clauses) split case forms along the lines of her proposals. As Woolford argues, non-structural case is assigned early in the derivation, namely at the v P-structure in Θ-positions, hence at EM. The loci of this first Merger-operations differ, however, in that the more idiosyncratic lexical case 3.3 Splitting Nominal Concord 207 is assigned inside VP, while predictable inherent case is licensed in (a possibly additional) v P. The application of licensing ‘at the v P-structure’ (EM) is nevertheless identical. Contrasting this with the observations on structural case made above, nonstructural case ‘comes free’ in that no match for φ or any other feature category poses a precondition for its assignment in addition to the structural configuration established by External Merge of the nominal into the verbal domain. This begs the question whether the deviance in precondition moreover bears consequences for the succeeding assignment. For SLIs bearing the probing root node [φ], the system predicts a uniform assignment of non-structural case, parallel to the derivations exemplified in the last section to the exclusion of the probing operation for φ-features from the extended verbal domain (i.e. T 0 or v 0 ). This is exactly what we find with D and Q S . Observe the derivational snapshots below parallel to the three stages discussed in the last section: The nominal domain is derived on a parallel workspace in (76) a., subsequently inserted into the verbal spine in (76) b. and thereon assigned case by the lexical verbal head V 0 in (76) c., depicted here by the dotted arrow, as is exemplified with a nominal hierarchy of projections headed by DP. 22 (76) Nominal Concord & Definiteness I NFERENCE 150 b. Non-Structural Case As H&R (fn. 32) note, the basic distinction of a hypothetical case-node in their feature geometry would have to be made on the basis of structural vs. local or oblique cases. Woolford (2006) proposes that despite the apparent difference e.g. in predictability/ regularity (inherent case is predictable, while lexical case is idiosyncratic), inherent and lexical cases (i.e. genitive and dative in German) should be bundled as non-structural case forms in opposition to the structural ones under analysis in the last section (i.e. nominative and accusative). She reviews a number of tests for structural case-hood and concludes that the reliable ones (e.g. preservation under A-movement in passive and raising, preservation on subjects in tensed clauses) split case forms along the lines of her proposals. As Woolford argues, non-structural case is assigned early in the derivation, namely at the v P-structure in Θ-positions, hence at EM. The loci of this first Merger-operations differ, however, in that the more idiosyncratic lexical case is assigned inside VP, while predictable inherent case is licensed in (a possibly additional) v P. The application of licensing ‘at the v P-structure’ (EM) is nevertheless identical. Contrasting this with the observations on structural case made above, non-structural case ‘comes free’ in that no match for φ or any other feature category poses a precondition for its assignment in addition to the structural configuration established by External Merge of the nominal into the verbal domain. This begs the question whether the deviance in precondition moreover bears consequences for the succeeding assignment. For SLIs bearing the probing root node [φ], the system predicts a uniform assignment of non-structural case, parallel to the derivations exemplified in the last section to the exclusion of the probing operation for φ-features from the extended verbal domain (i.e. T 0 or v 0 ). This is exactly what we find with D and Q S . Observe the derivational snapshots below parallel to the three stages discussed in the last section: The nominal domain is derived on a parallel workspace in 0 a., subsequently inserted into the verbal spine in 0 b. and thereon assigned case by the lexical verbal head V 0 in 0 c., depicted here by the dotted arrow, as is exemplified with a nominal hierarchy of projections headed by DP. 22 (76) a. [D > N] [φ[*m]] [φ[m]] b. V 0 > [D > N] [φ[*m]] [φ[m]] c. V 0 > [D > N] [φ[*m]] [φ[m]] [*gen/ *dat] [*gen/ *dat] The derivational processes sketched in 0 readily carry over to the second element bearing the probing root [φ], i.e. Q S . No Agree relation is proposed between the case-assigner and the nominal domain prior to the transmission of case. Next, let us take a closer look at non-structural case-bearing V 0 in concatenation with nominal domains headed by one of those SLIs categorized as probing for [φ[f]] above, namely Poss PRO and Q W : The φ-structure of the semi-lexical item reduced to partial default in the isolated derivation of the nominal domain with masculine or neuter head nouns since Match succeeded, but Value failed. This is exemplified in 0, with the outcome parallel to 0 a. above: (77) a. [SLI > N] [φ[f]]  [φ[m]] b. [SLI > N] [φ] [φ[m]] The complex nominal structure, bearing two distinct φ-values, is then inserted into the verbal domain in the complement position of the lexical verbal head V 0 (see fn. 22 above). No probing between the case-assigner and the complex lexical element is established as a precondition for case-assignment, which is initiated instantly under c-command. 22 In what follows I will exemplify non-structural case-assignment as assignment from the lexical verbal head to its complement, following Chomsky’s (1986: 293) inherent case-assignment. Nothing hinges on this, however, as long as the assignment takes place inside the verbal domain at EM. The derivational processes sketched in (76) readily carry over to the second element bearing the probing root [φ], i.e. Q S . No Agree relation is proposed between the case-assigner and the nominal domain prior to the transmission of case. Next, let us take a closer look at non-structural case-bearing V 0 in concatenation with nominal domains headed by one of those SLIs categorized as probing for [φ[f]] above, namely Poss PRO and Q W : The φ-structure of the semilexical item reduced to partial default in the isolated derivation of the nominal 22 In what follows I will exemplify non-structural case-assignment as assignment from the lexical verbal head to its complement, following Chomsky’s (1986: 293) inherent case-assignment. Nothing hinges on this, however, as long as the assignment takes place inside the verbal domain at EM. 208 3 INFERENCE - Nominal Concord & Definiteness domain with masculine or neuter head nouns since Match succeeded, but Value failed. This is exemplified in (77), with the outcome parallel to (76) a. above: (77) Nominal Concord & Definiteness I NFERENCE 150 b. Non-Structural Case As H&R (fn. 32) note, the basic distinction of a hypothetical case-node in their feature geometry would have to be made on the basis of structural vs. local or oblique cases. Woolford (2006) proposes that despite the apparent difference e.g. in predictability/ regularity (inherent case is predictable, while lexical case is idiosyncratic), inherent and lexical cases (i.e. genitive and dative in German) should be bundled as non-structural case forms in opposition to the structural ones under analysis in the last section (i.e. nominative and accusative). She reviews a number of tests for structural case-hood and concludes that the reliable ones (e.g. preservation under A-movement in passive and raising, preservation on subjects in tensed clauses) split case forms along the lines of her proposals. As Woolford argues, non-structural case is assigned early in the derivation, namely at the v P-structure in Θ-positions, hence at EM. The loci of this first Merger-operations differ, however, in that the more idiosyncratic lexical case is assigned inside VP, while predictable inherent case is licensed in (a possibly additional) v P. The application of licensing ‘at the v P-structure’ (EM) is nevertheless identical. Contrasting this with the observations on structural case made above, non-structural case ‘comes free’ in that no match for φ or any other feature category poses a precondition for its assignment in addition to the structural configuration established by External Merge of the nominal into the verbal domain. This begs the question whether the deviance in precondition moreover bears consequences for the succeeding assignment. For SLIs bearing the probing root node [φ], the system predicts a uniform assignment of non-structural case, parallel to the derivations exemplified in the last section to the exclusion of the probing operation for φ-features from the extended verbal domain (i.e. T 0 or v 0 ). This is exactly what we find with D and Q S . Observe the derivational snapshots below parallel to the three stages discussed in the last section: The nominal domain is derived on a parallel workspace in 0 a., subsequently inserted into the verbal spine in 0 b. and thereon assigned case by the lexical verbal head V 0 in 0 c., depicted here by the dotted arrow, as is exemplified with a nominal hierarchy of projections headed by DP. 22 (76) a. [D > N] [φ[*m]] [φ[m]] b. V 0 > [D > N] [φ[*m]] [φ[m]] c. V 0 > [D > N] [φ[*m]] [φ[m]] [*gen/ *dat] [*gen/ *dat] The derivational processes sketched in 0 readily carry over to the second element bearing the probing root [φ], i.e. Q S . No Agree relation is proposed between the case-assigner and the nominal domain prior to the transmission of case. Next, let us take a closer look at non-structural case-bearing V 0 in concatenation with nominal domains headed by one of those SLIs categorized as probing for [φ[f]] above, namely Poss PRO and Q W : The φ-structure of the semi-lexical item reduced to partial default in the isolated derivation of the nominal domain with masculine or neuter head nouns since Match succeeded, but Value failed. This is exemplified in 0, with the outcome parallel to 0 a. above: (77) a. [SLI > N] [φ[f]]  [φ[m]] b. [SLI > N] [φ] [φ[m]] The complex nominal structure, bearing two distinct φ-values, is then inserted into the verbal domain in the complement position of the lexical verbal head V 0 (see fn. 22 above). No probing between the case-assigner and the complex lexical element is established as a precondition for case-assignment, which is initiated instantly under c-command. 22 In what follows I will exemplify non-structural case-assignment as assignment from the lexical verbal head to its complement, following Chomsky’s (1986: 293) inherent case-assignment. Nothing hinges on this, however, as long as the assignment takes place inside the verbal domain at EM. The complex nominal structure, bearing two distinct φ-values, is then inserted into the verbal domain in the complement position of the lexical verbal head V 0 (see fn. 22 above). No probing between the case-assigner and the complex lexical element is established as a precondition for case-assignment, which is initiated instantly under c-command. (78) Nominal Concord & Definiteness I NFERENCE 151 (78) V 0 > [SLI > N] [φ] [φ[m]] At this point, recall the considerations on the locus of case-features in the feature geometrical set-up above. I therein hinted at [φ] as a potential mother node. More generally, I conjectured on the implementation of case into the geometry to return one interwoven, more complex feature structure. I will get back to these speculations immediately. Returning to the derivational snapshot under discussion, since there is no prerequisite for case-assignment with nonstructural case, the partially reduced SLI as well as the head noun receive a value from V 0 under c-command, as illustrated in 0: (79) V 0 > [SLI > N] [φ] [φ[m]] [*gen/ *dat] [*gen/ *dat] See that the elements making up the nominal domain hence share a feature at this point. This is where the observations surrounding 0 become crucial: Let us assume that case is indeed part of the feature geometrical specification, a dependent of the RE or φ-node, hence of a form similar to [φ[number/ gender][case]]. The assignment of case to the SLI would hence specify the item for the latter, but not for the former. However, case now constitutes a shared feature of the head noun and the SLI. The remaining structure of the root [φ] therefore likewise becomes part of the shared feature structure. Note that at the point of case-assignment, there is no conflict between the φ-specification of the noun and the SLI since the latter already bears a partially default root node. Hence, the residual specification of φ can readily be shared between the two LIs as well, resulting in a uniform feature structure inside the nominal domain as depicted in the schematic in 0 below: (80) V 0 > [SLI > N] [φ[*m][*gen/ *dat]] [φ[m][*gen/ *dat]] Let me discuss the second alternative from the preceding section in which case is coded apart from the feature geometry proper. Again, we start out at the derivational snapshot in 0. Again, case is a shared feature between the LIs that make up the nominal domain. Now, recall from chapter 0 that I followed Bošković (2009, 2011) in the admission of in-situ 2 nd cycle Agree-operations into the current system but additionally argued for the activation of such operations in that the probe itself has to be triggered to find a suitable goal. I want argue that this is what happens here. Again, the SLI is trivially specified for any φ-specification since it has been reduced to partially default before in-situ 2 nd cycle Agree is triggered by case-assignment. The semi-lexical head hence probes and finds a suitable goal in the head noun, with which Match and Value succeed this time, resulting in a shared φ-feature on both elements making up the nominal domain. The resulting feature configuration is illustrated in 0 below: (81) V 0 > [SLI > N] [φ[*m]] [φ[m]] [*gen/ *dat] [*gen/ *dat] As can be deduced from 0 and 0 above, the options are identical as to the outcome on the surface: All lexical elements in the nominal hierarchy of projections are fully and moreover uniformly specified for φas well as case-features. In what follows, I will assume that 0 is the result of a 2 nd cycle of probing for φ-features, since this latter option desists from referring to the additional premise of case-feature coding in the feature-geometry while returning the same explanatory power as the first alternative (but cf. chapter 3.3.4 below). That something along these lines actually underlies agreement of SLIs probing for [φ[f]] in non-structural case environments can be derived from the overt form of inflection, which resembles that of e.g. demonstratives, i.e. a φ-probing element which has been taken to always bear a value for this feature. The inflectional morpheme on the SLI is hence a portmanteau-morpheme in these cases, coding both φ and case. As I will demonstrate later in this chapter, these proposals are moreover backed up by properties of adjectival inflection. We have then encountered the first instance of an in-situ 2 nd cycle of Agree, initiated by the ‘free’ assignment of nonstructural case to SLIs bearing the partially default root node [φ]. The processes elaborated above naturally carry over to derivations with neuter HNs. Note that for derivations including feminine and plural noun phrases (i.e. derivations At this point, recall the considerations on the locus of case-features in the feature geometrical set-up above. I therein hinted at [φ] as a potential mother node. More generally, I conjectured on the implementation of case into the geometry to return one interwoven, more complex feature structure. I will get back to these speculations immediately. Returning to the derivational snapshot under discussion, since there is no prerequisite for case-assignment with non-structural case, the partially reduced SLI as well as the head noun receive a value from V 0 under c-command, as illustrated in (79): (79) Nominal Concord & Definiteness I NFERENCE 151 (78) V 0 > [SLI > N] [φ] [φ[m]] At this point, recall the considerations on the locus of case-features in the feature geometrical set-up above. I therein hinted at [φ] as a potential mother node. More generally, I conjectured on the implementation of case into the geometry to return one interwoven, more complex feature structure. I will get back to these speculations immediately. Returning to the derivational snapshot under discussion, since there is no prerequisite for case-assignment with nonstructural case, the partially reduced SLI as well as the head noun receive a value from V 0 under c-command, as illustrated in 0: (79) V 0 > [SLI > N] [φ] [φ[m]] [*gen/ *dat] [*gen/ *dat] See that the elements making up the nominal domain hence share a feature at this point. This is where the observations surrounding 0 become crucial: Let us assume that case is indeed part of the feature geometrical specification, a dependent of the RE or φ-node, hence of a form similar to [φ[number/ gender][case]]. The assignment of case to the SLI would hence specify the item for the latter, but not for the former. However, case now constitutes a shared feature of the head noun and the SLI. The remaining structure of the root [φ] therefore likewise becomes part of the shared feature structure. Note that at the point of case-assignment, there is no conflict between the φ-specification of the noun and the SLI since the latter already bears a partially default root node. Hence, the residual specification of φ can readily be shared between the two LIs as well, resulting in a uniform feature structure inside the nominal domain as depicted in the schematic in 0 below: (80) V 0 > [SLI > N] [φ[*m][*gen/ *dat]] [φ[m][*gen/ *dat]] Let me discuss the second alternative from the preceding section in which case is coded apart from the feature geometry proper. Again, we start out at the derivational snapshot in 0. Again, case is a shared feature between the LIs that make up the nominal domain. Now, recall from chapter 0 that I followed Bošković (2009, 2011) in the admission of in-situ 2 nd cycle Agree-operations into the current system but additionally argued for the activation of such operations in that the probe itself has to be triggered to find a suitable goal. I want argue that this is what happens here. Again, the SLI is trivially specified for any φ-specification since it has been reduced to partially default before in-situ 2 nd cycle Agree is triggered by case-assignment. The semi-lexical head hence probes and finds a suitable goal in the head noun, with which Match and Value succeed this time, resulting in a shared φ-feature on both elements making up the nominal domain. The resulting feature configuration is illustrated in 0 below: (81) V 0 > [SLI > N] [φ[*m]] [φ[m]] [*gen/ *dat] [*gen/ *dat] As can be deduced from 0 and 0 above, the options are identical as to the outcome on the surface: All lexical elements in the nominal hierarchy of projections are fully and moreover uniformly specified for φas well as case-features. In what follows, I will assume that 0 is the result of a 2 nd cycle of probing for φ-features, since this latter option desists from referring to the additional premise of case-feature coding in the feature-geometry while returning the same explanatory power as the first alternative (but cf. chapter 3.3.4 below). That something along these lines actually underlies agreement of SLIs probing for [φ[f]] in non-structural case environments can be derived from the overt form of inflection, which resembles that of e.g. demonstratives, i.e. a φ-probing element which has been taken to always bear a value for this feature. The inflectional morpheme on the SLI is hence a portmanteau-morpheme in these cases, coding both φ and case. As I will demonstrate later in this chapter, these proposals are moreover backed up by properties of adjectival inflection. We have then encountered the first instance of an in-situ 2 nd cycle of Agree, initiated by the ‘free’ assignment of nonstructural case to SLIs bearing the partially default root node [φ]. The processes elaborated above naturally carry over to derivations with neuter HNs. Note that for derivations including feminine and plural noun phrases (i.e. derivations See that the elements making up the nominal domain hence share a feature at this point. This is where the observations surrounding (78) become crucial: Let us assume that case is indeed part of the feature geometrical specification, a dependent of the re or φ-node, hence of a form similar to [φ[number/ gender] [case]]. The assignment of case to the SLI would hence specify the item for the latter, but not for the former. However, case now constitutes a shared feature 3.3 Splitting Nominal Concord 209 of the head noun and the SLI. The remaining structure of the root [φ] therefore likewise becomes part of the shared feature structure. Note that at the point of case-assignment, there is no conflict between the φ-specification of the noun and the SLI since the latter already bears a partially default root node. Hence, the residual specification of φ can readily be shared between the two LIs as well, resulting in a uniform feature structure inside the nominal domain as depicted in the schematic in (80) below: (80) Nominal Concord & Definiteness I NFERENCE 151 (78) V 0 > [SLI > N] [φ] [φ[m]] At this point, recall the considerations on the locus of case-features in the feature geometrical set-up above. I therein hinted at [φ] as a potential mother node. More generally, I conjectured on the implementation of case into the geometry to return one interwoven, more complex feature structure. I will get back to these speculations immediately. Returning to the derivational snapshot under discussion, since there is no prerequisite for case-assignment with nonstructural case, the partially reduced SLI as well as the head noun receive a value from V 0 under c-command, as illustrated in 0: (79) V 0 > [SLI > N] [φ] [φ[m]] [*gen/ *dat] [*gen/ *dat] See that the elements making up the nominal domain hence share a feature at this point. This is where the observations surrounding 0 become crucial: Let us assume that case is indeed part of the feature geometrical specification, a dependent of the RE or φ-node, hence of a form similar to [φ[number/ gender][case]]. The assignment of case to the SLI would hence specify the item for the latter, but not for the former. However, case now constitutes a shared feature of the head noun and the SLI. The remaining structure of the root [φ] therefore likewise becomes part of the shared feature structure. Note that at the point of case-assignment, there is no conflict between the φ-specification of the noun and the SLI since the latter already bears a partially default root node. Hence, the residual specification of φ can readily be shared between the two LIs as well, resulting in a uniform feature structure inside the nominal domain as depicted in the schematic in 0 below: (80) V 0 > [SLI > N] [φ[*m][*gen/ *dat]] [φ[m][*gen/ *dat]] Let me discuss the second alternative from the preceding section in which case is coded apart from the feature geometry proper. Again, we start out at the derivational snapshot in 0. Again, case is a shared feature between the LIs that make up the nominal domain. Now, recall from chapter 0 that I followed Bošković (2009, 2011) in the admission of in-situ 2 nd cycle Agree-operations into the current system but additionally argued for the activation of such operations in that the probe itself has to be triggered to find a suitable goal. I want argue that this is what happens here. Again, the SLI is trivially specified for any φ-specification since it has been reduced to partially default before in-situ 2 nd cycle Agree is triggered by case-assignment. The semi-lexical head hence probes and finds a suitable goal in the head noun, with which Match and Value succeed this time, resulting in a shared φ-feature on both elements making up the nominal domain. The resulting feature configuration is illustrated in 0 below: (81) V 0 > [SLI > N] [φ[*m]] [φ[m]] [*gen/ *dat] [*gen/ *dat] As can be deduced from 0 and 0 above, the options are identical as to the outcome on the surface: All lexical elements in the nominal hierarchy of projections are fully and moreover uniformly specified for φas well as case-features. In what follows, I will assume that 0 is the result of a 2 nd cycle of probing for φ-features, since this latter option desists from referring to the additional premise of case-feature coding in the feature-geometry while returning the same explanatory power as the first alternative (but cf. chapter 3.3.4 below). That something along these lines actually underlies agreement of SLIs probing for [φ[f]] in non-structural case environments can be derived from the overt form of inflection, which resembles that of e.g. demonstratives, i.e. a φ-probing element which has been taken to always bear a value for this feature. The inflectional morpheme on the SLI is hence a portmanteau-morpheme in these cases, coding both φ and case. As I will demonstrate later in this chapter, these proposals are moreover backed up by properties of adjectival inflection. We have then encountered the first instance of an in-situ 2 nd cycle of Agree, initiated by the ‘free’ assignment of nonstructural case to SLIs bearing the partially default root node [φ]. The processes elaborated above naturally carry over to derivations with neuter HNs. Note that for derivations including feminine and plural noun phrases (i.e. derivations Let me discuss the second alternative from the preceding section in which case is coded apart from the feature geometry proper. Again, we start out at the derivational snapshot in (79). Again, case is a shared feature between the LIs that make up the nominal domain. Now, recall from chapter 3.1.2 that I followed Bošković (2009, 2011) in the admission of in-situ 2 nd cycle Agree-operations into the current system but additionally argued for the activation of such operations in that the probe itself has to be triggered to find a suitable goal. I want argue that this is what happens here. Again, the SLI is trivially specified for any φ-specification since it has been reduced to partially default before in-situ 2 nd cycle Agree is triggered by case-assignment. The semi-lexical head hence probes and finds a suitable goal in the head noun, with which Match and Value succeed this time, resulting in a shared φ-feature on both elements making up the nominal domain. The resulting feature configuration is illustrated in (81) below: (81) Nominal Concord & Definiteness I NFERENCE 151 (78) V 0 > [SLI > N] [φ] [φ[m]] At this point, recall the considerations on the locus of case-features in the feature geometrical set-up above. I therein hinted at [φ] as a potential mother node. More generally, I conjectured on the implementation of case into the geometry to return one interwoven, more complex feature structure. I will get back to these speculations immediately. Returning to the derivational snapshot under discussion, since there is no prerequisite for case-assignment with nonstructural case, the partially reduced SLI as well as the head noun receive a value from V 0 under c-command, as illustrated in 0: (79) V 0 > [SLI > N] [φ] [φ[m]] [*gen/ *dat] [*gen/ *dat] See that the elements making up the nominal domain hence share a feature at this point. This is where the observations surrounding 0 become crucial: Let us assume that case is indeed part of the feature geometrical specification, a dependent of the RE or φ-node, hence of a form similar to [φ[number/ gender][case]]. The assignment of case to the SLI would hence specify the item for the latter, but not for the former. However, case now constitutes a shared feature of the head noun and the SLI. The remaining structure of the root [φ] therefore likewise becomes part of the shared feature structure. Note that at the point of case-assignment, there is no conflict between the φ-specification of the noun and the SLI since the latter already bears a partially default root node. Hence, the residual specification of φ can readily be shared between the two LIs as well, resulting in a uniform feature structure inside the nominal domain as depicted in the schematic in 0 below: (80) V 0 > [SLI > N] [φ[*m][*gen/ *dat]] [φ[m][*gen/ *dat]] Let me discuss the second alternative from the preceding section in which case is coded apart from the feature geometry proper. Again, we start out at the derivational snapshot in 0. Again, case is a shared feature between the LIs that make up the nominal domain. Now, recall from chapter 0 that I followed Bošković (2009, 2011) in the admission of in-situ 2 nd cycle Agree-operations into the current system but additionally argued for the activation of such operations in that the probe itself has to be triggered to find a suitable goal. I want argue that this is what happens here. Again, the SLI is trivially specified for any φ-specification since it has been reduced to partially default before in-situ 2 nd cycle Agree is triggered by case-assignment. The semi-lexical head hence probes and finds a suitable goal in the head noun, with which Match and Value succeed this time, resulting in a shared φ-feature on both elements making up the nominal domain. The resulting feature configuration is illustrated in 0 below: (81) V 0 > [SLI > N] [φ[*m]] [φ[m]] [*gen/ *dat] [*gen/ *dat] As can be deduced from 0 and 0 above, the options are identical as to the outcome on the surface: All lexical elements in the nominal hierarchy of projections are fully and moreover uniformly specified for φas well as case-features. In what follows, I will assume that 0 is the result of a 2 nd cycle of probing for φ-features, since this latter option desists from referring to the additional premise of case-feature coding in the feature-geometry while returning the same explanatory power as the first alternative (but cf. chapter 3.3.4 below). That something along these lines actually underlies agreement of SLIs probing for [φ[f]] in non-structural case environments can be derived from the overt form of inflection, which resembles that of e.g. demonstratives, i.e. a φ-probing element which has been taken to always bear a value for this feature. The inflectional morpheme on the SLI is hence a portmanteau-morpheme in these cases, coding both φ and case. As I will demonstrate later in this chapter, these proposals are moreover backed up by properties of adjectival inflection. We have then encountered the first instance of an in-situ 2 nd cycle of Agree, initiated by the ‘free’ assignment of nonstructural case to SLIs bearing the partially default root node [φ]. The processes elaborated above naturally carry over to derivations with neuter HNs. Note that for derivations including feminine and plural noun phrases (i.e. derivations As can be deduced from (80) and (81) above, the options are identical as to the outcome on the surface: All lexical elements in the nominal hierarchy of projections are fully and moreover uniformly specified for φas well as case-features. In what follows, I will assume that (81) is the result of a 2 nd cycle of probing for φ-features, since this latter option desists from referring to the additional premise of case-feature coding in the feature-geometry while returning the same explanatory power as the first alternative (but cf. chapter 3.3.4 below). That something along these lines actually underlies agreement of SLIs probing for [φ[f]] in non-structural case environments can be derived from the overt form of inflection, which resembles that of e.g. demonstratives, i.e. a φ-probing 210 3 INFERENCE - Nominal Concord & Definiteness element which has been taken to always bear a value for this feature. The inflectional morpheme on the SLI is hence a portmanteau-morpheme in these cases, coding both φ and case. As I will demonstrate later in this chapter, these proposals are moreover backed up by properties of adjectival inflection. We have then encountered the first instance of an in-situ 2 nd cycle of Agree, initiated by the ‘free’ assignment of non-structural case to SLIs bearing the partially default root node [φ]. The processes elaborated above naturally carry over to derivations with neuter HNs. Note that for derivations including feminine and plural noun phrases (i.e. derivations incorporating NumP), no mismatch is proposed between the φ-features of the nominal and the SLI, parallel to the uniform derivations of all number/ gender-specifications with D and Q S . Once again, I have implemented labels instead of lexical items in the derivational outlines of the analysis above to ease comprehension of the processes involved. Let me now present the paradigms of weak quantifiers and possessive pronouns in all structural and non-structural cases with feminine and plural as well as masculine and neuter head nouns. I will once again defer commenting on the optionality of inflection on the former SLI and simply build on the contrast between possible (optional as well as obligatory) and impossible realizations of inflection here: (82) a. Q W fem pl nom viel(-e) Milch viel-e Freunde gen viel(-er) Milch viel-er Freunde dat viel(-er) Milch viel-en Freunden acc viel(-e) Milch viel-e Freunde b. Q W masc neut nom viel Wein viel Bier gen viel(-en) Weins viel(-en) Biers dat viel(-em) Wein viel(-em) Bier acc viel Wein viel Bier While optional inflection is observed in all cases in concatenation of the weak quantifier with feminine nominals, it is uniformly obligatory based on linearizations with plural head nouns (cf. (83) a.). With masculine and neuter nominals, however, the cut between optional and impossible inflections mirrors the one 3.3 Splitting Nominal Concord 211 between structural ((83) b.) and non-structural ((83) c.) cases, hence the cut between φ-complete and partial default status of the SLI as elaborated above. (83) Nominal Concord & Definiteness I NFERENCE 152 incorporating NumP), no mismatch is proposed between the φ-features of the nominal and the SLI, parallel to the uniform derivations of all number/ gender-specifications with D and Q S . Once again, I have implemented labels instead of lexical items in the derivational outlines of the analysis above to ease comprehension of the processes involved. Let me now present the paradigms of weak quantifiers and possessive pronouns in all structural and non-structural cases with feminine and plural as well as masculine and neuter head nouns. I will once again defer commenting on the optionality of inflection on the former SLI and simply build on the contrast between possible (optional as well as obligatory) and impossible realizations of inflection here: (82) a. Q W fem pl nom viel(-e) Milch viel-e Freunde gen viel(-er) Milch viel-er Freunde dat viel(-er) Milch viel-en Freunden acc viel(-e) Milch viel-e Freunde b. Q W masc neut nom viel Wein viel Bier gen viel(-en) Weins viel(-en) Biers dat viel(-em) Wein viel(-em) Bier acc viel Wein viel Bier While optional inflection is observed in all cases in concatenation of the weak quantifier with feminine nominals, it is uniformly obligatory based on linearizations with plural head nouns (cf. 0 a.). With masculine and neuter nominals, however, the cut between optional and impossible inflections mirrors the one between structural (0 b.) and non-structural (0 c.) cases, hence the cut between φ-complete and partial default status of the SLI as elaborated above. (83) a. viel(-e) Milch [Q W > Num] [φ[f]] [φ[f]] [*nom] [*nom] b. viel-ø Wein [Q W > N] [φ] [φ[m]] [*nom] c. viel-em Wein [Q W > N] [φ[*m]] [φ[m]] [*dat] [*dat] Turning to the possessive pronoun, a parallel picture emerges: (84) a. Poss PRO fem pl nom mein-e Milch mein-e Freunde gen mein-er Milch mein-er Freunde dat mein-er Milch mein-en Freunden acc mein-e Milch mein-e Freunde b. Poss PRO masc neut nom mein Wein mein Bier gen mein-es Weins mein-es Biers dat mein-em Wein mein-em Bier acc mein-en Wein mein Bier As can be derived from 0 a., the possessive pronoun uniformly displays φ-inflection in both structural as well as nonstructural cases when concatenated with feminine and plural nominals. Concentrating on 0 b., however, a clear cut Turning to the possessive pronoun, a parallel picture emerges: (84) a. Poss PRO fem pl nom mein-e Milch mein-e Freunde gen mein-er Milch mein-er Freunde dat mein-er Milch mein-en Freunden acc mein-e Milch mein-e Freunde b. Poss PRO masc neut nom mein Wein mein Bier gen mein-es Weins mein-es Biers dat mein-em Wein mein-em Bier acc mein-en Wein mein Bier As can be derived from (84) a., the possessive pronoun uniformly displays φ-inflection in both structural as well as non-structural cases when concatenated with feminine and plural nominals. Concentrating on (84) b., however, a clear cut between these case forms is observable in that no inflection can be found in structural cases other than accusative masculine, while φ-agreement is 212 3 INFERENCE - Nominal Concord & Definiteness observed in all non-structural cases with masculine and neuter head nominals. Recall from chapter 2.3.3 that adjectival inflection with the possessive pronoun in accusative masculine is homonymous between the strong and the weak inflectional pattern. Therein, it was concluded that this form can hence not be readily attributed to either inflectional pattern. With the analysis elaborated above, we can find a hint that the inflectional morpheme on the SLI itself does not encode φ-features (i.e. does not constitute a portmanteau-morpheme but simple case inflection, cf. fn. 29, fn. 33 below). The link between these observations will be formally made below in the analysis of adjectival inflection in German. For now, let me just state that the occurrence of an inflectional suffix on Poss PRO does not readily undermine the analysis put forth in this section. Let me again bring together the concatenations of the SLI and HN with the featural set-up proposed in the analysis above: (85) Nominal Concord & Definiteness I NFERENCE 153 between these case forms is observable in that no inflection can be found in structural cases other than accusative masculine, while φ-agreement is observed in all non-structural cases with masculine and neuter head nominals. Recall from chapter 2.3.3 that adjectival inflection with the possessive pronoun in accusative masculine is homonymous between the STRONG and the WEAK inflectional pattern. Therein, it was concluded that this form can hence not be readily attributed to either inflectional pattern. With the analysis elaborated above, we can find a hint that the inflectional morpheme on the SLI itself does not encode φ-features (i.e. does not constitute a portmanteau-morpheme but simple case inflection, cf. fn. 30, fn. 34 below). The link between these observations will be formally made below in the analysis of adjectival inflection in German. For now, let me just state that the occurrence of an inflectional suffix on Poss PRO does not readily undermine the analysis put forth in this section. Let me again bring together the concatenations of the SLI and HN with the featural setup proposed in the analysis above: (85) a. mein-e Milch [Poss PRO > Num] [φ[f]] [φ[f ]] [*nom] [*nom] b. mein Wein [Poss PRO > N] [φ] [φ[m]] [*nom] c. mein-em Wein [Poss PRO > N] [φ[*m]] [φ[m]] [*dat] [*dat] In this section, I have taken the case-paradigm of German into account in deriving a stringent approach to the φ-featural set-up of semi-lexical items in the nominal domain. As has been demonstrated, the means of case-assignment in the clausal domain play a crucial role for the establishment of shared features between LIs making up the nominal spine: Structural case is assigned on the basis of Match/ Value of φ-features on the case-‘bearer’ T (and v ) from the nominal domain. T therein probes for φ-complete goals, hence skipping φ-defective (i.e. default) elements in the nominal spine parallel to the verbal domain with solely person-bearing expletives. Non-structural case - in contrast - is assigned at external Merger and hence ‘comes free’. SLIs probing for the root node [φ] are trivially compatible with any nominal φ-specification; they therefore form uniform domains with any head nominal w.r.t. φ-specification and hence partake in structural as well as non-structural case-assignment. On the other hand, SLIs which have been taken to match only a subset of the possible φ-specifications on nominals in chapter 0 are ‘ignored’ in the assignment of structural case. This is reflected in the absence of inflection on the surface (to the exclusion of the accusative masculine with possessive pronouns, to be analyzed in detail below). With non-structural case, however, the element is probed by the case-assigner V 0 (and possibly v 0 , cf. Woolford 2006) and can hence initiate a 2 nd cycle of probing for φ-features, this time likewise trivially specified for any φconfiguration on par with D and Q S . Therefore, we find a clear cut in the overt inflectional pattern, which will be associated with the idiosyncratic adjectival inflection that accompanies the (forms of the) elements below. c. 2 nd cycle Agreement by Internal Merge Before turning to more complex nominal domains incorporating more than one of the SLIs classified above, I want to end the discussion of nominal hierarchies headed by a single SLI in the analysis of the second configuration triggering 2 nd cycle Agree, namely probing of a head into its specifier position (cf. Rezac 2003) by IM of an appropriate goal (cf. Bejar 2003). It will be shown that in contrast to the original proposals for the unavailability of the successful establishment of a probegoal relation, in which an intervener matches the probe for the appropriate feature but fails to value it, the mismatch is attributed to the featural incompatibility of LIs in the nominal domain in German. In what follows, Agree will hence apply from the same probe to the same goal successively, but in two diverging structural configurations (the domain of the probe and - what Rezac calls - the extended search space, i.e. its specifier). More has to be said on these processes below in the analysis of the extended pattern of SLI concord; for now, I will only present one set of data from mono-SLI nominal domains which is readily analyzable along these lines. Recall from the analysis of structural case and nominals headed by SLIs bearing a partially default φ-configuration that these nominal hierarchies are raised in toto to the probing head’s specifier even though only one element inside the domain might match the probing for a φ-complete feature structure. The uniform movement was attributed to the phasal status of the SLI omitted in the establishment of Match (and later, the assignment of case). The inability of phasal complements to move from beneath a phase head, due to conflicting In this section, I have taken the case-paradigm of German into account in deriving a stringent approach to the φ-featural set-up of semi-lexical items in the nominal domain. As has been demonstrated, the means of case-assignment in the clausal domain play a crucial role for the establishment of shared features between LIs making up the nominal spine: Structural case is assigned on the basis of Match/ Value of φ-features on the case-‘bearer’ T (and v ) from the nominal domain. T therein probes for φ-complete goals, hence skipping φ-defective (i.e. default) elements in the nominal spine parallel to the verbal domain with solely person-bearing expletives. Non-structural case - in contrast - is assigned at external Merger and hence ‘comes free’. SLIs probing for the root node [φ] are trivially compatible with any nominal φ-specification; they therefore form uniform domains with any head nominal w.r.t. φ-specification and hence partake in structural as well as non-structural case-assignment. On the other hand, SLIs 3.3 Splitting Nominal Concord 213 which have been taken to match only a subset of the possible φ-specifications on nominals in chapter 3.2.1 are ‘ignored’ in the assignment of structural case. This is reflected in the absence of inflection on the surface (to the exclusion of the accusative masculine with possessive pronouns, to be analyzed in detail below). With non-structural case, however, the element is probed by the case-assigner V 0 (and possibly v 0 , cf. Woolford 2006) and can hence initiate a 2 nd cycle of probing for φ-features, this time likewise trivially specified for any φ-configuration on par with D and Q S . Therefore, we find a clear cut in the overt inflectional pattern, which will be associated with the idiosyncratic adjectival inflection that accompanies the (forms of the) elements below. c. 2 nd cycle Agreement by Internal Merge Before turning to more complex nominal domains incorporating more than one of the SLIs classified above, I want to end the discussion of nominal hierarchies headed by a single SLI in the analysis of the second configuration triggering 2 nd cycle Agree, namely probing of a head into its specifier position (cf. Rezac 2003) by IM of an appropriate goal (cf. Bejar 2003). It will be shown that in contrast to the original proposals for the unavailability of the successful establishment of a probe-goal relation, in which an intervener matches the probe for the appropriate feature but fails to value it, the mismatch is attributed to the featural incompatibility of LIs in the nominal domain in German. In what follows, Agree will hence apply from the same probe to the same goal successively, but in two diverging structural configurations (the domain of the probe and - what Rezac calls - the extended search space, i.e. its specifier). More has to be said on these processes below in the analysis of the extended pattern of SLI concord; for now, I will only present one set of data from mono-SLI nominal domains which is readily analyzable along these lines. Recall from the analysis of structural case and nominals headed by SLIs bearing a partially default φ-configuration that these nominal hierarchies are raised in toto to the probing head’s specifier even though only one element inside the domain might match the probing for a φ-complete feature structure. The uniform movement was attributed to the phasal status of the SLI omitted in the establishment of Match (and later, the assignment of case). The inability of phasal complements to move from beneath a phase head, due to conflicting requirements of the PIC and Anti-Locality, was one of the testing grounds for phasehood established in chapter 2.2. The requirement is however vastly relaxed, if not nullified by the addition of a focus particle like nur ‘only’ at the left periphery of the nominal spine. Note (86), in which the nominal Kekse ‘cookies’ is able to move from below the rigid phase head D once the latter is concatenated with the particle: 214 3 INFERENCE - Nominal Concord & Definiteness (86) a. Ich habe [nur dies-e Kekse] gegessen I have [ only these cookies ] eaten b. Kekse i habe ich [nur dies-e t i ] gegessen. cookies i have I [ only these t i ] eaten ‘I have only eaten these cookies.’ Extraction is not dependent on the phasal status of the raised element (recall from chapter 2.3 that Num[f[pl]] ceases to bear said property with EM of any head of the higher phasal cycle); as can be deduced from (87), nominal heads bearing e.g. [φ[n]] might likewise move into the clausal spine. (87) a. Ich habe [nur dies-es Bier] getrunken. I have [ only this beer ] drunk b. Bier i habe ich [nur dies-es t i ] getrunken. beer i have I [ only this t i ] drunk ‘I have only drunk this beer.’ Recall that the demonstrative has been categorized as probing for the root [φ] above and hence successfully matches and agrees with any specification on Num/ N, resulting in a shared feature structure established in the isolated derivation of DP: (88) Nominal Concord & Definiteness I NFERENCE 154 requirements of the PIC and Anti-Locality, was one of the testing grounds for phasehood established in chapter 2.2. The requirement is however vastly relaxed, if not nullified by the addition of a focus particle like nur ‘only’ at the left periphery of the nominal spine. Note 0, in which the nominal Kekse ‘cookies’ is able to move from below the rigid phase head D once the latter is concatenated with the particle: (86) a. Ich habe [nur dies-e Kekse] gegessen. I have [ only these cookies ] eaten b. Kekse i habe ich [nur dies-e t i ] gegessen. cookies i have I [ only these t i ] eaten ‘I have only eaten these cookies.’ Extraction is not dependent on the phasal status of the raised element (recall from chapter 2.3 that Num[f[pl]] ceases to bear said property with EM of any head of the higher phasal cycle); as can be deduced from 0, nominal heads bearing e.g. [φ[n]] might likewise move into the clausal spine. (87) a. Ich habe [nur dies-es Bier] getrunken. I have [ only this beer ] drunk b. Bier i habe ich [nur dies-es t i ] getrunken. beer i have I [ only this t i ] drunk ‘I have only drunk this beer.’ Recall that the demonstrative has been categorized as probing for the root [φ] above and hence successfully matches and agrees with any specification on Num/ N, resulting in a shared feature structure established in the isolated derivation of DP: (88) dies-es > Bier D N [φ[*n]] [φ[n]] Match:  Value:  As we know by now, there are nominal domains that do not exhibit such a uniform feature configuration at the moment when they externally merge into the verbal domain, namely those headed by the subgroup of SLIs probing for [φ[f]], i.e. Poss PRO and Q W , to be reduced to partial default in the isolated derivation with masculine and neuter head nouns prior to External Merge into the verbal spine: 23 (89) mein > Bier Poss PRO N i. [φ[f]]  [φ[n]] ii. [φ] [φ[n]] Match:  Value:  Observe the morphological reflexes on these φ-arbitrary nominal domains when inserted into the testing environment presented above. (90) a. Ich habe [nur mein Bier] getrunken. I have [ only my beer ] drunk b. Bier i habe ich [nur mein-es t i ] getrunken. beer i have I [ only mine t i ] drunk ‘I have only drunk this beer.’ As can be deduced from 0, the possessive pronoun exhibits overt inflection once the HN is moved out of the complement. This effect is uniformly observable with neuter as well as masculine nominals; however, recall from the discussion above 23 Observe that the co-occurrence of the focus particle ‘only’ with the weak quantifier is ill-formed for semantic reasons. I hence restrict myself to extraction from below the possessive pronoun for now. As we know by now, there are nominal domains that do not exhibit such a uniform feature configuration at the moment when they externally merge into the verbal domain, namely those headed by the subgroup of SLIs probing for [φ[f]], i.e. Poss PRO and Q W , to be reduced to partial default in the isolated derivation with masculine and neuter head nouns prior to External Merge into the verbal spine: 23 23 Observe that the co-occurrence of the focus particle ‘only’ with the weak quantifier is ill-formed for semantic reasons. I hence restrict myself to extraction from below the possessive pronoun for now. 3.3 Splitting Nominal Concord 215 (89) Nominal Concord & Definiteness I NFERENCE 154 requirements of the PIC and Anti-Locality, was one of the testing grounds for phasehood established in chapter 2.2. The requirement is however vastly relaxed, if not nullified by the addition of a focus particle like nur ‘only’ at the left periphery of the nominal spine. Note 0, in which the nominal Kekse ‘cookies’ is able to move from below the rigid phase head D once the latter is concatenated with the particle: (86) a. Ich habe [nur dies-e Kekse] gegessen. I have [ only these cookies ] eaten b. Kekse i habe ich [nur dies-e t i ] gegessen. cookies i have I [ only these t i ] eaten ‘I have only eaten these cookies.’ Extraction is not dependent on the phasal status of the raised element (recall from chapter 2.3 that Num[f[pl]] ceases to bear said property with EM of any head of the higher phasal cycle); as can be deduced from 0, nominal heads bearing e.g. [φ[n]] might likewise move into the clausal spine. (87) a. Ich habe [nur dies-es Bier] getrunken. I have [ only this beer ] drunk b. Bier i habe ich [nur dies-es t i ] getrunken. beer i have I [ only this t i ] drunk ‘I have only drunk this beer.’ Recall that the demonstrative has been categorized as probing for the root [φ] above and hence successfully matches and agrees with any specification on Num/ N, resulting in a shared feature structure established in the isolated derivation of DP: (88) dies-es > Bier D N [φ[*n]] [φ[n]] Match:  Value:  As we know by now, there are nominal domains that do not exhibit such a uniform feature configuration at the moment when they externally merge into the verbal domain, namely those headed by the subgroup of SLIs probing for [φ[f]], i.e. Poss PRO and Q W , to be reduced to partial default in the isolated derivation with masculine and neuter head nouns prior to External Merge into the verbal spine: 23 (89) mein > Bier Poss PRO N i. [φ[f]]  [φ[n]] ii. [φ] [φ[n]] Match:  Value:  Observe the morphological reflexes on these φ-arbitrary nominal domains when inserted into the testing environment presented above. (90) a. Ich habe [nur mein Bier] getrunken. I have [ only my beer ] drunk b. Bier i habe ich [nur mein-es t i ] getrunken. beer i have I [ only mine t i ] drunk ‘I have only drunk this beer.’ As can be deduced from 0, the possessive pronoun exhibits overt inflection once the HN is moved out of the complement. This effect is uniformly observable with neuter as well as masculine nominals; however, recall from the discussion above 23 Observe that the co-occurrence of the focus particle ‘only’ with the weak quantifier is ill-formed for semantic reasons. I hence restrict myself to extraction from below the possessive pronoun for now. Observe the morphological reflexes on these φ-arbitrary nominal domains when inserted into the testing environment presented above. (90) a. Ich habe [nur mein Bier] getrunken I have [ only my beer ] drunk b. Bier i habe ich [nur mein-es t i ] getrunken. beer i have I [ only mine t i ] drunk ‘I have only drunk my beer.’ As can be deduced from (90), the possessive pronoun exhibits overt inflection once the HN is moved out of the complement. This effect is uniformly observable with neuter as well as masculine nominals; however, recall from the discussion above that the pronoun does exhibit inflection in the accusative masculine, the case of the internal argument, to be further analyzed below: (91) Ich habe [nur mein-en Wein] getrunken. I have [ only my wine ] ACC drunk When extracting the nominal out of the domain merged into the external argument (i.e. nominative) position, however, the effect is observable once again: (92) a. Mir schmeckt [nur mein Wein]. I like [ only my wine ] NOM b. Wein i schmeckt mir [nur mein-er t i ]. wine i like I [ only mine t i ] NOM Under the theory elaborated here, the analysis is straightforward: The φ-feature of the pronoun is trivially specified for any value after reduction to partial default in the course of the isolated derivation of the nominal domain and is still active when the nominal complex is inserted into the clausal structure. By subsequent movement of HN into the specifier of the formerly unsuccessful 216 3 INFERENCE - Nominal Concord & Definiteness probe, Poss PRO can initiate a 2 nd cycle of Agree applying at the extended search space, i.e. its specifier. Observe that the mechanisms proposed here induce a violation of Anti-Locality in that the complement of the possessive finds an intermediary landing-site in the specifier of the same phrase. In the identification of contextual phase heads in the nominal domain in chapter 2.2, this restriction has been assumed to apply universally at every phrase-level; it has been indeed independently argued for both in terms of Anti-Locality (cf. Abels 2003, already mentioned in the context of chapter 2.2 above) as well as in terms of the ban on phraseinternal movement (cf. e.g. Kayne 2005, also references in Boeckx 2008: 106). Above, the conflicting restrictions of Anti-Locality and the Phase Impenetrability Condition (PIC) were furthermore held responsible for the inability of phasal complements to extract: While the PIC forces movement to proceed through the specifier of the phase head, i.e. the edge, Anti-Locality forbids phrase-internal/ too short movement from the complement to the specifier of the ph(r)ase-head. The latter is hence more restrictive. To account for the data above in terms of the analysis laid out so far in this chapter, modifications to the system employed in chapter 2.2 above do indeed have to target Anti-Locality rather than the PIC since movement through the specifier position constitutes exactly the path we want to propose in these cases. With this in mind, note that Grohmann’s (2000, 2003a) original claims about too short movement were made on the basis of the notion (prolific) domain, consisting of several phrases, rather than on the basis of single maximal projections. Movement therein was categorized as too short if no boundary between two domains was crossed. The relevant notions are given below: (93) (= Grohmann 2003a: ch. 2, (31)) Prolific Domain Let a Prolific Domain ΠΔ be a contextually defined part of C HL : (i) Each ΠΔ spells out its context information and (ii) Spell Out feeds the PF and LF interface levels. (94) (= Grohmann 2003a: ch. 2, (30)) Clausal Tripartition: What’s in a Prolific Domain (ΠΔ) (i) Θ-Domain: part of derivation where thematic relations are created (ii) Φ-Domain: part of derivation where agreement properties are licensed (iii) Ω-Domain: part of derivation where discourse information is established 3.3 Splitting Nominal Concord 217 (95) (= Grohmann 2003a: ch. 2, (35)) Condition on Domain Exclusivity (CDE) For a given Prolific Domain ΠΔ, an object O in the phrase-marker must receive an exclusive interpretation at the interfaces, unless duplicity of O yields a drastic effect on the output of that ΠΔ. With (93)-(95) in mind, let us turn to the nominal domain. Grohmann & Haegeman (2003) as well as Grohmann & Panagiotidis (2005) attempt to transpose the system sketched above to the nominal hierarchy of projections. Focusing on the tripartition established in (94), both argue with Abney (1987) for a parallel set-up of the nominal hierarchy with that of the clausal domain. However, as can be deduced from chapter one and mentioned by the authors themselves, agreement on the number and the structural and featural set-up of the phrases in the former domain are much more debatable than the generally agreed upon tripartition of the latter. Moreover, it is problematic for the authors to assign those phrase heads they identify in the nominal hierarchy to the respective prolific domains presented above. Grohmann & Panagiotidis (2005: 254), then, take agreement with HN in Modern Greek, the language under investigation in their analysis, to be a property of elements externally merged in the medial (i.e. Φ-)Domain of the nominal spine. This property thus bundles adjectives together with demonstratives. DP, to which the demonstrative might move, however, is taken as the equivalent of the CP-domain in the clause (in line with Abney 1987 and the finer articulation of Rizzi 1997; see also the analysis in Grohmann & Haegeman 2003 as well as chapter 2.3.4 above) without further comment. Another phrasal proposal that is accompanied by yet a different set-up of nominal prolific domains can be found in Ticio’s (2010) analysis of the nominal hierarchy in Spanish. The proposed nominal spine that her analysis is based on is exemplified below: (96) (following Ticio 2010: (56)) [FocP > DP] Ω > [AgrP] Φ > [ n P > NP] Θ Several comments are in order here. Note first that the little n P, parallel to Larson’s (1988) v P-shell, is taken to host the generic merging-site for agent arguments, i.e. Spec, n P. Likewise, possessors are said to externally merge in the specifier of the next higher phrase, AgrP. Ticio departs from the rationale of inflected (S)LIs as part of the medial domain; instead, the Φ-Domain solely consists of this agreement phrase. As the author further elaborates, it replaces “NumP, GenP, PossP and others […] where all agreement-based relations are established” (Ticio 2010: 60). Her Φ-Domain therein focusses on agreement of 218 3 INFERENCE - Nominal Concord & Definiteness the nominal head itself rather than agreement of the left periphery with said head, i.e. nominal concord. Concentrating on semi-lexical items, matters are not as uniform as one might deduce from the phrasal set-up presented above. To wit, the author utilizes the split between strong and weak determiners on the basis of Milsark’s (1977) findings, discussed in chapter one, to argue for diverging merging-sites for these classes of elements, namely DP and AgrP. Ticio motivates this by referring amongst others to Zamparelli’s ([1995] 2000) system, also laid out in the course of chapter one above. For reasons discussed there, the system under elaboration here incorporates a further division of Zamparelli’s original split into the categories strong and weak Ds on conceptual as well as empirical grounds. Hence, transposing Ticio’s proposals to the system elaborated here, Q S P and DP would be considered phrases of the Ω-Domain while AgrP would split into Poss PRO P and NumP. Consequently, Q W P, positioned between the latter two, would complete the Φ-Domain. Alternatively, the property of agreement with HN, as utilized by Grohmann & Panagiotidis (2005), bundles all SLIs under investigation, which would consequently be taken to constitute the complete Φ-Domain. All findings on illicit movements reported in chapter 2.2 would thus readily follow from the interaction of the PIC and the ban on (prolific) domain-internal movement. Let us follow this line of reasoning. Above, it was shown that extraction is possible from beneath a phasal SLI when it is itself dominated by a focus particle. The latter can be safely assumed to constitute an element of the Ω-Domain, i.e. Foc 0 . How, then, can the presence of the particle alter the set-up of the nominal domain in a way to legalize a formerly illicit movement? Returning to Grohmann’s original proposals in (93)-(95) above, we find that prolific domains are contextually defined, a notion based on propositional properties and hence not to be confused with contextuality in the sense of structural environment employed in the discussion of variable phase heads in chapter 2.2. Observe that the focus particle and the SLI form a semantic unit as to the restriction of the set introduced by the HN. Taking this observation as first tentative evidence for their co-occurrence in the same prolific domain (at least in these cases), intermediate movement through the edge of the phase Poss PRO is licit since it originates from below the Ω-Domain and proceeds into the clausal spine. Therefore, Anti-Locality is not violated here. As has been shown above, the same movement is also observed with focused DP configurations. It has, however, to be noted that prolific domains are rigid notions for Grohmann himself. Since these do not form the focus of the analysis presented here, I will refrain from going into detail but only note that the abandonment of a strict phrasal assignment to prolific domains legalizes the movement patterns observed above. 3.3 Splitting Nominal Concord 219 One last phenomenon, however, that has to be mentioned in this context is quantifier stranding (QST), specifically concerning those configurations presented at the end of chapter 2.2.1 above. Therein, the complex of strong quantifier and demonstrative or possessive pronoun was stranded together, with a plural head nominal moving to sentence initial position. Under the relaxed account of Anti-Locality, these data, too, fall into place. Alternatively, an analysis parallel to the treatment of Modern Greek demonstratives in Grohmann & Panagiotidis (2005), reported above, externally merging the possessive pronoun in the Φ-Domain followed by subsequent movement to the Ω-Domain would yield an equally acceptable output. In any case, under the assumption that both the PIC (by means of phasal status) as well as Anti-Locality (by means of membership in a specific prolific domain) might fail to apply, stranding of the strong quantifier might be traced to the membership of Q S in the Ω-Domain, as hinted at above. I believe that this proposal does not trivialize the establishment of the phasal hierarchy employed in chapter 2.2. To see this, observe that the second testing ground from said chapter, i.e. the property concerning ellipsis, identified in Bošković (2014) and reprinted below, still holds with Q S and non-phasal complements: (97) (= Bošković 2014: (56) a.) Only phases and complements of phase heads can undergo ellipsis. (98) Ich habe Fredericks Wein getrunken und du hast allen <Wein> getrunken. I have F.‘s wine drunk and you have all < wine > drunk ‘I have drunk Frederick’s wine and you drank all.’ Floating from beneath focus particles and strong quantifiers might thence be taken as a reflex of their shared membership in the Ωrather than the Φ-Domain, with the former arguably altering the affiliation of the head of its sister either contextually or by movement. 24 Observe that with D and Poss PRO , both taken to constitute heads of the latter domain by default, such movement is ruled out since in these cases the PIC and Anti-Locality do indeed pose contradicting restrictions. Recall that D-Poss PRO concatenations are licit with Num, i.e. nominal domains bearing [φ[f]], like the one below. Still, movement of the lower SLI with the lexical domain is ruled out: 24 Cf. Etxeberria (2012: 88f.) and references therein for further arguments to structurally group said LIs together. 220 3 INFERENCE - Nominal Concord & Definiteness (99) * [Mein-e Milch] i habe ich dies-e t i getrunken. [ my milk ] i have I this t i drunk As mentioned above, I will leave the exact establishment of prolific domains in the German nominal hierarchy for further research and focus on the properties of nominal concord in what follows. Returning to the initial object of investigation in this section, in sentential topicalizations with focused nominal domains, Match as well as Value successfully apply at the second attempt, resulting in φ-agreement of the stranded possessive pronoun in a structural case-position. In in-situ Poss PRO -N configurations in structural case positions, on the contrary, the pronoun is not triggered to probe again (neither by movement nor by Match/ Value) and hence surfaces as default, viz. bearing zero-morphology. As I hope to have demonstrated, the straightforward accountability for this idiosyncratic property of the possessive pronoun in terms of the theory elaborated here poses a strong argument in favor of the proposals made at the beginning of this chapter. Further instances of extraction, especially QST of full phases from beneath Q S , will be incorporated into the analysis of more complex nominal domains to which I will turn now. As has been demonstrated, nominal concord turns out a highly derivational phenomenon under the analysis elaborated above. In what follows, I will show that the arguments for a narrow-syntactic account on nominal inflection becomes even more favorable once complex nominal hierarchies are taken into consideration, in which various semi-lexical items are concatenated with one another. Therein, it will be shown that their properties, elaborated in the previous subchapters and proven solid in the preceding analysis above, can be extended to gain further insight into the derivational workings of the composition of complex nominal domains. 3.3.3 The Extended Pattern of SLI Concord Up until now, we have encountered two probing patterns of φ-features, namely [φ] and [φ[f]], whose selectional restrictions are either met by the most embedded heads N and Num, resulting in successful Match and Value and a uniform, shared feature structure on all elements making up the nominal spine, or incompatible, resulting in the successful application of Match but the failure to Value and subsequent reduction of the probing feature structure to partial default: the (still active) root [φ]. The proposed derivational character underlying nominal concord, as advocated above in the discussion of former analyses, however, foremost relies on 3.3 Splitting Nominal Concord 221 the reconstructability of interactions of the LIs involved. To this end, I want to discuss the interactions of SLIs and their respective featural compositions in this section. Therein, further arguments for a successive-cyclic derivation of the nominal domain will surface in the identification of patterns of mismatch, intervention and 2 nd cycle Agree. Recall once more the summary of probing φ-configurations of SLIs and their ability to reduce to partial default, hence to surface with default agreement, carved out in the analysis in the previous subchapters: (100) Q S D Poss PRO Q W Probing φ [φ] [φ] [φ[f]] [φ[f]] Default Agreement ? ?   Parallel to chapter 3.2.1 above, I commence my analysis by focusing on the weak quantifier, this time in co-occurrence with elements of the higher phasal cycle, and subsequently work my way up the nominal spine to cover all logically possible and syntactically licit combinations. a. Co-Occurrence with Q w As has been elaborated above, the featural set-up of the weak quantifier results in the successful application of both Match and Value when merged with a nominal domain headed by NumP, i.e. when the nominal structure either bears [φ[f]] or the more complex [φ[f[pl]]] as the first goal for the φ-probing of the SLI (recall from chapter 2.3 that N below Num might itself bear further specifications for φ which are however inaccessible to probing operations initiated from above the latter). The result is a shared feature structure between Q W and Num concerning φ. In contrast, derivations merging Q W with N specified for the incompatible values masculine or neuter ([φ[m]], [φ[n]] respectively) result in successful Match, evaluated at the root node, but an unsuccessful application of Value and hence the subsequent reduction of the probe’s φ-featural set-up. The result is zero-morphology on the surface, while successful applications of Value may result in either optional or obligatory inflection on the quantifier. Optionality itself will be one of the topics of chapter four below and will hence be omitted in what follows; the important opposition therefore is general availability rather than obligatory realization. Assignment of non-structural case triggers a second cycle of φ-Agree of the reduced probe (due to the shared case-feature and possibly the interdependence of the feature classes or the probing operation targeting the weak quantifier itself, cf. section 3.3.4 below) and hence the 222 3 INFERENCE - Nominal Concord & Definiteness availability of agreement-morphology in genitive and dative with all number/ gender-values in German. Underlying this mechanism is the ‘free’ assignment of non-structural cases in opposition to structural case which is ascribed on the basis of φ-complete Match with the case-‘bearer’. Below, I give the structural configuration for a successful as well as an unsuccessful application of Value between Q W and the lexical domain (i.e. N, Num) in the (extended) nominal projection, exemplified with a feminine and a masculine head noun respectively, employing the - by now - familiar format: (101) Nominal Concord & Definiteness I NFERENCE 158 Default Agreement ? ?   Parallel to chapter 0 above, I commence my analysis by focusing on the weak quantifier, this time in co-occurrence with elements of the higher phasal cycle, and subsequently work my way up the nominal spine to cover all logically possible and syntactically licit combinations. a. Co-Occurrence with Q W As has been elaborated above, the featural set-up of the weak quantifier results in the successful application of both Match and Value when merged with a nominal domain headed by NumP, i.e. when the nominal structure either bears [φ[f]] or the more complex [φ[f[pl]]] as the first goal for the φ-probing of the SLI (recall from chapter 2.3 that N below Num might itself bear further specifications for φ which are however inaccessible to probing operations initiated from above the latter). The result is a shared feature structure between Q W and Num concerning φ. In contrast, derivations merging Q W with N specified for the incompatible values masculine or neuter ([φ[m]], [φ[n]] respectively) result in successful Match, evaluated at the root node, but an unsuccessful application of Value and hence the subsequent reduction of the probe’s φ-featural set-up. The result is zero-morphology on the surface, while successful applications of Value may result in either optional or obligatory inflection on the quantifier. Optionality itself will be one of the topics of chapter four below and will hence be omitted in what follows; the important opposition therefore is general availability rather than obligatory realization. Assignment of non-structural case triggers a second cycle of φ-Agree of the reduced probe (due to the shared case-feature and possibly the interdependence of the feature classes or the probing operation targeting the weak quantifier itself, cf. section 3.3.4 below) and hence the availability of agreement-morphology in genitive and dative with all number/ gender-values in German. Underlying this mechanism is the ‘free’ assignment of non-structural cases in opposition to structural case which is ascribed on the basis of φ-complete Match with the case-‘bearer’. Below, I give the structural configuration for a successful as well as an unsuccessful application of Value between Q W and the lexical domain (i.e. N, Num) in the (extended) nominal projection, exemplified with a feminine and a masculine head noun respectively, employing the - by now - familiar format: (101) a. viel-e Milch Q W > Num [φ[f]] [φ[f]] b. viel Wein Q W > N i. [φ[f]]  [φ[m]] ii. [φ] [φ[m]] In the analysis of mono-SLI nominal domains above, 0 a. and b. ii. posed the status of the nominal hierarchy of projections at the time when they were merged into the verbal domain. Now, I want to show that the Merger of further SLIs does have derivational consequences. I will start out by concatenating the structures above with an element probing for the root node [φ], i.e. the demonstrative D. Recall that the partially reduced root node on the quantifier poses an active feature beyond the time of EM in the verbal spine since it is accessible for 2 nd cycle Agree after the assignment of case. The structures resulting from the next applications of EM are sketched in 0: (102) a. dies- viel-e Milch D > Q W > Num [φ] [φ[f]] [φ[f]] b. dies- viel Wein D > Q W > N [φ] [φ] [φ[m]] In both cases, D probes and Match successfully applies to the root of Q W ; in 0 a., Value applies and copies the dependent structure of the root from Q W onto D, resulting in a shared feature structure with three instances: Num, Q W and D. This is depicted in 0. In the analysis of mono-SLI nominal domains above, (101) a. and b. ii. posed the status of the nominal hierarchy of projections at the time when they were merged into the verbal domain. Now, I want to show that the Merger of further SLIs does have derivational consequences. I will start out by concatenating the structures above with an element probing for the root node [φ], i.e. the demonstrative D. Recall that the partially reduced root node on the quantifier poses an active feature beyond the time of EM in the verbal spine since it is accessible for 2 nd cycle Agree after the assignment of case. The structures resulting from the next applications of EM are sketched in (102): (102) Nominal Concord & Definiteness I NFERENCE 158 Default Agreement ? ?   Parallel to chapter 0 above, I commence my analysis by focusing on the weak quantifier, this time in co-occurrence with elements of the higher phasal cycle, and subsequently work my way up the nominal spine to cover all logically possible and syntactically licit combinations. a. Co-Occurrence with Q W As has been elaborated above, the featural set-up of the weak quantifier results in the successful application of both Match and Value when merged with a nominal domain headed by NumP, i.e. when the nominal structure either bears [φ[f]] or the more complex [φ[f[pl]]] as the first goal for the φ-probing of the SLI (recall from chapter 2.3 that N below Num might itself bear further specifications for φ which are however inaccessible to probing operations initiated from above the latter). The result is a shared feature structure between Q W and Num concerning φ. In contrast, derivations merging Q W with N specified for the incompatible values masculine or neuter ([φ[m]], [φ[n]] respectively) result in successful Match, evaluated at the root node, but an unsuccessful application of Value and hence the subsequent reduction of the probe’s φ-featural set-up. The result is zero-morphology on the surface, while successful applications of Value may result in either optional or obligatory inflection on the quantifier. Optionality itself will be one of the topics of chapter four below and will hence be omitted in what follows; the important opposition therefore is general availability rather than obligatory realization. Assignment of non-structural case triggers a second cycle of φ-Agree of the reduced probe (due to the shared case-feature and possibly the interdependence of the feature classes or the probing operation targeting the weak quantifier itself, cf. section 3.3.4 below) and hence the availability of agreement-morphology in genitive and dative with all number/ gender-values in German. Underlying this mechanism is the ‘free’ assignment of non-structural cases in opposition to structural case which is ascribed on the basis of φ-complete Match with the case-‘bearer’. Below, I give the structural configuration for a successful as well as an unsuccessful application of Value between Q W and the lexical domain (i.e. N, Num) in the (extended) nominal projection, exemplified with a feminine and a masculine head noun respectively, employing the - by now - familiar format: (101) a. viel-e Milch Q W > Num [φ[f]] [φ[f]] b. viel Wein Q W > N i. [φ[f]]  [φ[m]] ii. [φ] [φ[m]] In the analysis of mono-SLI nominal domains above, 0 a. and b. ii. posed the status of the nominal hierarchy of projections at the time when they were merged into the verbal domain. Now, I want to show that the Merger of further SLIs does have derivational consequences. I will start out by concatenating the structures above with an element probing for the root node [φ], i.e. the demonstrative D. Recall that the partially reduced root node on the quantifier poses an active feature beyond the time of EM in the verbal spine since it is accessible for 2 nd cycle Agree after the assignment of case. The structures resulting from the next applications of EM are sketched in 0: (102) a. dies- viel-e Milch D > Q W > Num [φ] [φ[f]] [φ[f]] b. dies- viel Wein D > Q W > N [φ] [φ] [φ[m]] In both cases, D probes and Match successfully applies to the root of Q W ; in 0 a., Value applies and copies the dependent structure of the root from Q W onto D, resulting in a shared feature structure with three instances: Num, Q W and D. This is depicted in 0. In both cases, D probes and Match successfully applies to the root of Q W ; in (102) a., Value applies and copies the dependent structure of the root from Q W onto 3.3 Splitting Nominal Concord 223 D, resulting in a shared feature structure with three instances: Num, Q W and D. This is depicted in (103). (103) Nominal Concord & Definiteness I NFERENCE 159 (103) dies-e viel-e Milch D > Q W > Num [φ[*f]] [φ[f]] [φ[f]] The uniform structure is externally merged into the verbal domain and surfaces with overt φ-inflection on both SLIs, as can be read off the first row above. Turning to the probing of D in 0 b., Match once again identifies Q W as a suitable goal. Recall that the partially reduced feature on Q W is still active when Match applies from D. Q W is hence forced to probe again into its own complement, this time successfully acquiring masculine specification since the reduced probe is compatible with the structure on the head noun in the second cycle. This stage is depicted in 0, the doubled line representing Match: (104) dies- viel-e Wein D > Q W > N [φ] [φ[*m]] [φ[m]] Subsequently, Value applies between the quantifier and the demonstrative, again resulting in three instances of the φstructure in the nominal domain as can be read off of the inflectional properties of the items in the first row in 0 below: (105) dies-er viel-e Wein D > Q W > N [φ[*m]] [φ[*m]] [φ[m]] Observe that inflection on Q W is not optional in these contexts. We have hence encountered a second instance of in-situ 2 nd cycle Agree: The reduced feature on the weak quantifier is triggered when targeted by Match of a higher head to probe for an appropriate structure in its own complement. This observation readily carries over to structures coordinating Q W under the second SLI probing for the root node, viz. the strong quantifier Q S , as demonstrated below. I will switch the gender of the HN to the second instance with which the original configuration on Q W , [φ[f]], is incompatible, namely neuter, for reasons to be elaborated further in the following subchapter on Adjectival Impoverishment and the analysis of the strong quantifier therein (see fn. 36 below). See that the mechanisms of nominal concord are not distorted by this switch. (106) all-es viel-e Bier Q S > Q W > N [φ[*n]] [φ[*n]] [φ[n]] One additional comment is in order concerning the morphological form of inflection on the weak quantifier in 0 above. Recall from the last section that I hinted at the featural content encoded by the inflection of the possessive pronoun in concatenation with masculine and neuter head nouns in non-structural case as being a portmanteau-morpheme for φ and case by its homonymy with the morphological shape of the determiner, which is assumed here to always bear a specification of φ. Applying this rationale to the structures in 0 and 0 above, the inflectional form of the weak quantifier surprisingly deviates from the expected realization, er with masculine nominals in 0 and es with neuter HNs in 0. A full analysis of this fact will be postponed until the next subchapter. Returning to the derivation of nominal concord, we have thus far seen that the weak quantifier is triggered to induce an in-situ 2 nd cycle of Agree by probing from a higher SLI for [φ]. There has, however, been another element identified bearing the exact same featural set-up as Q W , viz. the possessive pronoun. Let us focus on the interaction of these heads next. I once again start out with the derivational snapshot at EM of the second SLI: (107) a. mein- viel-e Milch Poss PRO > Q W > Num [φ[f]] [φ[f]] [φ[f]] The uniform structure is externally merged into the verbal domain and surfaces with overt φ-inflection on both SLIs, as can be read off the first row above. Turning to the probing of D in (102) b., Match once again identifies Q W as a suitable goal. Recall that the partially reduced feature on Q W is still active when Match applies from D. Q W is hence forced to probe again into its own complement, this time successfully acquiring masculine specification since the reduced probe is compatible with the structure on the head noun in the second cycle. This stage is depicted in (104), the doubled line representing Match: (104) Nominal Concord & Definiteness I NFERENCE 159 (103) dies-e viel-e Milch D > Q W > Num [φ[*f]] [φ[f]] [φ[f]] The uniform structure is externally merged into the verbal domain and surfaces with overt φ-inflection on both SLIs, as can be read off the first row above. Turning to the probing of D in 0 b., Match once again identifies Q W as a suitable goal. Recall that the partially reduced feature on Q W is still active when Match applies from D. Q W is hence forced to probe again into its own complement, this time successfully acquiring masculine specification since the reduced probe is compatible with the structure on the head noun in the second cycle. This stage is depicted in 0, the doubled line representing Match: (104) dies- viel-e Wein D > Q W > N [φ] [φ[*m]] [φ[m]] Subsequently, Value applies between the quantifier and the demonstrative, again resulting in three instances of the φstructure in the nominal domain as can be read off of the inflectional properties of the items in the first row in 0 below: (105) dies-er viel-e Wein D > Q W > N [φ[*m]] [φ[*m]] [φ[m]] Observe that inflection on Q W is not optional in these contexts. We have hence encountered a second instance of in-situ 2 nd cycle Agree: The reduced feature on the weak quantifier is triggered when targeted by Match of a higher head to probe for an appropriate structure in its own complement. This observation readily carries over to structures coordinating Q W under the second SLI probing for the root node, viz. the strong quantifier Q S , as demonstrated below. I will switch the gender of the HN to the second instance with which the original configuration on Q W , [φ[f]], is incompatible, namely neuter, for reasons to be elaborated further in the following subchapter on Adjectival Impoverishment and the analysis of the strong quantifier therein (see fn. 36 below). See that the mechanisms of nominal concord are not distorted by this switch. (106) all-es viel-e Bier Q S > Q W > N [φ[*n]] [φ[*n]] [φ[n]] One additional comment is in order concerning the morphological form of inflection on the weak quantifier in 0 above. Recall from the last section that I hinted at the featural content encoded by the inflection of the possessive pronoun in concatenation with masculine and neuter head nouns in non-structural case as being a portmanteau-morpheme for φ and case by its homonymy with the morphological shape of the determiner, which is assumed here to always bear a specification of φ. Applying this rationale to the structures in 0 and 0 above, the inflectional form of the weak quantifier surprisingly deviates from the expected realization, er with masculine nominals in 0 and es with neuter HNs in 0. A full analysis of this fact will be postponed until the next subchapter. Returning to the derivation of nominal concord, we have thus far seen that the weak quantifier is triggered to induce an in-situ 2 nd cycle of Agree by probing from a higher SLI for [φ]. There has, however, been another element identified bearing the exact same featural set-up as Q W , viz. the possessive pronoun. Let us focus on the interaction of these heads next. I once again start out with the derivational snapshot at EM of the second SLI: (107) a. mein- viel-e Milch Poss PRO > Q W > Num [φ[f]] [φ[f]] [φ[f]] Subsequently, Value applies between the quantifier and the demonstrative, again resulting in three instances of the φ-structure in the nominal domain as can be read off of the inflectional properties of the items in the first row in (105) below: (105) Nominal Concord & Definiteness I NFERENCE 159 (103) dies-e viel-e Milch D > Q W > Num [φ[*f]] [φ[f]] [φ[f]] The uniform structure is externally merged into the verbal domain and surfaces with overt φ-inflection on both SLIs, as can be read off the first row above. Turning to the probing of D in 0 b., Match once again identifies Q W as a suitable goal. Recall that the partially reduced feature on Q W is still active when Match applies from D. Q W is hence forced to probe again into its own complement, this time successfully acquiring masculine specification since the reduced probe is compatible with the structure on the head noun in the second cycle. This stage is depicted in 0, the doubled line representing Match: (104) dies- viel-e Wein D > Q W > N [φ] [φ[*m]] [φ[m]] Subsequently, Value applies between the quantifier and the demonstrative, again resulting in three instances of the φstructure in the nominal domain as can be read off of the inflectional properties of the items in the first row in 0 below: (105) dies-er viel-e Wein D > Q W > N [φ[*m]] [φ[*m]] [φ[m]] Observe that inflection on Q W is not optional in these contexts. We have hence encountered a second instance of in-situ 2 nd cycle Agree: The reduced feature on the weak quantifier is triggered when targeted by Match of a higher head to probe for an appropriate structure in its own complement. This observation readily carries over to structures coordinating Q W under the second SLI probing for the root node, viz. the strong quantifier Q S , as demonstrated below. I will switch the gender of the HN to the second instance with which the original configuration on Q W , [φ[f]], is incompatible, namely neuter, for reasons to be elaborated further in the following subchapter on Adjectival Impoverishment and the analysis of the strong quantifier therein (see fn. 36 below). See that the mechanisms of nominal concord are not distorted by this switch. (106) all-es viel-e Bier Q S > Q W > N [φ[*n]] [φ[*n]] [φ[n]] One additional comment is in order concerning the morphological form of inflection on the weak quantifier in 0 above. Recall from the last section that I hinted at the featural content encoded by the inflection of the possessive pronoun in concatenation with masculine and neuter head nouns in non-structural case as being a portmanteau-morpheme for φ and case by its homonymy with the morphological shape of the determiner, which is assumed here to always bear a specification of φ. Applying this rationale to the structures in 0 and 0 above, the inflectional form of the weak quantifier surprisingly deviates from the expected realization, er with masculine nominals in 0 and es with neuter HNs in 0. A full analysis of this fact will be postponed until the next subchapter. Returning to the derivation of nominal concord, we have thus far seen that the weak quantifier is triggered to induce an in-situ 2 nd cycle of Agree by probing from a higher SLI for [φ]. There has, however, been another element identified bearing the exact same featural set-up as Q W , viz. the possessive pronoun. Let us focus on the interaction of these heads next. I once again start out with the derivational snapshot at EM of the second SLI: (107) a. mein- viel-e Milch Poss PRO > Q W > Num [φ[f]] [φ[f]] [φ[f]] Observe that inflection on Q W is not optional in these contexts. We have hence encountered a second instance of in-situ 2 nd cycle Agree: The reduced feature on the weak quantifier is triggered when targeted by Match of a higher head to probe for an appropriate structure in its own complement. This observation readily carries over to structures coordinating Q W under the second SLI probing for the root node, viz. the strong quantifier Q S , as demonstrated below. I will switch the gender of the HN to the second instance with which the original configuration on Q W , [φ[f]], is incompatible, namely neuter, for reasons to be elaborated further in the following subchapter on Adjectival Impoverishment and the analysis of the strong quantifier therein (see fn. 34 below). See that the mechanisms of nominal concord are not distorted by this switch. 224 3 INFERENCE - Nominal Concord & Definiteness (106) Nominal Concord & Definiteness I NFERENCE 159 (103) dies-e viel-e Milch D > Q W > Num [φ[*f]] [φ[f]] [φ[f]] The uniform structure is externally merged into the verbal domain and surfaces with overt φ-inflection on both SLIs, as can be read off the first row above. Turning to the probing of D in 0 b., Match once again identifies Q W as a suitable goal. Recall that the partially reduced feature on Q W is still active when Match applies from D. Q W is hence forced to probe again into its own complement, this time successfully acquiring masculine specification since the reduced probe is compatible with the structure on the head noun in the second cycle. This stage is depicted in 0, the doubled line representing Match: (104) dies- viel-e Wein D > Q W > N [φ] [φ[*m]] [φ[m]] Subsequently, Value applies between the quantifier and the demonstrative, again resulting in three instances of the φstructure in the nominal domain as can be read off of the inflectional properties of the items in the first row in 0 below: (105) dies-er viel-e Wein D > Q W > N [φ[*m]] [φ[*m]] [φ[m]] Observe that inflection on Q W is not optional in these contexts. We have hence encountered a second instance of in-situ 2 nd cycle Agree: The reduced feature on the weak quantifier is triggered when targeted by Match of a higher head to probe for an appropriate structure in its own complement. This observation readily carries over to structures coordinating Q W under the second SLI probing for the root node, viz. the strong quantifier Q S , as demonstrated below. I will switch the gender of the HN to the second instance with which the original configuration on Q W , [φ[f]], is incompatible, namely neuter, for reasons to be elaborated further in the following subchapter on Adjectival Impoverishment and the analysis of the strong quantifier therein (see fn. 36 below). See that the mechanisms of nominal concord are not distorted by this switch. (106) all-es viel-e Bier Q S > Q W > N [φ[*n]] [φ[*n]] [φ[n]] One additional comment is in order concerning the morphological form of inflection on the weak quantifier in 0 above. Recall from the last section that I hinted at the featural content encoded by the inflection of the possessive pronoun in concatenation with masculine and neuter head nouns in non-structural case as being a portmanteau-morpheme for φ and case by its homonymy with the morphological shape of the determiner, which is assumed here to always bear a specification of φ. Applying this rationale to the structures in 0 and 0 above, the inflectional form of the weak quantifier surprisingly deviates from the expected realization, er with masculine nominals in 0 and es with neuter HNs in 0. A full analysis of this fact will be postponed until the next subchapter. Returning to the derivation of nominal concord, we have thus far seen that the weak quantifier is triggered to induce an in-situ 2 nd cycle of Agree by probing from a higher SLI for [φ]. There has, however, been another element identified bearing the exact same featural set-up as Q W , viz. the possessive pronoun. Let us focus on the interaction of these heads next. I once again start out with the derivational snapshot at EM of the second SLI: (107) a. mein- viel-e Milch Poss PRO > Q W > Num [φ[f]] [φ[f]] [φ[f]] One additional comment is in order concerning the morphological form of inflection on the weak quantifier in (106) above. Recall from the last section that I hinted at the featural content encoded by the inflection of the possessive pronoun in concatenation with masculine and neuter head nouns in non-structural case as being a portmanteau-morpheme for φ and case by its homonymy with the morphological shape of the determiner, which is assumed here to always bear a specification of φ. Applying this rationale to the structures in (105) and (106) above, the inflectional form of the weak quantifier surprisingly deviates from the expected realization, er with masculine nominals in (105) and es with neuter HNs in (106). A full analysis of this fact will be postponed until the next subchapter. Returning to the derivation of nominal concord, we have thus far seen that the weak quantifier is triggered to induce an in-situ 2 nd cycle of Agree by probing from a higher SLI for [φ]. There has, however, been another element identified bearing the exact same featural set-up as Q W , viz. the possessive pronoun. Let us focus on the interaction of these heads next. I once again start out with the derivational snapshot at EM of the second SLI: (107) Nominal Concord & Definiteness I NFERENCE 159 (103) dies-e viel-e Milch D > Q W > Num [φ[*f]] [φ[f]] [φ[f]] The uniform structure is externally merged into the verbal domain and surfaces with overt φ-inflection on both SLIs, as can be read off the first row above. Turning to the probing of D in 0 b., Match once again identifies Q W as a suitable goal. Recall that the partially reduced feature on Q W is still active when Match applies from D. Q W is hence forced to probe again into its own complement, this time successfully acquiring masculine specification since the reduced probe is compatible with the structure on the head noun in the second cycle. This stage is depicted in 0, the doubled line representing Match: (104) dies- viel-e Wein D > Q W > N [φ] [φ[*m]] [φ[m]] Subsequently, Value applies between the quantifier and the demonstrative, again resulting in three instances of the φstructure in the nominal domain as can be read off of the inflectional properties of the items in the first row in 0 below: (105) dies-er viel-e Wein D > Q W > N [φ[*m]] [φ[*m]] [φ[m]] Observe that inflection on Q W is not optional in these contexts. We have hence encountered a second instance of in-situ 2 nd cycle Agree: The reduced feature on the weak quantifier is triggered when targeted by Match of a higher head to probe for an appropriate structure in its own complement. This observation readily carries over to structures coordinating Q W under the second SLI probing for the root node, viz. the strong quantifier Q S , as demonstrated below. I will switch the gender of the HN to the second instance with which the original configuration on Q W , [φ[f]], is incompatible, namely neuter, for reasons to be elaborated further in the following subchapter on Adjectival Impoverishment and the analysis of the strong quantifier therein (see fn. 36 below). See that the mechanisms of nominal concord are not distorted by this switch. (106) all-es viel-e Bier Q S > Q W > N [φ[*n]] [φ[*n]] [φ[n]] One additional comment is in order concerning the morphological form of inflection on the weak quantifier in 0 above. Recall from the last section that I hinted at the featural content encoded by the inflection of the possessive pronoun in concatenation with masculine and neuter head nouns in non-structural case as being a portmanteau-morpheme for φ and case by its homonymy with the morphological shape of the determiner, which is assumed here to always bear a specification of φ. Applying this rationale to the structures in 0 and 0 above, the inflectional form of the weak quantifier surprisingly deviates from the expected realization, er with masculine nominals in 0 and es with neuter HNs in 0. A full analysis of this fact will be postponed until the next subchapter. Returning to the derivation of nominal concord, we have thus far seen that the weak quantifier is triggered to induce an in-situ 2 nd cycle of Agree by probing from a higher SLI for [φ]. There has, however, been another element identified bearing the exact same featural set-up as Q W , viz. the possessive pronoun. Let us focus on the interaction of these heads next. I once again start out with the derivational snapshot at EM of the second SLI: (107) a. mein- viel-e Milch Poss PRO > Q W > Num [φ[f]] [φ[f]] [φ[f]] Nominal Concord & Definiteness I NFERENCE 160 b. mein- viel Wein Poss PRO > Q W > N [φ[f]] [φ] [φ[m]] Obviously, the derivation following 0 a. does not pose any surprises at this point: Match and Value both successfully apply and the uniform φ-structure is readily shared between all elements, as can be deduced from the inflectional uniformity in 0. (108) mein-e viel-e Milch Poss PRO > Q W > Num [φ[f]] [φ[f]] [φ[f]] How about 0 b., then? Again, we should expect a 2 nd cycle Agree-operation on Q W triggered by the probing of the higher SLI, resulting in obligatory inflection on the quantifier, parallel to the derivations laid out above. I propose that this is exactly what happens in these cases. However, note that this time, the higher SLI bears a restricted specification for φ itself. At this point, the highly derivational character of SLI concord becomes most visible: Q W is trivially specified to be valued by any φ-specification at this point, while Poss PRO is externally merged with its original φ-configuration [φ[f]] still undisrupted by a failed attempt to become valued. Match identifies the root node on Q W as an appropriate goal, which therefore probes into its own complement once again, as above. But now, observe that the new feature structure on the quantifier (an instance of the newly shared feature between N and Q W ) is incompatible with the feature of the initial probing operation on Poss PRO : (109) mein- viel-er Wein Poss PRO > Q W > N [φ[f]] [φ[*m]] [φ[m]] Even though Q W has acquired a complex structure by this second probing operation, the feature on Poss PRO , the initiator of Q W ’s subsequent probing efforts, becomes reduced itself in the failure of Value from the newly specified Q W to the possessive pronoun. This scenario is supported by the inflectional idiosyncrasies of the SLIs involved: (110) mein viel-er Wein Poss PRO > Q W > N [φ] [φ[*m]] [φ[m]] Therein, we have come across a truly successive derivational set-up of nominal concord in German: The default realization (i.e. zero-morphology) has ‘skipped’ to the highest element on the surface, with Q W now specified, but Poss PRO reduced to partial default. Again, I want to end the discussion on this specific configuration by once more directing the attention to the morphological form of the inflection on the weak quantifier. Observe that in 0, we find a realization parallel to the appropriate forms of those elements probing for the root, i.e. Q S and D, when Q W is coordinated below the possessive pronoun. Once again, I postpone the analysis of this observation until the following subchapter. I have concentrated on the co-occurrence of the most deeply embedded SLI under consideration with the three elements making up the highest phasal cycle in the nominal domain. As has been laid out therein, the derivation of morphological form rests on the selectional restrictions of the elements involved: With D and Q S , probing for the root node, a second cycle of φ-Agree ultimately returns a uniform nominal domain with all LIs equally specified for φ by sharing of the appropriate featural configuration; with the highest SLI itself probing only for a subset of the possible feature configurations, the second cycle is triggered by the same mechanisms as before, yet the newly acquired value on the lower SLI induces a mismatch in the original inter-SLI application of Value and hence reduces the highest probe to partial default. Further observations concerning the morphological form of the inflection have been noted, but their analysis has been postponed for now. Instead, further patterns of SLI co-occurrence will be presented in what follows and analyzed along the lines of the theory elaborated in the former sections and put to use above. b. Co-Occurrence with Poss PRO Following the modus operandi of chapter 0, I will continue the analysis by focusing on the next higher element, the first head of the highest phrasal cycle in the nominal domain, i.e. Poss PRO . As elaborated above, this head likewise probes for Obviously, the derivation following (107) a. does not pose any surprises at this point: Match and Value both successfully apply and the uniform φ-structure is readily shared between all elements, as can be deduced from the inflectional uniformity in (108). (108) Nominal Concord & Definiteness I NFERENCE 160 b. mein- viel Wein Poss PRO > Q W > N [φ[f]] [φ] [φ[m]] Obviously, the derivation following 0 a. does not pose any surprises at this point: Match and Value both successfully apply and the uniform φ-structure is readily shared between all elements, as can be deduced from the inflectional uniformity in 0. (108) mein-e viel-e Milch Poss PRO > Q W > Num [φ[f]] [φ[f]] [φ[f]] How about 0 b., then? Again, we should expect a 2 nd cycle Agree-operation on Q W triggered by the probing of the higher SLI, resulting in obligatory inflection on the quantifier, parallel to the derivations laid out above. I propose that this is exactly what happens in these cases. However, note that this time, the higher SLI bears a restricted specification for φ itself. At this point, the highly derivational character of SLI concord becomes most visible: Q W is trivially specified to be valued by any φ-specification at this point, while Poss PRO is externally merged with its original φ-configuration [φ[f]] still undisrupted by a failed attempt to become valued. Match identifies the root node on Q W as an appropriate goal, which therefore probes into its own complement once again, as above. But now, observe that the new feature structure on the quantifier (an instance of the newly shared feature between N and Q W ) is incompatible with the feature of the initial probing operation on Poss PRO : (109) mein- viel-er Wein Poss PRO > Q W > N [φ[f]] [φ[*m]] [φ[m]] Even though Q W has acquired a complex structure by this second probing operation, the feature on Poss PRO , the initiator of Q W ’s subsequent probing efforts, becomes reduced itself in the failure of Value from the newly specified Q W to the possessive pronoun. This scenario is supported by the inflectional idiosyncrasies of the SLIs involved: (110) mein viel-er Wein Poss PRO > Q W > N [φ] [φ[*m]] [φ[m]] Therein, we have come across a truly successive derivational set-up of nominal concord in German: The default realization (i.e. zero-morphology) has ‘skipped’ to the highest element on the surface, with Q W now specified, but Poss PRO reduced to partial default. Again, I want to end the discussion on this specific configuration by once more directing the attention to the morphological form of the inflection on the weak quantifier. Observe that in 0, we find a realization parallel to the appropriate forms of those elements probing for the root, i.e. Q S and D, when Q W is coordinated below the possessive pronoun. Once again, I postpone the analysis of this observation until the following subchapter. I have concentrated on the co-occurrence of the most deeply embedded SLI under consideration with the three elements making up the highest phasal cycle in the nominal domain. As has been laid out therein, the derivation of morphological form rests on the selectional restrictions of the elements involved: With D and Q S , probing for the root node, a second cycle of φ-Agree ultimately returns a uniform nominal domain with all LIs equally specified for φ by sharing of the appropriate featural configuration; with the highest SLI itself probing only for a subset of the possible feature configurations, the second cycle is triggered by the same mechanisms as before, yet the newly acquired value on the lower SLI induces a mismatch in the original inter-SLI application of Value and hence reduces the highest probe to partial default. Further observations concerning the morphological form of the inflection have been noted, but their analysis has been postponed for now. Instead, further patterns of SLI co-occurrence will be presented in what follows and analyzed along the lines of the theory elaborated in the former sections and put to use above. b. Co-Occurrence with Poss PRO Following the modus operandi of chapter 0, I will continue the analysis by focusing on the next higher element, the first head of the highest phrasal cycle in the nominal domain, i.e. Poss PRO . As elaborated above, this head likewise probes for 3.3 Splitting Nominal Concord 225 How about (107) b., then? Again, we should expect a 2 nd cycle Agree-operation on Q W triggered by the probing of the higher SLI, resulting in obligatory inflection on the quantifier, parallel to the derivations laid out above. I propose that this is exactly what happens in these cases. However, note that this time, the higher SLI bears a restricted specification for φ itself. At this point, the highly derivational character of SLI concord becomes most visible: Q W is trivially specified to be valued by any φ-specification at this point, while Poss PRO is externally merged with its original φ-configuration [φ[f]] still undisrupted by a failed attempt to become valued. Match identifies the root node on Q W as an appropriate goal, which therefore probes into its own complement once again, as above. But now, observe that the new feature structure on the quantifier (an instance of the newly shared feature between N and Q W ) is incompatible with the feature of the initial probing operation on Poss PRO : (109) Nominal Concord & Definiteness I NFERENCE 160 b. mein- viel Wein Poss PRO > Q W > N [φ[f]] [φ] [φ[m]] Obviously, the derivation following 0 a. does not pose any surprises at this point: Match and Value both successfully apply and the uniform φ-structure is readily shared between all elements, as can be deduced from the inflectional uniformity in 0. (108) mein-e viel-e Milch Poss PRO > Q W > Num [φ[f]] [φ[f]] [φ[f]] How about 0 b., then? Again, we should expect a 2 nd cycle Agree-operation on Q W triggered by the probing of the higher SLI, resulting in obligatory inflection on the quantifier, parallel to the derivations laid out above. I propose that this is exactly what happens in these cases. However, note that this time, the higher SLI bears a restricted specification for φ itself. At this point, the highly derivational character of SLI concord becomes most visible: Q W is trivially specified to be valued by any φ-specification at this point, while Poss PRO is externally merged with its original φ-configuration [φ[f]] still undisrupted by a failed attempt to become valued. Match identifies the root node on Q W as an appropriate goal, which therefore probes into its own complement once again, as above. But now, observe that the new feature structure on the quantifier (an instance of the newly shared feature between N and Q W ) is incompatible with the feature of the initial probing operation on Poss PRO : (109) mein- viel-er Wein Poss PRO > Q W > N [φ[f]] [φ[*m]] [φ[m]] Even though Q W has acquired a complex structure by this second probing operation, the feature on Poss PRO , the initiator of Q W ’s subsequent probing efforts, becomes reduced itself in the failure of Value from the newly specified Q W to the possessive pronoun. This scenario is supported by the inflectional idiosyncrasies of the SLIs involved: (110) mein viel-er Wein Poss PRO > Q W > N [φ] [φ[*m]] [φ[m]] Therein, we have come across a truly successive derivational set-up of nominal concord in German: The default realization (i.e. zero-morphology) has ‘skipped’ to the highest element on the surface, with Q W now specified, but Poss PRO reduced to partial default. Again, I want to end the discussion on this specific configuration by once more directing the attention to the morphological form of the inflection on the weak quantifier. Observe that in 0, we find a realization parallel to the appropriate forms of those elements probing for the root, i.e. Q S and D, when Q W is coordinated below the possessive pronoun. Once again, I postpone the analysis of this observation until the following subchapter. I have concentrated on the co-occurrence of the most deeply embedded SLI under consideration with the three elements making up the highest phasal cycle in the nominal domain. As has been laid out therein, the derivation of morphological form rests on the selectional restrictions of the elements involved: With D and Q S , probing for the root node, a second cycle of φ-Agree ultimately returns a uniform nominal domain with all LIs equally specified for φ by sharing of the appropriate featural configuration; with the highest SLI itself probing only for a subset of the possible feature configurations, the second cycle is triggered by the same mechanisms as before, yet the newly acquired value on the lower SLI induces a mismatch in the original inter-SLI application of Value and hence reduces the highest probe to partial default. Further observations concerning the morphological form of the inflection have been noted, but their analysis has been postponed for now. Instead, further patterns of SLI co-occurrence will be presented in what follows and analyzed along the lines of the theory elaborated in the former sections and put to use above. b. Co-Occurrence with Poss PRO Following the modus operandi of chapter 0, I will continue the analysis by focusing on the next higher element, the first head of the highest phrasal cycle in the nominal domain, i.e. Poss PRO . As elaborated above, this head likewise probes for Even though Q W has acquired a complex structure by this second probing operation, the feature on Poss PRO , the initiator of Q W ’s subsequent probing efforts, becomes reduced itself in the failure of Value from the newly specified Q W to the possessive pronoun. This scenario is supported by the inflectional idiosyncrasies of the SLIs involved: (110) Nominal Concord & Definiteness I NFERENCE 160 b. mein- viel Wein Poss PRO > Q W > N [φ[f]] [φ] [φ[m]] Obviously, the derivation following 0 a. does not pose any surprises at this point: Match and Value both successfully apply and the uniform φ-structure is readily shared between all elements, as can be deduced from the inflectional uniformity in 0. (108) mein-e viel-e Milch Poss PRO > Q W > Num [φ[f]] [φ[f]] [φ[f]] How about 0 b., then? Again, we should expect a 2 nd cycle Agree-operation on Q W triggered by the probing of the higher SLI, resulting in obligatory inflection on the quantifier, parallel to the derivations laid out above. I propose that this is exactly what happens in these cases. However, note that this time, the higher SLI bears a restricted specification for φ itself. At this point, the highly derivational character of SLI concord becomes most visible: Q W is trivially specified to be valued by any φ-specification at this point, while Poss PRO is externally merged with its original φ-configuration [φ[f]] still undisrupted by a failed attempt to become valued. Match identifies the root node on Q W as an appropriate goal, which therefore probes into its own complement once again, as above. But now, observe that the new feature structure on the quantifier (an instance of the newly shared feature between N and Q W ) is incompatible with the feature of the initial probing operation on Poss PRO : (109) mein- viel-er Wein Poss PRO > Q W > N [φ[f]] [φ[*m]] [φ[m]] Even though Q W has acquired a complex structure by this second probing operation, the feature on Poss PRO , the initiator of Q W ’s subsequent probing efforts, becomes reduced itself in the failure of Value from the newly specified Q W to the possessive pronoun. This scenario is supported by the inflectional idiosyncrasies of the SLIs involved: (110) mein viel-er Wein Poss PRO > Q W > N [φ] [φ[*m]] [φ[m]] Therein, we have come across a truly successive derivational set-up of nominal concord in German: The default realization (i.e. zero-morphology) has ‘skipped’ to the highest element on the surface, with Q W now specified, but Poss PRO reduced to partial default. Again, I want to end the discussion on this specific configuration by once more directing the attention to the morphological form of the inflection on the weak quantifier. Observe that in 0, we find a realization parallel to the appropriate forms of those elements probing for the root, i.e. Q S and D, when Q W is coordinated below the possessive pronoun. Once again, I postpone the analysis of this observation until the following subchapter. I have concentrated on the co-occurrence of the most deeply embedded SLI under consideration with the three elements making up the highest phasal cycle in the nominal domain. As has been laid out therein, the derivation of morphological form rests on the selectional restrictions of the elements involved: With D and Q S , probing for the root node, a second cycle of φ-Agree ultimately returns a uniform nominal domain with all LIs equally specified for φ by sharing of the appropriate featural configuration; with the highest SLI itself probing only for a subset of the possible feature configurations, the second cycle is triggered by the same mechanisms as before, yet the newly acquired value on the lower SLI induces a mismatch in the original inter-SLI application of Value and hence reduces the highest probe to partial default. Further observations concerning the morphological form of the inflection have been noted, but their analysis has been postponed for now. Instead, further patterns of SLI co-occurrence will be presented in what follows and analyzed along the lines of the theory elaborated in the former sections and put to use above. b. Co-Occurrence with Poss PRO Following the modus operandi of chapter 0, I will continue the analysis by focusing on the next higher element, the first head of the highest phrasal cycle in the nominal domain, i.e. Poss PRO . As elaborated above, this head likewise probes for Therein, we have come across a truly successive derivational set-up of nominal concord in German: The default realization (i.e. zero-morphology) has ‘skipped’ to the highest element on the surface, with Q W now specified, but Poss PRO reduced to partial default. Again, I want to end the discussion on this specific configuration by once more directing the attention to the morphological form of the inflection on the weak quantifier. Observe that in (110), we find a realization parallel to the appropriate forms of those elements probing for the root, i.e. Q S and D, when Q W is coordinated below the possessive pronoun. Once again, I postpone the analysis of this observation until the following subchapter. 226 3 INFERENCE - Nominal Concord & Definiteness I have concentrated on the co-occurrence of the most deeply embedded SLI under consideration with the three elements making up the highest phasal cycle in the nominal domain. As has been laid out therein, the derivation of morphological form rests on the selectional restrictions of the elements involved: With D and Q S , probing for the root node, a second cycle of φ-Agree ultimately returns a uniform nominal domain with all LIs equally specified for φ by sharing of the appropriate featural configuration; with the highest SLI itself probing only for a subset of the possible feature configurations, the second cycle is triggered by the same mechanisms as before, yet the newly acquired value on the lower SLI induces a mismatch in the original inter-SLI application of Value and hence reduces the highest probe to partial default. Further observations concerning the morphological form of the inflection have been noted, but their analysis has been postponed for now. Instead, further patterns of SLI co-occurrence will be presented in what follows and analyzed along the lines of the theory elaborated in the former sections and put to use above. b. Co-Occurrence with Poss pro Following the modus operandi of chapter 3.2.1, I will continue the analysis by focusing on the next higher element, the first head of the highest phrasal cycle in the nominal domain, i.e. Poss PRO . As elaborated above, this head likewise probes for the complex subset [φ[f]]. We hence expect further instances of 2 nd cycle Agree-operations in the coordination of this head under the root-probing D and Q S . Let me once again start out by concatenating the element under consideration with the former. The derivational snapshots in concatenation with N and Num prior to EM of the demonstrative are laid out in (111): (111) Nominal Concord & Definiteness I NFERENCE 161 the complex subset [φ[f]]. We hence expect further instances of 2 nd cycle Agree-operations in the coordination of this head under the root-probing D and Q S . Let me once again start out by concatenating the element under consideration with the former. The derivational snapshots in concatenation with N and Num prior to EM of the demonstrative are laid out in 0: (111) a. mein-e Milch Poss PRO > Num [φ[f]] [φ[f]] b. mein Wein Poss PRO > N i. [φ[f]]  [φ[m]] ii. [φ] [φ[m]] Starting once again from 0 a. and 0 b. ii., we arrive at the structural configurations below, parallel to those incorporating the weak quantifier in the last section: (112) a. dies- mein-e Milch D > Poss PRO > Num [φ] [φ[f]] [φ[f]] b. dies- mein Wein D > Poss PRO > N [φ] [φ] [φ[m]] Also in parallel to the analysis in the preceding section, the derivations incorporating Num (i.e. feminine and plural HNs) pose no problem here: The specification of the φ-node is readily accessible by D for Match as well as Value, returning the uniform structure in 0. (113) dies-e mein-e Milch D > Poss PRO > Num [φ[*f]] [φ[f]] [φ[f]] The derivation following 0 b., however, is different. Observe that this is the first time in the analysis of complex concord phenomena that we face an ungrammatical combination of SLIs on the surface: (114) a. * dies-er mein Wein this my wine masc b. * dies-es mein Bier this my beer neut How can the theory elaborated here account for the ungrammaticality if the selectional restrictions, which so far have been taken to be the sole initiator of interactions between SLIs, are identical to those in the configurations analyzed in the previous section? Let us begin by sketching the derivation along the lines of said analysis above. Again, Match is taken to succeed, but this time no 2 nd cycle Agree is triggered at the lower SLI. Following Bejar’s original proposals, Value still succeeds but since no feature structure is copied from goal to probe, I claimed in chapter 0 that no sharing of the feature applies and no chain is formed from Poss PRO to D, leaving two independent root nodes on the SLIs: (115) dies- mein- Wein D > Poss PRO > N [φ] [φ] [φ[m]] Recall from chapter 0 as well as from the beginning of this chapter that I took those elements probing for [φ[f]] to also exhibit default agreement, i.e. zero-morphology, brought about when the value for φ is reduced to partially default: the root node. I stayed agnostic at that time concerning the availability of surfacing as a default form for the remaining two Starting once again from (111) a. and (111) b. ii., we arrive at the structural configurations below, parallel to those incorporating the weak quantifier in the last section: 3.3 Splitting Nominal Concord 227 (112) Nominal Concord & Definiteness I NFERENCE 161 the complex subset [φ[f]]. We hence expect further instances of 2 nd cycle Agree-operations in the coordination of this head under the root-probing D and Q S . Let me once again start out by concatenating the element under consideration with the former. The derivational snapshots in concatenation with N and Num prior to EM of the demonstrative are laid out in 0: (111) a. mein-e Milch Poss PRO > Num [φ[f]] [φ[f]] b. mein Wein Poss PRO > N i. [φ[f]]  [φ[m]] ii. [φ] [φ[m]] Starting once again from 0 a. and 0 b. ii., we arrive at the structural configurations below, parallel to those incorporating the weak quantifier in the last section: (112) a. dies- mein-e Milch D > Poss PRO > Num [φ] [φ[f]] [φ[f]] b. dies- mein Wein D > Poss PRO > N [φ] [φ] [φ[m]] Also in parallel to the analysis in the preceding section, the derivations incorporating Num (i.e. feminine and plural HNs) pose no problem here: The specification of the φ-node is readily accessible by D for Match as well as Value, returning the uniform structure in 0. (113) dies-e mein-e Milch D > Poss PRO > Num [φ[*f]] [φ[f]] [φ[f]] The derivation following 0 b., however, is different. Observe that this is the first time in the analysis of complex concord phenomena that we face an ungrammatical combination of SLIs on the surface: (114) a. * dies-er mein Wein this my wine masc b. * dies-es mein Bier this my beer neut How can the theory elaborated here account for the ungrammaticality if the selectional restrictions, which so far have been taken to be the sole initiator of interactions between SLIs, are identical to those in the configurations analyzed in the previous section? Let us begin by sketching the derivation along the lines of said analysis above. Again, Match is taken to succeed, but this time no 2 nd cycle Agree is triggered at the lower SLI. Following Bejar’s original proposals, Value still succeeds but since no feature structure is copied from goal to probe, I claimed in chapter 0 that no sharing of the feature applies and no chain is formed from Poss PRO to D, leaving two independent root nodes on the SLIs: (115) dies- mein- Wein D > Poss PRO > N [φ] [φ] [φ[m]] Recall from chapter 0 as well as from the beginning of this chapter that I took those elements probing for [φ[f]] to also exhibit default agreement, i.e. zero-morphology, brought about when the value for φ is reduced to partially default: the root node. I stayed agnostic at that time concerning the availability of surfacing as a default form for the remaining two Also in parallel to the analysis in the preceding section, the derivations incorporating Num (i.e. feminine and plural HNs) pose no problem here: The specification of the φ-node is readily accessible by D for Match as well as Value, returning the uniform structure in (113). (113) Nominal Concord & Definiteness I NFERENCE 161 the complex subset [φ[f]]. We hence expect further instances of 2 nd cycle Agree-operations in the coordination of this head under the root-probing D and Q S . Let me once again start out by concatenating the element under consideration with the former. The derivational snapshots in concatenation with N and Num prior to EM of the demonstrative are laid out in 0: (111) a. mein-e Milch Poss PRO > Num [φ[f]] [φ[f]] b. mein Wein Poss PRO > N i. [φ[f]]  [φ[m]] ii. [φ] [φ[m]] Starting once again from 0 a. and 0 b. ii., we arrive at the structural configurations below, parallel to those incorporating the weak quantifier in the last section: (112) a. dies- mein-e Milch D > Poss PRO > Num [φ] [φ[f]] [φ[f]] b. dies- mein Wein D > Poss PRO > N [φ] [φ] [φ[m]] Also in parallel to the analysis in the preceding section, the derivations incorporating Num (i.e. feminine and plural HNs) pose no problem here: The specification of the φ-node is readily accessible by D for Match as well as Value, returning the uniform structure in 0. (113) dies-e mein-e Milch D > Poss PRO > Num [φ[*f]] [φ[f]] [φ[f]] The derivation following 0 b., however, is different. Observe that this is the first time in the analysis of complex concord phenomena that we face an ungrammatical combination of SLIs on the surface: (114) a. * dies-er mein Wein this my wine masc b. * dies-es mein Bier this my beer neut How can the theory elaborated here account for the ungrammaticality if the selectional restrictions, which so far have been taken to be the sole initiator of interactions between SLIs, are identical to those in the configurations analyzed in the previous section? Let us begin by sketching the derivation along the lines of said analysis above. Again, Match is taken to succeed, but this time no 2 nd cycle Agree is triggered at the lower SLI. Following Bejar’s original proposals, Value still succeeds but since no feature structure is copied from goal to probe, I claimed in chapter 0 that no sharing of the feature applies and no chain is formed from Poss PRO to D, leaving two independent root nodes on the SLIs: (115) dies- mein- Wein D > Poss PRO > N [φ] [φ] [φ[m]] Recall from chapter 0 as well as from the beginning of this chapter that I took those elements probing for [φ[f]] to also exhibit default agreement, i.e. zero-morphology, brought about when the value for φ is reduced to partially default: the root node. I stayed agnostic at that time concerning the availability of surfacing as a default form for the remaining two The derivation following (112) b., however, is different. Observe that this is the first time in the analysis of complex concord phenomena that we face an ungrammatical combination of SLIs on the surface: (114) a. * dies-er mein Wein this my wine masc b. * dies-es mein Bier this my beer neut How can the theory elaborated here account for the ungrammaticality if the selectional restrictions, which so far have been taken to be the sole initiator of interactions between SLIs, are identical to those in the configurations analyzed in the previous section? Let us begin by sketching the derivation along the lines of said analysis above. Again, Match is taken to succeed, but this time no 2 nd cycle Agree is triggered at the lower SLI. Following Bejar’s original proposals, Value still succeeds but since no feature structure is copied from goal to probe, I claimed in chapter 3.1.2 that no sharing of the feature applies and no chain is formed from Poss PRO to D, leaving two independent root nodes on the SLIs: 228 3 INFERENCE - Nominal Concord & Definiteness (115) diesmein- Wein D > Poss PRO > N [φ] [φ] [φ[m]] Recall from chapter 3.2.1 as well as from the beginning of this chapter that I took those elements probing for [φ[f]] to also exhibit default agreement, i.e. zero-morphology, brought about when the value for φ is reduced to partially default: the root node. I stayed agnostic at that time concerning the availability of surfacing as a default form for the remaining two SLIs, namely D and Q S . However, I frequently indicated a split in the morphological form of all SLIs between the root and the inflection by means of a hyphen in the examples above. Returning to the demonstrative, note that it indeed provides an inflectionless form: dies . The property of surfacing without overt inflection, however, does not suffice to constitute a default form; per definitionem it has to exhibit a wider combinatorial applicability w.r.t. a certain grammatical category than its specified counterpart. In the case of the inflectionless demonstrative w.r.t. φ, this is not the case. 25 The form exhibiting zero-morphology can be substituted only for those inflected forms likewise ending in s , to wit, the neuter nominative and accusative: (116) a. neut nom dies Bier dies-es Bier acc dies Bier dies-es Bier b. masc nom *dies Wein dies-er Wein acc *dies Wein dies-en Wein 25 As noted by Roehrs (in prep.: 64f.), referring to Alexiadou et al. (2011), the inflectionless demonstrative does indeed exhibit a wider range of combinatorial applicability, namely as an anaphora for both nouns as well as clauses (whereas the inflected form can only act as an anaphora for the former). Roehrs speculates that there are two distinct LIs or rather diverging null complements at play. Alternatively, one might indeed view this state of affairs as one inflected and one default instance of the demonstrative, but not along the lines of φ-agreement as with other nominal SLIs under consideration here. This latter option hence still falls out of the scope of the current analysis and can thus be ignored. 3.3 Splitting Nominal Concord 229 c. fem nom *dies Milch dies-e Milch acc *dies Milch dies-e Milch Additionally, and more compellingly, recall from the beginning of this subchapter that the SLIs under discussion preceding the weak quantifier do influence the inflection of the latter in ways not made explicit before: (117) a. dies-es viel-e Bier this many beer neut b. all-es viel-e Bier all many beer neut c. mein-ø viel-es Bier my many beer neut The inflectionless form, however, coincides with the inflected forms of D and Q S rather than uninflected Poss PRO in this respect: (118) dies-ø viel-e Bier this many beer neut Tracing the variation of quantifier inflection back to the status of φ-completeness of the dominating SLI for now, I will disregard the uninflected demonstrative as a case of default agreement 26 and follow Roehrs (2009: ch. 4, fn. 33; referring to Gallmann 2004: fn. 3) in its treatment as a phonologically reduced form by means of late deletion, a haplological effect. This side note has further consequences for our classification of SLIs: Since φ-features are not the content of inflectional morphemes on D but the latter does pattern with SLIs which do have default forms for φ and hence do code these features by means of inflection (i.e. Q S ), in that they evoke an identical pattern of inflection on the weak quantifier in contrast with the quantifier inflection accompanied by non-inflected SLIs (i.e. Poss PRO ), φ-features are arguably present in the root or the stem and hence D does not possess a default form w.r.t. this inflectional category. This claim is further backed up by the morphological 26 Cf. Wiltschko (2009) for a parallel proposal, treating gender and number as inherent to D in German. Her concepts of default vs. unmarked values, however, vary; see e.g. Klinge (2008: 244) and references therein for analyses treating the form of the demonstrative as the bare stem with φ-features specified by suffixation. 230 3 INFERENCE - Nominal Concord & Definiteness form of the other instantiation of the D-head, the determiner, which constitutes a monomorphemic LI in German. Of course, determiners (and demonstratives) do show reflexes of φ-agreement on the surface; these are hence also not coded on the bare root d - (in line with Bernstein’s 2008 analysis of the English th -root, cf. also Greenberg 1978: 77), but on the stem. The classification of SLIs can then be updated to (119): (119) Q S D Poss PRO Q W Probing φ [φ] [φ] [φ[f]] [φ[f]] Default Agreement ?    We have hence found the first instance of an SLI unable to surface with a bare root node [φ]. Returning to the initial observation above, the ungrammaticality of the datum becomes evident therein since both Match as well as Value apply between D and Poss PRO , preventing the former to Agree with HN and thereby to acquire an articulate feature structure. D’s selectional restrictions are not met since it cannot surface with a default value. Poss PRO , when coupled with D, is hence a true intervener in concatenation with non-feminine HNs, another indication for the highly derivational nature of nominal concord. *16 A legitimate question to be asked at this point concerns the availability of insitu 2 nd cycle Agree in the configuration under discussion. We have seen above that all three SLIs dominating the weak quantifier trigger a second attempt to probe by Match applying to the hitherto reduced root node on Q W . Likewise, I argued for Poss PRO itself being able to initiate a second instance of the probing operation when assigned case in chapter 3.2.2 above. However, close examination of these cases reveals a significant parallel to the exclusion of the configuration at hand. Second cycle Agree of Q W as well as Poss PRO above is triggered by probing of a head from the next phasal cycle: Q S / D/ Poss PRO in the former, V 0 in the latter (depicted by the dotted line below). (120) Nominal Concord & Definiteness I NFERENCE 163 backed up by the morphological form of the other instantiation of the D-head, the determiner, which constitutes a monomorphemic LI in German. Of course, determiners (and demonstratives) do show reflexes of φ-agreement on the surface; these are hence also not coded on the bare root d - (in line with Bernstein’s 2008 analysis of the English th -root, cf. also Greenberg 1978: 77), but on the stem. The classification of SLIs can then be updated to 0: (119) Q S D Poss PRO Q W Probing φ [φ] [φ] [φ[f]] [φ[f]] Default Agreement ?    We have hence found the first instance of an SLI unable to surface with a bare root node [φ]. Returning to the initial observation above, the ungrammaticality of the datum becomes evident therein since both Match as well as Value apply between D and Poss PRO , preventing the former to Agree with HN and thereby to acquire an articulate feature structure. D’s selectional restrictions are not met since it cannot surface with a default value. Poss PRO , when coupled with D, is hence a true intervener in concatenation with non-feminine HNs, another indication for the highly derivational nature of nominal concord. 27 A legitimate question to be asked at this point concerns the availability of in-situ 2 nd cycle Agree in the configuration under discussion. We have seen above that all three SLIs dominating the weak quantifier trigger a second attempt to probe by Match applying to the hitherto reduced root node on Q W . Likewise, I argued for Poss PRO itself being able to initiate a second instance of the probing operation when assigned case in chapter 0 above. However, close examination of these cases reveals a significant parallel to the exclusion of the configuration at hand. Second cycle Agree of Q W as well as Poss PRO above is triggered by probing of a head from the next phasal cycle: Q S / D/ Poss PRO in the former, V 0 in the latter (depicted by the dotted line below). (120) a. Q S / D/ Poss PRO > Q W > N b. V > Poss PRO > N [φ] [φ] [φ] D and Poss PRO , however, are context sensitive phase heads of the same cycle, viz. the highest nominal phase. I will therefore propose that the triggering element is too close in these configurations and that in-situ 2 nd cycle Agree has to be initiated from a dominating phasal cycle. Note that therein, an additional argument can be found in favour of the split of the alleged category of ein-words discussed in chapter one. Under the ein-word-class approach, deconstruction of the singular possessive pronouns in first, second as well as (masculine and neuter) pronouns in the third person into a possessive prefix and the ein -stem suggests itself: (121) 1 st m-ein 27 Observe that the conjunction of these SLIs becomes acceptable in concatenation with masculine HNs in nominative case when the pronoun displays first or second person plural possessor agreement: i. dies-er unser Hund [this our dog masc ] NOM ii. dies-er euer Hund [this your pl dog masc ] NOM This effect, however, is rooted in the morphological form of the possessives as indicated by the omission of hyphens in the data above: The pronouns surface as bare, i.e. suffixless stems, parallel to the remaining forms in the prardigm; however, the stem-final -er mirrors the appropriate inflection on the demonstrative which facilitates the reanalysis of the form, thus its acceptability. This approach gains support from parallel configurations based on neuter HNs in both structural cases, which are unacceptable for the full paradigm of D-Poss PRO concatenations. The only diverging factor on the surface is the morphological dissimilation of inflection on the demonstrative and the final syllable of the possessive stem: iii. * dies-es unser Pferd [this our horse neut ] NOM/ ACC iv. * dies-es euer Pferd [this your pl horse neut ] NOM/ ACC D and Poss PRO , however, are context sensitive phase heads of the same cycle, viz. the highest nominal phase. I will therefore propose that the triggering element is too close in these configurations and that in-situ 2 nd cycle Agree has to be initiated from a dominating phasal cycle. 3.3 Splitting Nominal Concord 231 Note that therein, an additional argument can be found in favour of the split of the alleged category of ein-words discussed in chapter one. Under the einword-class approach, deconstruction of the singular possessive pronouns in first, second as well as (masculine and neuter) pronouns in the third person into a possessive prefix and the ein -stem suggests itself: (121) 1 st m-ein 2 nd d-ein masc neut 3 rd s-ein s-ein In light of the current proposals, the ein -root should then behave in a parallel manner to the possessive pronoun in the configuration(s) under discussion, to wit, ein might surface without overt inflection, due to possible selectional mismatches, and is moreover predicted to cause an intervention effect when concatenated with D 0 , due to its structural position. The first proposal is exactly what we find (note that the concatenation with plural HNs is illicit due to semantic reasons). Departing from the data above, I make use of common (count) nouns here, since the article/ numeral interpretation of ein enforces countability, even if the combination with mass nominals is not starred in German. (122) a. (m-)ein-e Straße ( poss -)ein street fem b. (m-)ein-ø Garten ( poss -)ein garden masc c. (m-)ein-ø Haus ( poss -)ein house neut Elaborating on (122) b. and c., coordination with D 0 should then result in illicit configurations. This is however not what we find: 27 27 Observe that the introduction of the D-head into these structures is accompanied by a stressed intonation and an emphatic reading of the article/ numeral. Above, I proposed the identity of these elements solely on the basis of morphological form; see Roehrs (in prep.) for a detailed analysis of possible ein-word categories in which the stressed numeral is structurally split from the composite and article uses/ functions, cf. also chapter 1.4.4 above. 232 3 INFERENCE - Nominal Concord & Definiteness (123) a. dieser ein-e Garten (= (122) b., [modified]) this one garden masc b. dieses ein-e Haus (= (122) c., [modified]) this one house neut The data in (123) are instead regular w.r.t. the proposals of in-situ 2 nd cycle Agree, as stated above, and can hence easily be subsumed into the theory elaborated here if the possessive and the ein -root do indeed differ in their structural position. The proposal has further implications for the last remaining SLI cooccurrence with Poss PRO , namely the concatenation with the strong quantifier. I will turn to this configuration now. Q S is the second SLI probing for the bare root [φ]. Starting again at the φ-uniform and simple combination of Poss PRO with feminine HNs, we arrive at (124) at the time of EM of all -. (124) Nominal Concord & Definiteness I NFERENCE 164 2 nd d-ein masc neut 3 rd s-ein s-ein In light of the current proposals, the ein -root should then behave in a parallel manner to the possessive pronoun in the configuration(s) under discussion, to wit, ein might surface without overt inflection, due to possible selectional mismatches, and is moreover predicted to cause an intervention effect when concatenated with D 0 , due to its structural position. The first proposal is exactly what we find (note that the concatenation with plural HNs is illicit due to semantic reasons). Departing from the data above, I make use of common (count) nouns here, since the article/ numeral interpretation of ein enforces countability, even if the combination with mass nominals is not starred in German. (122)a. (m-)ein-e Straße ( poss -)ein street fem b. (m-)ein-ø Garten ( poss -)ein garden masc c. (m-)ein-ø Haus ( poss -)ein house neut Elaborating on 0 b. and c., coordination with D 0 should then result in illicit configurations. This is however not what we find: 28 (123) a. dieser ein-e Garten (= 0 b., [modified]) this one garden masc b. dieses ein-e Haus (= 0 c., [modified]) this one house neut The data in 0 are instead regular w.r.t. the proposals of in-situ 2 nd cycle Agree, as stated above, and can hence easily be subsumed into the theory elaborated here if the possessive and the ein -root do indeed differ in their structural position. The proposal has further implications for the last remaining SLI co-occurrence with Poss PRO , namely the concatenation with the strong quantifier. I will turn to this configuration now. Q S is the second SLI probing for the bare root [φ]. Starting again at the φ-uniform and simple combination of Poss PRO with feminine HNs, we arrive at 0 at the time of EM of all -. (124) all- mein-e Milch Q S > Poss PRO > Num [φ] [φ[f]] [φ[f]] Again, there is nothing curious about these cases; subsequently, Q S probes for φ, Match as well as Value apply between the quantifier and the possessive pronoun valuing the former as [φ[f]] and returning a uniform nominal domain w.r.t. φfeature structure. Again, this is also visible on the surface since Q S marks these features by means of (optional) suffixation: (125) all(-e) mein-e Milch Q S > Poss PRO > Num [φ[*f]] [φ[f]] [φ[f]] A more interesting derivation is built on the initial mismatch between the lower SLI and the nominal. As has been argued above, this first mismatch forms the basis for ungrammatical linearizations of the form D-Poss PRO -HN masc/ neut since D is structurally too close to the intervener to trigger a second cycle of Agree on Poss PRO while also being unable to surface with a default root node. By the hypothesis that in-situ 2 nd cycle Agree can only be triggered from outside the respective phasal cycle, Q S , too, should be unable to initiate probing of Poss PRO since they also constitute context sensitive phase 28 Observe that the introduction of the D-head into these structures is accompanied by a stressed intonation and an emphatic reading of the article/ numeral. Above, I proposed the identity of these elements solely on the basis of morphological form; see Roehrs (in prep.) for a detailed analysis of possible ein-word categories in which the stressed numeral is structurally split from the composite and article uses/ functions, cf. also chapter 1.4.4 above. Again, there is nothing curious about these cases; subsequently, Q S probes for φ, Match as well as Value apply between the quantifier and the possessive pronoun valuing the former as [φ[f]] and returning a uniform nominal domain w.r.t. φ-feature structure. Again, this is also visible on the surface since Q S marks these features by means of (optional) suffixation: (125) Nominal Concord & Definiteness I NFERENCE 164 2 nd d-ein masc neut 3 rd s-ein s-ein In light of the current proposals, the ein -root should then behave in a parallel manner to the possessive pronoun in the configuration(s) under discussion, to wit, ein might surface without overt inflection, due to possible selectional mismatches, and is moreover predicted to cause an intervention effect when concatenated with D 0 , due to its structural position. The first proposal is exactly what we find (note that the concatenation with plural HNs is illicit due to semantic reasons). Departing from the data above, I make use of common (count) nouns here, since the article/ numeral interpretation of ein enforces countability, even if the combination with mass nominals is not starred in German. (122)a. (m-)ein-e Straße ( poss -)ein street fem b. (m-)ein-ø Garten ( poss -)ein garden masc c. (m-)ein-ø Haus ( poss -)ein house neut Elaborating on 0 b. and c., coordination with D 0 should then result in illicit configurations. This is however not what we find: 28 (123) a. dieser ein-e Garten (= 0 b., [modified]) this one garden masc b. dieses ein-e Haus (= 0 c., [modified]) this one house neut The data in 0 are instead regular w.r.t. the proposals of in-situ 2 nd cycle Agree, as stated above, and can hence easily be subsumed into the theory elaborated here if the possessive and the ein -root do indeed differ in their structural position. The proposal has further implications for the last remaining SLI co-occurrence with Poss PRO , namely the concatenation with the strong quantifier. I will turn to this configuration now. Q S is the second SLI probing for the bare root [φ]. Starting again at the φ-uniform and simple combination of Poss PRO with feminine HNs, we arrive at 0 at the time of EM of all -. (124) all- mein-e Milch Q S > Poss PRO > Num [φ] [φ[f]] [φ[f]] Again, there is nothing curious about these cases; subsequently, Q S probes for φ, Match as well as Value apply between the quantifier and the possessive pronoun valuing the former as [φ[f]] and returning a uniform nominal domain w.r.t. φfeature structure. Again, this is also visible on the surface since Q S marks these features by means of (optional) suffixation: (125) all(-e) mein-e Milch Q S > Poss PRO > Num [φ[*f]] [φ[f]] [φ[f]] A more interesting derivation is built on the initial mismatch between the lower SLI and the nominal. As has been argued above, this first mismatch forms the basis for ungrammatical linearizations of the form D-Poss PRO -HN masc/ neut since D is structurally too close to the intervener to trigger a second cycle of Agree on Poss PRO while also being unable to surface with a default root node. By the hypothesis that in-situ 2 nd cycle Agree can only be triggered from outside the respective phasal cycle, Q S , too, should be unable to initiate probing of Poss PRO since they also constitute context sensitive phase 28 Observe that the introduction of the D-head into these structures is accompanied by a stressed intonation and an emphatic reading of the article/ numeral. Above, I proposed the identity of these elements solely on the basis of morphological form; see Roehrs (in prep.) for a detailed analysis of possible ein-word categories in which the stressed numeral is structurally split from the composite and article uses/ functions, cf. also chapter 1.4.4 above. A more interesting derivation is built on the initial mismatch between the lower SLI and the nominal. As has been argued above, this first mismatch forms the basis for ungrammatical linearizations of the form D-Poss PRO -HN masc/ neut since D is structurally too close to the intervener to trigger a second cycle of Agree on Poss PRO while also being unable to surface with a default root node. By the hypothesis that in-situ 2 nd cycle Agree can only be triggered from outside the respective phasal cycle, Q S , too, should be unable to initiate probing of Poss PRO 3.3 Splitting Nominal Concord 233 since they also constitute context sensitive phase heads of the same (viz the highest) cycle. In this configuration, Poss PRO should therefore surface without inflection. This is again exactly what we find. Diverging from the structure above, however, Q S , coding φ by means of inflection rather than at the stem, is able to surface as default, i.e. without said inflection, resulting in the grammatical (126): (126) all mein Wein Q S > Poss PRO > N [φ] [φ] [φ[m]] Observe that default inflection on the highest SLI in (126) above is not taken to arise from a selectional mismatch, but from a successful application of both Match and Value. Like with the D-head above, all probing for the root node is taken to find a successful goal in the likewise reduced possessive pronoun, following Bejar’s (2003) proposals. The data differ in that the means of φ-coding allow for the quantifier to surface without a value in the case at hand. Reduced features can hence come about in two ways: either by reduction to partial default due to mismatch (successful application of Match, unsuccessful application of Value) or by successful probing of two root nodes (successful application of both Match as well as Value). I did, however, propose above that the latter option does not result in a shared feature since no structure can be said to be copied from goal to probe (following P&T’s preconditions for feature sharing), the reasons for which will become evident at the end of this section. I take both variants of the root node to constitute (still) active features and hence to pose as possible probes for further 2 nd cycle Agree-operations. For now, let us focus on two key-properties derivable from (126): (i) As proposed above, the strong quantifier is taken to be unable to trigger a second cycle of Agree on the possessive pronoun since it is structurally too close (i.e. it is part of the same phasal cycle); moreover, (ii) all is able to surface without an overt instantiation of φ-feature structure by suffixation, i.e. Q S does exhibit a default form. The latter once again represents a reflex of the intervention effect brought about by the possessive pronoun and can hence be taken as another argument for the general derivational nature of nominal concord in German. Incorporating these findings into the classification of SLIs, we thus arrive at (127): (127) Q S D Poss PRO Q W Probing φ [φ] [φ] [φ[f]] [φ[f]] Default Agreement     234 3 INFERENCE - Nominal Concord & Definiteness The two highest SLIs are hence featurally distinguished by their (in-)ability to surface with default agreement, while the lower items, diverging from them in their selectional properties, are identical to one another in this respect. In the ensuing subchapter, I will elaborate on a third property that completes the classification above in finally dividing all SLIs under consideration in the analysis of adjectival inflection. Before we turn to that, let me present the last logical possibility of SLI co-occurrence and its respective analysis, namely the concatenation of the strong quantifier with the two instances of the category D. c. Co-Occurrence with D At this point, only one concatenation of two SLIs has not been analyzed in detail, namely the co-occurrence of the D-head with the single element externally merged above it, viz. the strong quantifier Q S . I am nevertheless positive that the idiosyncrasies of this configuration justify the establishment of a new section at the end of this subchapter. Preanalytic considerations might lead the reader to disagree at first since the elements under consideration are those SLIs which both probe for the root node [φ] and will hence be trivially compatible with every specification for the complex feature structure that can be found on N as well as Num. The diverging ability to surface with default agreement should hence be not of our concern in what follows. By now, a featurally uniform nominal domain is consistently expected at the end of the isolated nominal derivation headed by the Q S -D complex. Once again starting with possibly complex lexical domains, I exemplify the two onsets for derivations based on N and Num with a masculine and a feminine HN respectively. (128) a. Wein N [φ[m]] b. Milch Num [φ[f]] 3.3 Splitting Nominal Concord 235 As conjectured above, EM of the determiner poses no surprises: D probes and finds a goal in N (a.) or Num (b.) respectively. Both Match as well as Value succeed, resulting in a uniform φ-structure: (129) Nominal Concord & Definiteness I NFERENCE 166 [φ[f]] As conjectured above, EM of the determiner poses no surprises: D probes and finds a goal in N (a.) or Num (b.) respectively. Both Match as well as Value succeed, resulting in a uniform φ-structure: (129) a. der Wein D > N [φ[*m]] [φ[m]] b. die Milch D > Num [φ[*f]] [φ[f]] Likewise, the subsequent EM of the quantifier should proceed along the lines just sketched, with Q S establishing Match and Value with D irrespective of the φ-specification of the D-N complex. However, this is not what we find. Instead, (optional) inflection on the quantifier is only observable with feminine and plural nominals, not in derivations on the basis of masculine and neuter nominals; therefore, only when Num is part of the derivation does Q S display (the ability to bear) inflection. This is unexpected, given that the simple paradigm of Q S -N configurations presented in 3.3.1 above exhibits the quantifier to obligatorily inflect for all number/ gender-values. The question, then, is what conclusions can be drawn from this state of affairs for the featural set-up of the strong quantifier: (130) a. all(*-er) der Wein Q S > D > N [? ] [φ[*m]] [φ[m]] b. all(-e) die Milch Q S > D > Num [φ[*f]] [φ[*f]] [φ[f]] Taking the second instantiation of D, the demonstrative, into account, however, we arrive at an even more puzzling picture. Observe that the counterpart to the ungrammatical datum 0 b. above with a masculine HN turns out acceptable when the strong quantifier is concatenated with the complex Dem-N masc . (131) a. all(-er) dies-er Wein Q S > D > N [[φ[*m]] [φ[*m]] [φ[m]] b. all(-e) dies-e Milch Q S > D > Num [φ[*f]] [φ[*f]] [φ[f]] Above, I have argued at length that the determiner and the demonstrative pose two instantiations of the same category; an axiom of the theory elaborated here that seems questionable by the facts just presented. But this is not necessarily so. To see this, let us take a closer look at the full paradigm of Q S -Det/ Dem forms with masculine and neuter head nominals: (132) a. Det-N masc nom all der Wein gen all des Weins dat all dem Wein acc all den Wein b. Det-N neut nom all das Bier gen all des Biers Likewise, the subsequent EM of the quantifier should proceed along the lines just sketched, with Q S establishing Match and Value with D irrespective of the φ-specification of the D-N complex. However, this is not what we find. Instead, (optional) inflection on the quantifier is only observable with feminine and plural nominals, not in derivations on the basis of masculine and neuter nominals; therefore, only when Num is part of the derivation does Q S display (the ability to bear) inflection. This is unexpected, given that the simple paradigm of Q S -N configurations presented in 3.3.1 above exhibits the quantifier to obligatorily inflect for all number/ gender-values. The question, then, is what conclusions can be drawn from this state of affairs for the featural set-up of the strong quantifier: (130) Nominal Concord & Definiteness I NFERENCE 166 [φ[f]] As conjectured above, EM of the determiner poses no surprises: D probes and finds a goal in N (a.) or Num (b.) respectively. Both Match as well as Value succeed, resulting in a uniform φ-structure: (129) a. der Wein D > N [φ[*m]] [φ[m]] b. die Milch D > Num [φ[*f]] [φ[f]] Likewise, the subsequent EM of the quantifier should proceed along the lines just sketched, with Q S establishing Match and Value with D irrespective of the φ-specification of the D-N complex. However, this is not what we find. Instead, (optional) inflection on the quantifier is only observable with feminine and plural nominals, not in derivations on the basis of masculine and neuter nominals; therefore, only when Num is part of the derivation does Q S display (the ability to bear) inflection. This is unexpected, given that the simple paradigm of Q S -N configurations presented in 3.3.1 above exhibits the quantifier to obligatorily inflect for all number/ gender-values. The question, then, is what conclusions can be drawn from this state of affairs for the featural set-up of the strong quantifier: (130) a. all(*-er) der Wein Q S > D > N [? ] [φ[*m]] [φ[m]] b. all(-e) die Milch Q S > D > Num [φ[*f]] [φ[*f]] [φ[f]] Taking the second instantiation of D, the demonstrative, into account, however, we arrive at an even more puzzling picture. Observe that the counterpart to the ungrammatical datum 0 b. above with a masculine HN turns out acceptable when the strong quantifier is concatenated with the complex Dem-N masc . (131) a. all(-er) dies-er Wein Q S > D > N [[φ[*m]] [φ[*m]] [φ[m]] b. all(-e) dies-e Milch Q S > D > Num [φ[*f]] [φ[*f]] [φ[f]] Above, I have argued at length that the determiner and the demonstrative pose two instantiations of the same category; an axiom of the theory elaborated here that seems questionable by the facts just presented. But this is not necessarily so. To see this, let us take a closer look at the full paradigm of Q S -Det/ Dem forms with masculine and neuter head nominals: (132) a. Det-N masc nom all der Wein gen all des Weins dat all dem Wein acc all den Wein b. Det-N neut nom all das Bier gen all des Biers Taking the second instantiation of D, the demonstrative, into account, however, we arrive at an even more puzzling picture. Observe that the counterpart to the ungrammatical datum (130) a. above with a masculine HN turns out acceptable when the strong quantifier is concatenated with the complex Dem-N masc . 236 3 INFERENCE - Nominal Concord & Definiteness (131) Nominal Concord & Definiteness I NFERENCE 166 [φ[f]] As conjectured above, EM of the determiner poses no surprises: D probes and finds a goal in N (a.) or Num (b.) respectively. Both Match as well as Value succeed, resulting in a uniform φ-structure: (129) a. der Wein D > N [φ[*m]] [φ[m]] b. die Milch D > Num [φ[*f]] [φ[f]] Likewise, the subsequent EM of the quantifier should proceed along the lines just sketched, with Q S establishing Match and Value with D irrespective of the φ-specification of the D-N complex. However, this is not what we find. Instead, (optional) inflection on the quantifier is only observable with feminine and plural nominals, not in derivations on the basis of masculine and neuter nominals; therefore, only when Num is part of the derivation does Q S display (the ability to bear) inflection. This is unexpected, given that the simple paradigm of Q S -N configurations presented in 3.3.1 above exhibits the quantifier to obligatorily inflect for all number/ gender-values. The question, then, is what conclusions can be drawn from this state of affairs for the featural set-up of the strong quantifier: (130) a. all(*-er) der Wein Q S > D > N [? ] [φ[*m]] [φ[m]] b. all(-e) die Milch Q S > D > Num [φ[*f]] [φ[*f]] [φ[f]] Taking the second instantiation of D, the demonstrative, into account, however, we arrive at an even more puzzling picture. Observe that the counterpart to the ungrammatical datum 0 b. above with a masculine HN turns out acceptable when the strong quantifier is concatenated with the complex Dem-N masc . (131) a. all(-er) dies-er Wein Q S > D > N [[φ[*m]] [φ[*m]] [φ[m]] b. all(-e) dies-e Milch Q S > D > Num [φ[*f]] [φ[*f]] [φ[f]] Above, I have argued at length that the determiner and the demonstrative pose two instantiations of the same category; an axiom of the theory elaborated here that seems questionable by the facts just presented. But this is not necessarily so. To see this, let us take a closer look at the full paradigm of Q S -Det/ Dem forms with masculine and neuter head nominals: (132) a. Det-N masc nom all der Wein gen all des Weins dat all dem Wein acc all den Wein b. Det-N neut nom all das Bier gen all des Biers Above, I have argued at length that the determiner and the demonstrative pose two instantiations of the same category; an axiom of the theory elaborated here that seems questionable by the facts just presented. But this is not necessarily so. To see this, let us take a closer look at the full paradigm of Q S -Det/ Dem forms with masculine and neuter head nominals: (132) a. Det-N masc nom all der Wein gen all des Weins dat all dem Wein acc all den Wein b. Det-N neut nom all das Bier gen all des Biers dat all dem Bier acc all das Bier (133) a. Dem-N masc nom all(-er) dies-er Wein gen all dies-es Weins dat all dies-em Wein acc all(-en) dies-en Wein 3.3 Splitting Nominal Concord 237 b. Dem-N neut nom all dies-es Bier gen all dies-es Biers dat all dies-em Bier acc all dies-es Bier It thus becomes apparent that optional inflection is not uniformly available with Dem, but restricted to two exceptional cases, to wit, masculine nominative and accusative. Still, the diverging patterns of optional inflection would cast doubt on the categorical identity of the determiner and demonstrative if the source would be traced back to their featural content. Let me continue by presenting the paradigm of the Q S -Det/ Dem complexes with feminine HNs: (134) a. Dem-N fem nom all(-e) dies-e Milch gen all(-er) dies-er Milch dat all(-er) dies-er Milch acc all(-e) dies-e Milch b. Det-N fem nom all(-e) die Milch gen all der Milch dat all der Milch acc all(-e) die Milch As can be derived from the data above, optional inflection of the strong quantifier is not uniformly observed with feminine nominals either; the observation in (134) b. is due to Müller (1986, cf. also Dudenredaktion 2007: 54). Contrasting the forms allowing for optional inflection across the gender values leaves us with yet another interesting categorization. I additionally incorporate the plural paradigms below: (135) Det masc neut fem pl nom all der Wein all das Bier all(-e) die Milch all(-e) die Menschen gen all des Weins all des Biers all der Milch all der Menschen dat all dem Wein all dem Bier all der Milch all den Menschen acc all den Wein all das Bier all(-e) die Milch all(-e) die Menschen 238 3 INFERENCE - Nominal Concord & Definiteness (136) Dem masc neut fem pl nom all(-er) dies-er Wein all dies-es Bier all(-e) dies-e Milch all(-e) dies-e Menschen gen all dies-es Weins all dies-es Biers all(-er) dies-er Milch all(-er) dies-er Menschen dat all dies-em Wein all dies-em Bier all(-er) dies-er Milch all(-en) dies-en Menschen acc all(-en) dies-en Wein all dies-es Bier all(-e) dies-e Milch all(-e) dies-e Menschen I have marked the forms under consideration by greyscale coloring. With feminine nominals, we can observe that only the structural cases permit optional inflection on Q S , coordinated under the determiner, while the full paradigm of structural and non-structural cases allows it with the demonstrative. Additionally, the structural case forms also liscense inflection with the latter when concatenated with a masculine HN, whereas it is barred throughout the full paradigm in derivations incorporating neuter nominals. How, then, can these data be analyzed in the theory under elaboration? Recall from the introduction to chapter 3.2 that I took nominal concord to not be fully derivable from operations applying at Narrow Syntax. I therein advocated the split of adjectival inflection to be analyzed as a post-syntactic mechanism applying at Morphology. Likewise, I hinted at further morpho-phonological processes partaking in the composition of the phenomenon. We are now at the point to make these ideas precise: What I want to argue for is that the pattern of optional inflection on the strong quantifier preceding determiners and demonstratives becomes coherent once the focus is shifted from formal features to morphological form. Therein, the featural set-up of the particular instantiations of the D-head will be left untouched, a welcome effect. Let us begin by focusing on the similarities of the forms allowing inflection on Q S in the paradigms above. Beginning with the demonstrative and taking the uniform feminine and plural paradigms in (136) as the point of departure, we find three diverging overt inflections on the demonstrative, viz. e (feminine and plural nominative and accusative), er (feminine dative as well as feminine and plural genitive) and en (plural dative). The forms of the paradigm of said semi-lexical items disallowing inflection, on the contrary, show two distinct morphological realizations, namely es (neuter nominative and accusative as well as masculine and neuter genitive) and em (masculine and neuter dative). There is hence no intersection between morphological forms of the demonstrative accompanying and 3.3 Splitting Nominal Concord 239 disallowing optional inflection on the strong quantifier. With this in mind, let us turn to the forms allowing optional inflection with masculine HNs. As we see, the morphological shapes resemble those of the group allowing inflection in concatenation with feminine and plural HNs, namely er (in the nominative, parallel to feminine dative as well as feminine and plural genitive) and en (in accusative, equal to the plural dative). Can this observation be made formal in the axioms of the analysis elaborated here? I believe it can, but that the derivation of an appropriate account requires some preliminary considerations to which I turn forthwith. It is a widely known phenomenon that languages try to avoid word-internal phonetic reduplication by change or deletion of one of the duplicates. 28 The latter option has been termed ‘haplology’. Several authors proposed this mechanism to also apply above the word-level to (parts of) adjacent LIs. For instance, in the German example given in (137), due to Radford (1977, 1979) and discussed in Neeleman & Van de Koot (2006), the phonological identity of the comparative and temporal prepositions ( als ) forces a diverging phonetic realization of the first LI ( denn ) when surfacing in adjacent positions (the authors take suppletition to be one of the mechanisms beside haplology to prevent - what they refer to as - accidental reduplication): (137) (= Neeleman & Van de Koot 2006: (30)) Goethe ist bekannter als Schriftsteller denn/ *als als Naturwissenschaftler. Goethe is better-known as writer than/ as as scientist ‘Goethe is better known as a writer than as a scientist.’ Such strategies might be taken to constitute filters at PF, however, application beyond the word-boundary also allows for a possible relocation of the mechanism to C HL . Thus, Szabolcsi’s (1994) notion of haplology in the treatment of complex nominal domains in Hungarian includes phonetic as well as syntactic mechanisms (viz. deletion and movement) in the prevention of illegal lexical strings: (138) (= Szabolcsi 1994: (69)) Haplology a. The co-occurrence of D and Det is grammatical if they are linearly separated by some intervener. 28 At the same time, regular reduplication does pose a strategy for both inflectional morphology as well as word formation in the languages of the world, cf. Bussmann (1996), Rubino (2013). 240 3 INFERENCE - Nominal Concord & Definiteness b. Contiguous strings of the type D Det, or D D, are ungrammatical. Ungrammaticality can be eliminated either by deleting a(z) of D in Phonetic Form, or by moving the constituent that contains Det or the second D. Likewise, the evaluation of conflicting material might also not necessarily be part of the PF-interface as well. Therefore, Barbiers (2005) is able to base said evaluation on formal features in his Haplology Rule in the analysis of complex NPs in Dutch dialects and English, given below: (139) (= Barbiers 2005: (23)) Haplology Rule Leave D empty at PF when D and Num are adjacent and the features of D are a subset of the features of Num. Neeleman & Van de Koot (2006) likewise base their evaluation of conflicting ‘material’ on formal features and are hence able to include the application of haplology and suppletition to non-identical phonetic representations of identical featural content. Consequently, the approach is termed ‘Syntactic Haplology’. Haplology therein hence constitutes a joint effect whose cause might originate in Narrow Syntax (feature identity) and PF (adjacency, rather than hierarchy) while the effect is solely generated at PF (deletion). The authors moreover stress that their analysis deals with accidental repetitions (i.e. adjacency on the surface) and that the application of preventive strategies is idiosyncratic in that they are “language-specific, construction-specific and morpheme-specific” (ibidem: 708). Hence, no universal rule is enforced; even more so, they acknowledge that repetitions will be tolerated in most cases and therefore phenomena of agreement are free from haplological filtering. I want to argue that the paradigms of Q S -D conjunctions presented above should be analyzed parallel to the processes just presented, i.e. as a phonological effect. Like the syntactic accounts above, this applies to adjacent but dedicated (semi-)lexical items but is in itself an effect of PF and hence evaluated on the basis of phonological form rather than featural content. A switch in perspective is vital to this account: It is not the case that PF deletes lexical material to save conflicting linearizations in the cases at hand, rather, PF allows repetitive lexical material if the linearization meets PF-restrictions. Therein, the account advocated is an instance of so-called ‘dittology’ which the Routledge Dictionary of Language and Linguistics (Bussmann 1996) defines as the “accidental or conventionalized repetition of a syllable.” Understanding the data presented at the beginning of this section in the light of this rationale allows us to uncover a regularity that is easily overlooked when focusing on the featural set-up of the SLIs involved. The reasoning behind a dittological approach to inflection 3.3 Splitting Nominal Concord 241 on strong quantifiers is as follows: Focussing again on the paradigm of Q S -Dem conjunctions first, recall that there is no intersection between the inflectional suffix forms on the D-head that allow for inflection on the quantifier and those which prohibit it. On the basis of licit inflections in all case forms of the [φ[f]] paradigms (feminine and plural), observe that the additional licit masculine nominative and accusative forms constitute the exhaustive group of phonological shapes homonymous with forms otherwise only found in the [φ[f]] paradigms, namely er (in the masculine nominative, parallel to genitive feminine and plural as well as dative feminine) and en (in masculine accusative, parallel to dative plural). Taking [φ[f]] as the phonological yardstick to measure grammatical conjuncts allowing for inflection on Q S , the rationale of the dittological account hence predicts that all forms that are homonymous with the licit paradigm also turn out grammatical. Call this the ‘inclusive paradigm’. When taking this idea to the paradigm of Q S -Det conjunctions, the varying pattern also falls into place since it constitutes the second logically possible application of said rationale: Here, the only acceptable cases of optional inflection on Q S are found in the nominative and accusative cases of the [φ[f]] paradigms. Focusing on morphological shape of the determiner, we uncover that these constitute the full range of forms, exclusive to said paradigms (i.e. all occurrences of the form die ), while the remaining ones der (feminine genitive and dative, as well as the plural genitive) and den (plural dative) are also found elsewhere, to wit, in the configurations prohibiting optional inflection. Call this the ‘exclusive paradigm’. A PF-account to optional quantifier inflection hence captures the diverging paradigms with determiners and demonstratives without recurrence to the featural content of the two instantiations of the category D: With demonstratives, all forms parallel to the licit [φ[f]] patterns turn grammatical by analogy, while with determiners, only the forms exclusive to the [φ[f]] paradigms are licit. For obvious reasons, I labeled these the inclusive and exclusive paradigm respectively. These idiosyncrasies remain mysterious under a narrow-syntactic approach to the featural content of the SLIs involved. The dittological account, in contrast, allows us to maintain a uniform feature structure of complex nominal domains headed by Q S -D concatenations. Q S always bears a complex φ-specification but is prohibited to surface with the appropriate suffixial inflection by filters applying after NS. (140) Nominal Concord & Definiteness I NFERENCE 169 Neeleman & Van de Koot (2006) likewise base their evaluation of conflicting ‘material’ on formal features and are hence able to include the application of haplology and suppletition to non-identical phonetic representations of identical featural content. Consequently, the approach is termed ‘Syntactic Haplology’. Haplology therein hence constitutes a joint effect whose cause might originate in Narrow Syntax (feature identity) and PF (adjacency, rather than hierarchy) while the effect is solely generated at PF (deletion). The authors moreover stress that their analysis deals with accidental repetitions (i.e. adjacency on the surface) and that the application of preventive strategies is idiosyncratic in that they are “languagespecific, construction-specific and morpheme-specific” (ibidem: 708). Hence, no universal rule is enforced; even more so, they acknowledge that repetitions will be tolerated in most cases and therefore phenomena of agreement are free from haplological filtering. I want to argue that the paradigms of Q S -D conjunctions presented above should be analyzed parallel to the processes just presented, i.e. as a phonological effect. Like the syntactic accounts above, this applies to adjacent but dedicated (semi- )lexical items but is in itself an effect of PF and hence evaluated on the basis of phonological form rather than featural content. A switch in perspective is vital to this account: It is not the case that PF deletes lexical material to save conflicting linearizations in the cases at hand, rather, PF allows repetitive lexical material if the linearization meets PF-restrictions. Therein, the account advocated is an instance of so-called ‘dittology’ which the Routledge Dictionary of Language and Linguistics (Bussmann 1996) defines as the “accidental or conventionalized repetition of a syllable.” Understanding the data presented at the beginning of this section in the light of this rationale allows us to uncover a regularity that is easily overlooked when focusing on the featural set-up of the SLIs involved. The reasoning behind a dittological approach to inflection on strong quantifiers is as follows: Focussing again on the paradigm of Q S -Dem conjunctions first, recall that there is no intersection between the inflectional suffix forms on the D-head that allow for inflection on the quantifier and those which prohibit it. On the basis of licit inflections in all case forms of the [φ[f]] paradigms (feminine and plural), observe that the additional licit masculine nominative and accusative forms constitute the exhaustive group of phonological shapes homonymous with forms otherwise only found in the [φ[f]] paradigms, namely er (in the masculine nominative, parallel to genitive feminine and plural as well as dative feminine) and en (in masculine accusative, parallel to dative plural). Taking [φ[f]] as the phonological yardstick to measure grammatical conjuncts allowing for inflection on Q S , the rationale of the dittological account hence predicts that all forms that are homonymous with the licit paradigm also turn out grammatical. Call this the ‘inclusive paradigm’. When taking this idea to the paradigm of Q S -Det conjunctions, the varying pattern also falls into place since it constitutes the second logically possible application of said rationale: Here, the only acceptable cases of optional inflection on Q S are found in the nominative and accusative cases of the [φ[f]] paradigms. Focusing on morphological shape of the determiner, we uncover that these constitute the full range of forms, exclusive to said paradigms (i.e. all occurrences of the form die ), while the remaining ones der (feminine genitive and dative, as well as the plural genitive) and den (plural dative) are also found elsewhere, to wit, in the configurations prohibiting optional inflection. Call this the ‘exclusive paradigm’. A PF-account to optional quantifier inflection hence captures the diverging paradigms with determiners and demonstratives without recurrence to the featural content of the two instantiations of the category D: With demonstratives, all forms parallel to the licit [φ[f]] patterns turn grammatical by analogy, while with determiners, only the forms exclusive to the [φ[f]] paradigms are licit. For obvious reasons, I labeled these the inclusive and exclusive paradigm respectively. These idiosyncrasies remain mysterious under a narrow-syntactic approach to the featural content of the SLIs involved. The dittological account, in contrast, allows us to maintain a uniform feature structure of complex nominal domains headed by Q S -D concatenations. Q S always bears a complex φ-specification but is prohibited to surface with the appropriate suffixial inflection by filters applying after NS. (140) all(*-er) der Wein Q S > D > N [φ[*m]] [φ[*m]] [φ[m]] When turning to the concatenation of the strong quantifier with the possessive pronoun, however, the narrow-syntactic derivation, advocated above, readily accounts for the full paradigm, i.e. the quantifier may take optional inflection iff the 242 3 INFERENCE - Nominal Concord & Definiteness When turning to the concatenation of the strong quantifier with the possessive pronoun, however, the narrow-syntactic derivation, advocated above, readily accounts for the full paradigm, i.e. the quantifier may take optional inflection iff the possessive pronoun likewise does so, hence if it forms a feature chain with the lexical domain of the nominal hierarchy of projections (i.e. N or Num): 29 (141) Poss masc neut fem pl nom all mein-ø N all mein-ø N all(-e) mein-e N all(-e) mein-e N gen all(-es) mein-es N all(-es) mein-es N all(-er) mein-er N all(-er) mein-er N dat all(-em) mein-em N all(-em) mein-em N all(-er) mein-er N all(-en) mein-en N acc all mein-en N all mein-ø N all(-e) mein-e N all(-e) mein-e N An open question remains as to why it is the [φ[f]] paradigms that constitute the basis for the evaluation of grammaticality concerning optional inflection on the quantifier when conjoined with elements of category D even though both elements involved are proposed to probe for the root node [φ]. I will, however, have to leave this question open for now and return to it in the course of the discussion on Adjectival Impoverishment in the next subchapter. We have, then, seen that nominal domains headed by the conjunction of determiner/ demonstrative with the strong quantifier are not captured in full by the derivational paths argued for in the preceding sections of this chapter. Deviating from the predictions of uniform feature structures, optional inflection on the quantifier is only allowed in a subset of the paradigms under consideration. Even worse, the licit forms vary with the two instantiations of the category D: With determiners, only non-structural case forms in feminine and plural number/ gender allow quantifier inflection on the surface, while with demonstratives, the full paradigm of feminine and plural forms as well as structural cases in masculine number/ gender may permit it. This is an unwelcome state of affairs, given that (i) both SLIs involved in the respective structures are proposed to probe for the root node [φ] (as backed up by their inflectional properties in 29 Observe that the accusative masculine form falls out of this generalization on the surface but therein supports the claims made about the featural content of the inflection in chapter 3.2.2 above, namely that this form does not pose an instantiation of φ-coding but rather an idiosyncratic case-suffix. I will return to this observation again in the next subchapter; cf. also fn. 33 below. 3.3 Splitting Nominal Concord 243 isolation, cf. chapter 3.2.1) and should therefore always display uniform inflection as well as (ii) that the two instances of the category D moreover exhibit diverging sets of case forms allowing inflection. Switching the viewpoint from featural content to phonological form, however, allowed to uncover regularities in the scattered and heterogeneous patterns: Based on the paradigms of [φ[f]] agreement, the inclusive paradigm permits inflection in all and only those forms that resemble realizations of the former on the surface (Q S -Dem), while the exclusive paradigm allows for inflection on those realizations that are solely found therein (Q S -Det). The exclusive and inclusive paradigms exhaust the logical possibilities of a dittological approach measured against a core pattern (i.e. [φ[f]] above). The question remains why the evaluation applies w.r.t. this paradigm, an issue that I have to postpone for now. We will see later in the analysis of Adjectival Impoverishment, that the choice is motivated in the selectional restrictions of the SLIs involved and is therefore regular. Before we turn to this, however, I want to close the subchapter on the extended pattern of SLI inflection and nominal concord in parallel to the one on case-assignment above in the discussion of further applications of 2 nd cycle Agree by IM. d. Quantifier Float, Split Topicalization and 2 nd Cycle Agree Recall from chapter 2.2 that I identified non-phasehood of the strong (and, by analogy, the weak) quantifier to pose a precondition for quantifier float, in line with Bošković’s (2014) proposed properties of phase heads. Moreover, I took the obligatory inflection on the possessive pronoun to constitute a reflex of movement through its specifier in the analysis of focused possessive nominal domains at the end of the last subchapter. With the detailed presentation of the extended pattern of SLI inflection and nominal concord in place, we are now in the position to unite these approaches into a coherent analysis of quantifier float in German. I will begin with the more frequent and well-known case of strong quantifier float (QST). As elaborated in detail above, Q S surfaces as a non-phase when conjoined with either D or Poss PRO , i.e. one of the more prominent contextual phase heads of the highest nominal phasal cycle. In the resulting configurations, we have observed heterogeneous patterns of optional inflection on the quantifier above due to intervention effects (Poss PRO ) or PF filtering (D). Merchant (1996) analyzes floating (strong) quantifier constructions in German along proposals originally made by Shlonsky (1991) for the same phenomenon in Hebrew. Therein, also granting phrasal status of the quantifier above DP, he takes the inflection on all - (which, for him, solely depicts caseagreement) to constitute the presence of a feature [F], neither strong nor weak in Chomsky’s (MPLT) terms. The optionality of QST relies in the overt vs. covert 244 3 INFERENCE - Nominal Concord & Definiteness checking of [F], which - following the state of the theory - can only apply in the canonical Spec-head configuration. Movement succeeds through Spec,QP, which Merchant, following Georgi & Longobardi (1991 for Spec,NP; cf. also Merchant 1996: fn. 3), takes as a necessary intermediary landing side in extraction from the nominal domain. It is the checking of [F] that motivates IM to this position. The presence of [F], and thereby inflection itself, is thus a precondition for floating constructions. Bare all does not bear [F] and hence never surfaces in a floating construction while inflected all might still be derived by covert operations. Therein, Merchant is able to capture both the optional occurrence of inflection in in-situ configurations as well as the obligatory inflection of the quantifier in QST-environments. Aiming for a unified account of floating phenomena, however, we immediately observe that inflection (qua presence of [F], transposed to [φ] in the theory elaborated here) does not universally pose a precondition for movement since focused possessive nominal domains allow for movement from below Poss PRO even in structural cases with non-feminine HNs, i.e. the environments in which the possessive pronoun surfaces as bare mein . Even more pressing, quantifier float is not restricted to those cases, presented in the previous section, in which optional inflection is allowed if one extends the paradigm to all gender and case forms (cf. ch. 4.2.2 below for a thorough discussion of optional and obligatory inflection including the proposals of Merhcant 1996). Following the rationale of chapter 3.2.2, I want to argue that obligatory inflection on the quantifier rather poses a reflex of movement through its extended search space, Spec,Q S P. The analysis is in most parts parallel to the one presented above but rests on a more complex phrasal set-up of the nominal domain. Working backwards through the configurations analyzed above, let me begin here with the conjunction of the quantifier with the two possible instantiations of DP respectively. Recall that Q S has been taken to bear a shared feature with the latter in all possible N and Num configurations, with restrictions concerning its overt realization applying later in the course of the derivation at PF. Abstracting away from specific φ-values, the stage of the derivation in which the nominal domain merges into the verbal spine is exemplified in (142): (142) Nominal Concord & Definiteness I NFERENCE 171 which - following the state of the theory - can only apply in the canonical Spec-head configuration. Movement succeeds through Spec,QP, which Merchant, following Georgi & Longobardi (1991 for Spec,NP; cf. also Merchant 1996: fn. 3), takes as a necessary intermediary landing side in extraction from the nominal domain. It is the checking of [F] that motivates IM to this position. The presence of [F], and thereby inflection itself, is thus a precondition for floating constructions. Bare all does not bear [F] and hence never surfaces in a floating construction while inflected all might still be derived by covert operations. Therein, Merchant is able to capture both the optional occurrence of inflection in in-situ configurations as well as the obligatory inflection of the quantifier in QST-environments. Aiming for a unified account of floating phenomena, however, we immediately observe that inflection (qua presence of [F], transposed to [φ] in the theory elaborated here) does not universally pose a precondition for movement since focused possessive nominal domains allow for movement from below Poss PRO even in structural cases with non-feminine HNs, i.e. the environments in which the possessive pronoun surfaces as bare mein . Even more pressing, quantifier float is not restricted to those cases, presented in the previous section, in which optional inflection is allowed if one extends the paradigm to all gender and case forms (cf. ch. 4.2.2 below for a thorough discussion of optional and obligatory inflection including the proposals of Merhcant 1996). Following the rationale of chapter 0, I want to argue that obligatory inflection on the quantifier rather poses a reflex of movement through its extended search space, Spec,Q S P. The analysis is in most parts parallel to the one presented above but rests on a more complex phrasal set-up of the nominal domain. Working backwards through the configurations analyzed above, let me begin here with the conjunction of the quantifier with the two possible instantiations of DP respectively. Recall that Q S has been taken to bear a shared feature with the latter in all possible N and Num configurations, with restrictions concerning its overt realization applying later in the course of the derivation at PF. Abstracting away from specific φ-values, the stage of the derivation in which the nominal domain merges into the verbal spine is exemplified in 0: (142) Q S > D > Num/ N [φ[* val ]] [φ[* val ]] [φ[ val ]] Already bearing an instance of the shared φ-structure from the initial probing in the course of the isolated derivation on a parallel workspace, Q S is φ-specified when the complex is externally merged into the verbal domain. Let us, for the sake of exposition, take 0 to be merged into the external argument position. Subsequent probing from T hence matches φ-complete [φ[* val ]]. At this point, however, two diverging derivational paths are available: Either T 0 moves Q S and its complement, i.e. all elements bearing an instance of [φ[* val ]], thus the complete nominal domain internally merges in the specifier or T attracts the most minimal string bearing said specification, which - in this case - differs in that it is the phase head D 0 and its complement which are able to move. Focusing on the latter option, Q S in base position and the nominal domain in the specifier of TP still share the φ-specification with one another as well as with T 0 . Therefore, case-assignment to the nominal (as a modification of the complex shared feature structure) also affects the featural content and hence the morphological shape of the (inflection of) the quantifier. This is exemplified below, with case again marked apart from the φ-specification for ease of exposition: (143) [D > Num/ N] i > T 0 > [Q S t i ] > v 0 [φ[* val ]] [φ[ val ]] [φ[* val ]] [φ[* val ]] [*nom] [*nom] [*nom] With the linear configuration that constitutes the basis for the application of dittology at PF split up in NS, we additionally expect the phonological deletion of quantifier inflection to vanish in QST-contexts. This prediction is borne out; see the floating construction below involving a nominal domain on the basis of a neuter HN in direct object position. Recall from the last section that inflection on the quantifier never surfaces in conjunction with instantiations of D heading neuter nominal hierarchies. (144)a. Ich habe [all-ø das/ dies-es Wasser] getrunken. I have [ all the/ this water ] drunk b. [Das/ Dies-es Wasser] i habe ich all-es t i getrunken. [ the/ this water ] i have I all t i drunk ‘I have drunk all the/ this water.’ As can be deduced from the data presented above, inflection on the quantifier is not a precondition for movement out of the nominal domain in conjunctions with the determiner/ demonstrative but rather a side effect of the elimination of the Already bearing an instance of the shared φ-structure from the initial probing in the course of the isolated derivation on a parallel workspace, Q S is φ-specified when the complex is externally merged into the verbal domain. Let us, for the 3.3 Splitting Nominal Concord 245 sake of exposition, take (142) to be merged into the external argument position. Subsequent probing from T hence matches φ-complete [φ[* val ]]. At this point, however, two diverging derivational paths are available: Either T 0 moves Q S and its complement, i.e. all elements bearing an instance of [φ[* val ]], thus the complete nominal domain internally merges in the specifier or T attracts the most minimal string bearing said specification, which - in this case - differs in that it is the phase head D 0 and its complement which are able to move. Focusing on the latter option, Q S in base position and the nominal domain in the specifier of TP still share the φ-specification with one another as well as with T 0 . Therefore, case-assignment to the nominal (as a modification of the complex shared feature structure) also affects the featural content and hence the morphological shape of the (inflection of) the quantifier. This is exemplified below, with case again marked apart from the φ-specification for ease of exposition: (143) Nominal Concord & Definiteness I NFERENCE 171 which - following the state of the theory - can only apply in the canonical Spec-head configuration. Movement succeeds through Spec,QP, which Merchant, following Georgi & Longobardi (1991 for Spec,NP; cf. also Merchant 1996: fn. 3), takes as a necessary intermediary landing side in extraction from the nominal domain. It is the checking of [F] that motivates IM to this position. The presence of [F], and thereby inflection itself, is thus a precondition for floating constructions. Bare all does not bear [F] and hence never surfaces in a floating construction while inflected all might still be derived by covert operations. Therein, Merchant is able to capture both the optional occurrence of inflection in in-situ configurations as well as the obligatory inflection of the quantifier in QST-environments. Aiming for a unified account of floating phenomena, however, we immediately observe that inflection (qua presence of [F], transposed to [φ] in the theory elaborated here) does not universally pose a precondition for movement since focused possessive nominal domains allow for movement from below Poss PRO even in structural cases with non-feminine HNs, i.e. the environments in which the possessive pronoun surfaces as bare mein . Even more pressing, quantifier float is not restricted to those cases, presented in the previous section, in which optional inflection is allowed if one extends the paradigm to all gender and case forms (cf. ch. 4.2.2 below for a thorough discussion of optional and obligatory inflection including the proposals of Merhcant 1996). Following the rationale of chapter 0, I want to argue that obligatory inflection on the quantifier rather poses a reflex of movement through its extended search space, Spec,Q S P. The analysis is in most parts parallel to the one presented above but rests on a more complex phrasal set-up of the nominal domain. Working backwards through the configurations analyzed above, let me begin here with the conjunction of the quantifier with the two possible instantiations of DP respectively. Recall that Q S has been taken to bear a shared feature with the latter in all possible N and Num configurations, with restrictions concerning its overt realization applying later in the course of the derivation at PF. Abstracting away from specific φ-values, the stage of the derivation in which the nominal domain merges into the verbal spine is exemplified in 0: (142) Q S > D > Num/ N [φ[* val ]] [φ[* val ]] [φ[ val ]] Already bearing an instance of the shared φ-structure from the initial probing in the course of the isolated derivation on a parallel workspace, Q S is φ-specified when the complex is externally merged into the verbal domain. Let us, for the sake of exposition, take 0 to be merged into the external argument position. Subsequent probing from T hence matches φ-complete [φ[* val ]]. At this point, however, two diverging derivational paths are available: Either T 0 moves Q S and its complement, i.e. all elements bearing an instance of [φ[* val ]], thus the complete nominal domain internally merges in the specifier or T attracts the most minimal string bearing said specification, which - in this case - differs in that it is the phase head D 0 and its complement which are able to move. Focusing on the latter option, Q S in base position and the nominal domain in the specifier of TP still share the φ-specification with one another as well as with T 0 . Therefore, case-assignment to the nominal (as a modification of the complex shared feature structure) also affects the featural content and hence the morphological shape of the (inflection of) the quantifier. This is exemplified below, with case again marked apart from the φ-specification for ease of exposition: (143) [D > Num/ N] i > T 0 > [Q S t i ] > v 0 [φ[* val ]] [φ[ val ]] [φ[* val ]] [φ[* val ]] [*nom] [*nom] [*nom] With the linear configuration that constitutes the basis for the application of dittology at PF split up in NS, we additionally expect the phonological deletion of quantifier inflection to vanish in QST-contexts. This prediction is borne out; see the floating construction below involving a nominal domain on the basis of a neuter HN in direct object position. Recall from the last section that inflection on the quantifier never surfaces in conjunction with instantiations of D heading neuter nominal hierarchies. (144)a. Ich habe [all-ø das/ dies-es Wasser] getrunken. I have [ all the/ this water ] drunk b. [Das/ Dies-es Wasser] i habe ich all-es t i getrunken. [ the/ this water ] i have I all t i drunk ‘I have drunk all the/ this water.’ As can be deduced from the data presented above, inflection on the quantifier is not a precondition for movement out of the nominal domain in conjunctions with the determiner/ demonstrative but rather a side effect of the elimination of the With the linear configuration that constitutes the basis for the application of dittology at PF split up in NS, we additionally expect the phonological deletion of quantifier inflection to vanish in QST-contexts. This prediction is borne out; see the floating construction below involving a nominal domain on the basis of a neuter HN in direct object position. Recall from the last section that inflection on the quantifier never surfaces in conjunction with instantiations of D heading neuter nominal hierarchies. (144) a. Ich habe [all-ø das/ dies-es Wasser] getrunken. I have [ all the / this water ] drunk b. [Das/ Dies-es Wasser] i habe ich all-es t i getrunken. [ the / this water ] i have I all t i drunk ‘I have drunk all the/ this water.’ As can be deduced from the data presented above, inflection on the quantifier is not a precondition for movement out of the nominal domain in conjunctions with the determiner/ demonstrative but rather a side effect of the elimination of the linearization that forces a dittological reduction after Narrow Syntax at PF. The specification of φ can hence freely surface on the quantifier. With the possessive pronoun, however, no additional filtering strategies are observed. Inflection (qua φ-specification) surfaces on Q S iff it also surfaces on 246 3 INFERENCE - Nominal Concord & Definiteness Poss PRO due to intervention of the latter in the attempt of the former to probe in non-feminine nominal domains. These configurations prior to EM in the verbal domain are exemplified below. (145) Nominal Concord & Definiteness I NFERENCE 172 linearization that forces a dittological reduction after Narrow Syntax at PF. The specification of φ can hence freely surface on the quantifier. With the possessive pronoun, however, no additional filtering strategies are observed. Inflection (qua φ-specification) surfaces on Q S iff it also surfaces on Poss PRO due to intervention of the latter in the attempt of the former to probe in nonfeminine nominal domains. These configurations prior to EM in the verbal domain are exemplified below. (145) a. all-e mein-e Milch Q S > Poss PRO > Num [φ[*f]] [φ[f]] [φ[f]] b. all mein Wein Q S > Poss PRO > N [φ] [φ] [φ[m]] Recall from chapter 0 as well as from the analyses above, that the root node on the strong quantifier in 0 b. does not constitute the result of a mismatch, hence reduction to partial default, but rather the outcome of a successful application of Value (following Bejar 2003) without the subsequent establishment of a shared feature on the basis of P&T’s analysis. Following the rationale of structural case-assignment made in chapter 0 above as well as the derivation sketched on the basis of Q S -D conjuncts, let us concentrate on the clausal derivation following EM of 0 b. in external argument position (i.e. Spec, v P). Once again, T merges and probes for a φ-complete goal, therein skipping both SLIs parallel to the derivation proposed for simple Poss PRO -N complexes above. N matches T and once again the derivation can succeed along two paths: Either the full nominal domain (i.e. Q S P) or the most minimal phase (i.e. Poss PRO P) is moved to the specifier of TP. Let us further pursue the latter option here. Movement is once again taken to succeed through the specifier of Q S P but this time the quantifier does not bear an instance of the shared φ-feature and is hence able to probe again into its extended search space. Observe that in this second probing the highest SLI finds a match in the φ-value of the nominal below the phase head Poss PRO , hence the intervention effect is no longer observed. We might therein have found an indication as to the timing of deactivation of formal features from partial to total default (cf. Bejar 2003: 79) in the derivation of nominal concord in German. However, I have not been able to pin down the timing further given that [φ] on Poss PRO probes a second time in-situ when assigned non-structural case by V (contra EM of the nominal domain) and by movement through its extended search space in focused NP-constructions (contra targeting from a structural case-‘bearer’). I therefore have to leave this as an open question here. Observe moreover that I take the root node on the quantifier, derived by successful applications of Match and Value, to initiate a 2 nd cycle of Agree parallel to the root node brought about by mismatch and reduction. Since no shared feature is established at the root, however, modifications of feature structure do not carry over to the original intervener Poss PRO in these cases. Hence, whatever causes the feature structure of an SLI to constitute the root node [φ], the structure is undifferentiated for means of 2 nd cycle Agree in the further course of the derivation before deactivation to total default. Below I give the structure for the stranded quantifier construction on the basis of 0 b. as discussed above. Observe that once again case-assignment successfully applies to the stranded SLI due to the shared feature with the head noun in Spec,TP as well as T 0 . (146) [Poss PRO > N] i > T 0 > [Q S t i ] > v 0 [φ] [φ[m]] [φ[*m]] [φ[*m]] [*nom] [*nom] The result on the surface is as expected by now: We have seen above that in in-situ configurations both SLIs surface with default agreement (i.e. suffixless), while in floating constructions, the second probing of Q S due to movement from below the quantifier into its extended search space and the succeeding implementation of the SLI into the feature chain results in overt inflection on the surface. In both cases, however, the possessive pronoun fails to Value on the first cycle, thus only bearing a root node feature to be successively spelled out as default agreement. Both cases are exemplified with a neuter HN, incompatible with the selectional restrictions of the pronoun, below. Recall from chapter 3.1.2 as well as from the analyses above, that the root node on the strong quantifier in (145) b. does not constitute the result of a mismatch, hence reduction to partial default, but rather the outcome of a successful application of Value (following Bejar 2003) without the subsequent establishment of a shared feature on the basis of P&T’s analysis. Following the rationale of structural case-assignment made in chapter 3.2.2 above as well as the derivation sketched on the basis of Q S -D conjuncts, let us concentrate on the clausal derivation following EM of (145) b. in external argument position (i.e. Spec, v P). Once again, T merges and probes for a φ-complete goal, therein skipping both SLIs parallel to the derivation proposed for simple Poss PRO -N complexes above. N matches T and once again the derivation can succeed along two paths: Either the full nominal domain (i.e. Q S P) or the most minimal phase (i.e. Poss PRO P) is moved to the specifier of TP. Let us further pursue the latter option here. Movement is once again taken to succeed through the specifier of Q S P but this time the quantifier does not bear an instance of the shared φ-feature and is hence able to probe again into its extended search space. Observe that in this second probing the highest SLI finds a match in the φ-value of the nominal below the phase head Poss PRO , hence the intervention effect is no longer observed. We might therein have found an indication as to the timing of deactivation of formal features from partial to total default (cf. Bejar 2003: 79) in the derivation of nominal concord in German. However, I have not been able to pin down the timing further given that [φ] on Poss PRO probes a second time in-situ when assigned non-structural case by V (contra EM of the nominal domain) and by movement through its extended search space in focused NP-constructions (contra targeting from a structural case-‘bearer’). I therefore have to leave this as an open question here. 3.3 Splitting Nominal Concord 247 Observe moreover that I take the root node on the quantifier, derived by successful applications of Match and Value, to initiate a 2 nd cycle of Agree parallel to the root node brought about by mismatch and reduction. Since no shared feature is established at the root, however, modifications of feature structure do not carry over to the original intervener Poss PRO in these cases. Hence, whatever causes the feature structure of an SLI to constitute the root node [φ], the structure is undifferentiated for means of 2 nd cycle Agree in the further course of the derivation before deactivation to total default. Below I give the structure for the stranded quantifier construction on the basis of (145) b. as discussed above. Observe that once again case-assignment successfully applies to the stranded SLI due to the shared feature with the head noun in Spec,TP as well as T 0 . (146) Nominal Concord & Definiteness I NFERENCE 172 linearization that forces a dittological reduction after Narrow Syntax at PF. The specification of φ can hence freely surface on the quantifier. With the possessive pronoun, however, no additional filtering strategies are observed. Inflection (qua φ-specification) surfaces on Q S iff it also surfaces on Poss PRO due to intervention of the latter in the attempt of the former to probe in nonfeminine nominal domains. These configurations prior to EM in the verbal domain are exemplified below. (145) a. all-e mein-e Milch Q S > Poss PRO > Num [φ[*f]] [φ[f]] [φ[f]] b. all mein Wein Q S > Poss PRO > N [φ] [φ] [φ[m]] Recall from chapter 0 as well as from the analyses above, that the root node on the strong quantifier in 0 b. does not constitute the result of a mismatch, hence reduction to partial default, but rather the outcome of a successful application of Value (following Bejar 2003) without the subsequent establishment of a shared feature on the basis of P&T’s analysis. Following the rationale of structural case-assignment made in chapter 0 above as well as the derivation sketched on the basis of Q S -D conjuncts, let us concentrate on the clausal derivation following EM of 0 b. in external argument position (i.e. Spec, v P). Once again, T merges and probes for a φ-complete goal, therein skipping both SLIs parallel to the derivation proposed for simple Poss PRO -N complexes above. N matches T and once again the derivation can succeed along two paths: Either the full nominal domain (i.e. Q S P) or the most minimal phase (i.e. Poss PRO P) is moved to the specifier of TP. Let us further pursue the latter option here. Movement is once again taken to succeed through the specifier of Q S P but this time the quantifier does not bear an instance of the shared φ-feature and is hence able to probe again into its extended search space. Observe that in this second probing the highest SLI finds a match in the φ-value of the nominal below the phase head Poss PRO , hence the intervention effect is no longer observed. We might therein have found an indication as to the timing of deactivation of formal features from partial to total default (cf. Bejar 2003: 79) in the derivation of nominal concord in German. However, I have not been able to pin down the timing further given that [φ] on Poss PRO probes a second time in-situ when assigned non-structural case by V (contra EM of the nominal domain) and by movement through its extended search space in focused NP-constructions (contra targeting from a structural case-‘bearer’). I therefore have to leave this as an open question here. Observe moreover that I take the root node on the quantifier, derived by successful applications of Match and Value, to initiate a 2 nd cycle of Agree parallel to the root node brought about by mismatch and reduction. Since no shared feature is established at the root, however, modifications of feature structure do not carry over to the original intervener Poss PRO in these cases. Hence, whatever causes the feature structure of an SLI to constitute the root node [φ], the structure is undifferentiated for means of 2 nd cycle Agree in the further course of the derivation before deactivation to total default. Below I give the structure for the stranded quantifier construction on the basis of 0 b. as discussed above. Observe that once again case-assignment successfully applies to the stranded SLI due to the shared feature with the head noun in Spec,TP as well as T 0 . (146) [Poss PRO > N] i > T 0 > [Q S t i ] > v 0 [φ] [φ[m]] [φ[*m]] [φ[*m]] [*nom] [*nom] The result on the surface is as expected by now: We have seen above that in in-situ configurations both SLIs surface with default agreement (i.e. suffixless), while in floating constructions, the second probing of Q S due to movement from below the quantifier into its extended search space and the succeeding implementation of the SLI into the feature chain results in overt inflection on the surface. In both cases, however, the possessive pronoun fails to Value on the first cycle, thus only bearing a root node feature to be successively spelled out as default agreement. Both cases are exemplified with a neuter HN, incompatible with the selectional restrictions of the pronoun, below. The result on the surface is as expected by now: We have seen above that in in-situ configurations both SLIs surface with default agreement (i.e. suffixless), while in floating constructions, the second probing of Q S due to movement from below the quantifier into its extended search space and the succeeding implementation of the SLI into the feature chain results in overt inflection on the surface. In both cases, however, the possessive pronoun fails to Value on the first cycle, thus only bearing a root node feature to be successively spelled out as default agreement. Both cases are exemplified with a neuter HN, incompatible with the selectional restrictions of the pronoun, below. (147) a. Ich habe [all-ø mein-ø Wasser] getrunken. I have [ all my water ] drunk b. [Mein-ø Wasser] i habe ich all-es t i getrunken. [ my water ] i have I all t i drunk ‘I have drunk all my water.’ We have thus incorporated the precondition for quantifier floating constructions identified in chapter two, i.e. the non-phasal status of the stranded element and the phasal status of its complement (cf. chapter 2, endnote *14), into the derivational system elaborated here. Taking these observations to the structural set-up of split-topicalizations, discussed in chapter 2.3, leaves us with those cases in which the selectional restrictions of the weak quantifier are necessarily met, since it probes for [φ[f]], while only Num-heads specified for plural, i.e. 248 3 INFERENCE - Nominal Concord & Definiteness [φ[f[pl]]], have been taken to constitute phases below Q W . Therefore, we do not observe cases of 2 nd cycle Agree with movement from the lower phasal cycle of the nominal domain. Before closing this section, however, I want to point to one additional datum of 2 nd cycle Agree whose outcome supports the claims concerning the featural set-up and location of φ-structures in demonstratives made in the course of the analysis of this subchapter. Recall from the end of subchapter 3.2.2 that focus particles facilitate floating of the nominal domain they dominate. I exemplified this with the possessive pronoun to give a first hint at the workings of 2 nd cycle Agree and the effect of obligatory inflection on the surface above. In the course of the discussion on the extended pattern of SLI concatenations and the resulting inflectional idiosyncrasies, it was moreover concluded that demonstratives (as well as determiners) do not employ default agreement; hence, they cannot surface with an unvalued instantiation of the root node [φ]. I briefly discussed the seeming counterexample of suffixial inflection on demonstratives and argued that φ-features are coded on the stem rather than the suffix, contrary to the general rationale applied to the remaining SLIs. The main argument was based on the combinatorial restrictions of the apparently uninflected form dies . As I laid out above, default forms per definitionem exhibit less restrictiveness in their selectional constraints and therefore dies should readily combine with several number/ gender-specifications if it indeed constitutes one such form (note that all default forms identified above do indeed surface with more than one φ-value specification). As we have seen, however, dies rather mirrors the restriction of dies es and was hence regarded as the outcome of phonological reduction. An additional datum pointing to this conclusion is found in focused demonstrative DPs and extraction of neuter NPs from this configuration. (148) a. Ich habe [nur dies-ø Wasser] getrunken. I have [ only this water ] drunk b. Wasser i habe ich [nur dies-ø t i ] getrunken. water i have I [ only this t i ] drunk ‘I have only drunk this water.’ In (148), HN is said to move from below the DP through its specifier into the verbal domain. Parallel to the focused construction incorporating the possessive pronoun above, we expect a 2 nd cycle Agree effect to be triggered by this movement operation. However, only in this case, no change of morphological form arises on the SLI: It still surfaces without a suffix. I take this fact to indicate that the demonstrative is indeed specified for φ in the in-situ configuration (148) a. 3.3 Splitting Nominal Concord 249 and hence no further application of Match/ Value takes place by movement of the nominal out of its complement in line with the facts presented above. I have then to end this section in the revision of the varying implementations of the operation Agree. As I mentioned in chapter 3.1.2 and reviewed again in chapter 3.2.1, several implementations of Match as well as Value lend themselves to the derivation of nominal concord while further insight into the workings of Agree was expected to be gained from the full set of SLI-conjunctions in the nominal domain. I believe that we are now in a position to discuss the alternatives against the full body of data presented above. The first point to be revised here concerns the notion of non-distinctness employed in the standard formulation of Match in Chomsky (DbP, OP). As I argued above, non-distinctness, evaluated against the full structure of the goal (in a complex feature structure-account neglecting the feature-value bipartition), predicts that the initial application of probing in simple SLI-NP configurations overlooks incompatible, complex feature structures and keeps probing further. The outcome was nevertheless taken to be identical since no goal was found beyond the lexical domain and Match hence fails. The respective example is reprinted here. (149) (= (52)) Nominal Concord & Definiteness I NFERENCE 174 (149) (= (52)) viel > Wein Q W N i. [φ[f]]  [φ[m]] ø ii. [φ] [φ[m]] Match:  Value:  After reviewing more complex nominal set-ups, however, we can conclude here that this notion of Match circumvents the trigger for in-situ 2 nd cycle Agree in the nominal domain: Recall from the analysis of Poss PRO -Q W configurations above that the root node [φ] on the weak quantifier is targeted by Match of the more highly specified possessive pronoun. Under the standard formulation of Match evaluated w.r.t. the complete structure on probe and goal, however, the root node is taken to be overlooked and hence probing continues beyond the quantifier into its complement. Subsequently, Poss PRO should likewise fail to find a goal since N is equally non-distinct from the specification on the pronoun than it is from that on the quantifier. Consequently, both elements are predicted to surface with default agreement, contrary to fact: (150) * mein > viel > Wein Poss PRO Q W N i. [φ[f]]  [φ] [φ[m]] ø ii. [φ] [φ] [φ[m]] Match:  Value:  Therein, one of the prime examples for a derivational approach to nominal concord remains unaccounted for. This critique readily carries over to Match on the basis of Starke’s (2001) Anti-Identity Principle, as discussed in chapter 0, since it likewise applies to full feature specifications rather than roots. The predicted outcomes of Match under said principles, as discussed by Bejar (2003: 55f.), are reprinted here for ease of exposition, with 0 an instance of 0 a. (151) (= Bejar 2003: ch. 2, (42)) PROBE GOAL MATCH a. [π[ PART ]] [π]  b. [π[ PART ]] [π[ PART ]]  c. [π[ PART ]] [π[ PART ] ADRESSEE ]  Taking ‘Chomskyan’ Match, incorporating non-distinctness, however, to apply at the root, parallel to Bejar’s implementation, leaves us with another interesting observation concerning the datum in 0 above. Observe that therein, Match of the possessive pronoun and the quantifier succeeds because the root is non-distinct. The trigger for the in-situ 2 nd cycle probing of Q W is thereby established. Successively, the quantifier becomes distinctly specified for masculine/ neuter. Mismatch between the two SLIs under this notion hence occurs at Match once both elements are specified. As I have elaborated above, however, the outcome of the failure of Match and Value is identical, resulting in the reduction to partial default: the root node [φ]. Thus, once again the standard notion and Bejar’s version of Match arrive at the same outcome by different derivations, parallel to cases of simple SLI-N concatenations discussed in chapter 0 above. For the sake of coherence, I will maintain the implementation of Bejar’s notion of Match, evaluated at the root in a system dispensing with feature-value bipartitions, for the analysis elaborated here but note that it is compatible to a modification of the standard notion of Agree. This section reviewed complex nominal phrases in German employing more than one SLI heading the lexical domain (Num-)N. As has been observed therein, the derivation of morphological form of the items involved can be traced back to their selectional restrictions and their timing of EM as well as the application of probing operations. Therefore, the inflectional idiosyncrasies of nominal concord in German turn out the result of a highly derivational composition of the nominal domain. As will be shown in the remainder of this chapter, the featural set-up of the SLIs under consideration and their respective combinations and interactions have deep consequences for the interpretation of the complete domain. As I suggested in the introduction to this chapter, the findings of the previous sections will be shown to constitute the basis of morphological definiteness in German. The route I take in demonstrating this state of affairs takes a close look at the derivation of adjectival inflection as a related phenomenon, as elaborated in chapter one. As will be demonstrated therein, morphological definiteness reduces to the locus of φ-structure in the nominal domain. I will present my analysis in the following subchapter. After reviewing more complex nominal set-ups, however, we can conclude here that this notion of Match circumvents the trigger for in-situ 2 nd cycle Agree in the nominal domain: Recall from the analysis of Poss PRO -Q W configurations above that the root node [φ] on the weak quantifier is targeted by Match of the more highly specified possessive pronoun. Under the standard formulation of Match evaluated w.r.t. the complete structure on probe and goal, however, the root node is taken to be overlooked and hence probing continues beyond the quantifier into its complement. Subsequently, Poss PRO should likewise fail to find a goal since N is equally non-distinct from the specification on the pronoun than it is from that on the quantifier. Consequently, both elements are predicted to surface with default agreement, contrary to fact: 250 3 INFERENCE - Nominal Concord & Definiteness (150) Nominal Concord & Definiteness I NFERENCE 174 (149) (= 0(52)) viel > Wein Q W N i. [φ[f]]  [φ[m]] ø ii. [φ] [φ[m]] Match:  Value:  After reviewing more complex nominal set-ups, however, we can conclude here that this notion of Match circumvents the trigger for in-situ 2 nd cycle Agree in the nominal domain: Recall from the analysis of Poss PRO -Q W configurations above that the root node [φ] on the weak quantifier is targeted by Match of the more highly specified possessive pronoun. Under the standard formulation of Match evaluated w.r.t. the complete structure on probe and goal, however, the root node is taken to be overlooked and hence probing continues beyond the quantifier into its complement. Subsequently, Poss PRO should likewise fail to find a goal since N is equally non-distinct from the specification on the pronoun than it is from that on the quantifier. Consequently, both elements are predicted to surface with default agreement, contrary to fact: (150) * mein > viel > Wein Poss PRO Q W N i. [φ[f]]  [φ] [φ[m]] ø ii. [φ] [φ] [φ[m]] Match:  Value:  Therein, one of the prime examples for a derivational approach to nominal concord remains unaccounted for. This critique readily carries over to Match on the basis of Starke’s (2001) Anti-Identity Principle, as discussed in chapter 0, since it likewise applies to full feature specifications rather than roots. The predicted outcomes of Match under said principles, as discussed by Bejar (2003: 55f.), are reprinted here for ease of exposition, with 0 an instance of 0 a. (151) (= Bejar 2003: ch. 2, (42)) PROBE GOAL MATCH a. [π[ PART ]] [π]  b. [π[ PART ]] [π[ PART ]]  c. [π[ PART ]] [π[ PART ] ADRESSEE ]  Taking ‘Chomskyan’ Match, incorporating non-distinctness, however, to apply at the root, parallel to Bejar’s implementation, leaves us with another interesting observation concerning the datum in 0 above. Observe that therein, Match of the possessive pronoun and the quantifier succeeds because the root is non-distinct. The trigger for the in-situ 2 nd cycle probing of Q W is thereby established. Successively, the quantifier becomes distinctly specified for masculine/ neuter. Mismatch between the two SLIs under this notion hence occurs at Match once both elements are specified. As I have elaborated above, however, the outcome of the failure of Match and Value is identical, resulting in the reduction to partial default: the root node [φ]. Thus, once again the standard notion and Bejar’s version of Match arrive at the same outcome by different derivations, parallel to cases of simple SLI-N concatenations discussed in chapter 0 above. For the sake of coherence, I will maintain the implementation of Bejar’s notion of Match, evaluated at the root in a system dispensing with feature-value bipartitions, for the analysis elaborated here but note that it is compatible to a modification of the standard notion of Agree. This section reviewed complex nominal phrases in German employing more than one SLI heading the lexical domain (Num-)N. As has been observed therein, the derivation of morphological form of the items involved can be traced back to their selectional restrictions and their timing of EM as well as the application of probing operations. Therefore, the inflectional idiosyncrasies of nominal concord in German turn out the result of a highly derivational composition of the nominal domain. As will be shown in the remainder of this chapter, the featural set-up of the SLIs under consideration and their respective combinations and interactions have deep consequences for the interpretation of the complete domain. As I suggested in the introduction to this chapter, the findings of the previous sections will be shown to constitute the basis of morphological definiteness in German. The route I take in demonstrating this state of affairs takes a close look at the derivation of adjectival inflection as a related phenomenon, as elaborated in chapter one. As will be demonstrated therein, morphological definiteness reduces to the locus of φ-structure in the nominal domain. I will present my analysis in the following subchapter. Therein, one of the prime examples for a derivational approach to nominal concord remains unaccounted for. This critique readily carries over to Match on the basis of Starke’s (2001) Anti-Identity Principle, as discussed in chapter 3.2.1, since it likewise applies to full feature specifications rather than roots. The predicted outcomes of Match under said principles, as discussed by Bejar (2003: 55f.), are reprinted here for ease of exposition, with (150) an instance of (151) a. (151) (= Bejar 2003: ch. 2, (42)) probe goal match a. [π[ part ]] [π]  b. [π[ part ]] [π[ part ]]  c. [π[ part ]] [π[ part [ adressee ]]  Taking ‘Chomskyan’ Match, incorporating non-distinctness, however, to apply at the root, parallel to Bejar’s implementation, leaves us with another interesting observation concerning the datum in (150) above. Observe that therein, Match of the possessive pronoun and the quantifier succeeds because the root is non-distinct. The trigger for the in-situ 2 nd cycle probing of Q W is thereby established. Successively, the quantifier becomes distinctly specified for masculine/ neuter. Mismatch between the two SLIs under this notion hence occurs at Match once both elements are specified. As I have elaborated above, however, the outcome of the failure of Match and Value is identical, resulting in the reduction to partial default: the root node [φ]. Thus, once again the standard notion and Bejar’s version of Match arrive at the same outcome by different derivations, parallel to cases of simple SLI-N concatenations discussed in chapter 3.2.1 above. For the sake of coherence, I will maintain the implementation of Bejar’s notion of Match, evaluated at the root in a system dispensing with feature-value bipartitions, for the analysis elaborated here but note that it is compatible to a modification of the standard notion of Agree. This section reviewed complex nominal phrases in German employing more than one SLI heading the lexical domain (Num-)N. As has been observed therein, the derivation of morphological form of the items involved can be traced back to their selectional restrictions and their timing of EM as well as the applica- 3.3 Splitting Nominal Concord 251 tion of probing operations. Therefore, the inflectional idiosyncrasies of nominal concord in German turn out the result of a highly derivational composition of the nominal domain. As will be shown in the remainder of this chapter, the featural set-up of the SLIs under consideration and their respective combinations and interactions have deep consequences for the interpretation of the complete domain. As I suggested in the introduction to this chapter, the findings of the previous sections will be shown to constitute the basis of morphological definiteness in German. The route I take in demonstrating this state of affairs takes a close look at the derivation of adjectival inflection as a related phenomenon, as elaborated in chapter one. As will be demonstrated therein, morphological definiteness reduces to the locus of φ-structure in the nominal domain. I will present my analysis in the following subchapter. 3.3.4 Adjectival Inflection in Morphology As announced in the introduction to subchapter 3.2, the derivation of the morphological form of prenominal attributive adjectives will have to be ‘outsourced’ from the syntactic derivational paths sketched above and treated as a subsequent process in the post-syntactic morphological component Morphology (cf. Chomsky MP: 229, cf. also ch. 2.3.3 above). A first argument, as presented above on the basis of data from Fanselow (2013), was the unaccountability of morphological form of adjectives in featural terms, if - as is done here - a derivational approach is advocated, since the feature structure of the adjective would have to be taken to switch twice in focused floating constructions as first discussed at the end of chapter 3.2.2: To wit, at the stage of EM, the adjective bears strong inflection as is observable on the surface in nominal domains consisting solely of the Adj-N complex. By EM of D or (in certain cases) Poss PRO , the form would have been taken to change to the weak pattern as can be derived from the respective in-situ configurations. Finally, the form (and hence the feature) must have changed a third time, presumably back to the original value, when the Adj- N complex moves from beneath the SLI. As I have argued, this scenario seems unlikely under the Boolean, the feature-value as well as the complex feature coding advocated here. Likewise, Fanselow (2013) concludes that morphological form can be taken to be derived after movement. Sauerland (1996, see also Roehrs 2009) takes the phenomenon at hand to reinforce his argument for the Late Insertion account of inflectional forms in Germanic. He argues that the weak pattern of adjectival inflection in German as well as Dutch, Norwegian and Icelandic can be derived from the strong pattern via post-syntactic Impoverishment of the feature structure of the respective LI (cf. Halle 1997, Noyer [1992] 1997, Halle & Marantz 1993). This process applies 252 3 INFERENCE - Nominal Concord & Definiteness w.r.t. the inflection of the SLI dominating/ preceding the adjective. I will follow this basic rationale in my analysis below. 30 However, several adjustments are necessary, to which I turn now. Firstly, Sauerland (as well as Roehrs 2009, 2012) follows standard assumptions about overt inflection on SLIs triggering Impoverishment on the adjective (cf. p. 27, p. 33f., see also e.g. Sternefeld 2006: 226). While I do by and large adopt this line of reasoning, I believe to have demonstrated above that for the category D, the applicability of such a decompositional approach is called into question since inflectionless demonstratives can still be taken to encode φ and hence not to pose a partial default form in the sense of Bejar (2003). This was demonstrated by means of combinatorial restrictions as well as suffixless overt forms in floating constructions. For the other instantiation of D, the determiner, I argued that this SLI also does not exhibit default agreement, which coincides with its monomorphemic form. In what follows, I hence want to shift the view for the trigger for - what will be called - ‘Adjectival Impoverishment’ (AI) from inflection to the underlying concept of φ-structure on the SLI. Although the predictions made on the basis of these two viewpoints are parallel in most cases, we will see later (as has already been done above, cf. fn. 29) that an account based on featural content, as elaborated here, fares better with some of the inflectional idiosyncrasies of complex nominal domains in German. Second, Sauerland bases his rationale of impoverished LIs on Harley’s (1994) feature hierarchy given in (152) below. (152) (= Sauerland 1996: (10), [modified from Harley 1994: (19)]) Nominal Concord & Definiteness I NFERENCE 175 3.3.4 Adjectival Inflection in Morphology As announced in the introduction to subchapter 0, the derivation of the morphological form of prenominal attributive adjectives will have to be ‘outsourced’ from the syntactic derivational paths sketched above and treated as a subsequent process in the post-syntactic morphological component Morphology (cf. Chomsky MP: 229, cf. also ch. 2.3.3 above). A first argument, as presented above on the basis of data from Fanselow (2013), was the unaccountability of morphological form of adjectives in featural terms, if - as is done here - a derivational approach is advocated, since the feature structure of the adjective would have to be taken to switch twice in focused floating constructions as first discussed at the end of chapter 0: To wit, at the stage of EM, the adjective bears STRONG inflection as is observable on the surface in nominal domains consisting solely of the Adj-N complex. By EM of D or (in certain cases) Poss PRO , the form would have been taken to change to the WEAK pattern as can be derived from the respective in-situ configurations. Finally, the form (and hence the feature) must have changed a third time, presumably back to the original value, when the Adj-N complex moves from beneath the SLI. As I have argued, this scenario seems unlikely under the Boolean, the feature-value as well as the complex feature coding advocated here. Likewise, Fanselow (2013) concludes that morphological form can be taken to be derived after movement. Sauerland (1996, see also Roehrs 2009) takes the phenomenon at hand to reinforce his argument for the Late Insertion account of inflectional forms in Germanic. He argues that the WEAK pattern of adjectival inflection in German as well as Dutch, Norwegian and Icelandic can be derived from the STRONG pattern via post-syntactic Impoverishment of the feature structure of the respective LI (cf. Halle 1997, Noyer [1992] 1997, Halle & Marantz 1993). This process applies w.r.t. the inflection of the SLI dominating/ preceding the adjective. I will follow this basic rationale in my analysis below. 31 However, several adjustments are necessary, to which I turn now. Firstly, Sauerland (as well as Roehrs 2009, 2012) follows standard assumptions about overt inflection on SLIs triggering Impoverishment on the adjective (cf. p. 27, p. 33f., see also e.g. Sternefeld 2006: 226). While I do by and large adopt this line of reasoning, I believe to have demonstrated above that for the category D, the applicability of such a decompositional approach is called into question since inflectionless demonstratives can still be taken to encode φ and hence not to pose a partial default form in the sense of Bejar (2003). This was demonstrated by means of combinatorial restrictions as well as suffixless overt forms in floating constructions. For the other instantiation of D, the determiner, I argued that this SLI also does not exhibit default agreement, which coincides with its monomorphemic form. In what follows, I hence want to shift the view for the trigger for - what will be called - ‘Adjectival Impoverishment’ (AI) from inflection to the underlying concept of φ-structure on the SLI. Although the predictions made on the basis of these two viewpoints are parallel in most cases, we will see later (as has already been done above, cf. fn. 30) that an account based on featural content, as elaborated here, fares better with some of the inflectional idiosyncrasies of complex nominal domains in German. Second, Sauerland bases his rationale of impoverished LIs on Harley’s (1994) feature hierarchy given in 0 below. (152) (= Sauerland 1996: (10), [modified from Harley 1994: (19)]) Person 3 participant Number ! 3 addressee plural Gender ! ! 3 inclusive dual feminine neuter The capitalized labels in Harley’s feature set-up designate organizing nodes. Sauerland is hence able to propose the clean removal of Gender by Impoverishment (cf. p. 31, p. 35) while the feature structure dominating said organizing node, namely Number and Person, can be assumed to be still present in WEAK inflected adjectives in German. Turning to the modified feature hierarchy elaborated in chapter 2.3 above, the proposal of a genderless but numbered adjectival form due 31 Of course, the contrasting option to take the WEAK inflectional pattern as the unmarked vantage point has also been explored in the literature, cf. e.g. Olsen (1990), who takes φ-features to be expressed morphologically different (viz. the STRONG pattern) in the domain of determiners (including possessives) in her review of nominal structures in German under Abney’s (1987) DP-Hypothesis. The capitalized labels in Harley’s feature set-up designate organizing nodes. Sauerland is hence able to propose the clean removal of Gender by Impoverishment (cf. p. 31, p. 35) while the feature structure dominating said organizing 30 Of course, the contrasting option to take the weak inflectional pattern as the unmarked vantage point has also been explored in the literature, cf. e.g. Olsen (1990), who takes φ-features to be expressed morphologically different (viz. the strong pattern) in the domain of determiners (including possessives) in her review of nominal structures in German under Abney’s (1987) DP-Hypothesis. 3.3 Splitting Nominal Concord 253 node, namely Number and Person, can be assumed to be still present in weak inflected adjectives in German. Turning to the modified feature hierarchy elaborated in chapter 2.3 above, the proposal of a genderless but numbered adjectival form due to impoverishment rules seems harder to derive under the interwoven view of the complex number/ gender category. To solve this, let us take a closer look at the weak inflectional pattern in German: (153) weak masc neut fem pl nom der gut-e Wein das gut-e Bier die gut-e Milch die gut-en Häuser gen des gut-en Weins des gut-en Biers der gut-en Milch der gut-en Häuser dat dem gut-en Wein dem gut-en Bier der gut-en Milch den gut-en Häusern acc den gut-en Wein das gut-e Bier die gut-e Milch die gut-en Häuser the good wine masc the good beer neut the good milk fem the good houses pl Sauerland’s proposal is backed up by the observation that the weak inflectional paradigm consists of a subset of those forms that make up the strong paradigm. Moreover, this subclass constitutes the least marked instances of said paradigm, following Halle (1994). Sauerland’s lexical insertion rules for both / e/ as well as (unmarked) / n/ are given below (note that he glosses over the vowel of the consonant-final inflectional forms above in the equal treatment of adjectival and SLI inflection): (154) (= Sauerland 1996: (40), [reduced]) [+ direct] → / e/ [-] → / n/ The feature [+ direct] is taken from Harris’ (1994) case-hierarchy, elaborated on the basis of Romance languages. It can be fully translated to the notion of structural case (nom, acc), employed above, in contrast to non-structural cases (gen, dat), following Woolford (2006). Additionally, the Impoverishment rules in (155) apply to weak positions in German: (155) (= Sauerland 1996: (43)) Impoverishment rules applying only in weak positions delete [Gender] 254 3 INFERENCE - Nominal Concord & Definiteness delete [accusative] delete [oblique] delete [Case] in the environment of +plural Thus, / e/ is inserted into structural case forms in the singular while the elsewhere form / n/ consequently surfaces in non-structural cases and in all forms of the plural paradigm. Once again, the form of the accusative masculine conflicts with this neat partition in taking / n/ rather than / e/ like the other structural cases in the singular. I have presented tentative evidence for the idiosyncratic status of this inflectional form on SLIs above. Let us likewise adopt this rationale for the adjectival form; Sauerland is forced to employ an additional impoverishment rule, for adjectives as well as SLIs to adjust it to the insertion rule for / n/ stated above: (156) (= Sauerland 1996: (41), [reduced]) Impoverishment Rules Applying to Determiners and Adjective Inflection delete Agr with [masc,+direct,+accusative] Thereby, complete symmetry w.r.t. structural vs. non-structural case forms is achieved. Hence, Sauerland’s account is equally unable to directly account for this form in (impoverished) featural terms. I nevertheless feel justified to continue its treatment apart from φ-featural content in what follows (see fn. 33 below). The rules of lexical insertion and impoverishment stated above doubtlessly constitute a maximally economical way to code the apparent idiosyncratic distribution of inflectional forms in nominal concord. In fact, Sauerland bases his account on such considerations in the first place. However, it might be useful to visualize what the Impoverishment rules in (155) mean for the featural set-up of adjectives in weak positions: The first rule of deletion is evaluated w.r.t. Harley’s (1994) feature hierarchy, given above, in that the most embedded organizing node (with all its dependents) is elided from the structure. There is hence no gender distinction in the featural set-up of weak inflected adjectives. The two subsequent rules impoverish case distinctions to the (minimally specified, cf. the second rule) structural vs. elsewhere dichotomy. This bipartition is then additionally banished from the plural paradigm, which is therefore completely caseless (cf. the fourth rule). From the viewpoint of φ-features (and continuing to gloss over the category person), weak inflected adjectives are hence only specified for number. Under the theory elaborated here, this state of affairs seems dubious since number and gender are interwoven into a complex category. Consider again the feature hierarchy elaborated in chapter 2.3 and proposed in full at the beginning of this chapter: 3.3 Splitting Nominal Concord 255 (157) (= (10)) Nominal Concord & Definiteness I NFERENCE 177 (157) (= (10)) INDIVIDUATION 9 [f] [sg] class ! 3 [pl] [m] [n] Therein, the monovalent character of features forces e.g. the presence of [f] when [pl] is active. Therefore, the proposal of featural reduction of WEAK adjectival forms to number and case (as well as person) to the exclusion of gender appears problematic. However, this is not necessarily so. Note that all ‘classical’ gender values constitute sisters of the [sg]-node. Impoverishment can hence universally target either masculine or neuter by their shared dominator, the class-node, as well as [f]. With plural nominals, then, impoverishment of gender on adjectives is harder to achieve. One might - at first - tend to refer to the same strategy advocated in chapter 2.3.3 above for non-feminized simple plural nominals once again, namely enhancement (cf. Noyer [1992] 1997: 56), to reduce the output from the predictable feature [f]. However, I cannot see how this strategy would bring about a distinct WEAK pattern, contrasting with the fully specified STRONG one. A more straightforward option can be found in the proposal that delinking of the feature [f] results in the simultaneous deletion of [pl], hence of the full [f[pl]] complex, from WEAK inflected adjectives after the syntactic derivation, leaving only the root node and the specification for case acquired therein. The contrast in featural structure between WEAK inflected singular and plural adjectives is no more radical than Sauerland’s proposal of case-sensitive singulars vs. unmarked plurals, presented above. Nothing really hinges on this for the theory elaborated here and I will hence stay agnostic as to the correct approach in what follows. For illustrational purposes, however, I will follow the latter option and mark [f] together with its dependent [pl] as the target of impoverishment. Focusing on the singular paradigm next, observe that the exclusion of the conflicting morphological form / n/ of the masculine accusative from derivational processes on the basis of φ-featural content leaves us with a homogenous pattern of inflection for the WEAK forms in the singular, with all structural case forms surfacing as / e/ and non-structural cases as / n/ ; hence a paradigm only inflecting for (some distinction of) case: (158) WEAK sg pl nom / e/ / n/ gen / n/ / n/ dat / n/ / n/ acc / e/ / n/ We can now tentatively answer the question left open in chapter 0 above (cf. the discussion surrounding 0 above), wether uninflected SLIs do underlyingly bear case. There, I remained agnostic as to this matter, even though theory-internal considerations favored to deny it, since structural case was assigned on the basis of φ-completeness to members of feature chains which were established to the exclusion of said uninflected items. With the WEAK inflection identified as caseinflection in the absence of φ-completeness in the singular and as pure case-inflection in the plural, the assumptions of the theory elaborated above are strengthened: Zero-morphology does indeed signify the absence of both φ as well as case. Furthermore, contrasting this state of affairs with the STRONG paradigm, where gender - along with case - constitutes a classifying feature, leads us to an interesting generalization: Although one finds case-inflection without φ(-complete) inflection, one never exhibits inflection for φ without case. A parallel rationale was implemented in chapter 2.3 to initially argue for the relocation of the feature [f] as the node dominating [pl]. Following this line of reasoning once again, a hint might have been uncovered as to H&R’s original conjecture of the coding of case-features in their feature geometry: Case dominates φ-features and hence the INDIVIDUATION -node in 0 above (again glossing over the structural location of person). The discussion concerning parallel versus integrated coding of case in complex feature structures from chapter 0 thence might shift to either [ RE [case][φ[number/ gender]]] (with case coded above, but apart from, the φ-node) or [ RE [case[φ[number/ gender]]]] (with [φ] a dependent of case). I leave this matter for further research and employ the less restrictive former representation in what follows. Observe that this move, in turn, has further consequences for the proposed 2 nd cycle Agree operation in non-structural cases, also presented there (cf. the discussion surrounding 0-0 in chapter 0 above), in that in these cases the partially reduced SLI might indeed gain its φ-specification by the shared feature structure with HN instead of a second in-situ probing operation into its complement. As noted in the discussion of the feature sharing account in chapter 0, subsequent changes to one instance of a shared feature structure carry over to all Therein, the monovalent character of features forces e.g. the presence of [f] when [pl] is active. Therefore, the proposal of featural reduction of weak adjectival forms to number and case (as well as person) to the exclusion of gender appears problematic. However, this is not necessarily so. Note that all ‘classical’ gender values constitute sisters of the [sg]-node. Impoverishment can hence universally target either masculine or neuter by their shared dominator, the class-node, as well as [f]. With plural nominals, then, impoverishment of gender on adjectives is harder to achieve. One might - at first - tend to refer to the same strategy advocated in chapter 2.3.3 above for non-feminized simple plural nominals once again, namely enhancement (cf. Noyer [1992] 1997: 56), to reduce the output from the predictable feature [f]. However, I cannot see how this strategy would bring about a distinct weak pattern, contrasting with the fully specified strong one. A more straightforward option can be found in the proposal that delinking of the feature [f] results in the simultaneous deletion of [pl], hence of the full [f[pl]] complex, from weak inflected adjectives after the syntactic derivation, leaving only the root node and the specification for case acquired therein. The contrast in featural structure between weak inflected singular and plural adjectives is no more radical than Sauerland’s proposal of case-sensitive singulars vs. unmarked plurals, presented above. Nothing really hinges on this for the theory elaborated here and I will hence stay agnostic as to the correct approach in what follows. For illustrational purposes, however, I will follow the latter option and mark [f] together with its dependent [pl] as the target of impoverishment. Focusing on the singular paradigm next, observe that the exclusion of the conflicting morphological form / n/ of the masculine accusative from derivational processes on the basis of φ-featural content leaves us with a homogenous pattern of inflection for the weak forms in the singular, with all structural case forms surfacing as / e/ and non-structural cases as / n/ ; hence a paradigm only inflecting for (some distinction of) case: 256 3 INFERENCE - Nominal Concord & Definiteness (158) weak sg pl nom / e/ / n/ gen / n/ / n/ dat / n/ / n/ acc / e/ / n/ We can now tentatively answer the question left open in chapter 3.2.2 above (cf. the discussion surrounding (73) above), wether uninflected SLIs do underlyingly bear case. There, I remained agnostic as to this matter, even though theory-internal considerations favored to deny it, since structural case was assigned on the basis of φ-completeness to members of feature chains which were established to the exclusion of said uninflected items. With the weak inflection identified as case-inflection in the absence of φ-completeness in the singular and as pure case-inflection in the plural, the assumptions of the theory elaborated above are strengthened: Zero-morphology does indeed signify the absence of both φ as well as case. Furthermore, contrasting this state of affairs with the strong paradigm, where gender - along with case - constitutes a classifying feature, leads us to an interesting generalization: Although one finds case-inflection without φ(-complete) inflection, one never exhibits inflection for φ without case. A parallel rationale was implemented in chapter 2.3 to initially argue for the relocation of the feature [f] as the node dominating [pl]. Following this line of reasoning once again, a hint might have been uncovered as to H&R’s original conjecture of the coding of case-features in their feature geometry: Case dominates φ-features and hence the individuation -node in (157) above (again glossing over the structural location of person). The discussion concerning parallel versus integrated coding of case in complex feature structures from chapter 3.2.2 thence might shift to either [ re [case][φ[number/ gender]]] (with case coded above, but apart from, the φ-node) or [ re [case[φ[number/ gender]]]] (with [φ] a dependent of case). I leave this matter for further research and employ the less restrictive former representation in what follows. Observe that this move, in turn, has further consequences for the proposed 2 nd cycle Agree operation in non-structural cases, also presented there (cf. the discussion surrounding (78)-(81) in chapter 3.2.2 above), in that in these cases the partially reduced SLI might indeed gain its φ-specification by the shared feature structure with HN instead of a second in-situ probing operation into its complement. As noted in the discussion of the feature sharing account in chapter 3.1.2, subsequent changes to one instance of a shared feature structure carry over to all remaining ones. However, since the second proposed case of 2 nd cycle Agree, i.e. subsequent probing of the 3.3 Splitting Nominal Concord 257 weak quantifier, is not reducible in the same way (banishing in-situ 2 nd cycle Agree from the system under elaboration), I will refrain from an appropriate recategorization here. Returning to (158), then, I will not take the morphological uniformity of the paradigm to indicate caselessness since, as we have seen, zero-morphology should in principle be available there, too. I acknowledge that thereby my account does suffer from a less economical coding than Sauerland’s original proposals. However, I believe the discussion above will justify this move. Lastly, before turning to AI in nominal concord, I want to discuss briefly the categorical status of the LIs subject to this post-syntactic process. For Sauerland, there exist two sets of rules for Impoverishment, those applying to all (semi-) lexical elements in the nominal domain and those only applicable to the category of adjectives in certain syntactic environments, i.e. Adjectival Impoverishment. With the former, Sauerland is able to derive e.g. the uniform omission of the gender distinction in the plural paradigm. The motivation for this approach is found in the nearly homonymous patterns of strong adjectival inflection with the morphological form of SLIs, mentioned above (cf. Sauerland 1996: 30). In the introduction of Sauerland’s proposals, I adopted the rationale that featural content is deducible from the overt form of inflection; however, the general derivation of inflection on both SLIs and adjectives as elaborated in the preceding subchapters has been solely located in Narrow Syntax. Observe now that these latter rules of AI are not necessarily bound to apply on the basis of categorical status but that their evaluation might likewise be located in C HL : While Sauerland focuses on adjectives, I want to argue that a subset of semi-lexical elements might also be subject to the second class of rules of Impoverishment under certain circumstances. Thus, this morphological process might have less to do with categorical classification than structural position (i.e. locus of EM). I will elaborate this claim in the course of the ensuing analysis. For now, let me introduce the property to undergo AI as one of the characteristics classifying SLIs along with their selectional restrictions and the ability to surface with default agreement, elaborated in the preceding subchapters: (159) Q S D Poss PRO Q W Adj Probing φ [φ] [φ] [φ[f]] [φ[f]] [φ] Default Agreement      Adjectival Impoverishment ? ? ? ?  The discussion above already allows us to include adjectives together with their full specification for all listed properties into the table: They agree with N/ Num 258 3 INFERENCE - Nominal Concord & Definiteness in all number/ gender-specifications and can hence be taken to probe for the root node [φ] in Narrow Syntax. They never surface with default, i.e. zero-, morphology and are moreover trivially subject to AI. The rest of this subchapter will be devoted to carve out the property of AI for the SLIs listed in (159) on the one hand, as well as formalizing the impact of these SLIs on AI themselves on the other. I will once again begin my analysis with the discussion of adjective-noun complexes headed by a single SLI and utilize the findings made therein to deduce the derivational patterns of more complex multi-SLI configurations. To this end, I start out by discussing co-occurrences with the category D since the properties of this SLI are the most straightforward ones in the classification above: D probes for the root node [φ] and is hence trivially compatible with all instantiations of Num/ N, moreover, it never surfaces as a reduced form. a. Co-Occurrence with D Let me proceed by discussing the effect of D on elements trivially undergoing AI, namely adjectives, before turning to the discussion of the influence of AI on the SLI itself. To this end, note first how the system under elaboration here derives strong inflected (i.e. non-impoverished) adjective-noun concatenations, i.e. linearizations not preceded by one of the SLIs in (159) above, given that I characterized these LIs as probing for the root [φ] themselves: (160) Nominal Concord & Definiteness I NFERENCE 179 (160) Adj > Num/ N [φ[ *val ]] [φ[ val ]] [ case ] [ case ] Since no further operations apply after TRANSFER, the adjective in 0 surfaces with its full featural specification, i.e. the STRONG inflectional suffix. By now, the reader should be able to predict the featural set-up of the completed nominal derivation from the probing properties of the (S)LIs involved. With the determiner/ demonstrative dominating the adjective, both of which probe for the root node, the derivation on the basis of all available feature configurations of Num and N arrives at 0 once the isolated erection of the nominal domain is completed: (161) D > Adj > Num/ N [φ[* val ]] [φ[ *val ]] [φ[ val ]] There is no intervention effect predicted to occur in any concatenation of D and Adj. Subsequently, the nominal domain is externally merged into the clausal spine and uniformly assigned case. Consequently, the feature structure of the adjective is impoverished at Morphology after TRANSFER of the phase the nominal domain is part of. Until now I have refrained from overtly displaying the value singular in the feature structure of SLIs. Since [sg] is a sister to both [f] as well as the class-node (which has been omitted below for ease of exposition), all three dependent on the root, its location in the notion of features elaborated above is hence coded as [φ[[sg][m/ n]]] and [φ[[sg][f]]] for masculine/ neuter and feminine respectively. Additionally, following the observations made in the last section, case will be located above the root node, hence we arrive at [[case][φ[[sg][m/ n]]]] and [[case][φ[[sg][f]]]]. Following the proposals by Sauerland (1996), presented above, all of these configurations impoverish by deletion of the sister node of [sg] to [[case][φ[sg]]], viz. the featural content of WEAK inflection. These stages are exemplified below. Due to the complexity of the feature structures, I will switch to indicate the acquisition of additional featural content by placing the asterisk outside of the most minimal bracket of the new feature (complex) in what follows (cf. fn. 9 above). In addition, boldface is utilized in the schematics below to highlight the incorporation of a number/ gender-value in the shared feature configuration between SLI and adjective, to be subsequently targeted by deletion on the latter indicated by double strikethrough: (162) Narrow Syntax: a. D > Adj > N [φ*[[sg][m]]] [φ*[[sg][m]]] [φ[[sg][m]]] b. [D > Adj > N] i > T 0 > t i [*[nom][φ*[[sg][m]]]] [*[nom][φ*[[sg][m]]]] [*[nom][φ[[sg][m]]]] [φ*[[sg][m]]] Morphology: c. dies-er gut-e Wein D > Adj > N [*[nom][φ*[[sg][m]]]] [*[nom][φ*[[sg][m]]]] [*[nom][φ[[sg][m]]]] As has been elaborated above, Impoverishment is triggered by precedence/ dominance rather than partaking in the establishment of φ-chains, in line with Sauerland (1996). Therefore, moving the adjective from below the demonstrative should result in STRONG adjective inflection on the surface. This prediction is borne out as can be derived from optional Adj-N and N topicalizations in focused DP-constructions. The effect is once more exemplified on the basis of a neuter HN to illustrate the switch in inflectional pattern in object position: (163)a. Ich mag [nur dies-es gut-e Bier]. I like [ only this good WEAK beer ] b. [Gut-es Bier] i mag ich [nur dies-es t i ]. [ good STRONG beer ] i like I [ only this t i ] c. Bier i mag ich [nur dies-es gut-e t i ]. beer i like I [ only this good WEAK t i ] ‘I only like this good beer.’ Since no further operations apply after TRANSFER, the adjective in (160) surfaces with its full featural specification, i.e. the strong inflectional suffix. By now, the reader should be able to predict the featural set-up of the completed nominal derivation from the probing properties of the (S)LIs involved. With the determiner/ demonstrative dominating the adjective, both of which probe for the root node, the derivation on the basis of all available feature configurations of Num and N arrives at (161) once the isolated erection of the nominal domain is completed: (161) Nominal Concord & Definiteness I NFERENCE 179 (160) Adj > Num/ N [φ[ *val ]] [φ[ val ]] [ case ] [ case ] Since no further operations apply after TRANSFER, the adjective in 0 surfaces with its full featural specification, i.e. the STRONG inflectional suffix. By now, the reader should be able to predict the featural set-up of the completed nominal derivation from the probing properties of the (S)LIs involved. With the determiner/ demonstrative dominating the adjective, both of which probe for the root node, the derivation on the basis of all available feature configurations of Num and N arrives at 0 once the isolated erection of the nominal domain is completed: (161) D > Adj > Num/ N [φ[* val ]] [φ[ *val ]] [φ[ val ]] There is no intervention effect predicted to occur in any concatenation of D and Adj. Subsequently, the nominal domain is externally merged into the clausal spine and uniformly assigned case. Consequently, the feature structure of the adjective is impoverished at Morphology after TRANSFER of the phase the nominal domain is part of. Until now I have refrained from overtly displaying the value singular in the feature structure of SLIs. Since [sg] is a sister to both [f] as well as the class-node (which has been omitted below for ease of exposition), all three dependent on the root, its location in the notion of features elaborated above is hence coded as [φ[[sg][m/ n]]] and [φ[[sg][f]]] for masculine/ neuter and feminine respectively. Additionally, following the observations made in the last section, case will be located above the root node, hence we arrive at [[case][φ[[sg][m/ n]]]] and [[case][φ[[sg][f]]]]. Following the proposals by Sauerland (1996), presented above, all of these configurations impoverish by deletion of the sister node of [sg] to [[case][φ[sg]]], viz. the featural content of WEAK inflection. These stages are exemplified below. Due to the complexity of the feature structures, I will switch to indicate the acquisition of additional featural content by placing the asterisk outside of the most minimal bracket of the new feature (complex) in what follows (cf. fn. 9 above). In addition, boldface is utilized in the schematics below to highlight the incorporation of a number/ gender-value in the shared feature configuration between SLI and adjective, to be subsequently targeted by deletion on the latter indicated by double strikethrough: (162) Narrow Syntax: a. D > Adj > N [φ*[[sg][m]]] [φ*[[sg][m]]] [φ[[sg][m]]] b. [D > Adj > N] i > T 0 > t i [*[nom][φ*[[sg][m]]]] [*[nom][φ*[[sg][m]]]] [*[nom][φ[[sg][m]]]] [φ*[[sg][m]]] Morphology: c. dies-er gut-e Wein D > Adj > N [*[nom][φ*[[sg][m]]]] [*[nom][φ*[[sg][m]]]] [*[nom][φ[[sg][m]]]] As has been elaborated above, Impoverishment is triggered by precedence/ dominance rather than partaking in the establishment of φ-chains, in line with Sauerland (1996). Therefore, moving the adjective from below the demonstrative should result in STRONG adjective inflection on the surface. This prediction is borne out as can be derived from optional Adj-N and N topicalizations in focused DP-constructions. The effect is once more exemplified on the basis of a neuter HN to illustrate the switch in inflectional pattern in object position: (163)a. Ich mag [nur dies-es gut-e Bier]. I like [ only this good WEAK beer ] b. [Gut-es Bier] i mag ich [nur dies-es t i ]. [ good STRONG beer ] i like I [ only this t i ] c. Bier i mag ich [nur dies-es gut-e t i ]. beer i like I [ only this good WEAK t i ] ‘I only like this good beer.’ There is no intervention effect predicted to occur in any concatenation of D and Adj. Subsequently, the nominal domain is externally merged into the clausal 3.3 Splitting Nominal Concord 259 spine and uniformly assigned case. Consequently, the feature structure of the adjective is impoverished at Morphology after TRANSFER of the phase the nominal domain is part of. Until now I have refrained from overtly displaying the value singular in the feature structure of SLIs. Since [sg] is a sister to both [f] as well as the class-node (which has been omitted below for ease of exposition), all three dependent on the root, its location in the notion of features elaborated above is hence coded as [φ[[sg][m/ n]]] and [φ[[sg][f]]] for masculine/ neuter and feminine respectively. Additionally, following the observations made in the last section, case will be located above the root node, hence we arrive at [[case][φ[[sg][m/ n]]]] and [[case][φ[[sg][f]]]]. Following the proposals by Sauerland (1996), presented above, all of these configurations impoverish by deletion of the sister node of [sg] to [[case][φ[sg]]], viz. the featural content of weak inflection. These stages are exemplified below. Due to the complexity of the feature structures, I will switch to indicate the acquisition of additional featural content by placing the asterisk outside of the most minimal bracket of the new feature (complex) in what follows (cf. fn. 9 above). In addition, boldface is utilized in the schematics below to highlight the incorporation of a number/ gender-value in the shared feature configuration between SLI and adjective, to be subsequently targeted by deletion on the latter indicated by double strikethrough: (162) Nominal Concord & Definiteness I NFERENCE 179 (160) Adj > Num/ N [φ[ *val ]] [φ[ val ]] [ case ] [ case ] Since no further operations apply after TRANSFER, the adjective in 0 surfaces with its full featural specification, i.e. the STRONG inflectional suffix. By now, the reader should be able to predict the featural set-up of the completed nominal derivation from the probing properties of the (S)LIs involved. With the determiner/ demonstrative dominating the adjective, both of which probe for the root node, the derivation on the basis of all available feature configurations of Num and N arrives at 0 once the isolated erection of the nominal domain is completed: (161) D > Adj > Num/ N [φ[* val ]] [φ[ *val ]] [φ[ val ]] There is no intervention effect predicted to occur in any concatenation of D and Adj. Subsequently, the nominal domain is externally merged into the clausal spine and uniformly assigned case. Consequently, the feature structure of the adjective is impoverished at Morphology after TRANSFER of the phase the nominal domain is part of. Until now I have refrained from overtly displaying the value singular in the feature structure of SLIs. Since [sg] is a sister to both [f] as well as the class-node (which has been omitted below for ease of exposition), all three dependent on the root, its location in the notion of features elaborated above is hence coded as [φ[[sg][m/ n]]] and [φ[[sg][f]]] for masculine/ neuter and feminine respectively. Additionally, following the observations made in the last section, case will be located above the root node, hence we arrive at [[case][φ[[sg][m/ n]]]] and [[case][φ[[sg][f]]]]. Following the proposals by Sauerland (1996), presented above, all of these configurations impoverish by deletion of the sister node of [sg] to [[case][φ[sg]]], viz. the featural content of WEAK inflection. These stages are exemplified below. Due to the complexity of the feature structures, I will switch to indicate the acquisition of additional featural content by placing the asterisk outside of the most minimal bracket of the new feature (complex) in what follows (cf. fn. 9 above). In addition, boldface is utilized in the schematics below to highlight the incorporation of a number/ gender-value in the shared feature configuration between SLI and adjective, to be subsequently targeted by deletion on the latter indicated by double strikethrough: (162) Narrow Syntax: a. D > Adj > N [φ*[[sg][m]]] [φ*[[sg][m]]] [φ[[sg][m]]] b. [D > Adj > N] i > T 0 > t i [*[nom][φ*[[sg][m]]]] [*[nom][φ*[[sg][m]]]] [*[nom][φ[[sg][m]]]] [φ*[[sg][m]]] Morphology: c. dies-er gut-e Wein D > Adj > N [*[nom][φ*[[sg][m]]]] [*[nom][φ*[[sg][m]]]] [*[nom][φ[[sg][m]]]] As has been elaborated above, Impoverishment is triggered by precedence/ dominance rather than partaking in the establishment of φ-chains, in line with Sauerland (1996). Therefore, moving the adjective from below the demonstrative should result in STRONG adjective inflection on the surface. This prediction is borne out as can be derived from optional Adj-N and N topicalizations in focused DP-constructions. The effect is once more exemplified on the basis of a neuter HN to illustrate the switch in inflectional pattern in object position: (163)a. Ich mag [nur dies-es gut-e Bier]. I like [ only this good WEAK beer ] b. [Gut-es Bier] i mag ich [nur dies-es t i ]. [ good STRONG beer ] i like I [ only this t i ] c. Bier i mag ich [nur dies-es gut-e t i ]. beer i like I [ only this good WEAK t i ] ‘I only like this good beer.’ As has been elaborated above, Impoverishment is triggered by precedence/ dominance rather than partaking in the establishment of φ-chains, in line with Sauerland (1996). Therefore, moving the adjective from below the demonstrative should result in strong adjective inflection on the surface. This prediction is borne out as can be derived from optional Adj-N and N topicalizations in 260 3 INFERENCE - Nominal Concord & Definiteness focused DP-constructions. The effect is once more exemplified on the basis of a neuter HN to illustrate the switch in inflectional pattern in object position: (163) a. Ich mag [nur dies-es gut-e Bier]. I like [ only this good weak beer ] b. [Gut-es Bier] i mag ich [nur dies-es t i ]. [ good strong beer ] i like I [ only this t i ] c. Bier i mag ich [nur dies-es gut-e t i ]. beer i like I [ only this good weak t i ] ‘I only like this good beer.’ We hence observe that adjectives vary in their inflectional pattern concerning their position on the surface w.r.t. the nominal phase head. This is expected given that both elements probe for the root node and hence establish a uniform feature chain across the nominal domain in all concatenations with N/ Num. Before turning to the discussion of the remaining SLIs, I shortly want to draw the attention to one of the configurations discussed in chapter 3.2.3 above, namely the co-occurrence of elements of the category D with the strong quantifier Q S . The latter constitutes the only contextual phase head in the nominal domain that is externally merged above the solely rigid one, the determiner/ demonstrative DP. This SLI, too, was classified as probing for the root node [φ] in the discussion of simple SLI-N concatenations in chapter 3.2.1 but identified as being subject to morpho-phonological (i.e. dittological) restrictions in its concatenation with instantiations of category DP, as discussed in chapter 3.2.3. However, if the applicability of AI to DP is to be tested, this is the sole linearization in which such an effect on D itself would emerge. 31 As we have seen above, however, no change in morphological form of either the monomorphemic determiner or the (somehow) inflected demonstrative is evoked by precedence/ dominance of the quantifier. Therefore, we can conclude that AI has no effect on the DP itself and update the table of properties of SLIs accordingly: (164) Q S D Poss PRO Q W Adj Probing φ [φ] [φ] [φ[f]] [φ[f]] [φ] Default Agreement      Adjectival Impoverishment ?  ? ?  31 It has to be noted that this argument tacitly relies on Q S triggering AI itself, a state of affairs that should be obvious from the preceding subchapters but will only be made formal in the remainder of this chapter. 3.3 Splitting Nominal Concord 261 I will next turn to the second most prominent contextual phase head in the highest cycle of the nominal domain, the possessive pronoun Poss PRO . b. Co-Occurrence with Poss pro The possessive pronoun is the first element under consideration in this subchapter not probing for the root node but a subset of the possible specifications of [φ] upon EM, namely [φ[f]]. Therefore, we do expect to find disruptions in the establishment of feature chains in the concatenation with the φ-uniform Adj-N complex. As has been discussed in detail in chapter 3.2.2, these disruptions concern those linearizations in which case is assigned on the basis of φ-completeness of the nominal domain with either masculine or neuter HNs, since case ‘comes free’ and therein initiates a 2 nd cycle of φ-probing in nonstructural cases on the one hand, while Poss PRO constitutes an instance of the shared feature structure with both feminine and plural HNs on the other. If - as I have argued above - the selection of adjectival inflection is evaluated on the basis of φ-completeness of the SLI dominating it, we thence expect the inflection to vary within the paradigms of non-feminine genders between structural and non-structural cases. This prediction is borne out. The featural configurations are schematized in (165): (165) Nominal Concord & Definiteness I NFERENCE 181 (165) a. feminine gender, structural case: WEAK adjective inflection mein-e gut-e Milch Poss PRO > Adj > N um [*[nom][φ[*[sg][f]]]] [*[nom][φ*[[sg][f]]]] [*[nom][φ[[sg][f]]]] b. feminine gender, non-structural case: WEAK adjective inflection mein-er gut-en Milch Poss PRO > Adj > N um [*[dat][φ[*[sg][f]]]] [*[dat][φ*[[sg][f]]]] [*[dat][φ[[sg][f]]]] c. non-feminine gender, structural case: STRONG adjective inflection 32 mein-ø gut-es Bier Poss PRO > Adj > N [φ] [*[nom][φ*[[sg][n]]]] [*[nom][φ[[sg][n]]]] d. non-feminine gender, non-structural case: WEAK adjective inflection mein-em gut-en Bier Poss PRO > Adj > N [*[dat][φ*[[sg][n]]]] [*[dat][φ*[[sg][n]]]] [*[dat][φ[[sg][n]]]] Expanding this fourfold distinction to a full paradigm of Poss PRO -Adj-N concatenations leaves us with exactly those forms that constitute the MIXED adjective pattern in German. We have therefore found a featural basis for the proposed disambiguation of homonymous forms in these paradigms as discussed in chapter 2.3.3. Moreover, we therein also encountered strong evidence to finally disambiguate the last homonymous adjectival forms in genitive of both nonfeminine gender patterns to WEAK , a question left open in the aforementioned chapter above. The appropriate data as well as the proposed underlying featural configurations are given in 0. 33 (166)a. masc nom mein-ø gut-er Wein [φ] [*[nom][φ*[[sg][m]]]] [*[nom][φ[[sg][m]]]] gen mein-es gut-en Weins [*[gen][φ*[[sg][m]]]] [*[gen][φ*[[sg][m]]]] [*[gen][φ[[sg][m]]]] dat mein-em gut-en Wein [*[dat][φ*[[sg][m]]]] [*[dat][φ*[[sg][m]]]] [*[dat][φ[[sg][m]]]] acc mein-en gut-en Wein [φ] [*[acc][φ*[[sg][m]]]] [*[acc][φ[[sg][m]]]] b. neut 32 Returning to the cases of homonymy between inflection on D in nominative masculine with stem-final -er of the possessive with first and second person plural possessor agreement, reported in fn. 27 above, observe that - at least for some speakers - the latter are acceptable with adjectives carrying STRONG as well as WEAK inflection in line with the reanalysis proposed above. 33 Concentrating on the morphological form of the SLI in accusative masculine singular for the last time, a hint as to the underlying featural content might be found in the homonymy with inflection on WEAK plural adjectives since the latter constitutes only case apart from the root node [φ], after impoverishment, while the former has been identified as a root node which surfaces only in a specific case-position. Both inflectional instances might then be assumed to constitute the exact same featural set-up: [[acc][φ]]. Thus, the conflicting form might be traced back to its idiosyncratic featural content after all. Observe, however, that its occurrence is still gender-specific and hence runs counter to the proposals for the narrow-syntactic treatment, advocated in the main text, which takes the form to only bear the root node throughout the syntactic derivation. 32 Returning to the cases of homonymy between inflection on D in nominative masculine with stem-final er of the possessive with first and second person plural possessor agreement, reported in endnote *16, observe that - at least for some speakers - the latter are 262 3 INFERENCE - Nominal Concord & Definiteness Expanding this fourfold distinction to a full paradigm of Poss PRO -Adj-N concatenations leaves us with exactly those forms that constitute the mixed adjective pattern in German. We have therefore found a featural basis for the proposed disambiguation of homonymous forms in these paradigms as discussed in chapter 2.3.3. Moreover, we therein also encountered strong evidence to finally disambiguate the last homonymous adjectival forms in genitive of both nonfeminine gender patterns to weak , a question left open in the aforementioned chapter above. The appropriate data as well as the proposed underlying featural configurations are given in (166). 33 (166) a. masc nom mein-ø gut-er Wein [φ] [*[nom][φ*[[sg][m]]]] [*[nom][φ[[sg][m]]]] gen mein-es gut-en Weins [*[gen][φ*[[sg][m]]]] [*[gen][φ*[[sg][m]]]] [*[gen][φ[[sg][m]]]] dat mein-em gut-en Wein [*[dat][φ*[[sg][m]]]] [*[dat][φ*[[sg][m]]]] [*[dat][φ[[sg][m]]]] acc mein-en gut-en Wein [φ] [*[acc][φ*[[sg][m]]]] [*[acc][φ[[sg][m]]]] b. neut nom mein-ø gut-es Bier [φ] [*[nom][φ*[[sg][n]]]] [*[nom][φ[[sg][n]]]] gen mein-es gut-en Biers [*[gen][φ*[[sg][n]]]] [*[gen][φ*[[sg][n]]]] [*[gen][φ[[sg][n]]]] acceptable with adjectives carrying strong as well as weak inflection in line with the reanalysis proposed above. 33 Concentrating on the morphological form of the SLI in accusative masculine singular for the last time, a hint as to the underlying featural content might be found in the homonymy with inflection on weak plural adjectives since the latter constitutes only case apart from the root node [φ], after impoverishment, while the former has been identified as a root node which surfaces only in a specific case-position. Both inflectional instances might then be assumed to constitute the exact same featural set-up: [[acc][φ]]. Thus, the conflicting form might be traced back to its idiosyncratic featural content after all. Observe, however, that its occurrence is still gender-specific and hence runs counter to the proposals for the narrow-syntactic treatment, advocated in the main text, which takes the form to only bear the root node throughout the syntactic derivation. 3.3 Splitting Nominal Concord 263 dat mein-em gut-en Bier [*[dat][φ*[[sg][n]]]] [*[dat][φ*[[sg][n]]]] [*[dat][φ[[sg][n]]]] acc mein-ø gut-es Bier [φ] [*[acc][φ*[[sg][n]]]] [*[acc][φ[[sg][n]]]] c. fem nom mein-e gut-e Milch [*[nom][φ[*[sg][f]]]] [*[nom][φ*[[sg][f]]]] [*[nom][φ[[sg][f]]]] gen mein-er gut-en Milch [*[gen][φ[*[sg][f]]]] [*[gen][φ*[[sg][f]]]] [*[gen][φ[[sg][f]]]] dat mein-er gut-en Milch [*[dat][φ[*[sg][f]]]] [*[dat][φ*[[sg][f]]]] [*[dat][φ[[sg][f]]]] acc mein-e gut-e Milch [*[acc][φ[*[sg][f]]]] [*[acc][φ*[[sg][f]]]] [*[acc][φ[[sg][f]]]] d. pl nom mein-e gut-en Häuser [*[nom][φ[[f*[pl]]]]] [*[nom][φ*[[f[pl]]]]] [*[nom][φ[[f[pl]]]]] gen mein-er gut-en Häuser [*[gen][φ[[f*[pl]]]]] [*[gen][φ*[[f[pl]]]]] [*[gen][φ[[f[pl]]]]] dat mein-en gut-en Häusern [*[dat][φ[[f*[pl]]]]] [*[dat][φ*[[f[pl]]]]] [*[dat][φ[[f[pl]]]]] acc mein-e gut-en Häuser [*[acc][φ[[f*[pl]]]]] [*[acc][φ*[[f[pl]]]]] [*[acc][φ[[f[pl]]]]] As can be deduced from (166), the theory elaborated here fully captures the mixed adjective pattern in German. Uniformity of weak adjectival inflection coincides with the successful probing of the conjoined SLI. Therefore, the pattern of mixed adjectival inflection surfaces when probing is disrupted by means of a selectional mismatch. Moreover, the analysis makes additional claims concerning (focused) floating constructions: Recall that movement from below the partially default possessive pronoun enforces a 2 nd cycle of Agree on said SLI while the proposals from the beginning of this subchapter predict the adjective in a floated Adj-N complex to pattern with the strong inflection since it is not dominated/ preceded by an SLI with appropriate feature-content. Combining 264 3 INFERENCE - Nominal Concord & Definiteness these observations hence arrives at a linearly disrupted but featurally uniform nominal domain. This prediction is once again borne out. (167) Nominal Concord & Definiteness I NFERENCE 182 nom mein-ø gut-es Bier [φ] [*[nom][φ*[[sg][n]]]] [*[nom][φ[[sg][n]]]] gen mein-es gut-en Biers [*[gen][φ*[[sg][n]]]] [*[gen][φ*[[sg][n]]]] [*[gen][φ[[sg][n]]]] dat mein-em gut-en Bier [*[dat][φ*[[sg][n]]]] [*[dat][φ*[[sg][n]]]] [*[dat][φ[[sg][n]]]] acc mein-ø gut-es Bier [φ] [*[acc][φ*[[sg][n]]]] [*[acc][φ[[sg][n]]]] c. fem nom mein-e gut-e Milch [*[nom][φ[*[sg][f]]]] [*[nom][φ*[[sg][f]]]] [*[nom][φ[[sg][f]]]] gen mein-er gut-en Milch [*[gen][φ[*[sg][f]]]] [*[gen][φ*[[sg][f]]]] [*[gen][φ[[sg][f]]]] dat mein-er gut-en Milch [*[dat][φ[*[sg][f]]]] [*[dat][φ*[[sg][f]]]] [*[dat][φ[[sg][f]]]] acc mein-e gut-e Milch [*[acc][φ[*[sg][f]]]] [*[acc][φ*[[sg][f]]]] [*[acc][φ[[sg][f]]]] d. pl nom mein-e gut-en Häuser [*[nom][φ[[f*[pl]]]]] [*[nom][φ*[[f[pl]]]]] [*[nom][φ[[f[pl]]]]] gen mein-er gut-en Häuser [*[gen][φ[[f*[pl]]]]] [*[gen][φ*[[f[pl]]]]] [*[gen][φ[[f[pl]]]]] dat mein-en gut-en Häusern [*[dat][φ[[f*[pl]]]]] [*[dat][φ*[[f[pl]]]]] [*[dat][φ[[f[pl]]]]] acc mein-e gut-en Häuser [*[acc][φ[[f*[pl]]]]] [*[acc][φ*[[f[pl]]]]] [*[acc][φ[[f[pl]]]]] As can be deduced from 0, the theory elaborated here fully captures the MIXED adjective pattern in German. Uniformity of WEAK adjectival inflection coincides with the successful probing of the conjoined SLI. Therefore, the pattern of MIXED adjectival inflection surfaces when probing is disrupted by means of a selectional mismatch. Moreover, the analysis makes additional claims concerning (focused) floating constructions: Recall that movement from below the partially default possessive pronoun enforces a 2 nd cycle of Agree on said SLI while the proposals from the beginning of this subchapter predict the adjective in a floated Adj-N complex to pattern with the STRONG inflection since it is not dominated/ preceded by an SLI with appropriate feature-content. Combining these observations hence arrives at a linearly disrupted but featurally uniform nominal domain. This prediction is once again borne out. (167) a. Ich habe [nur mein-ø kühl-es Bier] getrunken. I have [ only my cold STRONG beer ] drunk Poss PRO Adj N neut [φ] [*[acc][φ*[[sg][n]]]] [*[acc][φ[[sg][n]]]] b. [Kühl-es Bier] i habe ich [nur mein-es t i ] getrunken. [ cold STRONG beer ] i have I [ only mine t i ] drunk Adj N neut Poss PRO [*[acc][φ*[[sg][n]]]] [*[acc][φ[[sg][n]]]] [*[acc][φ*[[sg][n]]]] ‘I only drank my cold beer.’ One additional observation that has to be mentioned in this context is the absence of 2 nd cycle Agree effects when a residue of the nominal domain below the SLI remains in-situ: (168) Bier i habe ich [nur mein-ø gekühlt-es t i ] getrunken. beer i have I [ only my cooled STRONG t i ] drunk N Poss PRO Adj [*[acc][φ[[sg][n]]]] [φ] [*[acc][φ*[[sg][n]]]] One additional observation that has to be mentioned in this context is the absence of 2 nd cycle Agree effects when a residue of the nominal domain below the SLI remains in-situ: (168) Nominal Concord & Definiteness I NFERENCE 182 nom mein-ø gut-es Bier [φ] [*[nom][φ*[[sg][n]]]] [*[nom][φ[[sg][n]]]] gen mein-es gut-en Biers [*[gen][φ*[[sg][n]]]] [*[gen][φ*[[sg][n]]]] [*[gen][φ[[sg][n]]]] dat mein-em gut-en Bier [*[dat][φ*[[sg][n]]]] [*[dat][φ*[[sg][n]]]] [*[dat][φ[[sg][n]]]] acc mein-ø gut-es Bier [φ] [*[acc][φ*[[sg][n]]]] [*[acc][φ[[sg][n]]]] c. fem nom mein-e gut-e Milch [*[nom][φ[*[sg][f]]]] [*[nom][φ*[[sg][f]]]] [*[nom][φ[[sg][f]]]] gen mein-er gut-en Milch [*[gen][φ[*[sg][f]]]] [*[gen][φ*[[sg][f]]]] [*[gen][φ[[sg][f]]]] dat mein-er gut-en Milch [*[dat][φ[*[sg][f]]]] [*[dat][φ*[[sg][f]]]] [*[dat][φ[[sg][f]]]] acc mein-e gut-e Milch [*[acc][φ[*[sg][f]]]] [*[acc][φ*[[sg][f]]]] [*[acc][φ[[sg][f]]]] d. pl nom mein-e gut-en Häuser [*[nom][φ[[f*[pl]]]]] [*[nom][φ*[[f[pl]]]]] [*[nom][φ[[f[pl]]]]] gen mein-er gut-en Häuser [*[gen][φ[[f*[pl]]]]] [*[gen][φ*[[f[pl]]]]] [*[gen][φ[[f[pl]]]]] dat mein-en gut-en Häusern [*[dat][φ[[f*[pl]]]]] [*[dat][φ*[[f[pl]]]]] [*[dat][φ[[f[pl]]]]] acc mein-e gut-en Häuser [*[acc][φ[[f*[pl]]]]] [*[acc][φ*[[f[pl]]]]] [*[acc][φ[[f[pl]]]]] As can be deduced from 0, the theory elaborated here fully captures the MIXED adjective pattern in German. Uniformity of WEAK adjectival inflection coincides with the successful probing of the conjoined SLI. Therefore, the pattern of MIXED adjectival inflection surfaces when probing is disrupted by means of a selectional mismatch. Moreover, the analysis makes additional claims concerning (focused) floating constructions: Recall that movement from below the partially default possessive pronoun enforces a 2 nd cycle of Agree on said SLI while the proposals from the beginning of this subchapter predict the adjective in a floated Adj-N complex to pattern with the STRONG inflection since it is not dominated/ preceded by an SLI with appropriate feature-content. Combining these observations hence arrives at a linearly disrupted but featurally uniform nominal domain. This prediction is once again borne out. 167) a. Ich habe [nur mein-ø kühl-es Bier] getrunken. I have [ only my cold STRONG beer ] drunk Poss PRO Adj N neut [φ] [*[acc][φ*[[sg][n]]]] [*[acc][φ[[sg][n]]]] b. [Kühl-es Bier] i habe ich [nur mein-es t i ] getrunken. [ cold STRONG beer ] i have I [ only mine t i ] drunk Adj N neut Poss PRO [*[acc][φ*[[sg][n]]]] [*[acc][φ[[sg][n]]]] [*[acc][φ*[[sg][n]]]] ‘I only drank my cold beer.’ One additional observation that has to be mentioned in this context is the absence of 2 nd cycle Agree effects when a residue of the nominal domain below the SLI remains in-situ: (168) Bier i habe ich [nur mein-ø gekühlt-es t i ] getrunken. beer i have I [ only my cooled STRONG t i ] drunk N Poss PRO Adj [*[acc][φ[[sg][n]]]] [φ] [*[acc][φ*[[sg][n]]]] I will not speculate on the derivational processes underlying this construction here but argue in chapter 4.2.1 below on the basis of related data that variation in (the membership in) prolific domains influences the selection of intermediate landing sites of the moved nominal to avoid the specifier of the SLI in structures like (168) and thus bars probing of Poss PRO into its extended search space. To end this section, then, let me once again focus on possible effects of AI on the possessive pronoun itself. Both structurally higher SLIs in the nominal domain are part of the same context sensitive phasal cycle, namely D and Q S , the co-occurrence of each of which with Poss PRO has been discussed in detail in chapter 3.2.3 above. Therein, it was concluded that although the possessive does have an effect on the inflection of the higher SLIs, by means of intervention effects resulting in either default, i.e. zero-, morphology with Q S or ungrammaticality with D, none of these elements changes the morphological form of inflection on the possessive pronoun itself. Rather, the grammatical cases of D/ Q S -Poss PRO concatenation with overt inflection on all elements display the same inflectional patterns also found in the respective SLI-N configurations. Hence, 3.3 Splitting Nominal Concord 265 we can conclude that AI does not apply to the possessive pronoun and update the table of properties of SLIs in the nominal domain accordingly: (169) Q S D Poss PRO Q W Adj Probing φ [φ] [φ] [φ[f]] [φ[f]] [φ] Default Agreement      Adjectival Impoverishment ?   ?  By now, the two most prominent contextual phase heads of the highest cycle of the nominal domain have been tested for their influence on the morphological form of adjectives in their complement. As has been shown therein, adjectival inflection coincides with the featural content of the SLI that dominates it. Consequently, the mismatches occurring in probing for a subset of φ-feature configurations have consequences for the evaluation of surface structure of adjectives, namely the inapplicability of impoverishment of number/ gender-features in post-syntactic Morphology. This is the source of the so-called mixed adjectival inflection. In contrast, elements probing for the root node consequently cooccur with weak adjectival inflection in all φ/ case-configurations since they are trivially specified for Match and Value to succeed in any case. Moreover, we have observed that SLIs seem to be immune to the process of AI themselves. The only elements left in the table in (169), then, are the two quantifier phrases Q S and Q W , the former the least prominent phase head of the higher, the latter the least prominent phase head of the lower cycle. By descending further along the nominal spine, I will continue my analysis with a closer inspection of the effects of the lowest phase head on adjectives in its complement. c. Co-Occurrence with Q w The weak quantifier constitutes the second element under consideration, probing for the φ-feature subset [φ[f]]. Therefore, we expect mismatches in probing of the SLI. As has been noted above, inflection on Q W itself is optional under certain conditions to be analyzed in detail in the next chapter. In what follows, I will therefore limit myself to contrast the two paradigms of (i) complete inflectionless Q W and (ii) possible vs. impossible inflection and elaborate on their respective influence on AI before again turning to the potential application of said impoverishment to Q W itself. Let me begin by proposing that, in the fully inflectionless paradigm, the weak quantifier can be understood as not probing for φ-features at all. As would be expected by now, impoverishment effects on adjectival inflection are taken to be nonexistent in these cases. Hence, under the theory elaborated above, the 266 3 INFERENCE - Nominal Concord & Definiteness adjective is taken to uniformly surface with strong inflection. This prediction is borne out. (170) strong adjectival inflection with uninflected Q W masc neut fem nom viel gut-er Wein viel gut-es Bier viel gut-e Milch gen viel gut-en Weins viel gut-en Biers viel gut-er Milch dat viel gut-em Wein viel gut-em Bier viel gut-er Milch acc viel gut-en Wein viel gut-es Bier viel gut-e Milch Observe that inflectionless Q W is prohibited with plural HNs, and hence excluded from (170), for reasons to be elaborated further in the next chapter (cf. ch. 4.2.2). Additionally, I marked those forms homonymous between the strong and weak inflectional pattern in grayscale once again. Based on the uniformity of the unambiguous forms, however, I believe it is safe to conclude that inflectionless (i.e. unprobing) viel patterns with strong adjectival inflection in line with the proposals made above. Turning to the (potentially) inflected variant of the quantifier, we expect to observe a pattern of adjectival inflection parallel to that accompanying the possessive pronoun, in line with their proposed shared φ-probing configuration [φ[f]]. This pattern is indeed observable, cf. (171) below; once again, homonymous inflections on the adjective are marked in greyscale: (171) mixed adjectival inflection with inflected Q W masc neut fem pl nom viel-ø gut-er Wein viel-ø gut-es Bier viel-e gut-e Milch viel-e gut-e Häuser gen viel-es gut-en Weins viel-en gut-en Biers viel-er gut-en Milch viel-er gut-er Häuser dat viel-em gut-en Wein viel-em gut-en Bier viel-er gut-en Milch viel-en gut-en Häusern acc viel-ø gut-en Wein viel-ø gut-es Bier viel-e gut-e Milch viel-e gut-e Häuser However, matters are more complicated here: In fact, apart from the homonymous forms indicated in the tables above, one additionally finds optional weak endings in certain cases accompanying the ( strong ) inflected quantifier. Vice 3.3 Splitting Nominal Concord 267 versa, other cases optionally allow for parallel weak inflectional suffixes on the adjective and SLI itself, the latter a complete novelty in the current analysis. For clarification and illustrational purposes, I will first give the full paradigm in (172) followed by a descriptive table making use of the abbreviations already employed above: (172) Full adjectival inflections with inflected Q W masc neut fem pl nom viel-ø gut-er Wein viel-ø gut-es Bier viel-e gut-e Milch viel-e gut-e Häuser gen viel-es/ en gut-en Weins viel-es/ en gut-en Biers viel-er gut-en/ er Milch viel-er gut-er/ en Häuser dat viel-em gut-en Wein viel-em gut-en Bier viel-er gut-en/ er Milch viel-en gut-en Häusern acc viel-ø gut-en Wein viel-ø gut-es Bier viel-e gut-e Milch viel-e gut-e Häuser (173) Full adjectival inflections with inflected Q W - schematized masc neut fem pl nom Q-ø ADJstrong Q-ø ADJstrong Qhomonym ADJhomonym Qstrong ADJstrong gen Qsli / homonym ADJhomonym Qsli / homonym ADJhomonym Qstrong ADJstrong / weak Qstrong ADJstrong / weak dat Qstrong ADJweak Qstrong ADJweak Qstrong ADJstrong / weak Qhomonym ADJhomonym acc Q-ø ADJhomonym Q-ø ADJstrong Qhomonym ADJhomonym Qstrong ADJstrong I have coded the strong and weak endings on both the weak quantifier as well as the adjective following it. The notion of SLI inflection marked twice on the quantifier in (173) denotes the cases in which the morphological form of other SLIs, taken to agree in these cases (e.g. Poss PRO ), and adjectival inflection diverge from one another, which has been noted earlier by Karnowski & Pafel (2004: 169, 172f.). These forms are restricted to the genitive masculine and neuter (i.e. SLI: es , strong / weak : en ) and are always accompanied by homonymous strong / weak adjectival inflection. Furthermore, homonymous forms on the quantifier occur in both structural cases in the feminine paradigm (e ) and dative plural 268 3 INFERENCE - Nominal Concord & Definiteness (en ); the latter three, moreover, also constitute cases of homonymy of strong and weak adjectival inflection. The only case in which inflection on the adjective is homonymous between both paradigms without variation on the quantifier (by means of either variable realizations or a homonymous form) is found in the accusative masculine, in which the quantifier surfaces uninflected. How, then, can this vastly heterogeneous pattern be captured in terms of the analysis elaborated here? To begin, let us focus on the schematic representation in (173) from the point of view of the mixed adjectival inflection and the underlying Agree-operations proposed to partake in its particular derivations. To this end, I designate the partition of explicit and ambiguous forms of the full paradigm in (174). (174) Full adjectival inflections with inflected Q W - ambiguous forms masc neut fem pl nom Q-ø ADJstrong Q-ø ADJstrong Qhomonym adjhomonym Qstrong ADJstrong gen Qsli / homonym ADJhomonym Qsli / homonym adjhomonym Qstrong adjstrong / weak Qstrong adjstrong / weak dat Qstrong ADJweak Qstrong ADJweak Qstrong adjstrong / weak Qhomonym adjhomonym acc Q-ø adjhomonym Q-ø ADJstrong Qhomonym adjhomonym Qstrong ADJstrong Concentrating on the masculine and neuter patterns, we can observe that the partition of structural vs. non-structural cases still holds in that the former constitute the complete set of non-inflected forms of the quantifier, which are accompanied by strong adjectival inflection (glossing over the homonymous accusative masculine as justified above). Ambiguity is hence only observed with inflected forms of the quantifier. From these, we might further abstract away from regular homonymy in the structural cases of the feminine paradigm, for now, since these occur with all inflected SLIs. I will return to these forms below. Focusing on the non-structural cases, then, we find that the genitive pattern consistently exhibits some kind of ambiguity in all number/ gender-values, either by optional strong / weak -variation on the adjective or homonymy, while variation in the dative is observable only with [f]/ [pl] nominals. At this point, an interesting observation concerning the inflectional idiosyncrasies of the two seemingly distinct forms (dative masculine and neuter) is in order. As Roehrs (2009: ch. 4, fn. 17 and references therein; see also Roehrs in prep.: 31) observes, these forms exhibit the peculiar property of allowing AI 3.3 Splitting Nominal Concord 269 of the most embedded adjective in an Adj-Adj-N complex without concurrent dominance by an SLI. Observe (175) below, taken from Roehrs (in prep.: 31f.); Chakroun (1991) calls these inflectional configurations Stufung (‘gradation’), in contrast to the uniform pattern, termed Reihung (‘sequence’). He cites Duden (Dudenredaktion 1984), predominantly perceived as the prescriptive grammar of German, in the fact that gradatial inflection in these cases is outdated and does no longer hold. Likewise, the PONS guide for German (Balcik 2007: 26), yet another prescriptive grammar, labels gradial inflection in these cases as “nicht falsch […], sie gilt heute aber als veraltet.” (‘not wrong but counting as outdated today’ [MB]). Observe, however, that although Roehrs labels the options of sequential and gradatial inflection as vaying in acceptability with the speaker (his use of the percentage sign ‘%’ below, cf. Roehrs in prep.: 31), my informants greatly accept (175) b. over a., contrary to PONS’ and Duden’s claims; the remaining logical possibilities in c. and d. in contrast, labelled as odd by Roehrs, are completely ruled out. (175) (= Roehrs in prep.: ch. 2, (41)) a. % mit frischem schwarzem Kaffee with fresh (ST) black (ST) coffee b. % mit frischem schwarzen Kaffee with fresh (ST) black (WK) coffee c. ? ? mit frischen schwarzem Kaffee with fresh (WK) black (ST) coffee d. ? ? mit frischen schwarzen Kaffee with fresh (WK) black (WK) coffee ‘with fresh black coffee’ Returning to the full paradigm of inflected Q W with this observation in mind leads us to a uniform picture concerning the ambiguity of form and derivational properties of the weak quantifier: I want to propose that all three idiosyncrasies of inflectional form of Q W itself as well as of adjectival items it dominates, to wit, strong / weak -variation, homonymy and gradatial inflection, serve the sole purpose to cloud the derivational status of the SLI w.r.t. its ability to enforce AI at Morphology; thus, in none of the non-structural case forms can Q W be considered to unambiguously act as a semi-lexical item parallel to the ones discussed above. In the genitive masculine and neuter, both SLIas well as homonymous adjectival inflection can surface on Q W but the effects are not observable due to the homonymy of inflection on the adjective itself. In the dative masculine and neuter, sequences of adjectives and SLI-Adj complexes behave alike in that the former bears strong , the latter weak inflection. Turning to the non-structural 270 3 INFERENCE - Nominal Concord & Definiteness cases in the feminine paradigm, Q W may either be taken to trigger Impoverishment or inflect in parallel like another adjectival element (the Q STRONG -Adj WEAK , Q STRONG -Adj STRONG optionality), which is also observed in the genitive plural. The dative plural as well as both structural cases in the feminine paradigm, then, are freed from the optionality just outlined in that the respective morphological forms of the strong and weak patterns are homonymous on both the SLI as well as the adjective, anyway. Thence, only the structural cases of the plural paradigm do unambiguously show parallel strong inflection and hence no application of AI. However, even if ungrammatical for my informants, internet searches for structures of the form Q STRONG -Adj WEAK -N pl do at times return up to over 100.000 hits. Since the morphological forms of Q STRONG -Adj WEAK are homonymous in nominative and accusative while at the same time divergent from all other forms in the paradigm this number can be fully attributed to instances of the constructions under consideration. I would therefore conjecture that the parallel strategy of SLI coding on the weak quantifier, alongside the adjectival coding, becomes strengthened for reasons to be elaborated further below. Returning to the point of departure in (173) above, observe that ambiguity does not arise in the structural cases in the masculine and neuter paradigm. We can therefore conclude that the narrow-syntactic processes elaborated above do hold for the weak quantifier but that all (non-homonymous) ambiguity discussed here arises from the variable application of AI at Morphology: If the adjectival ending is homonymous between the two patterns, the quantifier might surface with either the same homonymous or a divergent SLI-suffix, if strong and weak inflectional suffixes diverge morphologically, the adjective surfaces with either of these while Q W bears strong inflection. The dative masculine and neuter are ambiguous w.r.t. the categorical status of the higher (S)LI since AI applies in any case. Under one interpretation of the homonymous forms above we have uncovered a paradigm not found elsewhere but consistent in the hypotheses of the analysis elaborated here: strong adjectival inflection on the quantifier, arising only in the inflected forms of the mixed paradigm resulting from the quantifier probing for a subpart of all available φ-configurations while at the same time failing to trigger AI and being interpreted as an adjective-like element. We have then fully derived the wide variety of morphological realizations of nominal domains headed by the SLI of the lower phasal cycle from the ambiguity of this element to optionally trigger Adjectival Impoverishment. *17 Following the approach of the previous sections, I will end the analysis of the weak quantifier in its concatenation with other SLIs, i.e. one of the elements constituting the higher phasal cycle of the nominal domain, to therein focus on the application of AI to Q W itself. These configurations have been discussed in 3.3 Splitting Nominal Concord 271 chapter 3.2.3 above; here, I repeat the critical configurations from (105), (106) and (110) as (176) a.-c. respectively. (176) Nominal Concord & Definiteness I NFERENCE 187 Following the approach of the previous sections, I will end the analysis of the weak quantifier in its concatenation with other SLIs, i.e. one of the elements constituting the higher phasal cycle of the nominal domain, to therein focus on the application of AI to Q W itself. These configurations have been discussed in chapter 0 above; here, I repeat the critical configurations from 0, 0 and 0 as 0 a.-c. respectively. (176) a. (= 105) dies-er viel-e Wein D > Q W > N [φ[*m]] [φ[*m]] [φ[m]] b. (= 106) all-es viel-e Bier Q S > Q W > N [φ[*n]] [φ[*n]] [φ[n]] c. (= 110) mein viel-er Wein Poss PRO > Q W > N [φ] [φ[*m]] [φ[m]] In 0, the quantifier is headed by all three contextual phase heads of the higher cycle, i.e. D, Q S and Poss PRO respectively. Due to the 2 nd cycle Agree effects, brought about in the successive probing from said domain, Q W is obligatorily inflected in the concatenation with masculine as well as neuter HNs in structural case. In the discussions surrounding these data above, I hinted at inter-SLI effects targeting the inflectional form of the weak quantifier. We are now in the position to analyze these in detail. At first, contrasting 0 a. and c., it becomes apparent that, although inflected in both configurations, the morphological form of the suffix on Q W differs with the SLI sharing the complex feature structure with HN and Q W (i.e. D) and the SLI bearing partial default (i.e. zero-)morphology, Poss PRO . These heterogeneous patterns fall into place once the weak quantifier is identified as an element itself undergoing AI, parallel to adjectives. The difference in inflection hence again reveals the contrast of full vs. impoverished feature structure. Consequently, one expects to find the full paradigm of DP-Q W concatenations to mirror the WEAK , the quantifier in Poss PRO -Q W concatenations to surface in MIXED inflection. These predictions are both borne out. I will present the appropriate data with the proposed featural set-ups of the weak quantifier (parallel to the adjective) below. (177) WEAK adjectival inflection of Q W with D masc neut fem nom der viel-e gut-e Wein das viel-e gut-e Bier die viel-e gut-e Milch [*[nom][φ*[[sg][m]]]] [*[nom][φ*[[sg][n]]]] [*[nom][φ*[[sg][f]]]] gen des viel-en gut-en Weins des viel-en gut-en Biers der viel-en gut-en Milch [*[gen][φ*[[sg][m]]]] [*[gen][φ*[[sg][n]]]] [*[gen][φ*[[sg][f]]]] dat dem viel-en gut-en Wein dem viel-en gut-en Bier der viel-en guten Milch [*[dat][φ*[[sg][m]]]] [*[dat][φ*[[sg][n]]]] [*[dat][φ*[[sg][f]]]] acc den viel-en gut-en Wein das viel-e gut-e Bier die viel-e gut-e Milch [*[acc][φ*[[sg][m]]]] [*[acc][φ*[[sg][n]]]] [*[acc][φ*[[sg][f]]]] (178) MIXED adjectival inflection of Q W with Poss PRO masc neut fem nom mein-ø viel-er gut-er Wein mein-ø viel-es gut-es Bier mein-e viel-e gut-e Milch [*[nom][φ*[[sg][m]]]] [*[nom][φ*[[sg][n]]]] [*[nom][φ*[[sg][f]]]] gen mein-es viel-en gut-en Weins mein-es viel-en gut-en Biers mein-er viel-en gut-en Milch [*[gen][φ*[[sg][m]]]] [*[gen][φ*[[sg][n]]]] [*[gen][φ*[[sg][f]]]] dat mein-em viel-en gut-en Wein mein-em viel-en gut-en Bier mein-er viel-en gut-en Milch [*[dat][φ*[[sg][m]]]] [*[dat][φ*[[sg][n]]]] [*[dat][φ*[[sg][f]]]] acc mein-en viel-en gut-en Wein mein-ø viel-es gut-es Bier mein-e viel-e gut-e Milch [*[acc][φ*[[sg][m]]]] [*[acc][φ*[[sg][n]]]] [*[acc][φ*[[sg][f]]]] In (176), the quantifier is headed by all three contextual phase heads of the higher cycle, i.e. D, Q S and Poss PRO respectively. Due to the 2 nd cycle Agree effects, brought about in the successive probing from said domain, Q W is obligatorily inflected in the concatenation with masculine as well as neuter HNs in structural case. In the discussions surrounding these data above, I hinted at inter-SLI effects targeting the inflectional form of the weak quantifier. We are now in the position to analyze these in detail. At first, contrasting (176) a. and c., it becomes apparent that, although inflected in both configurations, the morphological form of the suffix on Q W differs with the SLI sharing the complex feature structure with HN and Q W (i.e. D) and the SLI bearing partial default (i.e. zero-)morphology, Poss PRO . These heterogeneous patterns fall into place once the weak quantifier is identified as an element itself undergoing AI, parallel to adjectives. The difference in inflection hence again reveals the contrast of full vs. impoverished feature structure. Consequently, one expects to find the full paradigm of DP-Q W concatenations to mirror the weak , the quantifier in Poss PRO -Q W concatenations to surface in mixed inflection. These predictions are both borne out. I will present the appropriate data with the proposed featural set-ups of the weak quantifier (parallel to the adjective) below. 272 3 INFERENCE - Nominal Concord & Definiteness (177) weak adjectival inflection of Q W with D masc neut fem nom der viel-e gut-e Wein das viel-e gut-e Bier die viel-e gut-e Milch [*[nom][φ*[[sg][m]]]] [*[nom][φ*[[sg][n]]]] [*[nom][φ[*[sg][f]]]] gen des viel-en gut-en Weins des viel-en gut-en Biers der viel-en gut-en Milch [*[gen][φ*[[sg][m]]]] [*[gen][φ*[[sg][n]]]] [*[gen][φ[*[sg][f]]]] dat dem viel-en gut-en Wein dem viel-en gut-en Bier der viel-en guten Milch [*[dat][φ*[[sg][m]]]] [*[dat][φ*[[sg][n]]]] [*[dat][φ[*[sg][f]]]] acc den viel-en gut-en Wein das viel-e gut-e Bier die viel-e gut-e Milch [*[acc][φ*[[sg][m]]]] [*[acc][φ*[[sg][n]]]] [*[acc][φ[*[sg][f]]]] (178) mixed adjectival inflection of Q W with Poss PRO masc neut fem nom mein-ø viel-er gut-er Wein mein-ø viel-es gut-es Bier mein-e viel-e gut-e Milch [*[nom][φ*[[sg][m]]]] [*[nom][φ*[[sg][n]]]] [*[nom][φ[*[sg][f]]]] gen mein-es viel-en gut-en Weins mein-es viel-en gut-en Biers mein-er viel-en gut-en Milch [*[gen][φ*[[sg][m]]]] [*[gen][φ*[[sg][n]]]] [*[gen][φ[*[sg][f]]]] dat mein-em viel-en gut-en Wein mein-em viel-en gut-en Bier mein-er viel-en gut-en Milch [*[dat][φ*[[sg][m]]]] [*[dat][φ*[[sg][n]]]] [*[dat][φ[*[sg][f]]]] acc mein-en viel-en gut-en Wein mein-ø viel-es gut-es Bier mein-e viel-e gut-e Milch [*[acc][φ*[[sg][m]]]] [*[acc][φ*[[sg][n]]]] [*[acc][φ[*[sg][f]]]] Thus, all three inflectional patterns are observable with the weak quantifier depending on the structural environment in which it occurs: As the highest functional head we find both a probing and a non-probing variant, which leads to either mixed adjectival inflection (since Q W probes only for a subset of φ) or uniform strong inflection on coordinated adjectives (since AI does not occur in the absence of a φ-complete SLI) respectively. Furthermore, the former case shows ambiguity between SLI and adjectival status of the weak quantifier in that every inflected form exhibits both parallel (adjectival) as well as divergent (SLI) inflection either by optional variation or homonymy of the appropriate forms. Coordinating these instances under one of the contextual phase heads of 3.3 Splitting Nominal Concord 273 the highest cycle, viz. D/ Poss PRO / Q S , disambiguates the picture: 2 nd cycle Agree, triggered by probing of the latter phase heads, enforces the (partially default forms of the) weak quantifier to agree with every number/ gender instantiation of HN. Subsequently, this SLI is uniformly treated in parallel to adjectives in the application of AI at Morphology in accordance with the selectional restrictions of the SLI dominating it. As carved out in previous sections of this subchapter, weak inflection of Q W and adjectives is observed in all configurations coordinating said complex under D, while mixed inflection arises therein when coordinated under Poss PRO . We have thus uncovered the flipside of application of AI at the SLI-level to be parallel inflection with a coordinated adjectival head in the weak as well as the strong paradigm depending on the syntactic environment. This will become important below. For now, I summarize the inflectional properties of Q W and the syntactic environments in which they occur in (179). (179) variant of Q W : not inflected SLI • strong / SLI inflection • triggering AI adjectival • parallel inflection with adjectives • exhibiting AI non-probing    probing [φ[f]] [[nom/ acc][φ[m/ n]]]   probing [φ] by 2 nd cycle Agree (headed by D/ Poss PRO )    For the sake of completeness and uniformity, I once again want to address floating constructions at the end of this section. However, see that the by now standard testing ground of a nominal domain headed by the focus particle nur (‘only’) is semantically anomalous while we found in chapter 2.2 that only plural nominals constitute the phasal configuration capable of moving from beneath the weak quantifier if no further SLI is externally merged above Q W . With one of the contextual phase heads of the higher cycle dominating the configuration Q W -Num/ N, however, floating from below the phasal complement (the now non-phasal Q W ) becomes possible again. Therefore, the relevant subset of the testing environments utilized above would make no new predictions about the AI properties of the quantifier but rather reiterate the findings concerning such effects these higher SLIs induce on Q W and adjectives, coordinated below them. 274 3 INFERENCE - Nominal Concord & Definiteness Since these have been discussed in detail above, I will refrain from repeating them here. We can, then, close the section on the effects of the weak quantifier on adjectival inflection in the modification of the table of SLI-properties, parallel to the treatment of D and Poss PRO above. However, the outcome is not as distinct as it has been thus far with the two most prominent contextual phase heads of the higher cycle: On the one hand, Q W is able to force AI under one interpretation of the full paradigm (to which the grammar itself stays agnostic due to the strategies of leveling adjectival and SLI inflections in variable overt realizations on both elements as well as homonymy), on the other hand, it might itself undergo this process of impoverishment. The latter option is a novel property not observed with those SLIs under analysis in previous sections, while all these elements are able to trigger AI. We hence arrive at (180). (180) Q S D Poss PRO Q W Adj Probing φ [φ] [φ] [φ[f]] [φ[f]] [φ] Default Agreement      Adjectival Impoverishment ?     The only element left to be analyzed regarding its properties of triggering/ undergoing AI in the table of SLIs in (180) is the strong quantifier Q S . I will turn to this item now. d. Co-Occurrence with Q s Focusing on the selectional properties of the strong quantifier as stated in the first row of (180), we - once again - expect not to find any disruptions in the featural chains of the nominal domain and hence to exhibit uniform weak adjectival inflection in hierarchies of projection headed by Q S . Application of AI to the quantifier itself, however, seems hardly testable on hierarchical/ linearizational grounds, since none of the elements identified as enforcing said impoverishment thus far finds itself in a suitable structural position w.r.t. Q S to trigger the effect, i.e. a dominance/ precedence configuration. However, as we have seen above, a second property of elements exhibiting AI has been uncovered in the analysis of the inflectional properties of the weak quantifier in the last section, namely the parallel inflection with adjectives in the nominal domain. With this in mind, let us begin by focusing on concatenations of the strong quantifier with Adj-N complexes in all number/ gender configurations: 3.3 Splitting Nominal Concord 275 (181) a. all-e gut-en Häuser Q Sstrong Adj weak N pl b. all-e gut-e Milch Q Sstrong Adj weak N fem c. all-es gut-e Bier Q Sstrong Adj weak N neut d. all-er gut-er Wein Q Sstrong Adj strong N masc Note that inflection on the quantifier is obligatorily present in all the strings above. In what follows, I will once again abstract away from optionality of inflection on the quantifier; since it has been identified as probing for the root node [φ], inflection will be at least possible in all mono-SLI configurations. As can be deduced from (181), feminine/ plural (cf. (181) a., b.) as well as neuter (cf. (181) c.) hierarchies of projection pattern along the proposals made in the beginning of this section, whereas the adjective and the quantifier in masculine nominal hierarchies (cf. (181) d.) surprisingly inflect in parallel, i.e. sequentially. 34 Observe even more so that, diverging from cases involving the weak quantifier in the previous section, neither optional variation between strong and weak inflectional realization nor homonymy of these forms exists. The paradigm in (182) below illustrates this point with all remaining logically possible collocations of strong and weak inflectional suffixes on Q S and the adjective in concatenation with a masculine HN: (182) a. ## all-er gut-e Wein Q Sstrong Adj weak N masc b. * all-e gut-e Wein Q Sweak Adj weak N masc c. * all-e gut-er Wein Q Sweak Adj strong N masc Duden (Dudenredaktion 2007: 53) once again does not differentiate between number/ gender variations and simply states that strong adjectival inflection counts as “veraltet und kommt heute nur noch selten vor” (‘outdated and only 34 To this end, the switch of the HN from a masculine to a neuter LI was conducted in chapter 3.3.3 in the analysis of concatenations incorporating two quantifiers, see the discussion surrounding (106) above. 276 3 INFERENCE - Nominal Concord & Definiteness occurs rarely today’ [MB], cf. also Dudenredaktion 2010: 58) in the context of masculine HNs. Building on the observation above, let me present the full case-paradigm of Q S -Adj-N concatenations on the basis of masculine nominals. Again, I will mark homonymous forms in greyscale: (183) masc nom all-er gut-er Wein Q S-STRONG Adj STRONG gen all-en gut-en Weins Q S-HOMONYM Adj HOMONYM dat all-em gut-en Wein Q S-STRONG Adj WEAK acc all-en gut-en Wein Q S-HOMONYM Adj HOMONYM This pattern is - once again - highly heterogeneous. Note first that, diverging from the pattern observed in concatenations with the weak quantifier in the previous section, the form of the suffix on Q S in genitive is uniquely adjectival and does furthermore not exhibit the same degree of contrast to the corresponding SLI-suffix es (as was the case above): While Duden (Dudenredaktion 2010: 58) designates concatenations of SLI inflected Q S and masculine HN as “selten” (‘scarce’ [MB]), Duden elsewhere (Dudenredaktion 2016: 979) moreover describes parallel configuration incorporating adjectival inflection on the quantifier as “fast nur noch so” (‘almost exclusive’ [MB]). The adjectival suffix en , however, is itself homonymous between the strong and weak adjectival form. Proceeding further down, we find the single unambiguous concatenation patterning according to the proposals made throughout the course of this chapter with a strong ending on the quantifier and corresponding weak inflection on the coordinated adjective. The masculine accusative - as is well known by now - is again homonymous on both elements. These data leave us with a puzzle: It is not the case that the sequential inflection can be identified to apply in all cases since the dative inflects as expected. Therefore, we cannot disambiguate the homonymous forms to one paradigm or the other. Moreover, the split into structural vs. non-structural case forms would be equally stipulative since homonymy is observed in both case-domains. Instead, I believe that the idiosyncratic properties of all paradigms of the two quantifiers with adjectives are historically motivated. Let me lay out why this 3.3 Splitting Nominal Concord 277 is so: Observe that the characteristics that set these apart from the ‘pure’ SLI paradigms are equal to one another but more strongly articulated with Q W than the object under investigation here while the characterizing property apart from unequivocal SLI inflection, i.e. sequential inflection with adjectives, has been described as outdated by Duden. I want to argue that these properties indicate the drift of the quantificational LIs from their adjectival syntactic origin to the establishment of semi-lexical categories. The split is neatly observed with weak quantifiers to constitute the mid-state in which the grammar allows for strategies of both categories to be applicable when these diverge morphologically; with the strong quantifier, however, its structural position sets it apart much more readily today and hence only single idiosyncratic forms have survived until this stage (in cases where ambiguity does not cloud the partition). Therein, the high frequency in use of the nominative might further benefit the fossilization of the sequential inflection, while a reanalysis of the relatively seldom used dative is much more accessible. A hint as to the correctness of these proposals can be deduced from the artificial modelling of unambiguous SLI inflection on the quantifier with accompanying homonymous adjectival inflection on a conjoined lexical head in genitive case: (184) gen, masc # all-es gut-en Weins Q S-SLI Adj HOMONYM Such structures are not homonymous with any other occurrences of the masculine paradigm and are still not readily rejected by my informants but instead judged as slightly odd, but acceptable. I will, however, refrain from speculating on the reasons for the masculine paradigm to still exhibit sequential inflection. Note instead that, at least concerning the adjectival inflection in genitive, the argument fully carries over to neuter nominal hierarchies as well, including the acceptability of disambiguated SLI inflection on Q S : (185) gen, neut all-es gut-en Biers Q S-SLI Adj HOMONYM Furthermore, a hint might be found therein as to why it is on the basis of morphological realization of the [f]/ [pl] paradigms that inflection on the quantifier is licit with varying forms in concatenation with the two elements of category D, i.e. the determiner and the demonstrative. This question was left unanswered 278 3 INFERENCE - Nominal Concord & Definiteness in the analysis of the extended pattern of SLI concatenations in chapter 3.2.3 above. As will be shown below, there are no ambiguities in inflectional form exclusive to the paradigms of the two quantifier heads in either the feminine or plural paradigm. In the reanalysis of the categorical status of these elements, concatenations with the embedded feminine number/ gender-value [pl] doubtlessly underwent said change first due to the primary semantic connectivity. This arguably carries over easily to the dominating node [f], while the status of the daughters of the sister-node of [f], i.e. class: [m] and [n], is still in the process of reanalysis today, which accounts for the inflectional paradigms observable on the weak quantifier. As for Q S , the realization of SLI inflection in concatenation with the prime semi-lexical item D 0 is based on the unambiguous cases in which the strong quantifier has fully transcended into such a category, too: to wit, the [φ[f]] paradigms. Additionally, observe that this property of the strong quantifier in concatenations with masculine head nouns can be utilized to solve another question, postponed in the discussion of the extended pattern of Q W concatenations in chapter 3.2.3, namely the featural content of inflection of the weak quantifier in configurations dominated by further SLIs of the highest phasal cycle. Concerning this matter, I remained agnostic as to the featural set-up including φ-features apart from case and only spelled out the underlying rationale of Q W undergoing AI in the last section. Coordinating the weak quantifier under the strong one in a masculine nominal domain, however, results in the following linearization: (186) all-er viel-er gut-er Wein all many good wine Q S > Q W > Adj > N Q W therein surfaces with the strong inflectional form, parallel to the suffix of the adjective (as well as the strong quantifier). We can therefore conclude that, even in configurations headed by an element of the highest phasal cycle, Q W does indeed bear a full set of φ-features even with masculine and neuter HNs in structural cases, to be standardly impoverished after Narrow Syntax at Morphology: (187) Nominal Concord & Definiteness I NFERENCE 191 Additionally, observe that this property of the strong quantifier in concatenations with masculine head nouns can be utilized to solve another question, postponed in the discussion of the extended pattern of Q W concatenations in chapter 0, namely the featural content of inflection of the weak quantifier in configurations dominated by further SLIs of the highest phasal cycle. Concerning this matter, I remained agnostic as to the featural set-up including φ-features apart from case and only spelled out the underlying rationale of Q W undergoing AI in the last section. Coordinating the weak quantifier under the strong one in a masculine nominal domain, however, results in the following linearization: (186) all-er viel-er gut-er Wein all many good wine Q S > Q W > Adj > N Q W therein surfaces with the STRONG inflectional form, parallel to the suffix of the adjective (as well as the strong quantifier). We can therefore conclude that, even in configurations headed by an element of the highest phasal cycle, Q W does indeed bear a full set of φ-features even with masculine and neuter HNs in structural cases, to be standardly impoverished after Narrow Syntax at Morphology: (187) Q S > Q W > Adj > N [*[nom][φ*[[sg][m]]]] [*[nom][φ*[[sg][m]]]] [*[nom][φ*[[sg][m]]]] [*[nom][φ[[sg][m]]]] Let me then present the full paradigm of strong quantifier and adjective concatenations in all remaining number/ gendervalues. In addition, I also incorporated the by now familiar schematic illustrations in the table below: (188) neut fem pl nom all-es gut-e Bier all-e gut-e Milch all-e gut-en Häuser Q S-STRONG Adj WEAK Q S-HOMONYM Adj HOMONYM Q S-STRONG Adj WEAK gen all-en gut-en Biers all-er gut-en Milch all-er gut-en Häuser Q S-HOMONYM Adj HOMONYM Q S-STRONG Adj WEAK Q S-STRONG Adj WEAK dat all-em gut-en Bier all-er gut-en Milch all-en gut-en Häusern Q S-STRONG Adj WEAK Q S-STRONG Adj WEAK Q S-HOMONYM Adj HOMONYM acc all-es gut-e Bier all-e gut-e Milch all-e gut-en Häuser Q S-STRONG Adj WEAK Q S-HOMONYM Adj HOMONYM Q S-STRONG Adj WEAK All homonymous cases left in 0 and not discussed above (i.e. all cases of the fem/ pl paradigms) are part of the regular inflectional pattern, as can be deduced from the analysis of the weak quantifier above. Based on the uniformity of the unambiguous cases in 0 above, I feel hence justified to classify the paradigm of adjectival inflection with the strong quantifier as uniformly undergoing AI to the exclusion of the nominative masculine form on historical grounds, but in line with the proposal concerning its probing properties elaborated above. Once again, I want to briefly discuss focused floating constructions at the end of this section. Again, however, semantic restrictions bar the coordination of the quantifying LI under the particle, parallel to Q W as elaborated above. Furthermore, as elaborated in chapter 2.2, regular QST is restricted to those cases in which a more prominent contextual phase head (i.e. D, Poss PRO ) dominates the complement of Q S . Therefore, adjectival inflection will always impoverish w.r.t. the featural setup of the latter. Variation in AI on adjectives of the form ‘ STRONG when focalized, WEAK when coordinated’ can hence not be emulated with this SLI. However, there is one complex phrasal set-up worth mentioning in this context whose properties - once again - illustrate the derivational nature of nominal concord in the combined inner workings of the LIs involved, namely the concatenation of the strong quantifier, possessive pronoun and adjective. The lower domain of this complex configuration has been analyzed in detail in a previous section; likewise, the concatenation in the domain of SLIs (i.e. Q S -Poss PRO ) has been discussed in chapter 0 above. Here, I want to put emphasis on the derivational nature in bringing together the findings of both these sections: As has been demonstrated, Q S possesses the ability to surface with default (i.e. zero- )morphology which sets it apart from instances of category D. Poss PRO , on the other hand, patterns with the weak quantifier in probing for the φ-configuration [φ[f]] rather than the root, which brings about mismatches and, subsequently, Let me then present the full paradigm of strong quantifier and adjective concatenations in all remaining number/ gender-values. In addition, I also incorporated the by now familiar schematic illustrations in the table below: 3.3 Splitting Nominal Concord 279 (188) neut fem pl nom all-es gut-e Bier all-e gut-e Milch all-e gut-en Häuser Q S-STRONG Adj WEAK Q S-HOMONYM Adj HOMONYM Q S-STRONG Adj WEAK gen all-en gut-en Biers all-er gut-en Milch all-er gut-en Häuser Q S-HOMONYM Adj HOMONYM Q S-STRONG Adj WEAK Q S-STRONG Adj WEAK dat all-em gut-en Bier all-er gut-en Milch all-en gut-en Häusern Q S-STRONG Adj WEAK Q S-STRONG Adj WEAK Q S-HOMONYM Adj HOMONYM acc all-es gut-e Bier all-e gut-e Milch all-e gut-en Häuser Q S-STRONG Adj WEAK Q S-HOMONYM Adj HOMONYM Q S-STRONG Adj WEAK All homonymous cases left in (188) and not discussed above (i.e. all cases of the fem/ pl paradigms) are part of the regular inflectional pattern, as can be deduced from the analysis of the weak quantifier above. Based on the uniformity of the unambiguous cases in (188) above, I feel hence justified to classify the paradigm of adjectival inflection with the strong quantifier as uniformly undergoing AI to the exclusion of the nominative masculine form on historical grounds, but in line with the proposal concerning its probing properties elaborated above. Once again, I want to briefly discuss focused floating constructions at the end of this section. Again, however, semantic restrictions bar the coordination of the quantifying LI under the particle, parallel to Q W as elaborated above. Furthermore, as elaborated in chapter 2.2, regular QST is restricted to those cases in which a more prominent contextual phase head (i.e. D, Poss PRO ) dominates the complement of Q S . Therefore, adjectival inflection will always impoverish w.r.t. the featural set-up of the latter. Variation in AI on adjectives of the form ‘ strong when focalized, weak when coordinated’ can hence not be emulated with this SLI. However, there is one complex phrasal set-up worth mentioning in this context whose properties - once again - illustrate the derivational nature of nominal concord in the combined inner workings of the LIs involved, namely the concatenation of the strong quantifier, possessive pronoun and adjective. The lower domain of this complex configuration has been analyzed in detail in a previous section; likewise, the concatenation in the domain of SLIs (i.e. Q S -Poss PRO ) has been discussed in chapter 3.2.3 above. Here, I want to put emphasis on the derivational nature in bringing together the findings of both these 280 3 INFERENCE - Nominal Concord & Definiteness sections: As has been demonstrated, Q S possesses the ability to surface with default (i.e. zero-)morphology which sets it apart from instances of category D. Poss PRO , on the other hand, patterns with the weak quantifier in probing for the φ-configuration [φ[f]] rather than the root, which brings about mismatches and, subsequently, default morphology in concatenations with non-feminine HNs in structural case. In these configurations, we hence encounter two SLIs standardly able to enforce AI on coordinated adjectives both bearing only the partial default (i.e. root node) [φ]: Poss PRO by mismatch, Q S by the intervention effect of the former. It is in exactly these cases that we find coordinated adjectives to bear strong inflection in line with the theory elaborated here; the structures under consideration as well as the proposed featural set-ups of the elements involved are given below: (189) Nominal Concord & Definiteness I NFERENCE 192 default morphology in concatenations with non-feminine HNs in structural case. In these configurations, we hence encounter two SLIs standardly able to enforce AI on coordinated adjectives both bearing only the partial default (i.e. root node) [φ]: Poss PRO by mismatch, Q S by the intervention effect of the former. It is in exactly these cases that we find coordinated adjectives to bear STRONG inflection in line with the theory elaborated here; the structures under consideration as well as the proposed featural set-ups of the elements involved are given below: (189) a. all-ø mein-ø gut-es Bier all my good beer Q S > Poss PRO > Adj > N [φ] [φ] [*[nom][φ*[[sg][n]]]] [*[nom][φ[[sg][n]]]] b. all-ø mein-ø gut-er Wein all my good wine Q S > Poss PRO > Adj > N [φ] [φ] [*[nom][φ*[[sg][m]]]] [*[nom][φ[[sg][m]]]] Once again, floating of the highest contextual phase from beneath the strong quantifier is legal in these configurations. We again expect this movement to trigger a 2 nd cycle of Agree on the stranded element while at the same time AI is expected to not occur (since the phase head dominating the adjective still only bears the root [φ]). All these predictions are borne out. (190) [mein-ø gut-es Bier] i […] all-es t i my good beer all Poss PRO > Adj > N > Q S [φ] [*[nom][φ*[[sg][n]]]] [*[nom][φ[[sg][n]]]] [*[nom][φ*[[sg][n]]]] I take it as an argument in favor of the theory elaborated here that the morphological forms comprising these complex nominal configurations are readily explained in terms of the proposals made in the course of this chapter. Taking the findings of this section, we are finally in the position to complete the table of properties of SLIs, first installed in the discussion of the basic paradigm of SLI inflection in chapter 0 above: W.r.t. Adjectival Impoverishment, the strong quantifier therein constitutes the only element regularly divided in its application. (191) Q S D Poss PRO Q W Adj Probing φ [φ] [φ] [φ[f]] [φ[f]] [φ] Default Agreement      Adjectival Impoverishment [φ[m/ n]]:  /  [φ[f]]:      As can be deduced from the table in 0, all semi-lexical items under consideration in the analysis presented above diverge from one another in those three properties that build the basis of the approach to nominal concord put forward in this chapter. To reverse the viewpoint, all the properties listed in 0 suffice to account for the bulk of inflectional idiosyncrasies of the nominal domain in German, with exceptional realizations argued to also fall out of derivational processes in Narrow Syntax. Therein, the derivational approach to nominal concord based on a feature sharing account of Agree and the featural set-up argued for in chapter 2.3 and chapter 3.2.1 has proven to cover all successful as well as unsuccessful probing interactions of (semi-)lexical items and their morphological reflexes in the nominal domain in German. 3.4 Conclusion This chapter set out to derive morphological definiteness in the nominal domain in German. As I argued in the introduction above, I took the insight from chapter one, i.e. the equalization of morphological definiteness with the source for the STRONG / WEAK adjective inflection dichotomy, as the basis for my analysis to follow, to focus on agreement in the nominal domain, referred to as nominal concord. To this aim, I began my approach with the discussion of several accounts of φ-feature coding and featural interactions in previous analyses of agreement in general and nominal concord in particular. In the elaboration on findings in earlier chapters, probing features were equalized in their structural set-up with Once again, floating of the highest contextual phase from beneath the strong quantifier is legal in these configurations. We again expect this movement to trigger a 2 nd cycle of Agree on the stranded element while at the same time AI is expected to not occur (since the phase head dominating the adjective still only bears the root [φ]). All these predictions are borne out. (190) Nominal Concord & Definiteness I NFERENCE 192 default morphology in concatenations with non-feminine HNs in structural case. In these configurations, we hence encounter two SLIs standardly able to enforce AI on coordinated adjectives both bearing only the partial default (i.e. root node) [φ]: Poss PRO by mismatch, Q S by the intervention effect of the former. It is in exactly these cases that we find coordinated adjectives to bear STRONG inflection in line with the theory elaborated here; the structures under consideration as well as the proposed featural set-ups of the elements involved are given below: (189) a. all-ø mein-ø gut-es Bier all my good beer Q S > Poss PRO > Adj > N [φ] [φ] [*[nom][φ*[[sg][n]]]] [*[nom][φ[[sg][n]]]] b. all-ø mein-ø gut-er Wein all my good wine Q S > Poss PRO > Adj > N [φ] [φ] [*[nom][φ*[[sg][m]]]] [*[nom][φ[[sg][m]]]] Once again, floating of the highest contextual phase from beneath the strong quantifier is legal in these configurations. We again expect this movement to trigger a 2 nd cycle of Agree on the stranded element while at the same time AI is expected to not occur (since the phase head dominating the adjective still only bears the root [φ]). All these predictions are borne out. (190) [mein-ø gut-es Bier] i […] all-es t i my good beer all Poss PRO > Adj > N > Q S [φ] [*[nom][φ*[[sg][n]]]] [*[nom][φ[[sg][n]]]] [*[nom][φ*[[sg][n]]]] I take it as an argument in favor of the theory elaborated here that the morphological forms comprising these complex nominal configurations are readily explained in terms of the proposals made in the course of this chapter. Taking the findings of this section, we are finally in the position to complete the table of properties of SLIs, first installed in the discussion of the basic paradigm of SLI inflection in chapter 0 above: W.r.t. Adjectival Impoverishment, the strong quantifier therein constitutes the only element regularly divided in its application. (191) Q S D Poss PRO Q W Adj Probing φ [φ] [φ] [φ[f]] [φ[f]] [φ] Default Agreement      Adjectival Impoverishment [φ[m/ n]]:  /  [φ[f]]:      As can be deduced from the table in 0, all semi-lexical items under consideration in the analysis presented above diverge from one another in those three properties that build the basis of the approach to nominal concord put forward in this chapter. To reverse the viewpoint, all the properties listed in 0 suffice to account for the bulk of inflectional idiosyncrasies of the nominal domain in German, with exceptional realizations argued to also fall out of derivational processes in Narrow Syntax. Therein, the derivational approach to nominal concord based on a feature sharing account of Agree and the featural set-up argued for in chapter 2.3 and chapter 3.2.1 has proven to cover all successful as well as unsuccessful probing interactions of (semi-)lexical items and their morphological reflexes in the nominal domain in German. 3.4 Conclusion This chapter set out to derive morphological definiteness in the nominal domain in German. As I argued in the introduction above, I took the insight from chapter one, i.e. the equalization of morphological definiteness with the source for the STRONG / WEAK adjective inflection dichotomy, as the basis for my analysis to follow, to focus on agreement in the nominal domain, referred to as nominal concord. To this aim, I began my approach with the discussion of several accounts of φ-feature coding and featural interactions in previous analyses of agreement in general and nominal concord in particular. In the elaboration on findings in earlier chapters, probing features were equalized in their structural set-up with I take it as an argument in favor of the theory elaborated here that the morphological forms comprising these complex nominal configurations are readily explained in terms of the proposals made in the course of this chapter. Taking the findings of this section, we are finally in the position to complete the table of properties of SLIs, first installed in the discussion of the basic para- 3.4 Conclusion 281 digm of SLI inflection in chapter 3.2.1 above: W.r.t. Adjectival Impoverishment, the strong quantifier therein constitutes the only element regularly divided in its application. (191) Q S D Poss PRO Q W Adj Probing φ [φ] [φ] [φ[f]] [φ[f]] [φ] DefaultAgreement      AdjectivalImpoverishment [φ[m/ n]]:  /  [φ[f]]:      As can be deduced from the table in (191), all semi-lexical items under consideration in the analysis presented above diverge from one another in those three properties that build the basis of the approach to nominal concord put forward in this chapter. To reverse the viewpoint, all the properties listed in (191) suffice to account for the bulk of inflectional idiosyncrasies of the nominal domain in German, with exceptional realizations argued to also fall out of derivational processes in Narrow Syntax. Therein, the derivational approach to nominal concord based on a feature sharing account of Agree and the featural set-up argued for in chapter 2.3 and chapter 3.2.1 has proven to cover all successful as well as unsuccessful probing interactions of (semi-)lexical items and their morphological reflexes in the nominal domain in German. 3.4 Conclusion This chapter set out to derive morphological definiteness in the nominal domain in German. As I argued in the introduction above, I took the insight from chapter one, i.e. the equalization of morphological definiteness with the source for the strong / weak adjective inflection dichotomy, as the basis for my analysis to follow, to focus on agreement in the nominal domain, referred to as nominal concord. To this aim, I began my approach with the discussion of several accounts of φ-feature coding and featural interactions in previous analyses of agreement in general and nominal concord in particular. In the elaboration on findings in earlier chapters, probing features were equalized in their structural set-up with those on the heads of the lexical domain, N and Num. I additionally argued that nominal concord is a derivational phenomenon, derivable in specific terms of Agree proven to be applicable in all domains of Narrow Syntax and conceptually preferable over, or equal to, the standard account in Chomsky (DbP et seq.). Consequentially, the highly derivational system elaborated therein 282 3 INFERENCE - Nominal Concord & Definiteness took into account the findings on the phasal set-up and φ-feature relations carved out in previous sections of the analysis. Subsequently, the system was put to use in the derivation of agreement in simple SLI-N configurations to carve out the idiosyncratic restrictions in probing for φ-features of the respective SLIs. Therein, the first cases of mismatch in Agree were uncovered in the restricted probing of certain SLIs for only a subset of φ-values, utilizing the organization of features argued for in previous sections. As a result, inflectionless default forms surfaced in these configurations. The findings were recorded in a table of properties of SLIs. Following this establishment of a basic rationale, the full case-paradigms of the diverging classes of φ-probing SLIs were examined in more detail in the following section. Therein, fully φ-compatible items surfaced without zeromorphology while the first instances of 2 nd cycle Agree effects were uncovered with those items probing for only a subset of all available values (in SLI-N configurations with conflicting φ-values), which were triggered from outside the nominal domain proper by the ‘free’ assignment of non-structural cases to nominal domains externally merged in appropriate structural positions of the verbal spine. The properties were moreover backed up by data of multi-SLI concatenations in more complex nominal domains in the identification of further instances of agreement mismatches, 2 nd cycle Agree and intervention effects, as well as additional cases of default agreement on the surface. These were likewise included in the aforementioned table. Finally, the elaborate system of SLI inflection was analyzed w.r.t. the influence on the morphological form of adjectives. The phenomenon therein has been understood as a post-syntactic effect evaluated at Morphology by Impoverishment of the featural set-up of the latter elements and triggered by their structural relation to respectively specified SLIs in the nominal domain. The variation of morphological form between the strong and the weak paradigm thereby fully fell into place with the SLI’s status of φ-completeness: Impoverished adjectival inflection is always accompanied by a φ-complete SLI dominating the LI, while strong inflection occurs elsewhere, i.e. with featurally deprived SLIs as well as in nominal domains not dominated by any SLI at all. The listed properties of the respective SLIs, to wit, the (in-)ability to surface in default agreement, as well as their selectional restrictions might therein conspire to give the illusion of a categorical dependence of the strong / weak dichotomy, when underlyingly, these suffice to fully deduce all patterns of adjectival inflection. Even more so, note that these properties together with the third one elaborated above unequivocally split all semi-lexical items apart and moreover cluster them with structurally adjacent ones in the sharing of respective features: adjectives and weak quantifiers undergo AI, weak quantifiers and possessive pronouns probe 3.4 Conclusion 283 for [φ[f]] (and moreover may surface with default agreement), possessive pronouns and determiners/ demonstratives never undergo AI, items of category D and strong quantifiers probe for the root node [φ]. Therein, these properties mimic the phrasal configuration proposed above. Finally translating back from adjectival inflection into the initial notion, we can thence make strong predictions about morphological definiteness in that it is solely dependent on the status of φ-completeness of semi-lexical elements in the nominal domain. Neither a decided phrase nor a new feature has to be introduced to capture the phenomenon. The merit of understanding the data in this way is that contrasts like the one in (192) below readily fall into place: Recall that only indefinite nominal domains are licit in postverbal subject position of existential constructions. This testing environment has been used to divide [± definite] categories of SLIs above. However, the contrast below is unaccounted for under this rigid categorical view, while the φ-feature concordapproach elaborated in this chapter much more readily accounts for the variation in grammatical status: φ-complete SLIs render the nominal domain definite; definite nominal domains are excluded from ES-contexts. (192) a. Es gibt viel-ø Milch im Kühlschrank. b. * Es gibt viel-e Milch im Kühlschrank. ‚There is much milk in the fridge.’ Recall that the weak quantifier probes for the subset [φ[f]] of all available φ-values. Recall also that there exists a non-probing variant of the weak quantifier. The effect on the morphological form in simple SLI-N concatenations is identical with masculine and neuter HNs: Q W surfaces without φ-features either by reduction to partial default or by not probing at all. With feminine HNs like Milch in (192), however, the difference in featural configuration is crucial: Nominal domains headed by non-probing Q W are morphologically indefinite, hence (192) a. is licit; in contrast, probing Q W results in a morphologically definite nominal domain which is hence illicit in the ES-context in (192) b. It is then not the category but simply the featural content of the (unsuccessful or non-) probing SLI that is crucial here. I take this set of data as a strong argument in favor of the proposals at hand. Once again taking the findings of previous chapters into account, we see that the status of phasehood of the SLIs involved patterns with the status of definiteness of the complete domain in a way that can be captured in one simple restriction (but see below for a re-evaluation of this connection); the formulation in (193) correctly captures all dependencies carved out above: 284 3 INFERENCE - Nominal Concord & Definiteness (193) Morphological Definiteness in German (first version) The highest phase head must bear φ-features. 35 However, we are able to further narrow down the featureal content on which the effect of morphological definiteness rests upon in German: Up until now, I have focused on the contrast of partial default, i.e. reduced, versus - what has repeatedly been called - φ-complete LIs even though I concentrated on the complex feature number/ gender above; I additionally assumed with Sauerland that AI exclusively targets these exact nodes in the feature hierarchy (i.e. feminine/ plural as well as masculine/ neuter under the theory of complex number/ gender-values elaborated here). Therefore, I have de facto focused on reduced vs. number/ gender-specific, rather than φ-complete, LIs throughout the investigation. Howsoever, turning to Greenberg’s (1978) analysis on the emergence of systems of nominal classification, we find a diachronic and moreover language universal linkage between this exact φ-feature and the concept of definiteness. Greenberg sketches “the process by which a definite article ultimately may become a noun marker” (Greenberg 1978: 61) over time. The specific stages of this development, identified and discussed therein, are given in (194) below (note that due to Greenberg’s own uncontinuous numbering, I have employed letters rather than numbers in the itemization). (194) Stage A → Stage B → Stage C → Stage D → Stage E → Stage F demonstrative (subject/ object) pronouns definite determiner indefinite article in specific reading class prefix nominality marker Greenberg subsequently goes through these stages in detail. He concludes that given a classifying demonstrative [Stage A, MB] and the constant tendency of demonstratives to generate agreement, the result will be a gender system [Stage E, MB]. Further, if the demonstrative goes through all the stages outlined, the result will be a marker on the noun as well as on the agreeing element. (Greenberg 1978: 78) 35 The rationale of two separate structural domains and φ-featural distribution therein to account for the variability of inflectional patterns on adjectives has been implemented as early as Olsen (1990), employing Abney’s (1987) structural proposals to the German nominal domain (see also Haider 1992, who additionally extends the proposed DP-CP parallelism in his speculations on the feature [± definite], mirroring [± interrogative] in the CP). Her account is, however, neither derivationally motivated nor linked to the ‘deeper’ concept of morphological definiteness and hence misses out on all non-adjectival findings presented above (cf. also fn. 30). 3.4 Conclusion 285 For him, this process takes place almost inevitably. Without going into detail, then, observe that the φ-feature gender under the view elaborated therein is ultimately a reflex of deictic classification by demonstratives (Stage A), evolved into definiteness (Stage C) and specificity (Stage D) along the way. Observe furthermore that the sequence of these stages advocated therein runs counter to the common understanding that φ-features diacronically come into existence as nominal categories and subsequently spread to lexical material in the nominal’s vicinity over time. As Greenberg states, if his theses hold, this chain of events has to be reversed. It is therefore obvious that morphological definiteness might not be dependent on the overt occurrence of certain categories but rather on the overt coding of its diachronic successor in the nominal domain, i.e. gender. I therefore feel justified to tighten the link between the specific φ-feature and the phenomenon of morphological definiteness in German in line with Sauerland (1996) in the modification of (193) above to (195), as presented in the introduction to this chapter above: (195) Morphological Definiteness in German (final version) The highest phase head must bear a gender-value. Note that therein, the system furthermore conforms to the conclusions reached by Gallmann (1996) about the underlying nature of definiteness in German in contrast to other Germanic languages (cf. Gallmann 1996: ch. 4.3, 4.4). As he notes, morpho-syntactic definiteness has lost the status of being coded by a dedicated feature f in the functional layer of the extended nominal domain (i.e. the projection D) in this language. 36 Instead, the nominal features n and k (viz. number together with gender as well as case respectively; cf. Gallmann 1996: ch. 4.2) have supposedly adopted this function in German in the (direct or indirect) agreement relation from HN to D 0 (cf. Gallmann 1996: 304), i.e. morphological definiteness reduces to a label for the latter: to wit, f = ( n , k ), which hence again results in a regular, though heterogeneous, patterning of values for definiteness to overt occurrences of SLIs in the number/ gender and case-paradigm. Concentrating on the content of said replacements for the feature f , the foregone chapter likewise conflated the categories subsumed under n into a single feature number/ gender, while the treatment of case diverges between the proposals made above and those advocated by Gallmann: He takes k to percolate along the same lines as n from HN through the nominal spine (but cf. Gallmann 1996: 36 Gallmann deduces this from the disrupted pattern of adjectival agreement with various SLIs in German, while Dutch as well as the Nordic languages exhibit uniform patterns of agreement for definiteness and can hence be taken to categorize their SLIs accordingly. 286 3 INFERENCE - Nominal Concord & Definiteness fn. 26) and the two to consistently surface together in portmanteau-morphemes (cf. Gallmann 1996: ch. 4.2, principle 3). Therefore, n and k never occur as independent features in his system. The current chapter, however, treated case as imposed on the nominal domain from the dominating predicate and hence reserved the aforementioned path of percolation exclusively for the former feature structure. Thus, returning to Gallmann’s claims, removing the moot casefeature from the formula established above leaves us with the equalization of morphological definiteness with the feature number/ gender on SLIs (as well as the absence of it on weak inflected adjectives) in German: f = n , which is congruent with the findings arrived at here. Returning to (195), it has to be stressed at this point, however, that the neat patterning between phasehood and gender-valuedness is an accidental one (in terms of the analysis presented here) and does not follow from the premises employed above. To illustrate this, take once again the highest phasal cycle as the point of departure: In simple SLI-N concatenations, definiteness of the nominal domain indeed depends on the status of valuedness of the phasal head. However, once configurations involving multiple SLIs are taken into account, it becomes apparent that the valuedness of non-phasal SLIs does not matter for structural reasons: In concatenations of D and Poss PRO , D the phase head, the former can only be taken to bear φ-features when the latter non-phase does. There is plainly no grammatical case of φ-incomplete phasal D. Likewise, in either instance of Q S -D/ Poss PRO configurations, the featural content of the quantifier depends on the head of its complement but it is in itself irrelevant since another SLI, besides being the highest phase head, might already bear a gender-value. Here, cases of valued Q S preceding a featurally deprived second SLI simply do not exist. The ranking of contextual phasal prominence, then, forces the illusion of a connection between these concepts upon us. Nevertheless, the rule in (195) is stable as to the status of all phasal SLIs but - of course - to the exclusion of phasal Num[φ[f[pl]]] in the lower lexical domain. Recall that Num 0 itself has been taken to constitute a phase head of the lower cycle only when featurally specified accordingly, viz. [φ[f[pl]]]. Due to its structural position as the only phasal head below the adjectival domain (as well as the non-suffixial realization of its featural content), the test criteria utilized in the previous sections are not applicable here. Once again employing ES-contexts, we do however see that bare plurals are licit in postverbal subject position, hence classified as indefinite in the terms of the analysis presented here: 3.4 Conclusion 287 (196) Es gibt Hunde im Garten. it are dogs in.the garden ‘There are dogs in the garden.’ I will therefore restrict the application of (195) to elements acquiring their φ-specification by application(s) of Agree, signaled by suffixial realization, and exclude inherently φ-specified heads. I believe to have demonstrated that morphological definiteness fully reduces to φ-valuedness, more precisely gender in the extended nominal domain in German. Below I repeat the most crucial finding of the current chapter, merged with the ph(r)asal set-up elaborated in previous chapters. (197) Q S( * ) > D* > Poss PRO( * ) > Q W( * ) > Num ( * ) Probing: [φ] [φ] [φ[f]] [φ[f]] ø Default Agr:     ø AI:  /     ø Phase #: [3 1 2] [2 1] Cycle #: 1 2 There are, however, several lose strings left from the analysis in the chapters thus far, to wit, why and when is inflection (apart from featural content) optional? Do phase heads interact in the establishment of complex LIs? Do they interact on the basis of categorical status? How are the individual properties, proposed above, passed on to such complex elements? It will be the task of the ensuing chapter to tackle these open questions and establish a coherent picture of inter-SLI workings in the nominal domain in German. 4 IMPLICATIONS - Optionality & Complex Formations 4.1 Introduction The preceding chapters laid out an approach to morphological definiteness as gender-agreement in (complex) nominal domains in German. As I have demonstrated, nominal concord is a highly derivational phenomenon based on subsequent applications of Match and Value, subparts of the complex operation Agree, applying (i) w.r.t. a revised feature geometry, reinterpreting feminine gender as a numberless plural marker, as well as (ii) in a contextual phasal setup, employing a strict order of phase heads and non-phasal complements. All logically possible co-occurrences of semi-lexical items in the nominal domain and their featural constellations have been discussed therein and fully traced back to the attributes of the respective elements concerning (i) probing, (ii) default agreement and (iii) Adjectival Impoverishment. Hence, not only did grammatical and ungrammatical configurations split up on principled grounds, the strong / weak / mixed adjectival inflection tripartition was moreover accounted for. Although the current investigation thus prides itself on its completeness concerning the identification of these derivational paths, there are several morphosyntactic phenomena either (i) touched on in the course of the foregone analysis or (ii) influenced in their treatment by implications made therein. Instances of the former (i) can be found in both the optionality of inflection on quantifiers and of the overt dislocation of nominal domains apart from these heads (i.e. quantifier stranding, QST, and split topicalization), the latter (ii) include the formation of complex SLI concatenations in the nominal domain such as the possessor doubling construction, partitive quantifiers as well as the formation of the strong quantifying SOs ( ein -) jed - (‘every’) and d meist - (‘most’). The current chapter is devoted to these issues, viz. chapter 4.2 to the former (i) and chapter 4.3 to the latter (ii) respectively. Moreover, even though I have restricted myself to morphological definiteness in the analysis above, hence to the overt realization of φ-values in the nominal domain under the conclusions reached at the end of the previous chapter, I feel urged to incorporate some comments on the underlying semantic purpose of these derivational mechanisms. This will be touched upon with reference to the strategies of ‘contextual domain restric- 4.2 Optionality 289 tion’ foremost in Basque in the context of the second aforementioned complex quantifier in the concluding subchapter. The current chapter can hence be understood to tie up various strings left loose in the course of the analysis. It is due to this heterogeneous thematic composition that I will only be able to sketch preliminary paths for the analyses of the aforementioned phenomena and thus restrict myself to the presentation of tentative additional modifications to the system elaborated in this inquiry. Nevertheless, I will demonstrate that the phenomena to be discussed broadly align along the premises of the theory developed above. I take this as further evidence for the correctness of the proposals arrived at in the preceding chapters. 4.2 Optionality It is the general consensus in the literature that optional variation on the surface poses a problem for an economy-based framework of syntactic analysis like post-Move α generativism, i.e. the Minimalist Program, which takes overt operations to be rigidly triggered by the elimination of features as an inevitable strategy to circumvent a crash of the derivation. As laid out by Biberauer & Richards (2006), on this view, syntactic operations navigate the very narrow area of conflict between the principles ‘Last Resort’ and ‘Full Interpretation’ in that operations are either forced or forbidden by the computational system. 1 Therein, the output of syntactic computation becomes fully determined by the initial Numeration, the relation between these levels of computation is hence one-to-one. Of course, optional instances of Internal Merge - in the sense that they do not seem to adhere to Last Resort - are numerous across languages, identifiable by contrasting them with equally grammatical linearizations dispensing with the application under consideration. Consequently, such ‘additional’ instances have to be attributed to Full Interpretation in the system sketched above, hence to an effect on the output at SEM. This is the rationale behind the standard approach to optionality of Chomsky (MP: ch. 4, (76); DbP: 34), claiming that every operation must have an effect on the (semantic) output in accordance with inter alia Fox’s (1999: (108)) ‘Output Economy’. Therefore, this rationale has been termed the ‘Fox-Reinhart intuition’ in Biberauer et al. (2010). It predicts that optional 1 As the authors note (Biberauer & Richards 2006: fn. 2), the two constraints “roughly subsume the principles of derivational and representational economy,” cf. Chomsky (1991) for a relevant discussion; cf. moreover the conflict between ‘Greed’ and ‘Procrastinate’ in terms of costliness of overt operations in earlier instantiations of the program, cf. also the discussion in MI (p. 99) on economy conditions in the ‘derivational approach to L’. 290 4 IMPLICATIONS - Optionality & Complex Formations applications of IM induce additional so-called ‘discourse effects’. The argument is further modified in Chomsky (OP: 140f.), who splits the C-I interface along the lines of “generalized argument structure” vs. “discourse-related and scopal properties” (Chomsky OP: 141) with the respective forms of Merge exclusively serving one of these functions (i.e. EM the former, IM the latter). Chomsky himself subsumes “old information and specificity, along with scopal effects” (OP: 140) under the latter; furthermore, Yang (2008: 2) elaborates this list of ‘Deffects’ in detail to include “focus, topic, definiteness, specificity, highlighting, ‘aboutness’, givenness, and some other semantic effects like scope change.” The system outlined above therefore predicts that semantically vacuous optionality does not exist since it violates both Last Resort (i.e. it is not rigidly constrained in its application) as well as Full Interpretation (i.e. it is not sanctioned by effects at SEM). In what follows, I will refer to instances of semantically vacuous IM as ‘true optionality’, a term borrowed from Biberauer & Richards (2006: 37). This subchapter will focus on two instances of - what appears to be - true optionality with German quantifiers, namely optionality of movement, resulting in QST-configurations/ split-topicalization, as well as optionality of overt inflection, foremost on the strong quantifier all -. Both these notions surfaced numerous times during my analysis in the preceding chapters, fortunately, the system elaborated up to this point allows for the deduction of rather specific proposals concerning the derivational timing of application as well as interaction of the mechanisms carved out above. Therein, it will be shown that although cases of true optionality are arguably observable in German, their origin in the system elaborated here is congruent with that of Chomsky’s proposals, i.e. the Numeration (however slightly modified). The one-to-one relation from Numeration to Output can hence be retained. Furthermore, I believe the compatibility of the rather specific set of data with the system at hand to lend additional support to the claims made in the preceding chapters. 4.2.1 Of Movement In chapter 2.2 above I extensively discussed the paradigms of licit split-topicalization as well as QST-configurations from nominal domains in German to carve out the phasal status of various nominal SLIs. The current subchapter, then, focusses on the exact featural motivation and derivational implementation of the underlying mechanisms, proposed in the preceding chapters, into the system developed thus far. Such mechanisms have been the subject of a long line of research (cf. Bobaljik 2003 for an extensive overview of the research tradition as well as the current state of the inquiry; cf. also C&G, ch. 1.4.1 above). Beginning with Shlonsky 4.2 Optionality 291 (1991) and Giusti (1991), independently extending proposals from Sportiche (1988, arrived at mainly on the basis of French data, transposed to Hebrew and Italian respectively), QST is treated as a derivational phenomenon for the first time, moving a subpart of the extended nominal domain from below a dedicated quantifier head. The earliest transformational approaches to QST-configurations in German can be found in Link’s (1974) analysis of Quantoren-Floating (‘quantifier float’) with all - ‘all’ and beid - ‘both’, interpreting the disrupted linear order as derived by transformational rules moving the quantifier rather than the D/ Poss PRO P-remnant. Moreover, his analysis incorporates an additional rule inserting the definite article if neither the demonstrative nor a possessive pronoun heads the stranded complex (but cf. Vater 1980: ch. 6.1). As is obvious, this subsumes all the configurations incorporating a phase head of the highest cycle in the analysis elaborated here, hence all cases of QST discussed in chapter 2.2 above. Kniffka (1986: ch. 2) further elaborates on the so-called Distanzstellung (‘distant position’) with - what she calls - definite vs. indefinite nominal hierarchies. In the course of her analysis, she bundles weak quantifiers with numerals, adjectives as well as the ein-words indefinite article and nominal negation on the basis of their grammatical outcomes in stranding configurations with indefinite nominal hierarchies (apart from strong quantifiers which are in turn only grammatical in such configurations with definite ones). Turning to the ph(r)asal set-up elaborated above, elements making up the aforementioned bundle of (S)LIs have been uniformly taken to be externally merged below the higher phasal cycle. Furthermore, Kniffka (1986: 66) also notes a restriction concerning number in the nominal domain, left by the proposed movement of weak quantifiers, to be plural and thus fully merges the phenomena of QST and split-topicalization, as discussed in ch. 2.2.1, under her premises; a rationale at which I will likewise arrive below. The earliest derivational approach analyzing QST-configurations as derived by movement from below a quantifier head in German can be found in Merchant (1996), already mentioned in the course of ch. 3.3.3 above. Building on the aforementioned analysis of Shlonsky (1991) for Hebrew, Merchant’s main focus lies on the identification of various ‘scrambling-sites’ in the verbal domain in an articulated phrasal set-up of the functional layer incorporating agreement phrases for both the External and Internal Argument, encircling TP. Therein, he also touches on the preconditions for overt agreement on all -. Merchant is therefore predominantly concerned with the terminal locations of movement in the verbal spine, while I focus below on the origin of said applications of IM. Employing different adjunction sites (i.e. manner adverbs adjoined to VP, sentential adverbs adjoined to TP) to identify phrasal boundaries, Merchant proposes that scrambling in German consists of two decided movement operations, 292 4 IMPLICATIONS - Optionality & Complex Formations i.e. A-movement of the argument to its respective agreement-phrase (viz. EA to Spec,TP and Spec,AgrSP above TP, IA to Spec,AgrOP below TP) with subsequent adjunction to the outer layer of these projections, an instance of Ā-movement. Therein, Merchant proposes to capture the twofold nature of scrambling (viz. A and Ā). Turning to the treatment of QST in his approach, Merchant proposes that stranding is only licit from A-positions, which subsumes the specifiers of the aforementioned phrases making up the functional layer together with the θ-position in the lexical one below. The resulting system therefore predicts an asymmetry in the stranding configurations for EA and IA w.r.t. TP, made visible by their relative position to the aforementioned adverb (though Merchant himself notes that this criterion is not foolproof, cf. ibidem: ch. 3.2.3): QST from IA is not able to strand the quantifier to the left of sentential adverbs since its highest stranding-site is the specifier of the corresponding agreement phrase, AgrOP. On the contrary, AgrSP, dominating TP, readily allows stranding of the respective quantifier from EA in its specifier - and therefore to the right of said adverb - with subsequent adjunction of the nominal SO to the maximal projection. Even though many of Merchant’s actual proposals are hardly translatable into the system elaborated here, I will come back to his analysis several times over the course of the following subchapters since various tenets of my own work are already implicitly contained in it. With this much in place, let us turn to Internal Merge in recent terms of the theory as well as the proposed instances thereof in the phenomena under consideration. Optionality of syntactic operations is most prominently depicted and analyzed in the domain of relative constituent order. As has been exemplified several times above, German, in this respect, apart from optional object shift/ stylistic fronting/ scrambling allows for the stranding of strong and weak quantifiers in base position when movement applies to the nominal domain. 2 Since 2 The terms object shift/ stylistic fronting/ scrambling are used here in an intentionally vague, non-technical sense. I will leave the derivational mechanisms underlying such movement targeting the IA unspecified and thus refer to all applications of raising to preverbal, sentence-initial positions by SOs other than the external argument simply as ‘OS’ in what follows, since I restrict myself to transitive matrix clauses; cf. Lenerz (1977) for a thorough discussion of argument order variations and default configurations predominantly in the German ‘Middle Field’; cf. Fanselow (2002), Frey (2005) for an approach to German V2-configurations based on said order variations, coupled with subsequent attraction of the highest appropriate SO to the CP-domain. As for standard Minimalism, Chomsky (MI, cf. also MP: ch. 4.10.2) approaches such OS-configurations by raising of IA to an additional specifier of v P with subsequent movement of EA into the verbal domain, checking T’s EPP, thereby crossing IA which is inactive for A-movement by means of having its features checked v P-internally and barred from defective intervention by equidistance of EA and IA in the same domain (MI: (40)/ (41), cf. the discussion in Chomsky MI: 130, surrounding (50) bii.), while Chomsky (DbP: 26, cf. also DbP: fn. 48) likewise 4.2 Optionality 293 no deviation emerges between the readings of moved vs. stranded quantifier constructions, I take the latter to result from truly optional, i.e. semantically vacuous, operations. Let us, then, focus on the exact implementation of these operations in the system at hand. To begin, observe the schematic structures of uniform vs. QST-ed Q S -DP concatenations in external argument position (i.e. (1) a. and b. respectively) in the format familiar from chapter three above. The derivational snapshots depict the stage after IM of the nominal domain to the specifier of T: (1) Optionality & Complex Formations I MPLICATIONS 237 nature of scrambling (viz. A and Ā). Turning to the treatment of QST in his approach, Merchant proposes that stranding is only licit from A-positions, which subsumes the specifiers of the aforementioned phrases making up the functional layer together with the θ-position in the lexical one below. The resulting system therefore predicts an asymmetry in the stranding configurations for EA and IA w.r.t. TP, made visible by their relative position to the aforementioned adverb (though Merchant himself notes that this criterion is not foolproof, cf. ibidem: ch. 3.2.3): QST from IA is not able to strand the quantifier to the left of sentential adverbs since its highest stranding-site is the specifier of the corresponding agreement phrase, AgrOP. On the contrary, AgrSP, dominating TP, readily allows stranding of the respective quantifier from EA in its specifier - and therefore to the right of said adverb - with subsequent adjunction of the nominal SO to the maximal projection. Even though many of Merchant’s actual proposals are hardly translatable into the system elaborated here, I will come back to his analysis several times over the course of the following subchapters since various tenets of my own work are already implicitly contained in it. With this much in place, let us turn to Internal Merge in recent terms of the theory as well as the proposed instances thereof in the phenomena under consideration. Optionality of syntactic operations is most prominently depicted and analyzed in the domain of relative constituent order. As has been exemplified several times above, German, in this respect, apart from optional object shift/ stylistic fronting/ scrambling allows for the stranding of strong and weak quantifiers in base position when movement applies to the nominal domain. 2 Since no deviation emerges between the readings of moved vs. stranded quantifier constructions, I take the latter to result from truly optional, i.e. semantically vacuous, operations. Let us, then, focus on the exact implementation of these operations in the system at hand. To begin, observe the schematic structures of uniform vs. QST-ed Q S -DP concatenations in external argument position (i.e. 0 a. and b. respectively) in the format familiar from chapter three above. The derivational snapshots depict the stage after IM of the nominal domain to the specifier of T: (1) a. [Q S > D > N] i > T 0 > t i > v 0 [* val ] [* val ] [ val ] [* val ] b. [D > N] i > T 0 > [Q S > t i ] > v 0 [* val ] [ val ] [* val ] [* val ] The derivation arriving at 0 a. proceeds along well-known lines: The nominal domain, constructed in a parallel workspace, externally merges at the edge of the strong v P-phase. Subsequently, T 0 merges and probes for [φ] (i.e. [ val ] in the schematics above). Due to the DP Inclusion Assumption from ch. 3.3.2, the phasal nominal domain is fully present and T finds a goal in the feature structure uniformly shared between the three elements Q S , D and N, the successful application of Agree is followed by IM of the complex nominal domain with the former probe, thereby checking T’s EPP-feature in the structural (i.e. nominative) case position. This series of operations has been gone through with all logically possible and grammatical SLI configurations in the nominal domain in the course of the preceding chapter. Let us turn to the QSTconfiguration illustrated in 0 b., then. Likewise, linearizations of the type depicted there have surfaced in the course of the preceding discussion. The derivational path just outlined is altered w.r.t. the size of the moved SO from the nominal domain in these configurations, attracting the subpart D-N instead of the full nominal domain to check the EPP. In chapter 2.2, such linearizations were taken as an argument for the non-phasal status of the quantifier in concatenations with the determiner/ demonstrative or possessive pronoun. Since the stranded SLI nevertheless bears an instance of the shared 2 The terms object shift/ stylistic fronting/ scrambling are used here in an intentionally vague, non-technical sense. I will leave the derivational mechanisms underlying such movement targeting the IA unspecified and thus refer to all applications of raising to preverbal, sentence-initial positions by SOs other than the external argument simply as ‘OS’ in what follows, since I restrict myself to transitive matrix clauses; cf. Lenerz (1977) for a thorough discussion of argument order variations and default configurations predominantly in the German ‘Middle Field’; cf. Fanselow (2002), Frey (2005) for an approach to German V2-configurations based on said order variations, coupled with subsequent attraction of the highest appropriate SO to the CP-domain. As for standard Minimalism, Chomsky (MI, cf. also MP: ch. 4.10.2) approaches such OS-configurations by raising of IA to an additional specifier of v P with subsequent movement of EA into the verbal domain, checking T’s EPP, thereby crossing IA which is inactive for A-movement by means of having its features checked v P-internally and barred from defective intervention by equidistance of EA and IA in the same domain (MI: (40)/ (41), cf. the discussion in Chomsky MI: 130, surrounding (50) bii.), while Chomsky (DbP: 26, cf. also DbP: fn. 48) likewise locates the initiator of OS in this application of IM but subsequently (and counter-cyclically, cf. fn. 5) raises EA from the lower specifier of v P to Spec,TP, with the IA either Ā-moved to Spec,CP (cf. DbP: 26) or displaced by Disl(ocation), a phonological rule (cf. DbP: 30f., cf. moreover fn. 4 below on equidistance in this context). Since the two phenomena, OS and QST, are able to intersect, I will come back to the former in the course of my analysis below. The derivation arriving at (1) a. proceeds along well-known lines: The nominal domain, constructed in a parallel workspace, externally merges at the edge of the strong v P-phase. Subsequently, T 0 merges and probes for [φ] (i.e. [ val ] in the schematics above). Due to the DP Inclusion Assumption from ch. 3.3.2, the phasal nominal domain is fully present and T finds a goal in the feature structure uniformly shared between the three elements Q S , D and N, the successful application of Agree is followed by IM of the complex nominal domain with the former probe, thereby checking T’s EPP-feature in the structural (i.e. nominative) case position. This series of operations has been gone through with all logically possible and grammatical SLI configurations in the nominal domain in the course of the preceding chapter. Let us turn to the QST-configuration illustrated in (1) b., then. Likewise, linearizations of the type depicted there have surfaced in the course of the preceding discussion. The derivational path just outlined is altered w.r.t. the size of the moved SO from the nominal domain in these configurations, attracting the subpart D-N instead of the full nominal domain to check the EPP. In chapter 2.2, such linearizations were taken as an argument for the non-phasal status of the quantifier in concatenations with the determiner/ demonstrative or possessive pronoun. Since the stranded SLI nevertheless bears an instance of the shared feature structure, established in the initial and isolated locates the initiator of OS in this application of IM but subsequently (and counter-cyclically, cf. endnote *19) raises EA from the lower specifier of v P to Spec,TP, with the IA either Ā-moved to Spec,CP (cf. DbP: 26) or displaced by Disl(ocation), a phonological rule (cf. DbP: 30f., cf. moreover endnote *18 on equidistance in this context). Since the two phenomena, OS and QST, are able to intersect, I will come back to the former in the course of my analysis below. 294 4 IMPLICATIONS - Optionality & Complex Formations derivation of the nominal domain, further modifications to the moved SLIs - e.g. by case-assignment - also affect the featural composition of the quantifier. The rationale incorporated therein also surfaces in Biberauer & Richards’ (2006) syntactic approach to ‘true optionality’ already mentioned above. In their account, the authors concentrate on strategies of satisfying obligatory rather than optional rules in NS like the EPP/ OCC of non-phasal heads in current standard minimalist terms (cf. Biberauer & Richards 2006: 40-42). While optional rules are restricted to having an effect on the output, operations satisfying obligatory (hence ‘costless’) rules are liberated from this FI-restriction, with Σ/ SEM free to choose any copy for interpretation. Therefore, the only restrictions on their satisfaction are syntactic in nature (with a necessary reflex at the A-P system). For the authors, this is the domain where semantically vacuous narrow-syntactic optionality prevails: It arises whenever the computational system does not care how a rule is satisfied (within the limits of the language’s specific parameter-settings, i.e. I-Language). Optionality hence proceeds along two structural dimensions, viz. source and size, with the cases under consideration here approachable along the latter lines. Biberauer & Richards concentrate on the checking of T’s EPP (a [D]-feature in their approach) by varying lexical material from v P. Therein, variation at the size-dimension is taken as the choice between ‘spec-raising’ and ‘spec-piedpiping’, viz. movement of the full lower verbal phase along with the attracted EA into the specifier of TP. Transposing this rationale to QST-constructions, T’s EPP would be satisfied by movement of the full φ-feature chain, as depicted in (1) a., however, the grammar might allow for a more minimal SO to serve this purpose equally well, to wit, the nominal phase. As has been elaborated in detail in chapter 2.2 above, lexical material coordinated under phasal D/ Poss PRO is barred from movement into the verbal spine that strands the phase head with default interpretation. We have therein identified the minimal structural complexity of the SO movable for the purpose of EPP-satisfaction from purely narrow-syntactic reasons. Observe that it is the exact same feature structure that enters into an Agree relation with T in both cases so that these options can be taken to be indistinguishable from a syntactic point of view. It is, then, not a question of how much lexical material can be pied-piped by the attracted SO but rather how much of the lexical material has to be pied-piped with the agreeing feature structure, with the lower-bound being the nominal phase, the upper-bound the nominal domain in cases where these notions diverge. Moreover, the obligatory nature of the feature initiating movement in these cases furthermore accords with Biberauer & Richards’ proposals. However, the second instance of a complex nominal domain from which QST-configurations have been taken to emerge incorporates another premise of 4.2 Optionality 295 the preceding analysis which challenges the proposals just laid out. Observe (2) below, in which DP is swapped for the second-most prominent highest nominal phase head, viz. Poss PRO0 in uniform vs. QST-configurations with a non-feminine HN, once again depicted by N/ [ val ]: (2) Optionality & Complex Formations I MPLICATIONS 238 feature structure, established in the initial and isolated derivation of the nominal domain, further modifications to the moved SLIs - e.g. by case-assignment - also affect the featural composition of the quantifier. The rationale incorporated therein also surfaces in Biberauer & Richards’ (2006) syntactic approach to ‘true optionality’ already mentioned above. In their account, the authors concentrate on strategies of satisfying obligatory rather than optional rules in NS like the EPP/ OCC of non-phasal heads in current standard minimalist terms (cf. Biberauer & Richards 2006: 40-42). While optional rules are restricted to having an effect on the output, operations satisfying obligatory (hence ‘costless’) rules are liberated from this FI-restriction, with Σ/ SEM free to choose any copy for interpretation. Therefore, the only restrictions on their satisfaction are syntactic in nature (with a necessary reflex at the A-P system). For the authors, this is the domain where semantically vacuous narrow-syntactic optionality prevails: It arises whenever the computational system does not care how a rule is satisfied (within the limits of the language’s specific parameter-settings, i.e. I- Language). Optionality hence proceeds along two structural dimensions, viz. source and size, with the cases under consideration here approachable along the latter lines. Biberauer & Richards concentrate on the checking of T’s EPP (a [D]-feature in their approach) by varying lexical material from v P. Therein, variation at the size-dimension is taken as the choice between ‘spec-raising’ and ‘spec-piedpiping’, viz. movement of the full lower verbal phase along with the attracted EA into the specifier of TP. Transposing this rationale to QST-constructions, T’s EPP would be satisfied by movement of the full φ-feature chain, as depicted in 0 a., however, the grammar might allow for a more minimal SO to serve this purpose equally well, to wit, the nominal phase. As has been elaborated in detail in chapter 2.2 above, lexical material coordinated under phasal D/ Poss PRO is barred from movement into the verbal spine that strands the phase head with default interpretation. We have therein identified the minimal structural complexity of the SO movable for the purpose of EPP-satisfaction from purely narrow-syntactic reasons. Observe that it is the exact same feature structure that enters into an Agree relation with T in both cases so that these options can be taken to be indistinguishable from a syntactic point of view. It is, then, not a question of how much lexical material can be pied-piped by the attracted SO but rather how much of the lexical material has to be pied-piped with the agreeing feature structure, with the lower-bound being the nominal phase, the upper-bound the nominal domain in cases where these notions diverge. Moreover, the obligatory nature of the feature initiating movement in these cases furthermore accords with Biberauer & Richards’ proposals. However, the second instance of a complex nominal domain from which QST-configurations have been taken to emerge incorporates another premise of the preceding analysis which challenges the proposals just laid out. Observe 0 below, in which DP is swapped for the second-most prominent highest nominal phase head, viz. Poss PRO0 in uniform vs. QST-configurations with a non-feminine HN, once again depicted by N/ [ val ]: (2) a. [Q S > Poss PRO > N] i > T 0 > t i > v 0 [ val ] [* val ] b. [Poss PRO > N] i > T 0 > [t i > Q S > t i ] [ val ] [* val ] [* val ] The schematic in 0 a. represents a structurally uniform but featurally disrupted nominal domain brought about by subsequent unsuccessful applications of Value (i.e. incompatibility between [φ] on N and [φ[f]] on Poss PRO as well as subsequent intervention by Poss PRO concerning Agree between N and Q S respectively) followed by featural reduction of both SLIs to partial default in the isolated derivation prior to EM in the verbal domain. However, as laid out in the last chapter, probing of T 0 still successfully applies with the nominal core N apart from these elements and applications of Agree and IM hence proceed in parallel to 0 a. from this point onward. Turning to QST from this nominal configuration as exemplified in 0 b., observe that the quantifier in base position bears overt inflection for φ and case and is hence taken to carry an instance of the shared feature structure. Above, I argued for the application of 2 nd cycle Agree of Q S , probing into its extended search space (viz. specifier) when the nominal domain raises into Spec,TP (i.e. the trace marked in boldface in the schematic above) coupled with the deactivation of the feature structure on the possessive to total default (cf. ch. 3.3.3). 3 Thereby, a second intervention effect is circumvented and the highest SLI of the nominal domain shares its feature structure with T and the nominal core N to the exclusion of Poss PRO . Generalizing from this, Spec,Q S P was taken as uniformly posing an intermediate landing site for the nominal domain in all QST-constructions, i.e. domains headed by 3 Additionally, violation of Anti-Locality by movement of the complement of Q S to its specifier was excluded in parallel to nur -Dem/ Poss PRO stranding configurations with reference to Grohmann’s original notion of Prolific Domain (pace Abels 2003, cf. ch. 3.3.2). The schematic in (2) a. represents a structurally uniform but featurally disrupted nominal domain brought about by subsequent unsuccessful applications of Value (i.e. incompatibility between [φ] on N and [φ[f]] on Poss PRO as well as subsequent intervention by Poss PRO concerning Agree between N and Q S respectively) followed by featural reduction of both SLIs to partial default in the isolated derivation prior to EM in the verbal domain. However, as laid out in the last chapter, probing of T 0 still successfully applies with the nominal core N apart from these elements and applications of Agree and IM hence proceed in parallel to (1) a. from this point onward. Turning to QST from this nominal configuration as exemplified in (2) b., observe that the quantifier in base position bears overt inflection for φ and case and is hence taken to carry an instance of the shared feature structure. Above, I argued for the application of 2 nd cycle Agree of Q S , probing into its extended search space (viz. specifier) when the nominal domain raises into Spec,TP (i.e. the trace marked in boldface in the schematic above) coupled with the deactivation of the feature structure on the possessive to total default (cf. ch. 3.3.3). 3 Thereby, a second intervention effect is circumvented and the highest SLI of the nominal domain shares its feature structure with T and the nominal core N to the exclusion of Poss PRO . Generalizing from this, Spec,Q S P was taken as uniformly posing an intermediate landing site for the nominal domain in all QST-constructions, i.e. domains headed by Poss PRO as well as D (with the latter featurally uniform upon the first application of Agree with all φ-configurations, however, this intermediate Merger remains without reflex on the surface). Upon closer examination, however, the path of successive cyclic movement from base position to Spec,QP and further to Spec,TP proves problematic in 3 Additionally, violation of Anti-Locality by movement of the complement of Q S to its specifier was excluded in parallel to nur -Dem/ Poss PRO stranding configurations with reference to Grohmann’s original notion of Prolific Domain (pace Abels 2003, cf. ch. 3.3.2). 296 4 IMPLICATIONS - Optionality & Complex Formations the system carved out in the preceding chapters. To illustrate this point, let me sketch the derivational steps leading to the snapshot in (2) b. in detail: Starting out with EM of the featurally disrupted nominal domain in the specifier of v P, the strong phase head’s EPP/ OCC is checked. Subsequently, T 0 merges and probes for [φ], finding a suitable goal in the caseless, hence active, nominal domain. In chapter three above, I repeatedly hinted at the varying complexity of the nominal domain to be internally merged as the sister of the higher label T as the source for the optionality between uniform and QST-configurations in a Biberauer & Richards-style approach w.r.t. the size-dimension, as exemplified above. Therein, uniform movement would result in the structural configuration depicted in (2) a., whereas attraction of a structurally smaller complex, i.e. the phase DP/ Poss PRO P, serves the purpose equally well while resulting in the latter configuration (2) b. since probing of T skips partially default values in disrupted nominal domains. Moreover, the equally specified instances αP and its complement βP (i.e. αP [* val ] > βP [* val ]) in a uniform nominal domain in Spec, v P are ‘equidistant’ if this concept is based on the notion ‘minimal domain’ and coupled with the bare phrase structural approach employed above with reference to Rezac (2003), i.e. the sister of a terminal α 0 is itself a copy of the head β 0 .* 18 Attraction of said phasal category then proceeds along the lines sketched above. Note now, however, that at the derivational timing of attraction from T with subsequent application of IM, the proposed intermediate landing site Spec,QP does not constitute the root node but is buried in the extended verbal projection instead. Therefore, QP does not exhibit a specifier position at all. The combination of (i) the proposal of the initiation of movement of the nominal domain by T’s EPP-feature with (ii) the suggested uniform successive cyclic movement through the specifier of QP thus results in a counter-cyclic application of IM that violates both the Extension Condition (cf. Chomsky MP: 190) as well as the later No-Tamper Condition (Chomsky MI: 137, (59), AUGB: 8, OP: 138) by ‘tucking-in’ of the phasal DP/ Poss PRO P. As already noted by Epstein, Kitahara & Seely (henceforth EKS 2012, cf. also EKS 2013) as well as Branigan (2013, 2014), the same is true for a syntactic framework incorporating feature inheritance from phase heads as proposed by Richards (2007), laid out in chapter 2.2 above and incorporated in Chomsky (AUGB et seq.).* 19 Let me demonstrate why this is so: Starting out with a fully constructed lower phase v P under the predicateinternal subject hypothesis, EA is located at the edge when T 0 is successively merged. Since T possesses no uninterpretable features of its own, Agree and IM do not apply and the derivation continues with EM of the next higher phase head C 0 . Subsequently, C hands down its uninterpretable φand its EPP/ OCCfeature to T, which then probes to find a goal in EA in Spec, v P. Agree applies and IM copies and reattaches EA as the sister of the higher label T. But this position 4.2 Optionality 297 already exists with the initiator of this movement operation, C 0 , at this moment in the derivation. EA must hence be tucked-in between these heads, thereby violating the aforementioned conditions since (i) the operation does not apply to the root node (Extension Condition, EC) and (ii) does not leave the two SOs that are Merged unchanged (No-Tamper Condition, NTC). Therefore, all derivational operations other than EM are restricted to the phase level in this framework and are taken to apply simultaneously, hence counter-cyclically (cf. endnote *19). Both approaches mentioned above nevertheless incorporate feature inheritance and hence tucking-in, however, not without some modification to be made precise here: For one, EKS focus on the modification of the inheriting head’s sisterhood relation by IM. To visualize this state of affairs, consider the combined set-theoretic notions from EKS for simple EPP-raising: (3) a. (= EKS 2012: (5)) {C, T 1 } (where T 1 = [T [ v P Bill [ v [ VP ate rice]]]]) b. (= EKS 2012: (6), (7), [combined]) {C, T 2 } (where T 2 = [Bill [T [ v P Bill [ v [ VP ate rice]]]]]) The former relation of C and T 1 , depicted in (3) a., cannot be destroyed by the application of Internal Merge (EKS’s ‘Law of Conservation of Relations’, contrasted with the EC/ NTC by the authors). Therefore, the authors propose that “counter-cyclic IM […] will produce two distinct but intersecting set-theoretic syntactic objects, SOs, which happen to share a term, namely, T 1 ” (EKS 2012: 256, emphasis in the original). The result is a non-uniform, i.e. ‘two-peaked’ hierarchy. It goes without saying that such an SO can be regarded as poisonous at the interfaces under standard assumptions, however, as noted by the authors, such configurations pose problems much earlier (in derivational timing), i.e. at the next application of Merge to the root node, which must be unambiguously identifiable in NS. EKS, then, deduce the cyclic application of TRANSFER from counter-cyclic Merge in an effort to subsequently disambiguate the root, thereby moreover capturing the effects of the PIC. EKS (2012) additionally deduce the theory-internal proposal of the invisibility of Spec,TP to C, while EKS (2013) moreover derive that-trace effects as well as a number of related empirical phenomena across languages, therefrom. On the other hand, Branigan (2013, 2014) implements counter-cyclic IM by relaxing the EC/ NTC to only preserve the complement of the probe in Branigan (2013) and moreover by transforming it into a violable constraint, stating the preference for the ‘re-Merge’ subpart of IM to target the root node, in Branigan (2014). In both cases, EC/ NTC is contrasted with a second, likewise violable, 298 4 IMPLICATIONS - Optionality & Complex Formations restriction in favor of IM terminating as close to the probe as possible; the latter is termed ‘Approach the Probe Principle’ (APP) while the former is renamed in parallel in Branigan (2014): the ‘Approach the Root Principle’ (APR). EC/ NTC hence allows IM to terminate as a tucked-in specifier of the complement of the probe as well as in head-adjoined position in addition to the still viable Merge at the (possibly also tucked-in) sister of the probe. With these notions in place, Branigan is able to derive various phenomena inter alia employing argument inversion by subsequent nested path movement to one multi-probing head P, each terminating as close to it as possible in accordance with the APP (in configurations in which IM to the probe itself, resulting in a filled specifier position, is barred): Since the closest goal α is attracted first, it re-Merges as the specifier of probe P’s complement C; however, additional probing by P now raises a lower goal β to the nearest structural position, i.e. the outer specifier of the complement C between α and P, thus transforming the hierarchical order P > C > α > β into P > [ CP β > α > C]. Observe, however, that even within the notions just exemplified, (intermediate) movement of DP/ Poss PRO P to Spec,Q S P (induced by T 0 ) is not predicted since Q S is never the complement of T 0 but either its complement’s specifier (i.e. EA in Spec, v P) or buried even deeper in the verbal spine. Turning to the system elaborated here, I fully excluded the rationale of feature inheritance (one of the fundamental principles of the accounts just outlined) in chapter 3.3 in favor of the contextual nature of phasal status. Therefore, all nominal (phase and non-phase) heads are taken to enter the derivation equipped with their full featural specification. Therein, strict cyclicity is retained in the adherence to the Extension Condition in the system at hand. Moreover, nominal concord itself has been analyzed as a highly derivational and cyclic phenomenon in the previous chapter. If we choose to follow parallel routes here and exclude the instance of tucking-in while striving to retain the proposal of 2 nd cycle Agree in the extended search space of the quantifier, we can unambiguously identify the narrow-syntactic source for optional movement from below Q S in the featural composition of the quantifier itself, more precisely: Q S can bear optional EPP/ OCC. 4 Under this view, optional movement in the nominal domain is hence born in the Numeration while the QST-constructions, exemplified above, pose a joint effect of two unrelated (but arguably identical, cf. however the discussion below) EPP/ OCC-features, viz. that of the quantifier as well as that of the tense-head in the verbal domain. A parallel proposal can already be found in Merchant’s analysis mentioned in the introduction to this subchap- 4 This is unexpected given that optional EPP/ OCC-features have been taken to constitute a property of phase heads (in stages of the program not incorporating feature inheritance, cf. Chomsky MI: 109; cf. also DbP: 34) while it has been argued above that a precondition for QST itself is the non-phasal status of the quantifier. 4.2 Optionality 299 ter. There, a feature [F], neither strong nor weak, is proposed only for inflected instances of the quantifier in both DP in-situ as well as QST-configurations, to be checked anytime in the derivation but at the least at LF. By checking of [F] in NS, DP raises to the “escape hatch” (Merchant 1996: 184) Spec,QP from which further movement might apply to it. It is therefore a feature of inflection resulting in movement, while I argue here for the identification of a feature of movement resulting in inflection. I will come back to Merchant’s feature [F] in more detail in the course of chapter 4.2.2 below. Observe that the claim made above does by no means pose an exceptional case of IM; under the strict PIC discussed in ch. 3.3.2 above, all movement out of the interior (i.e. complement) of a phase is necessarily taken to be initially motivated by the respective phase head itself , i.e. by an optional EPP/ OCC. In the case of OS, the phrase raised to the edge might be taken to subsequently raise onwards by T’s own EPP. 5 Let us follow this rationale further. If two unrelated applications of IM do indeed underlie QST-constructions in German, we may ask if these also occur detached from one another. This is undoubtedly the case for the second application, viz. Spec, v P to Spec,TP, the classic EPP. This leaves the former for further analysis, i.e. movement of the phase in the complement of Q S to Spec,Q S . See (4) below: (4) a. Die Männer haben das Bier getrunken. the men have the beer drunk a'. Das Bier haben die Männer getrunken. the beer have the men drunk b. (All) Die Männer haben (all) das Bier getrunken. [( all ) the men ] have [( all ) the beer ] drunk b'. (All) Das Bier haben (all) die Männer getrunken. [( all ) the beer ] have [( all ) the men ] drunk 5 As seems to be deduced from the OS-configurations in (4) below, the properties of raising to Spec,TP and receiving structural, i.e. nominative case, tightly linked together in chapter 3.3.2 above (as well as the one between v and accusative case) with reference to Chomsky’s (DbP: 6) approach, are split here in the proposal of IM of accusative-marked IA to said position with nominative-marked EA in its EM-position on the surface. However, the simplified structural proposals applied herein only serve an explanatory purpose. Therefore, reference to T as the probe in OS-configurations denote actually a not further specified part of the extended verbal domain, cf. fn. 2 on the terminal positions in OSconfigurations proposed in Chomsky (IM, DbP). As mentioned therein, I will refrain from speculating on the exact workings of OS here; note, however, that all proposals made in the main text hold irrespectively of the exact nature of the probe. 300 4 IMPLICATIONS - Optionality & Complex Formations c. Die Männer haben all-e das Bier getrunken. the men have all NOM.pl the beer drunk c'. Das Bier haben die Männer all-es getrunken. the beer have the men all ACC.neut drunk d. Die Männer haben das Bier all-es getrunken. the men have the beer all ACC.neut drunk d'. Das Bier haben die Männer all-e getrunken. the beer have the men all NOM.pl drunk ‘(All) The men have drunken (all) the beer.’ Let me go through the data in detail: (4) a. constitutes a run-of-the-mill V2 sentence structure of German with an inflected auxiliary verb occupying the higher verbal slot and the external argument in sentence-initial position. Datum (4) a'. depicts the corresponding OS-variant with IA now in preverbal position and EA in base position preceding the predicate (and following the auxiliary). The data in (4) b. and b'. demonstrate the ability of both arguments in both structural positions to occur with a strong quantifier in their respective extended projections, thus the omission of inflection on said SLI in accordance with the findings from chapter three. The configuration in (4) c. depicts the QST-configuration building on (4) a. above. As should be expected by now, the stranded quantifier occupies the position between the auxiliary and the internal argument, i.e. T 0 and Spec,VP, with no further applications of head movement in the lower verbal phase with the participle. Observe that the quantifier bears obligatory inflection parallel to the subject, i.e. nominative plural in the cases above. A parallel QST-construction building on the OS-configuration incorporates a quantifier bearing the equivalent morpho-syntactic property, viz. obligatory inflection in accordance with the internal argument, viz. accusative (singular) neuter in (4) c'. above; observe furthermore that the two morphological forms differ from one another and moreover that the ‘stranding-site’ of the quantifier differs with EA and IA w.r.t. the position of the argument in base position, i.e. Q of EA left of IA, Q of IA right of EA, again in accordance with standard claims concerning EMsites.* 20 Finally, the colloquial (4) d./ d'. are most important for the point at hand. Observe again that the strong quantifier is stranded in pre-predicate position in both examples. However, taking its inflectional form into account, it bears obligatory accusative (singular) neuter (IA) and nominative plural (EA) agreement and hence relates to the unmoved argument respectively. Thus, the structural configuration brought about by EPP/ OCC of Q S , i.e. an extended nominal domain in the specifier of the Q S (an instance of the initial movement proposed above) arguably surfaces in these linearizations. Furthermore, observe that such 4.2 Optionality 301 configurations might also surface bearing non-structural case when conjoined with an appropriate predicate as in (5) below. There, the quantifier once again unambiguously signals its membership in the extended domain of the IA by its distinct inflectional pattern. (5) a. Er hat all den Männern vertraut. he has [ all the men ] DAT trusted b. Den Männern hat er all-en vertraut. [ the men ] DAT has he all DAT.pl trusted c. Er hat den Männern all-en vertraut. he has [ the men ] DAT all DAT.pl trusted ‘He trusted all the men.’ As already noted by Merchant (1996: 186), such configurations are “structurally ambiguous” since potential movement from the nominal into the verbal domain can be taken to be string vacuous. The claims made above are however supported by the fact that the nominal domain in these cases seems to form a constituent together with the right adjacent quantifier by the ability to be substituted together as well as to be coordinated/ contrasted, the latter is laid out below: the linearizations in (6) a. and a'. demonstrate the constituency of the quantified nominal domains in SVO and OS-linearizations parallel to (4) b. and b'. above, (6) b. and b'. show the constituency of their counterparts incorporating DP > Q S0 configurations on the basis of (4) d. and d'. The data in (6) c. and c'. mimic the a. vs. b. contrast for nominal domains bearing non-structural case following (5). 6,7 (6) a. Die Männer haben etwas Wein und all das Bier getrunken. the men have [ some wine ] and [ all the beer ] drunk a'. Das Bier haben der Arzt und all die Männer getrunken. the beer have [ the doctor ] and [ all the men ] drunk b. Die Männer haben etwas Wein und das Bier all-es getrunken. the men have [ some wine ] and [ the beer all ACC.neut ] drunk 6 The coordinating/ contrasting compound’s φ-values have been selected to diverge from those of the original data in (4)/ (5), viz. the quantificational nominal complexes in (6), to unambiguously mark the affiliation of the stranded quantifier with the latter. 7 Observe that an argument against the proposal of Branigan’s APP for the cases under consideration is found in their constituency status, which is unexpected if the DP was tucked-in as a higher specifier of v P rather than moved to the specifier of Q S P. 302 4 IMPLICATIONS - Optionality & Complex Formations b'. Das Bier haben der Arzt und die Männer all-e getrunken. the beer have [ the doctor ] and [ the men all NOM.pl ] drunk c. Er hat keiner Frau aber all den Männern vertraut. he has [ no woman ] but [ all the men ] DAT trusted c'. Er hat keiner Frau aber den Männern all-en vertraut. he has [ no woman ] but [ the men all DAT.pl ] trusted Another property tightly linked to this observation that can be employed to clarify the structural status of such configurations is stress assignment, as laid out by Merchant (1996: 187) with reference to Link (1974) and Vater (1979). The argument goes as follows: The nominal domain poses a prosodic unit with the prosodic head that bears main stress located inside DP. This holds irrelevant of the structural relation of its parts, i.e. with DP in-situ as well as in Spec,QP. Therefore, the quantifier itself does not bear stress in both these configurations. A stranded quantifier, however, constitutes a prosodic unit itself and hence necessarily bears main stress. Turning to the data at hand, observe that both d. as well as d'. in (4) above are grammatical with unstressed Q. 8 However, these nominal domains seem to lack another property of constituents observed with their ‘regular’ counterparts, i.e. movement. I have argued above that QST-constructions are the result of two independent movement operations of the nominal domain originally coordinated below the quantifying head: IM to Spec,Q S P and Spec,TP, the latter motivated by T’s EPP, regularly (perhaps universally, cf. Chomsky MI: 109) part of the derivation. Once the independent existence of DP > Q S0 configurations is established, we can acknowledge that T’s EPP is uniform and economic in that it is the first appropriate and complete SO encountered by probing that has to raise in all configurations (I once again abstract away from the apparent skipping of EA in said probing in OS-configurations, cf. fn. 2, endnote *19). (7) a. * Die Männer all-e haben das Bier getrunken. [ the men all NOM.pl ] have the beer drunk 8 One reviewer remarks that any form of discontinuous nominal configuration is accompanied by an intonation pattern that hints at pragmatic or textual markedness of the construction. It was hence left to be demonstrated, as he further notes, that an unmarked type exists at all. With the approach just laid out, however, divergences in intonation of discontinuous from uniform domains have been exempted from pragmatic and semantic import but linked to that very syntactic property instead. 4.2 Optionality 303 a'. * Das Bier all-es haben die Männer getrunken. [ the beer all ACC.neut ] have the men drunk Moreover, it is tempting to equalize the featural content responsible for movement to T and Q S respectively since these apply to the exact same structural configurations. Given the insights from chapter three, we should tend to identify it as [φ]. However, recall from chapter 3.2.1 above that raising related to the EPP/ OCC takes place on the basis of a previously established relation of Agree. 9 As was argued in the preceding chapter, T is able to probe beyond partially default SLIs in the extended nominal domain and hence to always find a goal at least as late as the nominal core. However, as has been extensively argued above, the story is different for elements externally merged in said domain since intervention effects originate in the cyclic composition of the nominal spine and bring about featurally disrupted domains. Therefore, skipping partially default specifications cannot be proposed for φ-probing operations in the extended nominal domain. However, raising of a nominal domain headed by an intervener (i.e. Poss PRO0 in non-feminine nominal domains) to the specifier of the strong quantifier phrase (and onwards to Spec,TP) has been proposed here as well as in the preceding chapter. Combining all these postulates leaves us with the result that probing for [φ] cannot be taken to motivate the initial movement of the nominal spine to Spec,QP. I will not go into these matters in more detail. 10 An additional point that has to be addressed in this section concerns QSTconfigurations that are not instances of true optionality as defined above in that they do indeed have an effect on interpretation and therein adhere to FI. I will only present a single set of data contrasting QST on the basis of OS in (4) c'. above with one such configuration employing a deviant stranding-site of the quantifier: (8) a. Das Bier haben die Männer all-es getrunken. (= (4) c'.) the beer have the men all ACC.neut drunk 9 Note that the specific implementation of EPP/ OCC employed above has not gone unrivaled (cf. ch. 2, fn. 2 above) and equalization of the features on T and Q S would go through without a problem, e.g. under a ‘strong D-feature’ approach to the EPP (cf. Chomsky MP); however, several key aspects of the analysis from chapter three would be lost in turn. 10 Observe furthermore the contradictory line of reasoning concerning the proposed instances of 2 nd cycle Agree configurations under this view: The nominal spine would be taken to raise due to the successful application of Agree w.r.t. φ, thereby establishing a structural configuration in which Q S is able to retry probing for φ into the extended search space after a first failed attempt; cf. however fn. 9 above. 304 4 IMPLICATIONS - Optionality & Complex Formations b. Das Bier haben all-es die Männer getrunken. the beer have all ACC.neut the men drunk As has been elaborated above, the position of the stranded quantifier in (8) a. corresponds to the proposed external merging-site of the IA and hence acts as a “visible trace” (Merchant 1996: 181) for the fronted SO. In contrast, the quantifier right adjacent to the EA in base position and following the auxiliary would tentatively best be located at the outer edge of the v P phase. I will not go into detail on the structural position here (cf. endnote *19 above). Turning to meaning, the semantic supplement of (8) b. defies an exact paraphrase; it tends to bring about a strong notion of ‘aboutness’ concerning the raised object and furthermore seems to imply the contrast to a second relation, formed by the predicate but involving a different agent and patient in the discourse. This latter effect, however, is readily cancelable as is expected of a discourse effect like the ones argued to be brought about by optional EPP/ OOC-features: Note the grammatical outcome resulting from the addition of (9) to (8) b. (9) …und sonst hat niemand etwas getrunken. …and else has no.one something drunk ‘… and apart from that, no one drank anything.’ We have encountered, then, an instance of (discourse-related) semantically relevant optional movement in QST in accordance with standard proposals as laid out above. These instances are based on a diverging stranding-site of the quantifier rather than a diverging landing-site of the nominal SO in the extended projection of the verb (but nevertheless to the relevant structural position, the outer edge of v P, cf. also Miyagawa 2011: 362) as can be deduced from its adjacency to the auxiliary on the one side and the EA on the other, while truly optional, i.e. semantically vacuous movement leaves the SLI in base position. One might therefore feel tempted to attribute the supplement in meaning to the structural position. Since I am not concerned with different readings but with the derivational mechanisms underlying linearizational variation, I will not speculate on these matters further. The last phenomenon that has to be addressed in the current section, then, concerns a paradigm of movement configurations first mentioned at the end of chapter 2.2.1 and discussed again in chapter 3.3.2, viz. stranding of the dual-SLI complex Q S -Dem/ Poss PRO . There, this phenomenon was equalized with movement from below the particle nur ‘only’. In what follows, I will refer to such cases as ‘complex stranding’ configurations for obvious reasons. Stranding of 4.2 Optionality 305 this sort can apply to all nominal φ-configurations from below both of the most prominent phase heads of the higher cycle as can be deduced from the exemplary collection of data in (10), compiled from the relevant discussion in the previous chapter. Here, plural, neuter and masculine HNs raise from below both the demonstrative as well as the possessive pronoun: (10) a. (= chapter 3, (86) b., [gloss modified]) Kekse i habe ich [nur dies-e t i ] gegessen. cookies pl have I [ only these ] eaten ‘I have only eaten these cookies.’ b. (= chapter 3, (87) b., [gloss modified]) Bier i habe ich [nur dies-es t i ] getrunken. beer neut have I [ only this ] drunk ‘I have only drunk this beer.’ c. (= chapter 3, (90) b., [gloss modified]) Bier i habe ich [nur mein-es t i ] getrunken. beer neut have I [ only mine ] drunk ‘I have only drunk my beer.’ d. (= chapter 3, (92) b., [gloss modified]) Wein i schmeckt mir [nur mein-er t i ]. wine masc enjoy I [ only mine ] NOM Above, the particle has been located in a focus position parallel to FocP as proposed in Ticio’s (2010) phrasal set-up of the nominal domain. Therein, said phrase is taken to constitute a part of the highest prolific domain, the Ω-Domain, together with DP. In chapter three above, however, I hinted at a parallel membership concerning prolific domains for Foc and Q S , thereby legalizing movement of the (Θ/ Φ-Domain) complement of the latter through its specifier position in QST-configurations since Anti-Locality was relaxed to its original ban on domain-internal movement (cf. Grohmann 2000, 2003a) rather than applying to every phrase-boundary (cf. Abels 2003). 11 11 Observe, however, that the parallelism is not perfect since Foc 0 , i.e. nur , in contrast to Q S cannot be stranded alone but necessarily co-occurs with a nominal phase head in-situ. 306 4 IMPLICATIONS - Optionality & Complex Formations Returning to complex Q S -Dem/ Poss PRO stranding configurations, note first that these likewise can be initiated from all nominal φ-configurations with both of the aforementioned contextual phase heads in parallel to the data in (10) above (note, moreover, that this paradigm cannot be captured by the means provided by Merchant’s 1996 theory). Turning to Grohmann’s characterization of the highest prolific domain we obtain (11) below: (11) (= Grohmann 2003a: ch. 2, (30) iii.) Clausal Tripartition: What’s in a Prolific Domain (ΠΔ) (i) Ω-Domain: part of derivation where discourse information is established Additionally, recall that I speculated on the contextual application of prolific domains in the discussion above to standardly assign phasal D/ Poss PRO to the lower, i.e. Θ/ Φ-, exceptionally to the higher Ω-Domain in complex stranding configurations to therein permit movement of the complement through its edge. Both positions are exemplified below: (12) a. [Q S ] Ω > [D/ Poss PRO > … ] Θ/ Φ b. [Q S > D/ Poss PRO ] Ω > [ … ] Θ/ Φ Putting all these proposals to work, if something along these lines does indeed enable complex stranding, the null hypothesis should be to expect such variation to also enforce an effect on the (semantic) output in accordance with FI. Indeed, complex stranding constructions once again carry additional implications concerning a contrasted set of entities in the discourse universe parallel to the QST-constructions presented previously in this chapter, this time, however, designated by HN. Thus, the quantifier acts as a semantic unit together with the phase head in referring to one partitive set of HN in the discourse domain in contrast to the exhaustive set designated in unmarked configurations. Even more so, and moreover in accordance with Grohmann’s characterization of the Ω-Domain, this implication is once again cancelable as is demonstrated below by the grammatical outcome of an apposition in conjunction with the Q S variant of (10) a.: (13) Kekse habe ich all diese gegessen [und sonst sind keine Kekse mehr da]. cookies have I all these eaten [ and else are no cookies more there ] ‘I have only eaten these cookies and there are no more cookies.’ Note moreover that all complex stranding configurations exhibit stress on the stranded phase head. Interestingly, once this pattern of stress assignment is 4.2 Optionality 307 applied to parallel in-situ nominal domains, the same implications also arise therein suggesting that it signals a change in the assignment of prolific domain of the phase head in these cases independent of linearizational properties. Most of the discussion of this subchapter readily carries over to structures incorporating the second structural position hosting quantifying elements in the extended nominal domain, Q W (e.g. 2 nd cycle Agree in Spec,Q W P with nonfeminine HNs, (at least partly) stranding in EA and IA position in colloquial registers) with the obvious structural reduction of coordinated/ floated complexes, supporting the view of underlyingly parallel phenomena of QST and split-topicalization. 12 Merchant (1996: ch. 3.3) argues against such a conflation on the basis of three criteria that distinguish these configurations, i.e. (i) occurrence with individuallevel predicates (henceforth IL; vs. stage-level, SL), (ii) varying landing-sites of movement (Spec,CP with split-topicalization, various sites with QST) and (iii) the structural complexity of the moved SO. I argued above that the latter property necessarily follows from the phrasal set-up of the nominal domain; in a related matter, the ungrammaticality described w.r.t. property (i) might be traced back to the semantic incompatibility of the unrestricted reference of the bare HN in combination with an individual-level predicate. Syntactically, however, note that Merchant (1996: (36), (37)) furthermore employs different types of predicates in his contrast, i.e. predicative adjectives in copula constructions for IL, locative PPs for SL respectively. Equalizing along both these lines, we find Merchant’s copula split-topicalization constructions ungrammatical across the SL/ IL distinction (cf. (14) a.) and those discussed above uniformly grammatical (cf. (14) b.): (14) a. (= Merchant 1996: (36) b., [modified]) * Schuhe sind viele wasserdicht/ nass. shoes are many waterproof IL / wet SL b. (= chapter 2, (19) a., [modified]) Freunde habe ich viele große/ asiatische/ wütende/ hungrige. friends i have I [ many tall IL / Asian IL / angry SL / hungry SL t i ] 12 The linearizational parallel of default EM-order and floating configurations, derivable therefrom, has already been noted by Kniffka (1986: 76f.) at least for the order of weak quantifier and prenominal adjectives. As I hope to have demonstrated here as well as in chapter 2.2 above, this correlation holds for all simple and complex grammatical splittopicalization and QST-constructions, lending additional weight to the structural claims made in chapter one. 308 4 IMPLICATIONS - Optionality & Complex Formations Moreover, observe that such copula constructions cannot be taken to be derived by the same applications of IM under analysis thus far in this subchapter: While copula split-topicalization constructions can be related to linearizations of the form [Q-N]-V-Adj, this is not possible for the split-topicalization constructions discussed above, i.e. those based on a transitive lexical verb; 13 (15) a. and (15) b. below represent the appropriate strings for (14) a. and (14) b. above respectively: (15) a. Viele Schuhe sind wasserdicht. many shoes are waterproof b. * Viele Freunde habe ich große. many friends have I big The last property to discuss (ii) is given with reference to Bayer & Kornfilt (1994). As far as I can tell, the observation does indeed pose a valid distinction of QST and split-topicalization constructions. Parallel to the third property above, it might simply follow from the featural and structural buildup of the nominal domain, e.g. the unique featural interaction between bare HN and a topical feature on C in contrast to the appropriateness of DP/ Poss PRO P with EPP/ OCC. I will not speculate on these matters here but hope to have shown that there is considerably less evidence to retain split-topicalization constructions and QST-configurations in German as unrelated phenomena than has been argued for by Merchant. Concentrating, then, on said parallelism, note that the recent proposals concerning prolific domains are redundant in the context of split-topicalizations, the reason being the default stranding rather than floating of higher phasal heads in co-occurrence with Q W : If a demonstrative or possessive pronoun is introduced in a floating construction, D/ Poss PRO -Q W conjunctions readily strand together in contrast to the default movement of D/ Poss PRO with the nominal core from Q S , in line with the phrasal set-up elaborated in chapter one. Hence, the phase-based restrictions on movement elaborated in chapter 2.2 suffice to capture all licit stranding configurations incorporating weak quantifiers. We thus expect one prolific domain to span over both D/ Poss PRO and Q W (and possibly even further below). What has to be stressed in this context is the parallel D-effect of nominal domains incorporating either complex structural configuration Q S -D or (D-)Q W , already mentioned in the discussion of the former one above, namely the implica- 13 Intransitive predicates leave us with another interesting observation in this respect: The corresponding split-topicalization constructions cluster with copula rather than transitive verb configurations. I will not attempt an explanation here. 4.2 Optionality 309 tion of partitivity, i.e. of further contrasted but unmentioned instances of HN in the discourse. As noted e.g. by Giusti (1991: 452), “[i]ndefinite quantifiers must have a partitive interpretation, universal quantifiers cannot.” Complex stranding configurations apparently pose counter-evidence to this generalization. As noted as early as Kniffka (1986), focusing on simple stranding configurations: Zu indefiniten Nominalphrasen können nur solche Elemente in Distanzstellung treten, die den Umfang einer Teilmenge genauer bestimmen oder sie qualifizieren, d.h. Elemente, die nicht eine Gesamtheit anzeigen. Kniffka (1986: 75) ‘With indefinite noun phrases, only items specifying or qualifying the range of a subset can occur in distant position, i.e. items not signaling totality.’ [translation my own, MB] We can now state that this also applies to complex stranding configurations incorporating multiple (in Kniffka’s sense ‘definite’) SLIs. In the analysis above, partitive interpretations have been traced back to the shared membership of two SLIs in one prolific domain, acting as a semantic unit. Let us follow this line of reasoning. Incorporating the findings concerning the weak quantifier, we thus arrive at (16), with a. the default, b. the marked distribution of SLIs to prolific domains. Observe that I stay agnostic to the nature of said domain itself. (16) (= (12), [modified]) a. [Q S ] > [D/ Poss PRO > Q W ] > [Num > N] b. [Q S > D/ Poss PRO > Q W ] > [Num > N] For Giusti (1991: 451), the absence of canonical QST-constructions with Q W (in contrast to Q S ) is the main reason to refrain from unifying the structures of - what she refers to as - definite and indefinite quantified nominals; contrariwise, I hope to have shown that the identification of non-canonical QST-configurations with strong quantifiers favors a unification of the underlying mechanisms. I will, then, not speculate on these any further but simply note that the system elaborated above accounts for all marked and unmarked simple and complex stranding configurations. Obviously, more research is needed on these points. This subchapter took another look at the optionality of movement from quantified nominal domains in German. As was argued on the basis of theoryinternal premises, semantically vacuous optional movement from below strong quantifiers is born in the Numeration in terms of an optional EPP/ OCC-feature on the quantifier itself forcing IM of its phasal complement to its own specifier position possibly coupled with subsequent raising to T in the course of the ongoing derivation. This option turned out as the only derivational path fully 310 4 IMPLICATIONS - Optionality & Complex Formations incorporating strong cyclicity and obeying EC/ NTC, in line with the overall rationale of the preceding analysis. The proposal was further backed up by the identification of surface configurations arguably incorporating the initial movement inside the nominal domain while dispensing with further instances of IM into the verbal spine, which are not accounted for under a single-movement approach. The analysis above, then, contrasted instances of true (i.e. semantically vacuous) optional movement with stranding configurations that indeed evoke an effect on the (semantic) output. It has been shown that these semantic supplements are readily cancelable, as expected from D-effects that can be traced back to optional EPP/ OCC-features, in line with the initial proposals of this section as well as standard minimalist claims. Likewise, semantic implications in complex stranding constructions seem to fall into this category as well given their parallel characteristics regarding cancelability. Concerning the derivational path argued for in chapter 3.3.2 and refined here, movement in such configurations necessarily proceeds through the specifier of the phase head rather than the quantifier as can be deduced from successful application of Agree in examples like (10) d. above, a point that we will return to in the ensuing subchapter. 14 Within the premises employed in this book, then, optional movement is once again necessarily born in the Numeration, i.e. the optional assignment of EPP/ OCC to the phase head. In the following subchapter, I will take another look at the second realm of true optionality with quantifying elements touched on several times in the course of chapter three above, i.e. overt inflection, to approach the identification of the structural factors enabling it. 4.2.2 Of Inflection The previous section laid out the mechanisms of and the trigger for optional applications of Internal Merge from below quantifying SLIs. As has been demonstrated, optional instances of IM are born in the Numeration, traceable to the presence of an additional feature in the composition of the respective quantifier (or coordinated phase head). The optionality under consideration in this section, viz. of inflection on quantifying LIs, has been analyzed in similar fashion. Even more so, these two domains of optionality have been proposed to be interrelated in a way that they constitute two surface effects of the same underlying 14 Observe that subsequent movement through both specifier positions is barred in these configurations since they are taken to be part of the same prolific domain in accordance with Grohmann’s less restrictive original version of Anti-Locality. 4.2 Optionality 311 mechanism, a rationale that traces back to Shlonsky’s (1991) analysis of QST primarily in Hebrew proposing (i) movement of the extracted nominal SO to proceed through the specifier of the dominating QP, parallel to the approach from the last section (cf. Shlonsky 1991: 164), coupled with (ii) Agree applying in Spec-head rather than head-complement configurations in Hebrew (cf. Shlonsky 1991: 165). Inflection on, and stranding of, the quantifier are hence tightly linked as captured by the biconditional rule of Merchant (1996: 182): “agreement iff overt raising.” As has been laid out in some detail in the last section (as well as in ch. 3.3.3 above), the latter author sets out to derive the paradigm of German stranding configurations and the accompanying restrictions on overt inflectional realization in a modification of said biconditional. He does so by stipulating an “inflectional agreement feature” (Merchant 1996: 183) [F] on the quantifier, a feature-theoretically abnormal entity in that it defies its classification regarding the bipartition of strength to the effect that checking via movement to the corresponding specifier Spec,QP is able to apply either overtly or covertly, with the resulting agreement morphology on the surface visible in either case. Merchant therein relaxes the tight linkage between movement and overt inflection from Shlonsky’s approach to maintain the relation between these surface effects in QST-configurations (overt checking of [F]) while additionally allowing overt inflection without movement (covert checking of [F]). The biconditional is hence loosened to a theorem omitting predictions about the occurrence of agreement morphology on the quantifier with unraised nominal SOs (but still carrying strong predictions about agreement in raising configurations, i.e. ‘overt raising if agreement’). As I have demonstrated elsewhere (cf. Benincasa 2014: ch. 4), Merchant’s modified proposal is, however, too relaxed to account for the full paradigm of quantifier inflection in German since even in structurally uniform nominal domains one finds cases of obligatory overt realization of inflection on Q S , namely with adjectives, numerals and weak quantifiers adjacent to Q S . Based on this observation, I therein divided the nominal domain into two subdomains of inherently [± definite] SLIs in a phrasal set-up roughly corresponding to the one elaborated over the course of chapter one above and employed throughout the analysis. The highest phasal cycle thereof subsumed the definite, the lower domain all indefinite items. Inflection on the quantifier was taken as the overt representation of (what has been termed grammatical) definiteness while the evaluation of morphological form was likewise located at Morphology and therein associated with the possession of (sets of) phonological features, [P] (cf. Nunes 2004). Evaluation was taken to follow the rule in (17) below: (17) (= Benincasa 2014: (16)) Realize [+def] on Q S if the first head of its complement which bears P does not bear [+def] 312 4 IMPLICATIONS - Optionality & Complex Formations With the remaining heads of the highest phasal cycle inherently definite, (17) restricted the obligatory occurrence of inflection in structurally uniform nominal domains to those linearizations laid out above. 15 Even though more successful in accounting for the German data presented thus far than the approach of Merchant (1996), the preceding chapter has unearthed complex nominal domains that challenge the evaluative mechanism above in favor of a strictly derivational view of these concord phenomena. Observe the nominal domains in (18) below, reprinted from chapter three above, in which the strong quantifier heads possessive non-feminine nominal domains. As has been laid out at length above, unsuccessful probing resulted in partial default (i.e. zero-)agreement on the possessive pronoun prior to EM of Q S . What is relevant for the current discussion are the inflectional options of the latter: (18) (= chapter 3, (189)) a. all-ø mein-ø gut-es Bier all my good beer Q S > Poss PRO > Adj > N [φ] [φ] [*[nom][φ*[[sg][n]]]]-[*[nom][φ[[sg][n]]]] b. all-ø mein-ø gut-er Wein all my good wine Q S > Poss PRO > Adj > N [φ] [φ] [*[nom][φ*[[sg][m]]]]-[*[nom][φ[[sg][m]]]] In contrast to the configurations discussed so far in this chapter, inflection on the highest semi-lexical item in (18) is neither optional nor obligatorily present but rather obligatorily absent, i.e. forbidden to occur. This is mysterious under (17) above if the possessive pronoun is taken to be both either inherently definite or indefinite: In the former case, inflection on Q S should be optional, in the latter case, it should be obligatory present. In contrast, as has been laid out throughout the previous chapter, under a derivational view to nominal concord, the data in (18) depict a regular pattern of successive applications of EM and Agree with partial default agreement posing as an intervener for further probing of subsequently externally merged SLIs. With this property accounted 15 Moreover, (17) was conjectured to be capable of deriving the occurrence of strong adjectival inflection from the default weak pattern in the equalization of the former notion with the value [+ definite], cf. Benincasa 2014: fn. 14. 4.2 Optionality 313 for under the premises of the current analysis, the remainder of this section will attempt to likewise derive the division of optional vs. obligatorily present inflectional morphology in the terms of the theory elaborated above for both classes of quantifying SLIs. Let me begin by bringing together all the facts about quantifier inflection gained throughout the course of the previous chapter to subsequently flesh out valid rules therefrom. Among the most obvious facts derivable from the preceding analysis are the specifications of φ-probing of the two classes of quantifying semi-lexical items, i.e. Q S probing for the root node [φ], Q W probing for the subset [φ[f]]. Based on this property, the former seems to be able to bear inflection in all concatenations with further (S)LIs in the nominal domain: i.e. obligatorily overt with bare nominals as well as adjectives, while inflection with possessive pronouns and SLIs of category D seems to be optional in principle. However, due to the aforementioned φ-probing specification, one expects intervention effects to arise with SLIs coordinated below the strong quantifier that bear the partial default value [φ] at the time of EM and subsequent probing of Q S , i.e. the configurations depicted in (18) above. The grammatical outcome of these derivations furthermore demonstrates that the strong quantifier is able to surface with default agreement. Departing from this regular narrow-syntactic accountability in turning to the remaining category of SLIs, concatenations of the two instantiations of the head D 0 with the strong quantifier uncovered two idiosyncratic patterns of optionally realized and obligatorily absent forms. The effect has been subsequently located at the phonological interface as the reflex of a dittological process. Q S was thus taken to always probe for [φ], the various instances of illicit overt realization traced back to either regular theory-internal intervention effects or irregular PF-effects. Contrary to that, the analysis of inflection of the second quantifying SLI firstly uncovered two decided patterns, i.e. a probing and a non-probing instance: Q W is therefore able to uniformly surface without overt inflection. The apocope is moreover facilitated by the occurrence of an adjectival LI coordinated below the quantifier and bearing strong inflection in these cases. This might be attributed to the restriction of overt realization qua visibility, i.e. recoverability of the features suffixially coded on the quantifier in the extended nominal domain (cf. Barbiers 2005 on parallel restrictions in noun ellipsis in English and Dutch). In what follows, I will focus on the probing instance(s) of the weak quantifier. As has been laid out at length in chapter 3.3.4 c. above, this SLI does not yield a uniform inflectional pattern but rather depicts several varieties that nevertheless all conform to the general tenets of inflection in the nominal domain identified above. The basis of all these varieties is found in the mixed pattern of inflection on coordinated adjectives in accordance with the quantifier’s proper- 314 4 IMPLICATIONS - Optionality & Complex Formations ties concerning φ-probing, i.e. [φ[f]]. Therein, cases of successful applications of Agree result in optional rather than obligatorily present inflection in nonstructural cases of the non-feminine paradigms as well as all case forms in the feminine paradigm. Moreover, all genitive and non-feminine dative concatenations as well as all overtly inflected feminine cases display ambiguous combinations of quantifier and adjectival form w.r.t. sequential or gradatial inflection (i.e. Q strong / weak -Adj homonym , Q strong -Adj strong / weak , Q homonym -Adj homonym ). In the same vein, in the rare cases in which the morphological form of SLI inflection diverges from the adjectival one (to wit, genitive masculine and neuter: SLI: es , strong / weak : en , cf. ch. 3.3.4 c. above, Karnowski & Pafel 2004: 169, 172f.), the weak quantifier is able to bear either instance. Above, I argued that these variations serve the purpose of clouding the categorical status of the weak quantifier between semilexical item and adjectival modifier. That this SLI varies between said categories has been furthermore proposed before by Karnowski & Pafel (2004: 174) for the variation between the inflected and uninflected instances, laid out above. However, once the weak quantifier is dominated by a head of the higher phasal cycle, inflection is disambiguated along the dimensions of optional vs. obligatory present and absent occurrence as well as SLI vs. adjectival form to fully conform to the pattern of adjectival inflection brought about by the dominating head, i.e. uniform weak with instances of D, mixed with the possessive pronoun derived by instances of 2 nd cycle Agree across the phasal boundary. Bundling all these observations, Q W optionally inflects in non-structural cases in the non-feminine number/ gender-paradigms masculine and neuter, while inflection is obligatorily absent in these paradigms in structural cases. Turning to the feminine number/ gender-paradigms, optionality of inflection is likewise observed across all cases in the less specified feminine paradigm [f], while surprisingly it is obligatorily present throughout all cases in the plural (and, moreover, unambiguously strong and sequentially inflecting with concatenated adjectives). This last observation has already been made in the previous chapter. There, I refrained from commenting on this deviation from the inflectional restrictions of the quantifier with the two number/ gender-values of NumP. I believe, however, that this insight may lead to a coherent and moreover unified account on the licensing of optional inflection on quantifying SLIs understood as the optional ellipsis of an available φ-specification in line with the SLI’s respective properties of φ-probing likewise elaborated above. The first step, then, might be to ask what differentiates the feminine paradigm of Q W -Num concatenations allowing optional ellipsis from its plural counterpart which does not. The answer can be found in the phasal status of SLIs in the lower cycle: As has been elaborated in chapter 2.3.3 above, the findings from chapter 2.2 on the phasal status of NumP had to be revised in the context of split-topicalization constructions to only 4.2 Optionality 315 apply to maximally specified instances, viz. Num[φ[f[pl]]]. Hence, the lexical domain of plural nominals constitutes a phase, while it does not for feminine (as well as masculine and neuter) HNs. Further following this line of reasoning, said configurations share a phasal cycle with the element under analysis here, i.e. Q W , which is the less prominent phase head of the cycle and should therefore surface as a non-phase in complex concatenations incorporating both phrases; however, phasal Q W heads non-phasal NumP, parallel to non-phasal masculine or neuter nominals, in cases where the latter head is not maximally specified. Returning to the notion under analysis in this section, these cases depict the exact instances in which optional inflection has been identified above (in the limits of the SLI’s own φ-probing properties, i.e. non-structural cases in nonfeminine number/ gender-paradigms as well as all cases in the feminine number/ gender). Optionality of inflection on the weak quantifier hence coincides with its phasal status in the extended nominal domain: Non-phasal Q W in the lower cycle bears obligatory inflection. Taking these insights to the analysis of the second phrase that hosts quantifying SLIs in the nominal domain, i.e. Q S P, it becomes evident that the corresponding contrast of optional vs. obligatory inflection, stated above, is likewise approachable in terms of phasehood; however, the exact implementation diverges from the one for the weak quantifier: Recall that inflection is obligatorily present on the strong quantifier when preceding bare nominals and adjectives, whereas it is optional when occurring with Poss PRO and D (with the resulting patterns themselves disrupted by means of intervention and dittological effects respectively). Recall furthermore that Q S constitutes the least prominent phase head of the highest cycle, thus posing as a phasal head in concatenations with all SLIs of the lower cycle, while surfacing as a non-phase in conjunction with any of the remaining items form the higher one. The partition of nominal SLIs arising therefrom fully conforms to the division just outlined: Phasal Q S bears obligatory inflection, while the non-phasal instance exhibits optionality (again within the limits of the general φ-probing properties of the semi-lexical items involved and their respective interactions). We have hence fully derived the paradigms of optional inflection on quantifying elements from theory-internal notions carved out earlier in the course of the current analysis. This leaves us with the question why these items behave in this narrowly restricted way and even more so in an opposing manner. I believe that a coherent proposal can be formulated on the basis of the observations made above. Recall that omission of inflection on the weak quantifier (qua employment of the non-probing variant) was reported to be facilitated by the occurrence of an accompanying adjective. There, I briefly speculated on the accessibility and recoverability of nominal (φ-)features for the ongoing (i.e. clausal) derivation (cf. 316 4 IMPLICATIONS - Optionality & Complex Formations also ch. 3.3.2 above). Turning to the theoretical model developed in the course of the current analysis, this rationale can be elaborated further: Concentrating first on the strong quantifier, recall that it bears overt inflection only if no other SLI is part of its phasal cycle, i.e. Q S constitutes a phase head. Recall moreover that I argued for the postponement of Spell-Out/ TRANSFER for this highest nominal cycle above for related reasons to apply with the verbal phase it merges with, the DP Inclusion Assumption from ch. 3.3.2. This mechanism does, however, not bar the nominal core to be spelled out prior to EM of the complex nominal domain into the verbal spine in cases in which maximally specified (i.e. plural) Num is present and an intervener additionally separates the phase heads. In this scenario, obligatory inflection on the strong quantifier would guarantee the availability of nominal features to be available at the verbal domain for the ongoing derivation. In contrast, once a more prominent phase head of the higher cycle is present, syntax does not care if Q S bears inflection since said nominal features will be available at its complement. Observe that this argument fortunately also holds for the case of the possessive pronoun, probing only for a subpart of all available φ-values, since phasal Num falls into this range of values, i.e. [pl], the dependent of the possessive’s probing configuration [φ[f]]. The picture differs with the weak quantifier. Recall that inflection of the probing instance of Q W is optional when the SLI constitutes a phase head and is obligatorily present when the quantifier does not pose as a phase. Recall furthermore that it constitutes the least prominent phase head of the lower cycle, thus only acts as such if the maximally specified instance of Num is not part of its complement and it is furthermore not dominated by a phase head of the higher cycle. In the last chapter, the uniform inflection of the quantifier in concatenation with the latter class of SLIs has been attributed to additional applications of Agree; these therefore cloud the workings of obligatory agreement to be elaborated here and can hence be ignored in what follows. Concentrating first on phasal Q W , note that the DP Inclusion Assumption always applies. Note furthermore that, in contrast to the situation described above, there is no lower nominal phasal cycle available. Nominal features are hence always available at the lexical domain, the N-Num complex, for further derivation at the verbal domain. This renders inflection on Q W moot. Turning to the single lower phasal configuration in which Q W is not a phase head due to competition with Num, the weak quantifier acts in parallel to the behavior attributed to Q S above, to wit, rescuing nominal features for the ensuing sentential derivation. It does so since at the point when plural Num merges with Q W , there is no guarantee that the resulting structure will constitute the highest phasal cycle and hence be subject to the DP Inclusion Assumption because non-phasal Q W creates an intervener necessary for two nominal phase heads to co-occur in one complex domain. 4.2 Optionality 317 Obligatory inflection therefore anticipates the possible Spell-Out/ TRANSFER of nominal features. The observations concerning the fitting configurations of the number/ gender-value on the phasal Num[φ[f[pl]]] and the probing [φ[f]] therein carry over from Poss PRO as noted above. The apparently idiosyncratic properties of optional vs. obligatory inflection on both classes of quantifying semi-lexical items have hence been shown to strictly follow under the set of theorems elaborated over the course of the current analysis: Obligatorily present inflection guarantees the availability of nominal features at the ongoing verbal derivation under the DP Inclusion Assumption and w.r.t. the status of contextual phasehood in the nominal domain, while optional inflection constitutes redundant instances of coding. As has been announced in the previous section, the last phenomenon that needs to be addressed in the context of the current subchapter concerns the properties of quantifier inflection in complex stranding configurations. As has been laid out above, movement out of the nominal domain (i) necessarily proceeds through the specifier of the phase head (either the quantifier itself or its complement) and (ii) is motivated by an additional EPP/ OCC-feature. Therein, the relevance of Spec-head configurations for agreement with the stranded SLI as well as the motivation for movement both carry over from the earliest derivational approaches to the phenomenon of QST in Germanic incorporating QP, namely Shlonsky (1991) on (i) and Merchant (1996) on (i) and (ii). As has been further laid out with reference to the latter, the strong quantifier in German exhibits obligatory inflection in QST-configurations, the modified biconditional of Merchant (1996): overt raising if agreement. This rule, however, has to be revised. Note (19) below, depicting instances of complex stranding on the basis of (18) above, i.e. masculine and neuter HNs moving from below the quantifier conjoined with the possessive pronoun: (19) Optionality & Complex Formations I MPLICATIONS 251 a. [Gut-es Bier] i habe ich [all-ø mein-es t i ] getrunken. [ good STRONG beer ] i have I [all mine t i ] d runk Adj N neut Poss PRO [*[acc][φ*[[sg][n]]]] [*[acc][φ[[sg][n]]]] [*[acc][φ*[[sg][n]]]] ‘I drank all my good beer.’ b. [Gut-en Wein] i habe ich [all-ø mein-en t i ] getrunken. [ good STRONG wine ] i have I [ all mine t i ] drunk Adj N masc Poss PRO [*[acc][φ*[[sg][m]]]] [*[acc][φ[[sg][m]]]] [*[acc][φ*[[sg][m]]]] ‘I drank all my good wine.’ Parallel configurations have been presented in chapter 3.3.4 above concatenating the possessive nominal domain with the second item from the highest prolific domain, the focus particle nur . In chapter three, it was also laid out that inflection on the quantifier turns uniform in QST-configurations even with featurally disrupted complements (e.g. Poss PRO with nonfeminine HNs) since 2 nd cycle Agree applies to the extended search space, i.e. Spec,QP. Above, however, the nominal domain-internal landing-site has been identified as the specifier of the phase head. Therefore, no 2 nd cycle Agree applies and Q S necessarily surfaces as reduced to total default (i.e. with obligatorily absent inflection), while inflection is indeed optional with both number/ gender-values coded on Num. Likewise, in complex stranding with Dem, the instance of the category D capable to partake in stranding (cf. the paradigm presented at the end of ch. 2.2.1.2 above), the quantifier can be taken to be always specified for [φ] since no intervention effects emerge in the isolated derivation of the nominal domain; however, the dittological restrictions are still active on the surface and thus the patterns of overt occurrence seem to be disrupted in the familiar fashion. 19 No additional applications of 2 nd cycle Agree probing into the complement or the extended search space are hence stipulated (cf. also fn. 17 above). This section has been concerned with the availability of optional vs. obligatorily present or absent inflection on quantifying elements in the nominal domain. Once again, the basic rationale took redundancy of operations as the source of optionality; coupled with the requirement to make nominal φ-features available at the verbal derivation, the contrasting restrictions concerning obligatorily present vs. optional inflection on the two quantifier phrases received a unified treatment. Subsequently, these findings resurfaced in the discussion of inflection in complex stranding configurations. Therein, the path of movement of the nominal SO from below the complex of multiple SLIs, sketched in the previous section, was held responsible for the absence of 2 nd cycle Agree effects, proposed for simple QST-configurations. Moreover, the rationale from early predecessors of the current analysis has proven correct in the revised terms of the theory elaborated here. The ensuing second part of this chapter, then, gathers a variety of topics, left loose in the course of the preceding analysis and tries to interlace these into the current system by revising them in the discussion of various complex formations in the nominal domain. As will be shown, this strategy will not only result in a coherent picture of said structural configurations but moreover lead to various revisions concerning the findings made in previous chapters. 4.3 Loose Ends & Complex Formations This book has been predominantly concerned with the featural interaction between terminal LIs in their EM-position in the extended nominal domain. Nevertheless, a decompositional rationale underlying the formation of morphologically complex SLIs as well as the derivational approach to fixed syntactic constructions have surfaced on various occasions throughout the analysis. 20 In this subchapter, I want to extend these lines of reasoning in the analysis of further complex nominal constructions as well as SLIs in German. Therein, both the phrasal set-up as well as the interactions between the proposed heads will gain further support in the accountability of syntactic phenomena apart from nominal concord. However, the reader is asked to bear in mind that, due to the diversity of the objects of investigation, the current subchapter can only constitute a first, tentative approach to the various inter-SLI formations and will hence be forced to remain superficial in a number of respects. The organization of this section is as follows: Chapter 0 will start out by shortly revisiting the structural set-up of complex SLI concatenations in light of the insights arrived at in the last chapters, namely the possessor doubling 19 A slight preference for overt inflection on Q S in complex stranding with the demonstrative from masculine (in contrast to neuter) HNs in accusative case has been noted by one of my informants in line with the inclusive paradigm from ch. 3.3.3 above. 20 Cf. ch. 1.4.4, ch 2.3.3 c. on the proposed structural complexity of the monomorphemic ein-words in the possessive and personal pronominal categories, moreover ch. 3.3.3 on the coding of φ-features in the root in inflectionless demonstratives as well as ch. 1.4.3.2, ch. 2.2.1.1 on the formation of possessor doubling constructions. 318 4 IMPLICATIONS - Optionality & Complex Formations Parallel configurations have been presented in chapter 3.3.4 above concatenating the possessive nominal domain with the second item from the highest prolific domain, the focus particle nur . In chapter three, it was also laid out that inflection on the quantifier turns uniform in QST-configurations even with featurally disrupted complements (e.g. Poss PRO with non-feminine HNs) since 2 nd cycle Agree applies to the extended search space, i.e. Spec,QP. Above, however, the nominal domain-internal landing-site has been identified as the specifier of the phase head. Therefore, no 2 nd cycle Agree applies and Q S necessarily surfaces as reduced to total default (i.e. with obligatorily absent inflection), while inflection is indeed optional with both number/ gender-values coded on Num. Likewise, in complex stranding with Dem, the instance of the category D capable to partake in stranding (cf. the paradigm presented at the end of ch. 2.2.1.2 above), the quantifier can be taken to be always specified for [φ] since no intervention effects emerge in the isolated derivation of the nominal domain; however, the dittological restrictions are still active on the surface and thus the patterns of overt occurrence seem to be disrupted in the familiar fashion. 16 No additional applications of 2 nd cycle Agree probing into the complement or the extended search space are hence stipulated (cf. also fn. 14 above). This section has been concerned with the availability of optional vs. obligatorily present or absent inflection on quantifying elements in the nominal domain. Once again, the basic rationale took redundancy of operations as the source of optionality; coupled with the requirement to make nominal φ-features available at the verbal derivation, the contrasting restrictions concerning obligatorily present vs. optional inflection on the two quantifier phrases received a unified treatment. Subsequently, these findings resurfaced in the discussion of inflection in complex stranding configurations. Therein, the path of movement of the nominal SO from below the complex of multiple SLIs, sketched in the previous section, was held responsible for the absence of 2 nd cycle Agree effects, proposed for simple QST-configurations. Moreover, the rationale from early predecessors of the current analysis has proven correct in the revised terms of the theory elaborated here. The ensuing second part of this chapter, then, gathers a variety of topics, left loose in the course of the preceding analysis and tries to interlace these into the current system by revising them in the discussion of various complex formations in the nominal domain. As will be shown, this strategy will not only result in a coherent picture of said structural configurations but moreover lead to various revisions concerning the findings made in previous chapters. 16 A slight preference for overt inflection on Q S in complex stranding with the demonstrative from masculine (in contrast to neuter) HNs in accusative case has been noted by one of my informants in line with the inclusive paradigm from ch. 3.3.3 above. 4.3 Loose Ends & Complex Formations 319 4.3 Loose Ends & Complex Formations This book has been predominantly concerned with the featural interaction between terminal LIs in their EM-position in the extended nominal domain. Nevertheless, a decompositional rationale underlying the formation of morphologically complex SLIs as well as the derivational approach to fixed syntactic constructions have surfaced on various occasions throughout the analysis. 17 In this subchapter, I want to extend these lines of reasoning in the analysis of further complex nominal constructions as well as SLIs in German. Therein, both the phrasal set-up as well as the interactions between the proposed heads will gain further support in the accountability of syntactic phenomena apart from nominal concord. However, the reader is asked to bear in mind that, due to the diversity of the objects of investigation, the current subchapter can only constitute a first, tentative approach to the various inter-SLI formations and will hence be forced to remain superficial in a number of respects. The organization of this section is as follows: Chapter 4.3.1 will start out by shortly revisiting the structural set-up of complex SLI concatenations in light of the insights arrived at in the last chapters, namely the possessor doubling construction (PDC) and simple (i.e. non-PP) partitives, revising and building on the observations from chapter 1.4.3.2 and chapter 1.3.3 respectively. These will be linked together in the proposed movement between categorically related phrases (i.e. Poss-to-Poss and Q-to-Q). Subsequently, I want to carve out various syntactic mechanisms as well as semantic reflexes accompanying said movement. These insights will be incorporated in the decompositional analysis of two complex quantificational SLIs of German, i.e. ‘( ein -) jed -’ as well as ‘ d meist -’ in the remainder of this chapter in section 4.3.2. As will be demonstrated therein, much of the syntactic and semantic idiosyncrasies of these complex configurations follow from the atomic, dedicated LIs involved and the way they are combined along the lines identified beforehand. Moreover, additional implications concerning the underlying structure of simple partitives will be uncovered in the analysis of the former (subsection 4.3.2 a.) while the semantic basis for compositionality of complex SLIs will be approached in focusing on the latter (subsection 4.3.2 b.): As will be argued predominantly on the basis of parallel data from Basque and Greek, the need to semantically restrict the denotation of HN w.r.t. to the context forces idiosyncratic patterns of SLI-occurrences across 17 Cf. ch. 1.4.4, ch. 2.3.3 c. on the proposed structural complexity of the monomorphemic ein-words in the possessive and personal pronominal categories, moreover ch. 3.3.3 on the coding of φ-features in the root in inflectionless demonstratives as well as ch. 1.4.3.2, ch. 2.2.1.1 on the formation of possessor doubling constructions. 320 4 IMPLICATIONS - Optionality & Complex Formations languages. Therein, the semantic motivation for the syntactic buildup of complex quantifying SLIs will be uncovered. 4.3.1 Complex SOs a. Possessor Doubling Constructions I will begin by revisiting complex possessives in German, termed possessor doubling constructions (PDCs) above, which can be found throughout Germanic languages and are of colloquial nature in most of these. 18 As laid out in detail in chapter 1.4.3.2 above, PDCs consist of a genitive- (i.e. possessive case-)marked pronoun, preceded by a possibly complex dative-marked nominal domain denoting the possessor. By analyzing these linearizations structurally as concatenations of Spec and head-positions, I followed standard assumptions above (cf. e.g. Corver 1990: ch. 7.3.3, Roehrs 2005, 2013, moreover Salzmann 2011: ch. 2.2 and references therein). As has been furthermore elaborated in chapter 1.4.4, the incorporated SLI exhibits dual agreement (in isolations as well as PDCs), namely φand case-agreement with the HN by means of suffixation as well as φ-agreement with the (pronounced or unpronounced) possessor via the form of the prefix, which has been taken as an argument for the underlying complexity of the monomorphemic LI itself throughout the current analysis (cf. ch. 3.3.3). Above, proposals concerning the derivation of these possessive complexes have been intentionally kept vague, particularly w.r.t. the assignment of the two non-structural cases.* 22 In chapter one, I moreover hinted at the option of all possessive items to be introduced in postnominal position (i.e. Poss LEX P) followed by obligatory movement in the case of pronominal elements. However, I rejected the approach with reference to those PDCs concatenating two pronominal elements which vary only in their case-values.* 23 Tying all these loose strings from the preceding chapters together, one might arrive at a system categorizing possessive LIs based on their case-value. To commence an analysis along this line of reasoning, I will present the paradigm of linearizations involving pronominal and lexical possessors in the extended nominal domain on the basis of a singular count nominal subsequently followed by listing a few generalizations that can be deduced from these: 18 Cf. Delsing (1998), Norde (1997), Julien (2005: ch. 6.4). 4.3 Loose Ends & Complex Formations 321 (20) a. sein Buch his GEN book b. Martins Buch Martin GEN book c. (dem) Martin sein Buch [( the ) Martin ] DAT his GEN book d. ihm sein Buch him DAT his GEN book e. * sein Buch Martins his book Martin GEN (21) a. Pronominal possessives occur prenominally. a'. Lexical possessives occur preand postnominally. b. Postnominal possessive elements bear genitive (i.e. possessive) case. b'. Prenominal possessive elements in isolation bear genitive (i.e. possessive) case. c. If two possessives co-occur, both have to surface prenominally. c'. If two possessives co-occur, the leftmost bears dative case. From the generalizations above, we can draw further conclusions for the proposed internal structure of possessives once we also take previous considerations into account: The divergence in linearizational properties depicted in (21) a., a'. have already been laid out in chapter 1.4.3.2 above while those in (21) b., b'. generalize over such differences in uniformly associating genitive case with SOs bearing a possessive relation to HN when they occur in isolation. It hence seems reasonable to associate these properties in the concept of a ‘possessive case’ following the proposal made above. In chapter three, I incorporated non-structural cases in the verbal domain (in the sense of Woolford 2006, i.e. genitive and dative in German) into my analysis in the concept of ‘free’ assignment by appropriately specified predicates to nominal domains at EM. With respect to the domain under consideration here, a parallel picture emerges: Case cannot be taken to be assigned in the verbal domain but rather from inside the domain of HN to which the nominal SO denoting the possessor externally merges, as can be deduced from the complex configurations above incorporating genitive case in the absence of a corresponding verbal case-assigner. Transposing Woolford’s typology to the current object of investigation, the proposed possessive (i.e. genitive) case in the nominal domain should be tentatively classified accordingly as 322 4 IMPLICATIONS - Optionality & Complex Formations an inherent case on the basis of the actual value as well as its high predictability of occurrence (cf. Woolford 2006: 112, but cf. Baker 2015: ch. 4.4.1 for arguments that genitive rather poses an unmarked, i.e. default case in the nominal domain, not assigned in the traditional sense). With a non-structural case assigned in the nominal as well as the verbal domain, the working hypothesis obviously takes their means of assignment to proceed along parallel lines. As has been hinted at in chapter one, possessives are widely understood as arguments of the nominal, hence parallel to bearers of verbal θ-roles; another property associated with inherent case by Woolford (2006: (4)) is therein transposed to nominal genitives. Its assignment should therefore likewise apply in the lexical (i.e. lowest) subpart of the extended nominal domain at EM. Focusing on the proposed hierarchy of projections, this is what we find with Poss LEX P in the complement of the head noun, which again parallels non-structural (however lexical, cf. Woolford 2006: (6)/ (15)) case in the verbal domain in being licensed by a lexical projection while this structural position has moreover been identified as the most embedded surface-position of the elements under consideration. Possessive case is thus categorized as non-structural case, i.e. freely assigned at EM. If, then, (i) nominally assigned genitive case is uniformly the marker for possessives irrespective of their surface position while (ii) case-assignment has been taken to apply in a uniform fashion parallel to the verbal domain, and (iii) the lowest surface position of any of these equally case-marked SOs constitutes that exact structural position while the others can be derived therefrom by movement, we have a strong case, I think, to argue in favor of the uniform merging-site for possessives in Poss LEX P as hinted at in chapter one. Possessive pronouns are then also taken to be base-generated in postnominal position even though they have to vacate this structural location in the course of the derivation in contrast to their lexical counterparts. That something along these lines seems to underlie the derivation of prenominal possessive pronouns can be deduced from (22) below. Observe that postnominal lexical possessives naturally conjoin with nominal hierarchies headed by D. In contrast, extended nominal projections incorporating the second-most prominent phase head of the highest nominal cycle are unable to surface both in parallel constructions as well as in co-occurrence with the aforementioned SLI. 19 19 The change in the number of HN from (20) to (22) is conducted to guarantee a uniform line of projection in the sense of chapter three and thus to exclude intervention as the source of ungrammaticality in the concatenation of the non-default agreement SLI Dem and the [φ[f]]-probing item Poss PRO . 4.3 Loose Ends & Complex Formations 323 (22) a. diese guten Bücher Martins these good books Martin GEN b. * seine guten Bücher Martins his good books Martin GEN c. * diese seine guten Bücher Martins these his good books Martin GEN In the premises carved out above, pronominal and lexical possessives stand in complementary distribution since the EM-trace/ copy of the former occupies the overt position of the latter. Furthermore, taking into account the decompositional approach to possessive pronouns (cf. ch. 1.4.4 above), one might propose that it is only a subpart of the complex SLI, namely that denoting the possessor in contrast to the possessee, that is externally merged there: As has been laid out above, possessive pronouns incorporate a decided φ-feature bundle and are inherently possessive case-marked while at the same time also agreeing for all of these features with HN. As has been further illustrated, the means of coding these diverging values neatly split along the lines of the proposed ein-word approach (cf. ch. 1.4.4) with the former coded by the morphological form of the prefix (at least in those case where the ein-root is identifiable on the surface, i.e. the first and second person singular as well as all non-feminine third person possessors, cf. ch. 1.4.4, ch. 3.3.3), while NP-internal agreement is expressed by suffixation parallel to all remaining SLIs on the basis of the root ein -. The high base position of possessive pronouns in contrast to lexical ones, then, would be accounted for in the need of the former (now identified to be a mere prefix) to find a host, i.e. to fuse with an appropriate LI (cf. Roehrs 2009, 2013, in prep.). Moreover, parallels in the surface position of possessive pronouns with other ‘article ein-words’ (in the sense of Roehrs in prep., cf. ch. 1.4.4 above) would receive a uniform treatment together with the dual nature of its featural specification. I have to leave this point for further research here. Having levelled genitive case within the nominal domain and the structural basis of its assignment uniformly with semantic possessivity begs the question concerning the status of the prenominal dative-marked nominal domain co-occurring with possessive pronouns in PDCs. As can be deduced from (22) above, these cannot be taken to be externally merged in the complement of HN; even more so, additional data point to the conclusion that these SOs are not genuinely associated with any of the structural positions related to possessivity. 324 4 IMPLICATIONS - Optionality & Complex Formations (23) a. (dem) Martin seine Bücher [( the ) Martin ] DAT his books b. # diese (dem) Martin seine Bücher (= chapter 2, (16) b., [modified]) these [( the ) Martin ] DAT his books c. alle (dem) Martin seine Bücher (= chapter 2, (10) a., [modified]) all [( the ) Martin ] DAT his books c'. # (dem) Martin all(-e) seine Bücher [( the ) Martin ] DAT all his books d. # (dem) Martin die Bücher 20 [( the ) Martin ] DAT the books (24) a. alle Martins Bücher (= chapter 2, endnote *10, i., [modified]) all Martin GEN books a'. * Martins alle Bücher Martin GEN all books b. diese Martins Bücher these Martin GEN books b'. * Martins diese Bücher Martin GEN these books I will go through these data in detail: (23) a. illustrates the by now well-known configuration of a possibly complex dative-marked nominal domain preceding the genitive possessive pronoun in prenominal position in line with the generalizations (21) c., c'. Structures of the form depicted in (23) b. above already surfaced in chapter 2.2.1 in the analysis of phasehood in the left periphery of the nominal domain. As can be deduced therefrom as well as from the subsequent (23) c., the possessor doubling construction can be preceded by both structurally higher contextual phase heads of the highest nominal cycle, i.e. D 0 and Q S0 . Building on the latter datum, however, (23) c'. demonstrates that the possessive complex does not constitute an exclusive concatenation: There, the dative-marked possessive has switched places with the strong quantifier on the surface which structurally translates to the occupation of its specifier in 20 The existence of the structures depicted in (23) d. has already been observed by Roehrs (2013: ch. 5, cf. also Salzmann 2011: fn. 44 and references therein on parallel observations for Lucerne German). Although acceptable to me, such linearizations are rejected by the majority of my informants, which is why the datum is marked as odd above. 4.3 Loose Ends & Complex Formations 325 the premises of the system outlined above. 21 Consequently, the last datum, the grammatical (23) d., fully leaves out the possessive pronoun in conjoining the dative nominal domain solely with the determiner contrary to the claim made by generalization (21) b'. above. Observe that in this configuration, the dativemarked nominal necessarily precedes the determiner of HN. Likewise, genitive case-marked lexical possessives are able to co-occur with both non-possessive SLIs of the highest phasal cycle as can be deduced from (24) a., b. However, in these configurations, the possessive necessarily follows both D as well as Q S (cf. (24) a'., b'.) analogous to the position of possessive pronouns. Turning to the structural basis underlying these data, the isolated genitivemarked nominal domain in (24) seems to be associated with the same surfaceposition as the possessive pronoun w.r.t. further semi-lexical items in the left periphery, while dative nominal hierarchies occur more freely in specifier positions of the extended nominal domain. We have, then, identified both featural as well as linearizational deviations of dative-marked nominal domains from lexical possessives in German while these SOs nevertheless seem to align w.r.t. their semantic import. If this is indeed the case, the (up to this point covert) linkage between structural position (in contrast to particular lexical items) and semantic contribution, integral to any approach incorporating a (categorically diverse) possessive phrase, has to be abandoned. Recall from chapter one, however, that these structures can be employed to code a variety of relations between two nominals apart from physical possession. It seems, however, that the same variability is not observed with isolated dative-marked prenominal DPs not co-occurring with the possessive pronoun. This discrepancy in grammatical status is even confirmed by speakers generally rejecting dative-marked nominal domains co-occurring left-adjacent of determiners (cf. fn. 20). (25) a. Martins Buch Martin GEN book b. (dem) Martin sein Buch [( the ) Martin ] DAT his book 21 I have marked the datum as odd since it is not accepted by all my informants rather than being marginally grammatical for the majority of them (the same use employed for the datum (16) b. in chapter two and reprinted as (23) b. above). Observe, however, that among those accepting (23) c'., some speakers prefer it over PDCs. Note that contrary to this, parallel linearizations locating the dative left-adjacent to D 0 are uniformly rejected. I will refrain from speculating on these matters here. 326 4 IMPLICATIONS - Optionality & Complex Formations c. # (dem) Martin das Buch [( the ) Martin ] DAT the book a'. Martins {Angst/ Wut} Martin GEN { fear / anger } b'. (dem) Martin seine {Angst/ Wut} 22 [( the ) Martin ] DAT his { fear / anger } c'. * (dem) Martin die {Angst/ Wut} [( the ) Martin ] DAT the { fear / anger } The dative DP hence seems to lack the semantic variability of possessives. We are therefore able to dissociate the genitive-marked from the dative nominal domain in that the former bears a thematic relation to the denotation of HN structurally resonated by its association with the appropriate position, while the latter does not. I will refrain from speculating on both the structural origin of the dative-marked nominal domain as well as the mechanisms underlying its case-assignment in what follows but conclude this section with a recapitulation of the proposed derivational steps involving ‘true’ (i.e. genitive-marked) possessive SLIs and SOs, instead. As has been argued for above, the composition of extended nominal domains incorporating prenominal pronominal and lexical possessives always involve an instance of category-internal IM, i.e. head-tohead movement in the case of pronouns, XP-to-Spec movement with genitive nominal domains; as has been furthermore speculated, the obligatory nature of the former application can be accounted for under a decompositional approach to possessive pronouns as an underlyingly complex configuration consisting of a possessive prefix and the root ein , while the latter is morphologically independent and hence not forced to attach to a host. Therefore, lexical possessives are not forced to leave their EM-site and may also occur postnominally. Moreover, the linkage between structural position and semantic import has been maintained since nominal domains deviating in case-value also fail to encode the broad variability of thematic relations standardly expressed by possessive elements. We thus observe uniform paths of derivation of semantically related constructions by means of movement between categorically related phrases. In the remainder of this section I will elaborate on these properties and mechanisms of SLI concatenation in the nominal domain in the revision of simple partitives and extract a number of characteristics of complex SOs in the nominal domain from 22 Data of the form (25) b'. are due to Salzmann (2011: 180) who employs them to provide a related yet different argument. 4.3 Loose Ends & Complex Formations 327 parallels with the foregone analysis before turning to multi-SLI concatenations in complex quantifying items in the remainder of chapter 4.3. b. Simple Partitive Constructions As has been laid out in chapter 1.3.3 above, simple partitive constructions are made up of the concatenation of a weak quantifier left-adjacent to a nominal domain marked with genitive case. The latter can be read off the morphological form of the SLI dominating it, either a possessive pronoun or an instantiation of the category D, i.e. the determiner or demonstrative. We therefore observe a deviation from the default SLI-order D/ Poss PRO -Q W in these conjunctions; hence, once again, variation in featural as well as linearizational properties of SLIs are accompanied by modified semantic import. The data in (26) illustrate these points with nominal domains on the basis of a plural HN. (26) a. {die/ dies-e/ mein-e} viel-en Bücher [{ the / these / my } many weak books ] NOM b. * {der/ dies-er/ mein-er} viel-en Bücher [{ the / these / my } many weak books ] GEN c. viel-e {der/ dies-er/ mein-er} Bücher many strong .NOM [{ the / these / my } books ] GEN d. * viel-e {die/ dies-e/ mein-e} Bücher many strong .NOM [{ the / these / my } books ] NOM The datum (26) a. depicts the default case of D/ Poss PRO -Q W concatenation with the SLIs of the higher phasal cycle preceding the weak quantifier in structural (i.e. nominative) case. In line with the observations made in the last chapter, Q W displays weak inflection in these concatenations signaling the application of Adjectival Impoverishment by the aforementioned SLIs in these featurally uniform configurations. In (26) b., the case of the complete extended nominal domain has been switched to genitive, which underlies simple partitives. The datum is, however, not grammatical as it stands. Observe that the nominal domain in (26) b. parallels the subpart of partitive structures depicting the full and exhaustive set from which the partitive set is extracted (the ‘ of the N’-subpart of the English counterpart, cf. ch. 1.3.3 above) and hence turns grammatical when prefixed with e.g. the strong quantifier. The reading of (26) c. is different in that semantically, the set from which the quantifier selects does itself denote the exhaustive set of the/ my books and a subpart is taken therefrom; henceforth the exhaustive first set will be termed the ‘range’, as opposed to the one denoted by the quantifier, the ‘variable’, following C&G (p. 64). Observe that in this case as 328 4 IMPLICATIONS - Optionality & Complex Formations well as in the aforementioned grammatical variant of (26) b. with prefixed Q S , the variable does not bear the same case as the range: While the latter exhibits genitive-marking, the former is structurally case-marked with nominative and moreover adheres to the case-assignment of the predicate it is externally merged with, which can be deduced from the concatenation of partitive nominal hierarchies with non-structural case-assigning predicates. (27) Wir vertrauen viel-en der Bücher. we trust many DAT [ the books ] GEN ‘We trust many of the books.’ The final datum of (26), then, demonstrates that this order is restricted to the specific configuration of case-assignment in that uniformly assigned structural case results in ungrammaticality. In parallel, uniformly assigned dative case in linearizations of the form depicted in (27) above likewise turns out ungrammatical. We can therefore tentatively conclude that simple partitives involve (i) a weak quantifier in non-canonical position on the surface and (ii) a genitivemarked nominal domain to the complexity of the highest contextual phasal cycle, i.e. DP/ Poss PRO P. C&G, reviewing analyses of quantified nominal domains across various stages of the generative enterprise, predominantly discuss complex partitive constructions (i.e. prepositional or clitic). As they note, beginning with Dean (1966), as reported in Jackendoff (1968: 430-432, cf. ch. 1.4.1 above), partitives, as well as simple quantified nominal domains under the ‘Hidden Partitive Hypothesis’ (cf. C&G: ch. 2.1.1), have been analyzed as underlyingly consisting of two nominal domains depicting variable (i.e. N coordinated under Q or D) and range (N coordinated under P) respectively (cf. C&G: ch. 2.3.1-2.3.2, ch. 3, moreover ch. 2.1.3). The nominal cores of these domains were taken to be subject to a ‘lexical nondistinctness requirement’ (with possible variation in their functional structure, i.e. in φand case-values, cf. C&G: ch. 3.3.4, furthermore ch. 3.1, 3.2). Moreover, the semantic relation of the domains was reflected structurally in that the range had been located in the complement of the variable. The basic rationale of these approaches is exemplified with Dean’s (1966) original proposal in (28) below, reprinted from Jackendoff (1968). 4.3 Loose Ends & Complex Formations 329 (28) (= Jackendoff 1968: (66)) Optionality & Complex Formations I MPLICATIONS 257 C&G, reviewing analyses of quantified nominal domains across various stages of the generative enterprise, predominantly discuss complex partitive constructions (i.e. prepositional or clitic). As they note, beginning with Dean (1966), as reported in Jackendoff (1968: 430-432, cf. ch. 1.4.1 above), partitives, as well as simple quantified nominal domains under the ‘Hidden Partitive Hypothesis’ (cf. C&G: ch. 2.1.1), have been analyzed as underlyingly consisting of two nominal domains depicting variable (i.e. N coordinated under Q or D) and range (N coordinated under P) respectively (cf. C&G: ch. 2.3.1-2.3.2, ch. 3, moreover ch. 2.1.3). The nominal cores of these domains were taken to be subject to a ‘lexical non-distinctness requirement’ (with possible variation in their functional structure, i.e. in φand case-values, cf. C&G: ch. 3.3.4, furthermore ch. 3.1, 3.2). Moreover, the semantic relation of the domains was reflected structurally in that the range had been located in the complement of the variable. The basic rationale of these approaches is exemplified with Dean’s (1966) original proposal in 0 below, reprinted from Jackendoff (1968). (28) (= Jackendoff 1968: (66)) some of the men: NP 1 9 Det 1 N 1 PP ! ! 1 some men of NP 2 3 Det 2 N 2 ! ! the men However, the articulated set-up of the nominal domain elaborated up until this point seems to provide options for a structurally more economic analysis of partitive constructions that maintains the semantic import of these earlier approaches. With respect to (i) above, then, taking the insights previously arrived at in this section into account, the null hypothesis should take the inversion of weak quantifier and determiner/ demonstrative or possessive pronoun to result from movement of the former. Since the system argued for above provides two decided phrases for quantifying material, a parallel proposal to the categorically related PossP-internal movement readily suggest itself in head-to-head movement of the partitive quantifier from Q W0 to Q S0 skipping the aforementioned projections within a single extended nominal projection. As has been noted by Lyons (1999: ch. 3.5), the position of strong quantifiers can structurally be viewed as “outside the complete noun phrase, possibly […] linked to it by a partitive construction” (Lyons 1999: 148), uniting the two aforementioned notions. Though Lyons only cites partitives on the basis of the strong quantifiers in this context (i.e. ‘ all ( of ) the N’ and ‘ most of the N’), generalizing from there to weakly quantified, i.e. true partitives seems straightforward. A first argument in favor of this proposal can be found in the inability to form a partitive construction with the range (the ‘ of the N’-subpart) introduced by the strong quantifier all -, the least-prominent phase head of the highest cycle, even though this set-theoretic selection does not seem to carry semantic oddness. 28 (29) * viel-e all-er Bücher many NOM [ all books ] GEN Recall that the analysis of possessive phrases touched upon the proposal of a linkage between semantic import and structural positions. Investigating further along the ideas just outlined, one therefore expects partitive quantifiers to share syntactic and semantic characteristics with the strong quantifier that reach beyond linearization (cf. Musan 1999). Indeed, Etxeberria (2009: ch. 4.4.2.2), analyzing strategies of contextual domain restriction in Basque nominal projections draws parallels between the two in the classification of partitives as “strongly interpreted weak quantifiers” (Etxeberria 2009: 28 As one reviewer remarks, grammaticality of the construction under consideration dramatically improves with the addition of a postnominal relative clause and concludes that an analysis in terms of complementation cannot be excluded. I do not see a conflict between this statement and my argument presented in the main text, however. Building e.g. on the structural proposals put forward by Kayne (1994: ch. 8.2), who analyzes relative clauses as DP-projections with CPcomplements, the strong quantifier, constituting a semantic unit together with the HN and relative clause corresponding to the range, would consequentially be taken to be base-generated along with these items inside the coordinated CP and to be subsequently fronted together with the nominal head. The weak quantifier corresponding to the variable, on the other hand, would be taken to be generated in the dominating functional layer. This structural basis would hardly be proposed for parallel linearizations lacking a relative clause (like the datum presented in the main text) under the premises of the current theory. The fact that two distinct structural foundations are thus put forward for partitives incorporating a universally quantified range with and without an attached relative clause does not trivialize my analysis. Rather, my claims are strengthened in the accompanying shift in grammaticality between the two constructions. However, the articulated set-up of the nominal domain elaborated up until this point seems to provide options for a structurally more economic analysis of partitive constructions that maintains the semantic import of these earlier approaches. With respect to (i) above, then, taking the insights previously arrived at in this section into account, the null hypothesis should take the inversion of weak quantifier and determiner/ demonstrative or possessive pronoun to result from movement of the former. Since the system argued for above provides two decided phrases for quantifying material, a parallel proposal to the categorically related PossP-internal movement readily suggest itself in head-to-head movement of the partitive quantifier from Q W0 to Q S0 skipping the aforementioned projections within a single extended nominal projection. As has been noted by Lyons (1999: ch. 3.5), the position of strong quantifiers can structurally be viewed as “outside the complete noun phrase, possibly […] linked to it by a partitive construction” (Lyons 1999: 148), uniting the two aforementioned notions. Though Lyons only cites partitives on the basis of the strong quantifiers in this context (i.e. ‘ all ( of ) the N’ and ‘ most of the N’), generalizing from there to weakly quantified, i.e. true partitives seems straightforward. A first argument in favor of this proposal can be found in the inability to form a partitive construction with the range (the ‘ of the N’-subpart) introduced by the strong quantifier all -, the least-prominent phase head of the highest cycle, even though this settheoretic selection does not seem to carry semantic oddness. 23 23 As one reviewer remarks, grammaticality of the construction under consideration dramatically improves with the addition of a postnominal relative clause and concludes that an analysis in terms of complementation cannot be excluded. I do not see a conflict between this statement and my argument presented in the main text, however. Building e.g. on the structural proposals put forward by Kayne (1994: ch. 8.2), who analyzes relative clauses as DP-projections with CP-complements, the strong quantifier, constituting a semantic unit together with the HN and relative clause corresponding to the range, would consequentially be taken to be base-generated along with these items inside the coordinated CP and to be subsequently fronted together with the nominal head. The 330 4 IMPLICATIONS - Optionality & Complex Formations (29) * viel-e all-er Bücher many NOM [ all books ] GEN Recall that the analysis of possessive phrases touched upon the proposal of a linkage between semantic import and structural positions. Investigating further along the ideas just outlined, one therefore expects partitive quantifiers to share syntactic and semantic characteristics with the strong quantifier that reach beyond linearization (cf. Musan 1999). Indeed, Etxeberria (2009: ch. 4.4.2.2), analyzing strategies of contextual domain restriction in Basque nominal projections draws parallels between the two in the classification of partitives as “strongly interpreted weak quantifiers” (Etxeberria 2009: 98). 24 In the light of these observations, the syntactic patterning of partitive nominal hierarchies with strongly quantified ones, laid out in chapter 1.3.3 above, is expected due to their parallel structural composition. The relevant environments are compiled below: (30) a. Definiteness Restriction (= chapter 1, (23) b., [modified]) * Es gibt [{viele/ wenige} [der Bäume]] im Garten. it is [{ many / few } [ the trees ] GEN ] in.the garden ‘There are [{many/ few} of the trees] in the garden.’ b. Preposed Adjective Structures (= chapter 1, (25) b., [modified]) Betrunken wie [viele [der Männer]] waren, sprachen wir ungern mit ihnen. drunk as [ many [ the men ] GEN ] were, talked we reluctantly with them ‘Drunk as many of the men were, we only talked reluctantly with them.’ weak quantifier corresponding to the variable, on the other hand, would be taken to be generated in the dominating functional layer. This structural basis would hardly be proposed for parallel linearizations lacking a relative clause (like the datum presented in the main text) under the premises of the current theory. The fact that two distinct structural foundations are thus put forward for partitives incorporating a universally quantified range with and without an attached relative clause does not trivialize my analysis. Rather, my claims are strengthened in the accompanying shift in grammaticality between the two constructions. 24 Etxeberria furthermore notes that the restriction of these quantifying elements to cooccur with DP holds across a variety of languages for reasons of contextual domain restriction. I will get back to these strategies and parallel phenomena in German in the course of chapter 4.3.2 b. below. 4.3 Loose Ends & Complex Formations 331 (31) a. Definiteness Restriction * Es gibt [alle Bäume] im Garten. it is [ all trees ] in.the garden ‘There are [all trees] in the garden.’ b. Preposed Adjective Structures* 23 Betrunken wie [alle Männer] waren, sprachen wir ungern mit ihnen. drunk as [ all men ] were, talked we reluctantly with them ‘Drunk as all of the men were, we only talked reluctantly with them.’ We can hence attribute at least some of the linearizational idiosyncrasies of strong quantifiers to properties associated with its structural position rather than the semi-lexical head itself since other SLIs, which are located elsewhere in the extended nominal domain by default and do not display such characteristics there do so when moved into the position under consideration. It is worth highlighting this point which was already hinted at in chapter one, i.e. a structural foundation for the linkage between deviating linearization and semantic import: Above, movement between categorically related phrases connects variation in semantic and syntactic behavior with the structural positions, occupied therein, rather than the moved LIs. Recalling the typology of quantifiers and quantity adjectives from C&G, laid out in chapter 1.4.1 above, their proposal of the variable categorical status of quantifying elements accompanied by the variation in syntactic behavior thus also applies language-internally via Internal Merge (cf. C&G: 45). Focusing on property (ii), then, recall that at the beginning of this section, genitive case, not assigned by a respective predicate, was classified as the casevalue fully associated with possessivity, which in turn was dissociated from dative-marked SOs in the nominal domain. In the light of the complex configuration under consideration here, we seem forced do dilute this notion in that the case-value must itself be dissociated from exclusively coding said semantic import. However, I do not think that the system elaborated above is disrupted by this move since we are still able to maintain the association between casemarking and semantic contribution. To demonstrate this, recall that I argued for the complement of N 0 as the EM-site for all possessives above to be uniformly marked with genitive in the derivation of the nominal domain. There, I also identified HN as the assigner of said case-value on the basis of the parallelism between verbal arguments and nominal possessives. If we return to extended nominal hierarchies incorporating complex possessive SOs in various verbal 332 4 IMPLICATIONS - Optionality & Complex Formations case environments, we can confirm that this proposal makes correct predictions concerning the case-value of the projection line of HN: All LIs dominating the complement of the noun uniformly display the case assigned by the verbal head as is demonstrated with various instances in (32) a.-c. below. (32) a. Ich mag all(-e) dies-e Bücher des Lehrstuhls für Linguistik. I like [ all these books ] ACC [[ of.the chair ] GEN in linguistics ] b. Ich vertraue all(-en) dies-en Bücher-n des Lehrstuhls für Linguistik. I trust [ all these books ] DAT [[ of.the chair ] GEN in linguistics ] c. Ich gedenke all(-er) dies-er Bücher des Lehrstuhls für Linguistik. I commemorate [ all these books ] GEN [[ of.the chair ] GEN in linguistics ] However, turning to the domain of simple partitives, these properties of assignment are altered. Below, I eliminated the postnominal lexical possessive in favor of the possessive pronoun in prenominal position in the examples (33) a.-c. (cf. from (22) b., c. above that lexical and pronominal (genitive) possessives are in complementary distribution). Hence, the nominal complements in the linearizations below exhibit a non-possessive reading and bear accusative case assigned by the preposition, parallel to the more embedded complements of the possessor in (32) above. The datum (33) d. demonstrates the pattern of case-assignment in nominal domains incorporating a postnominal, i.e. lexical, possessive. There, I have marked the full extended projection of HN including the nominal core as bearing genitive case even though no morpho-syntactic marking of case can be found on the latter (cf. ch. 2.3.3 b.). Nevertheless, observe that in contrast to (32) b. above, no dative marking is found on HN, which indicates an alternation in its case-value parallel to the SLIs accompanying it. (33) a. Ich mag viel-e dies-er mein-er Bücher über Linguistik. I like many ACC [[ these my books ] GEN about linguistics ] b. Ich vertraue viel-en dies-er mein-er Bücher über Linguistik. I trust many DAT [[ these my books ] GEN about linguistics ] c. Ich gedenke viel-er dies-er mein-er Bücher über Linguistik. I commemorate many GEN [[ these my books ] GEN about linguistics ] d. Ich mag viel-e dies-er Bücher des Lehrstuhls für Linguistik. I like many ACC [[ these books ] GEN [[ of.the chair ] GEN in linguistics ]] Various conclusions can be drawn from this contrast, the most obvious being the variation in the size of the genitive-marked subpart of the extended nominal 4.3 Loose Ends & Complex Formations 333 domain between the two constructions, i.e. left-adjacent to the nominal head and to the partitive quantifier respectively, which themselves inflect according to the dominating case-assigners’ demands. Generalizing from the mechanisms of case-assignment in the nominal domain identified earlier in this section, i.e. the assignment of case by heads to complements, we can identify a split between the two genitive structures, which are schematically given on the basis of (32) a.-c. as well as the complex configuration (33) d. employing three applications of case-assignment (to wit, one verbal and two nominal) in (34) a. and b. respectively. Observe that I abstract away from the intricacies of verbal (non-) structural case-assignment here to illustrate the parallels between verbal and nominal domains in this respect; cf. ch. 3.3.2 above. (34) Optionality & Complex Formations I MPLICATIONS 260 (33) a. Ich mag viel-e dies-er mein-er Bücher über Linguistik. I like many ACC [[ these my books ] GEN about linguistics ] b. Ich vertraue viel-en dies-er mein-er Bücher über Linguistik. I trust many DAT [[ these my books ] GEN about linguistics ] c. Ich gedenke viel-er dies-er mein-er Bücher über Linguistik. I commemorate many GEN [[ these my books ] GEN about linguistics ] d. Ich mag viel-e dies-er Bücher des Lehrstuhls für Linguistik. I like many ACC [[ these books ] GEN [[ of.the chair ] GEN in linguistics ] Various conclusions can be drawn from this contrast, the most obvious being the variation in the size of the genitivemarked subpart of the extended nominal domain between the two constructions, i.e. left-adjacent to the nominal head and to the partitive quantifier respectively, which themselves inflect according to the dominating case-assigners’ demands. Generalizing from the mechanisms of case-assignment in the nominal domain identified earlier in this section, i.e. the assignment of case by heads to complements, we can identify a split between the two genitive structures, which are schematically given on the basis of 0 a.-c. as well as the complex configuration 0 d. employing three applications of caseassignment (to wit, one verbal and two nominal) in 0 a. and b. respectively. Observe that I abstract away from the intricacies of verbal (non-)structural case-assignment here to illustrate the parallels between verbal and nominal domains in this respect; cf. ch. 3.3.2 above. (34) a. V > [Q S > D > N] > [D > N] [ val ] [ val ] [ val ] [ val ] [ val ] b. V > Q S > [D > N] > [D > N] [ val ] [ val ] [ val ] [ val ] [ val ] Hence, even though it is the same case-value that is assigned in the two constructions, the semantic import is guided by the assigning head with genitive assigned by N 0 resulting in possessive, genitive assigned by Q S0 in partitive constructions (i.e. possessive case and partitive case; cf. also C&G: ch. 3.3.1). In what follows, I will bring these mechanisms together in a derivational sketch of simple partitives in German to refine the proposals just outlined above. Starting out with the nominal core, either Num 0 or Q W0 merges in the case of plurals/ feminine mass nouns and non-feminine mass nominals respectively. In the case of the former, Q W0 is then subsequently merged with the complex nominal domain levelling the resulting structures w.r.t. the highest phrase. Concentrating on the property of phasehood in these nominal hierarchies, recall that all of these configurations will ultimately have to be headed by Q S0 to enable the formation of the simple partitive construction by movement of the weak quantifier there. Bearing this in mind, and starting out with non-feminine nominal hierarchies, observe that Q W will constitute a part of the interior of the single nominal phase with D 0 as well as Poss PRO0 externally merging with the nominal hierarchy either in isolation or succession resulting in a phase-plus-non-phase configuration in the former cases (P-N, i.e. D/ Poss PRO -Q W ) and in the concatenation of the phase head with two consecutive non-phasal items with the combination of both in the latter (P-N-N, i.e. D-Poss PRO -Q W ). Nominal domains incorporating NumP moreover have to be split up again into those bearing either number/ gendervalue [f] or [pl] which accords to the phasal status of this head (cf. ch. 2.3.3 above). Concerning the former, the weak quantifier is located at the interior of the nominal phase, part of the non-phase (N-N) concatenation Q w -Num, once one of the three highest SLIs successively merges in isolation. In the case of subsequent EM of both Poss PRO and D it is moreover part of an even larger non-phasal concatenation additionally also including the possessive phrase, i.e. Poss PRO -Q w -Num. Concerning the latter instantiation of Num, bearing [pl], Q W is taken to be located outside the lower nominal phase and is therefore sandwiched between two phasal heads. Hence, its position is pinpointed at the interior of the upper phase when any of the higher SLIs externally merge in isolation while again subsequent EM of Poss PRO as well as D results in the concatenation of two non-phases dominated by the rigid phase head (i.e. Poss PRO > Q w , cf. the discussion surrounding ch. 2.2.1, (25) c. above). The discussion above thus illustrates that the two quantifier phrases are at most separated by a single phasal boundary in all possible SLI concatenations. Therefore, when partitive Q S externally merges with any of these configurations it is uniformly able to attract the weak quantifier to its head position (in line with the weak version of the PIC) and subsequently assign case to the (S)LIs in its complement before the complex nominal SO externally merges into the verbal Hence, even though it is the same case-value that is assigned in the two constructions, the semantic import is guided by the assigning head with genitive assigned by N 0 resulting in possessive, genitive assigned by Q S0 in partitive constructions (i.e. possessive case and partitive case; cf. also C&G: ch. 3.3.1). In what follows, I will bring these mechanisms together in a derivational sketch of simple partitives in German to refine the proposals just outlined above. Starting out with the nominal core, either Num 0 or Q W0 merges in the case of plurals/ feminine mass nouns and non-feminine mass nominals respectively. In the case of the former, Q W0 is then subsequently merged with the complex nominal domain levelling the resulting structures w.r.t. the highest phrase. Concentrating on the property of phasehood in these nominal hierarchies, recall that all of these configurations will ultimately have to be headed by Q S0 to enable the formation of the simple partitive construction by movement of the weak quantifier there. Bearing this in mind, and starting out with non-feminine nominal hierarchies, observe that Q W will constitute a part of the interior of the single nominal phase with D 0 as well as Poss PRO0 externally merging with the nominal hierarchy either in isolation or succession resulting in a phase-plus-non-phase configuration in the former cases (P-N, i.e. D/ Poss PRO -Q W ) and in the concatena- 334 4 IMPLICATIONS - Optionality & Complex Formations tion of the phase head with two consecutive non-phasal items with the combination of both in the latter (P-N-N, i.e. D-Poss PRO -Q W ). Nominal domains incorporating NumP moreover have to be split up again into those bearing either number/ gender-value [f] or [pl] which accords to the phasal status of this head (cf. ch. 2.3.3 above). Concerning the former, the weak quantifier is located at the interior of the nominal phase, part of the non-phase (N-N) concatenation Q W -Num, once one of the three highest SLIs successively merges in isolation. In the case of subsequent EM of both Poss PRO and D it is moreover part of an even larger non-phasal concatenation additionally also including the possessive phrase, i.e. Poss PRO -Q W -Num. Concerning the latter instantiation of Num, bearing [pl], Q W is taken to be located outside the lower nominal phase and is therefore sandwiched between two phasal heads. Hence, its position is pinpointed at the interior of the upper phase when any of the higher SLIs externally merge in isolation while again subsequent EM of Poss PRO as well as D results in the concatenation of two non-phases dominated by the rigid phase head (i.e. Poss PRO > Q W , cf. the discussion surrounding ch. 2.2.1, (25) c. above). The discussion above thus illustrates that the two quantifier phrases are at most separated by a single phasal boundary in all possible SLI concatenations. Therefore, when partitive Q S externally merges with any of these configurations it is uniformly able to attract the weak quantifier to its head position (in line with the weak version of the PIC) and subsequently assign case to the (S)LIs in its complement before the complex nominal SO externally merges into the verbal domain. Case-assignment from Q S , however, cannot reach into the complement of HN for two reasons: (i) structurally, and taking the nominal complement to constitute a phase for obvious reasons, because once Q S of the main projection line is externally merged the partitive noun has already been shipped to the interfaces and furthermore (ii) concerning timing, since HN (as well as P 0 in non-possessive complements as in (33) a.-c. above) has been in an appropriate structural configuration with the partitive noun and its respective extended projection before EM of Q S and thus already assigned case to its complement as has been worked out above. The system outlined here is hence able to derive the uniform patterns of partitive case-marking of SLIs in the extended nominal domain down to the nominal compliment with all (NumPas well as NP-internal) number/ gender-values within the derivational model elaborated in previous chapters. However, an open question regarding the mechanisms just outlined concerns the case-feature on the nominal head itself. Observe that in (34) b., I have marked HN to constitute a part of the case-feature chain headed by the partitive quantifier, whereas the derivational paths sketched above predict the nominal 4.3 Loose Ends & Complex Formations 335 head not to be part of the derivation at the time when partitive case is taken to be assigned by Q S in those configurations which incorporate plural (qua phasal) Num 0 together with one or both of the two more prominent contextual phase heads of the highest cycle, viz. Poss PRO and/ or D. 25 This is so since in these configurations, there are two phase-boundaries separating the assigner and the assignee, hence the weak PIC predicts N 0 to have already undergone TRANSFER/ Spell-Out. This state of affairs is schematically illustrated in (35) below with the dotted lines representing the boundaries of phasal interior to be spelled out with EM of a higher phase head. (35) Optionality & Complex Formations I MPLICATIONS 261 domain. Case-assignment from Q S , however, cannot reach into the complement of HN for two reasons: (i) structurally, and taking the nominal complement to constitute a phase for obvious reasons, because once Q S of the main projection line is externally merged the partitive noun has already been shipped to the interfaces and furthermore (ii) concerning timing, since HN (as well as P 0 in non-possessive complements as in 0 a.-c. above) has been in an appropriate structural configuration with the partitive noun and its respective extended projection before EM of Q S and thus already assigned case to its complement as has been worked out above. The system outlined here is hence able to derive the uniform patterns of partitive case-marking of SLIs in the extended nominal domain down to the nominal compliment with all (NumPas well as NP-internal) number/ gender-values within the derivational model elaborated in previous chapters. However, an open question regarding the mechanisms just outlined concerns the case-feature on the nominal head itself. Observe that in 0 b., I have marked HN to constitute a part of the case-feature chain headed by the partitive quantifier, whereas the derivational paths sketched above predict the nominal head not to be part of the derivation at the time when partitive case is taken to be assigned by Q S in those configurations which incorporate plural (qua phasal) Num 0 together with one or both of the two more prominent contextual phase heads of the highest cycle, viz. Poss PRO and/ or D. 31 This is so since in these configurations, there are two phase-boundaries separating the assigner and the assignee, hence the weak PIC predicts N 0 to have already undergone TRANSFER/ Spell-Out. This state of affairs is schematically illustrated in 0 below with the dotted lines representing the boundaries of phasal interior to be spelled out with EM of a higher phase head. (35) Q S > D/ Poss PRO > Q W > Num > N [ val ] [ val ] [ val ]  [ val ] The problem might at first appear to be only apparent since nominal heads coordinated under Num do not exhibit a morpho-syntactic reflex of genitive case on the surface in contrast to SLIs. Therefore, the interior of the lower nominal phase might simply be taken not to be specified for case and hence reduced to default. However, the situation sketched above is related to another more problematic configuration. Recall from the presentation of concord with non-structural case-assigning predicates above that plural nominals bear a morpho-syntactic reflex of dative case, i.e. the suffix / n/ following plural inflection. This is also observable on heads partaking in the formation of the complex domains just outlined and thus taken to be already spelled out when case is assigned by the dominating predicate. The relevant example from the foregone discussion is repeated here for ease of exposition as 0 a., a schematic illustration of the problematic configurations based on a predicate’s non-structural case-assignment of predicates to complex nominal domains parallel to 0 above is given in 0 b. (36) a. Ich vertraue all(-en) dies-en Bücher-n. (= (32) b., [modified]) I trust [ all these books ] DAT b. V > D/ Poss PRO > Q W > Num > N [ val ] [ val ] [ val ]  [ val ] In the light of these parallels, it seems pressing to uniformly specify HN as bearing case even in environments where no morpho-syntactic reflex is observable. How, then, can the system under elaboration be modified to counter this apparent flaw? I want to argue that it is already capable to do so but that we are only now in a position to further specify the relevant notion involved: Recall from chapter 2.3.3 that I followed Ritter (1993) and Bernstein (1991) in the proposal of Nto-Num movement. There, I stayed agnostic as to the derivational timing of application of the operation, to be located either at Narrow Syntax or post-syntactic Morphology (viz. IM or m-merger respectively). The data above can be consulted to finally dissolve this issue in that only when N moves to Num in Narrow Syntax can all morpho-syntactic featural demands of the nominal head be met in that it is still part of the derivation when case is assigned to its domain. Observe that this conclusion in turn bears further consequences for the theory of Anti-Locality as modified in chapter 3.3.2 above: 31 Recall that in the system elaborated above, only one phasal boundary is taken to exist if either D or Poss PRO surfaces in isolation with NumP in either of its featural specifications. However, since Q W is necessarily also taken to be part of the phrasal set-up underlying partitives, the specification of co-occurrences in the main text suffices. The problem might at first appear to be only apparent since nominal heads coordinated under Num do not exhibit a morpho-syntactic reflex of genitive case on the surface in contrast to SLIs. Therefore, the interior of the lower nominal phase might simply be taken not to be specified for case and hence reduced to default. However, the situation sketched above is related to another more problematic configuration. Recall from the presentation of concord with non-structural caseassigning predicates above that plural nominals bear a morpho-syntactic reflex of dative case, i.e. the suffix / n/ following plural inflection. This is also observable on heads partaking in the formation of the complex domains just outlined and thus taken to be already spelled out when case is assigned by the dominating predicate. The relevant example from the foregone discussion is repeated here for ease of exposition as (36) a., a schematic illustration of the problematic configurations based on a predicate’s non-structural case-assignment to complex nominal domains parallel to (35) above is given in (36) b. 25 Recall that in the system elaborated above, only one phasal boundary is taken to exist if either D or Poss PRO surfaces in isolation with NumP in either of its featural specifications. However, since Q W is necessarily also taken to be part of the phrasal set-up underlying partitives, the specification of co-occurrences in the main text suffices. 336 4 IMPLICATIONS - Optionality & Complex Formations (36) 0 b. (36) a. Ich vertraue all(-en) dies-en Bücher-n. (= (32) b., [modified]) I trust [ all these books ] DAT b. V > D/ Poss PRO > Q W > Num > N [ val ] [ val ] [ val ]  [ val ] In the light of these parallels, it seems pressing to uniformly specify HN as bearing case even in environments where no morpho-syntactic reflex is observable. How, then, can the system under elaboration be modified to counter this apparent flaw? I want to argue that it is already capable to do so but that we are only now in a position to further specify the relevant notion involved: Recall from chapter 2.3.3 that I followed Ritter (1993) and Bernstein (1991) in the proposal of Nto-Num movement. There, I stayed agnostic as to the derivational timing of application of the operation, to be located either at Narrow Syntax or post-syntactic Morphology (viz. IM or m-merger respectively). The data above can be consulted to finally dissolve this issue in that only when N moves to Num in Narrow Syntax can all morpho-syntactic featural demands of the nominal head be met in that it is still part of the derivation when case is assigned to its domain. Observe that this conclusion in turn bears further consequences for the theory of Anti-Locality as modified in chapter 3.3.2 above: In the light of these parallels, it seems pressing to uniformly specify HN as bearing case even in environments where no morpho-syntactic reflex is observable. How, then, can the system under elaboration be modified to counter this apparent flaw? I want to argue that it is already capable to do so but that we are only now in a position to further specify the relevant notion involved: Recall from chapter 2.3.3 that I followed Ritter (1993) and Bernstein (1991) in the proposal of N-to-Num movement. There, I stayed agnostic as to the derivational timing of application of the operation, to be located either at Narrow Syntax or postsyntactic Morphology (viz. IM or m-merger respectively). The data above can be consulted to finally dissolve this issue in that only when N moves to Num in Narrow Syntax can all morpho-syntactic featural demands of the nominal head be met in that it is still part of the derivation when case is assigned to its domain. Observe that this conclusion in turn bears further consequences for the theory of Anti-Locality as modified in chapter 3.3.2 above: Non-feminine HNs will never be coordinated below two nominal phrasal cycles and are therefore always taken to receive case in base position. 26 As a last implication drawn from the contrast in (32) and (33) above, it would be fruitful to revisit the 2 nd cycle Agree-phenomena from chapter three in the light of case-assignment from inside the nominal domain (yet outside the phase of the triggered SLI) in the concatenation of the partitive quantifier with the possessive pronoun in the extended domain of a non-feminine HN. However, the combination of partitives with mass nominals is barred in general, i.e. with 26 While this proposal levels the inflectional idiosyncrasies between [[dat][φ[[f[pl]]]]] and [[dat][φ[[sg][m/ n]]]] (abstracting away from the optional and archaic character of overt inflection in the latter configuration), it opens a gap between the two NumP-configurations since dative inflection is not observed in the feminine paradigm. Moreover, returning to genitive inflection laid out in ch. 2.3.3 b. and discussed above, observe that the proposals in the main text would consequently take the nominal in both feminine derivations to be spelled out prior to case-assignment. Paradoxically, movement to Num 0 would be dependent on a case-value before said value has entered NS. Obviously, more research is needed in this area; nevertheless, the split in values identified in this book is still relevant here. 4.3 Loose Ends & Complex Formations 337 all high SLIs. 27 Nevertheless, some speakers do accept feminine and non-feminine singular count-nominals in simple partitives with all high SLIs, viz. determiner and demonstrative probing for [φ] as well as Poss PRO probing for the more restrictive [φ[f]]. 28 The latter data help to sharpen the notion of 2 nd cycle Agree: In chapter 3.3.3 above, probing from the highest nominal cycle to the weak quantifier Q W for [φ] was equalized with the assignment of non-structural case from the predicate to Poss PRO as triggers of a second attempt of the latter items to probe into their respective complements for [φ] (cf. the discussion surrounding (120), ch. 3.3.3). This configuration was split from those in which D co-occurs with the possessive pronoun in non-feminine environments, which turn out ungrammatical. The structural notion setting apart the grammatical Q S / D/ Poss PRO -Q W and V-Poss PRO from the ungrammatical D-Poss PRO concatenations has been taken to be that of shared contextual phasehood: While in the former configurations no shared phasal membership exists between the elements triggering and those undergoing 2 nd cycle Agree, it does so in the latter conjunction. 29 Returning to the simple partitives as described above, partitive Q S does trigger a second probing of Poss PRO into its complement in contrast to its non-partitive counterpart. 30 27 Interestingly, as has been pointed out to me by one informant, feminine mass nominals seem slightly more accessible in the configurations just outlined coordinated under an SLI probing for [φ], i.e. an instantiation of D. One might speculate that this deviance is related to the incorporation of the projection also bearing the plural value, the phrase NumP, under the proposals elaborated above. Nevertheless, such data still turn out ungrammatical and hence do not suffice to develop a stable testing environment. I hence have to leave this as an open matter. 28 We have hence found another featural deviation of simple partitives from standard derivations in the respective manner in which Q S is sensitive to the mass/ count distinction, i.e. combining with plural and singular nominals in the former and with mass as well as plural count nominals in the latter, i.e. excluding nominals without any number-feature in the former (cf. ch. 2.3 above) and one specific value in the latter. 29 Observe that it is not the notion of actual phasal barrier that is relevant here but the membership concerning phasal cycles as has been deduced from the parallel inability of D and Q S to trigger 2 nd cycle Agree of Poss PRO , the former a phase in their simultaneous occurrence, the latter a non-phase. 30 Observe furthermore that the weak quantifier in partitive Q S P is also specified for any value of [φ] even though this element has been classified as probing for the subpart [φ[f]] in chapter three above. This is in line with the proposed derivational path of the construction since Q W is taken to probe upon EM, hence to reduce to partial default in concatenations with masculine and neuter HNs. However, before movement to partitive Q S P, 2 nd cycle Agree has already been triggered by EM of one of the structurally lower SLIs of the highest cycle, i.e. D 0 / Poss PRO0 . 338 4 IMPLICATIONS - Optionality & Complex Formations (37) (37) a. viel-es mein-es Q S > Poss PRO > N [*m/ n] [*m/ n] [m/ n] [*gen] [*gen] b. all-ø mein-ø Q S > Poss PRO > N [φ] [φ] [m/ n] These SLIs, however, share the membership in the highest nominal cycle of contextual phasehood. What sets partitive Q These SLIs, however, share the membership in the highest nominal cycle of contextual phasehood. What sets partitive Q S apart from non-partitive Q S , then, is the category of the featural interaction: Default Q S probes for [φ] itself, parallel to D, while partitive Q S assigns non-structural case, viz. free assignment of genitive parallel to an appropriately specified predicate. We therefore have to split 2 nd cycle Agree w.r.t. the feature triggering it with [φ] locally restricted to apply beyond a contextual phasal cycle and non-structural case not limited in this way. These configurations are compiled in the table below: (38) probing φ assigning case cycle-internally * DP-Poss PRO Q S -Poss PRO Q S -Poss PRO cycle-externally DP-Q W Q S -Q W V-Poss PRO V-Q W We can conclude that partitive constructions parallel to the possessives under consideration at the beginning of this section couple (i) featural and (ii) linearizational divergences from the default buildup of the nominal spine with additional semantic import: Concerning the former (i), the domain-internal assignment of genitive case has been identified as marking a semantically coherent subpart in both complex configurations (i.e. the possessor and the ‘ of the N’-range of partitives), structurally corresponding to the complement of its assigner which is thereby identifiable, i.e. the nominal head and the partitive quantifier. Concerning the latter (ii), both constructions involve domain-internal movement operations between categorically related phrases, i.e. phrasal movement from complex postnominal possessives to the specifier of the prenominal possessive phrase as well as head movement of weak quantifiers to the higher quantifier phrase. The preceding discussion hence compiled a set of syntactic mechanisms that structurally sets apart these semantically enriched configurations from the default buildup of the nominal domain. Moreover, various refinements to notions established in the course of previous chapters could be developed therein 4.3 Loose Ends & Complex Formations 339 (e.g. the relation of possessivity and case as well as the featural trigger for 2 nd cycle Agree and the relevance of structural distance therein). In the remainder of this chapter, then, I will focus on complex items, rather than concatenations of independent SLIs and sharpen the proposals just outlined above in the analysis of the semantic and syntactic idiosyncrasies that accompany these items before turning to the semantic motivation for their exceptional structural set-up and for the mechanisms proposed above in general. 4.3.2 Complex LIs This subchapter is concerned with further rigid concatenations of semi-lexical items in the nominal domain; however, the current object of investigation departs from the constructions under analysis in the previous section in that their subparts are fixed not with respect to their categorical status but as unique lexical instances. Thus, while possessor doubling constructions might involve any coherent combination of lexical possessor plus appropriately φ-specified pronominal item (but cf. endnote *19) and partitive constructions might be based on various weak and even strong quantifiers, we observe tightly fixed co-occurrences of SLIs making up the configurations under investigation below. For this reason, they have been referred to as ‘complex lexical items’ and distinguished from those termed ‘complex syntactic objects’ above. As should be evident by now, I nevertheless believe a decompositional approach to these elements (as well as ultimately to their subparts) to pose as fruitful an approach as it proved in previous chapters and furthermore to gain insight into various properties of the nominal hierarchy of projections as well as their derivations by its application to the current object of investigation. a. ( ein -) jed - Let me, then, commence my analysis by focusing on the quantifying LI ( ein -) jed - ‘every’/ ‘each’. As can be readily observed, this item consists of two independent lexemes and hence incorporates a space in written form. As has been moreover indicated by the hyphens above, both terminal subparts of this fixed concatenation independently mark agreement, viz. they partake in the establishment of nominal concord via suffixation. As a last point concerning the initial notation given above, observe that the brackets enclosing the ein-word signal optionality, thus potentially leaving only the apparently terminal jed - (but cf. Roehrs 2012: ch. 3). I will come back to the status of both ein as well as jed - (and moreover its own internal structure) below. Before doing so, however, let me begin by introducing the syntactic and semantic properties of the complex SLI. As has already been indicated by the translation at the beginning of this paragraph, 340 4 IMPLICATIONS - Optionality & Complex Formations the item under investigation semantically selects exhaustive sets denoted by HN and possible modifiers, however, deviating from the universal quantifier all and parallel to English every and each , this item bears a selectional restriction for singular count nominals (cf. (39) a., b. below), adjacent to that identified in the context of simple partitives in the last section 31 which, as argued by Roehrs, is supported by the overt appearance of the optional ein-word subpart: “[A]lthough invoking a singularity assertion, under normal circumstances, this ein can only require the head noun to be countable” (Roehrs 2012: 21, emphasis in the original), resulting in semantic distributivity. 32 Concerning the syntactic criteria laid out in chapter one and employed in the previous section, this SLI can be categorized as a strong quantifier in patterning with all - (cf. (39) c., d. below, moreover ch. 1.3.2 above). Finally, and turning to linearizational properties, the structural position w.r.t. the remaining SLIs in the nominal spine reveals that co-occurrence of these is only licit when incorporating said item in a simple partitive construction (i.e. followed by a plural and partitive case-marked nominal domain) in which it necessarily occupies Q S , as has been elaborated above (cf. (39) e., f.; cf. moreover Fehlisch 1986: ch. 1.2 on the parallel syntactic and semantic behavior of the complex SLI and all in partitive contexts).* 24 All these properties are compiled below: (39) a. (ein) jed-es Buch (ein) every book b. * (ein-e) jed-e Bücher (ein) every books c. Definiteness Restriction * Es gibt [(ein-en) jed-en Baum] im Garten. it is [(ein) every tree ] in.the garden ‘There is [every tree] in the garden.’ d. Preposed Adjective Structures Betrunken wie [(ein) jed-er Mann] war, sprachen wir ungern mit ihnen. drunk as [(ein) every man ] was, talked we reluctantly with them ‘Drunk as every man was, we only talked reluctantly with them.’ 31 Cf. Fehlisch (1986: ch. 1.1.1), moreover ibidem (ch. 1.3.1.1 b.) on the reinterpretation of mass nominals as denoting subkinds in the concatenation with ( ein -) jed -. 32 Cf. Fehlisch (1986: ch. 1.1) and references therein for a detailed presentation of the semantic import of this LI, furthermore Pafel (2005: ch. 2.2.4). 4.3 Loose Ends & Complex Formations 341 e. * (ein) jed-es {das/ dies-es/ mein} Buch (ein) every [ { the / this / my } book ] NOM f. (ein) jed-es {der/ dies-er/ mein-er} Bücher (ein) every [{ the / this / my } books ] GEN A few comments are in order here. First, observe that the terminal subparts do not inflect in parallel across the data in (39) but rather split along the two inflectional patterns identified in chapter three, i.e. [φ] on jed and [φ[f]] on ein respectively. Furthermore, concentrating on the relation of this complex SLI to the number/ gender-feature in general, observe that syntactically singularmarked nominal domains nevertheless enforce semantic as well as pronominal syntactic plural reference: Observe (39) d., which turns ungrammatical with the replacement of the plural reflexive pronoun ihnen with the corresponding dative masculine singular form ihm . However, and somewhat contradictory, inflection on both terminal subparts in the partitive construction, exemplified in (39) f. as well as in more detail in (40) a.-c. below, exhibits agreement with the singular gender-value of HN (as is easily derivable from the form of the second terminal element probing for the root node [φ]) rather than the expected plural agreement,* 25 parallel to weakly quantified partitives discussed in the last section and exemplified again below in (40) d.: (40) a. (ein) jed-es der Bücher (Buch = [φ[[sg][n]]]) (ein) every neut [ the books ] GEN b. (ein) jed-er der Männer (Mann = [φ[[sg][m]]]) (ein) every masc [ the men ] GEN c. (ein-e) jed-e der Frauen (Frau = [φ[[sg][f]]]) (ein fem ) every fem [ the women ] GEN d. viel-e der Bücher/ Männer/ Frauen many pl [ the books / men / women ] GEN To continue the listing of unexpected properties, as Roehrs (2012) notes concerning Adjectival Impoverishment, ( ein -) jed not only seems to exhibit a certain variability between strong and weak adjectival inflection accompanying it (cf. (41) below) but is moreover able to display sequential as well as gradatial inflection between its terminal elements (cf. (42); cf. also ch. 3.3.4 above). Note 342 4 IMPLICATIONS - Optionality & Complex Formations that the marking of oddness in (41) and (42) parallels Roehrs’ categorization regarding relative frequency (cf. Roehrs 2012: fn. 5). (41) (= Roehrs 2012: (5) a., b., [modified]) a. ein jeder gute Student an every (NOM.ST) good (WK) student (MASC) b. # ein jeder guter Student an every (NOM.ST) good (ST) student (MASC) (42) (= Roehrs 2012: (12) a., b., [modified]) a. einer jeden Studentin an (DAT.ST) every (WK) student (FEM) b. # einer jeder {Saison/ Party/ Person} an (DAT.ST) every (ST) season / party / person (FEM) As Roehrs himself notes, sequential inflection between both the quantifier and the adjective as well as the subparts of the SLI themselves (i.e. (41) and (42) b. respectively) are judged as odd and are also reduced in number of occurrence but nevertheless observable in natural language use, which is confirmed by the internet search hits reported by the author. 33 Starting out with the premise that parallel inflection tends to indicate a shared lexical category and coupled with the intuition that the overt occurrence of ein favors SLI-internal gradial inflection, hence sequential adjectival one, Roehrs concludes that the right-hand terminal part of the complex item is categorically ambiguous between the semi-lexical class D and a “determiner-like quantifier with adjectival morpho-syntax” (Roehrs 2012: 6)/ “(determiner-like) adjective” (Roehrs 2012: 8, cf. also Karnowski & Pafel 2004: 172f., Pafel 2005: 45) and furthermore that the ein-word in their ‘composite’ (a term employed by Roehrs 2012: 9) co-occurrences reinforces the latter classification while semantically intensifying the aforementioned distributive reading. Thereby, both the heterogeneity of acceptable inflectional configurations as well as the existence of preferential patterns are accounted for if Adjectival Impoverishment is initiated by D. To formalize, the system classifies the ein-word as itself variable between phrasal intensifier vs. determiner-like intensifier (i.e. Spec,IntP vs. D 0 ), the right-hand element as either a phrasal quantifier or determiner (i.e. Spec,CardP vs. D 0 ), coupled with additional conditions restricting the various 33 I will ignore here another irregular pattern reported by Roehrs on the basis of internetsearch hits, e.g. those incorporating uninflected ein in [φ[f]]-contexts (cf. Roehrs 2012: 30). 4.3 Loose Ends & Complex Formations 343 co-occurrences thereof to result in a cluster of the four aforementioned licit inflectional configurations. In what follows, I will concentrate on those patterns classified as unmarked in (41) and (42) above, i.e. uniform gradational SLI-internal as well as adjectival inflection in line with the inflectional properties reported for isolated jed in ch. 1.3.2 above. The paradigm below demonstrates this with a neuter HN in non-structural case to evoke overt inflection on the ein-word probing for [φ[f]]: (43) a. mit gut-em Gefühl with [ good strong feeling ] DAT b. mit ein-em Gefühl with [ein strong feeling ] DAT c. mit jed-em Gefühl with [ every strong feeling ] DAT d. mit ein-em gut-en Gefühl with [ein strong good weak feeling ] DAT e. mit jed-em gut-en Gefühl with [ every strong good weak feeling ] DAT f. mit ein-em jed-en Gefühl with [ein strong every weak feeling ] DAT g. mit ein-em jed-en gut-en Gefühl with [ein strong every weak good weak feeling ] DAT Concentrating first on inter-SLI inflectional dependencies, Roehrs’ intuition to associate the right-hand subpart jed with the category adjective is correct insofar as it does indeed undergo AI triggered by the preceding ein (cf. also endnote *26). This item, in turn, probes for [φ[f]] parallel to the more complex possessive pronoun it constitutes a part of, under analysis in the last chapter. However, recall also from that chapter that the right-hand subpart shares this property not only with adjectives but also with conventional weak quantifiers (whose interaction with bare ein is not testable because of the obvious number mismatch involved). Abstracting from these facts, jed should be tentatively classified as underlyingly belonging to the class Q W while items enforcing AI have been in a position dominating the quantifier. To further narrow down the structural positions of the terminal SLIs involved, Roehrs reports that insertion of an adverbial modifier like fast (‘almost’) between the two terminal subparts of the complex SLI only results in a “slightly marked status” (Roehrs 2012: 19) while the unmarked position of such items is left-adjacent to the full nominal 344 4 IMPLICATIONS - Optionality & Complex Formations hierarchy. This is, however, not what my informants report, for whom (44) b. below is completely ruled out (data and glosses are taken from Roehrs 2012, grammaticality judgments are my own). (44) (= Roehrs 2012: (30) , [modified]) a. fast (# ein) jeder Student almost an every student b. * ein fast jeder Student an almost every student A parallel contrast incorporating the simplex strong quantifier all is obviously not reproducible. However, turning to the insertion of verbal interjections like sage ich (lit. say I, ‘I say’), leaves us with another interesting observation: (45) a. Alle [sage ich] diese Männer sind großartig! all [ say I ] these men are great b. * Ein [sage ich] jeder {Mann/ der Männer} ist großartig! ein [ say I ] every { man / [ of.the men ] GEN } is great If we accept the structural positions of the strong quantifier and the demonstrative, argued for in chapter one of this book and put to use throughout the previous analysis, the interjection arguably occupies an adjunct-position between these heads in (45) a., a structural position obviously not available between the terminal subparts of the complex quantifier in (45) b. which can thus be described as structurally ‘closer’ to one another. The latter intuition might then translate to the configuration Spec-head not allowing an intermediate adjunctposition. However, this analysis bears the implication that the left-adjacent subpart of the complex item ein has to exhibit phrasal status (cf. Roehrs 2012: ch. 5.1). Above, the root ein as well as its structurally complex derivatives have been classified as occupying head positions throughout the analysis, a uniformity that should be maintained here. Another option in accordance with the above findings, then, is to classify the concatenation in question as a single LI qua shared head status. This analysis is in line with Roehrs’ (2012: ch. 5.2) conclusions, classifying the complex SLI as a compound incorporating functional rather than lexical elements. However, I will argue that this ‘composite’ is born in NS (cf. Pafel 1994: 242). A hint as to the correctness of this proposal can be deduced from the fact that ein in concatenation with jed is barred from partaking in additional complex formations, as can be deduced from the inability to combine with either prefix coding possession or nominal negation (cf. also Roehrs 2012: 20f.). 4.3 Loose Ends & Complex Formations 345 (46) * {k-/ m-}ein jeder N { neg -/ poss -}ein every N Observe that while this analysis might initially appear ad hoc - in particular with respect to the indications given by the written form - it parallels proposals made on the basis of one of these latter ein-word formations, i.e. the possessive pronoun, in its key respects, to wit, both SLIs incorporate a terminal element bearing more than one instance of a single feature category. With possessive pronouns, the φ-values differed between possessor and possessee, as has been laid out at the beginning of this subchapter; with ( ein -) jed -, however, they must be taken to be identically specified in accordance with HN (and the idiosyncratic properties of probing). Below, I will once again sketch the outlines of a nominal derivation incorporating the composite before turning to the implications for simple partitives in German. To begin, I present the classification of the terminal subparts w.r.t. the relevant properties of nominal concord, carved out in chapter three above: (47) ein jed - 34 Probing φ [φ[f]] [φ] Default Agreement   Adjectival Impoverishment   Beginning with EM and subsequent probing of the weak quantifier into the extended nominal domain, then, jed trivially receives a value for [φ] while the subsequent sequence of identical operations for the ein-word only results in a specification for the feminine subgroup of φ-values and in default agreement in the remaining cases. Successively, the strong quantifier phrase externally merges and jed moves over the ein-word into head position via Q-to-Q movement to which the latter is subsequently attracted and head-adjoins in the first application of IM observed between categorically unrelated SLIs in the current analysis. The resulting structure splits the terminal position under the label of 34 Observe that jed- , classified as underlyingly an instance of the Q W based on the property to undergo AI, diverges in means of probing from the instances under analysis in previous chapters. Therefore, either those properties must be coded on each SLI separately or jedhas to be recategorized as Q S ; since I take the complex SLI to end up in the latter head position at Spell-Out, nothing really hinges on this here. 346 4 IMPLICATIONS - Optionality & Complex Formations the former (cf. also the discussions surrounding Roehrs 2012: (48), to a lesser extent Pafel 1994: (20), Pafel 2005: (1.61), (1.122)). 35 (48) Optionality & Complex Formations I MPLICATIONS 267 will argue that this ‘composite’ is born in NS (cf. Pafel 1994: 242). A hint as to the correctness of this proposal can be deduced from the fact that ein in concatenation with jed is barred from partaking in additional complex formations, as can be deduced from the inability to combine with either prefix coding possession or nominal negation (cf. also Roehrs 2012: 20f.). (46) * {k-/ m-}ein jeder N { neg -/ poss -}ein every N Observe that while this analysis might initially appear ad hoc - in particular with respect to the indications given by the written form - it parallels proposals made on the basis of one of these latter ein-word formations, i.e. the possessive pronoun, in its key respects, to wit, both SLIs incorporate a terminal element bearing more than one instance of a single feature category. With possessive pronouns, the φ-values differed between possessor and possessee, as has been laid out at the beginning of this subchapter; with ( ein -) jed -, however, they must be taken to be identically specified in accordance with HN (and the idiosyncratic properties of probing). Below, I will once again sketch the outlines of a nominal derivation incorporating the composite before turning to the implications for simple partitives in German. To begin, I present the classification of the terminal subparts w.r.t. the relevant properties of nominal concord, carved out in chapter three above: (47) ein jed - 42 Probing φ [φ[f]] [φ] Default Agreement   Adjectival Impoverishment   Beginning with EM and subsequent probing of the weak quantifier into the extended nominal domain, then, jed trivially receives a value for [φ] while the subsequent sequence of identical operations for the ein-word only results in a specification for the feminine subgroup of φ-values and in default agreement in the remaining cases. Successively, the strong quantifier phrase externally merges and jed moves over the ein-word into head position via Q-to-Q movement to which the latter is subsequently attracted and head-adjoins in the first application of IM observed between categorically unrelated SLIs in the current analysis. The resulting structure splits the terminal position under the label of the former (cf. also the discussions surrounding Roehrs 2012: (48), to a lesser extent Pafel 1994: (20), Pafel 2005: (1.61), (1.122)). 43 (48) Q S0 3 ein jed - Observe that in both the uniformly valued as well as the disrupted φ-configurations, the two sets of features are active as can be derived from their independent inflection for verbally assigned case (which moreover triggers 2 nd cycle Agree on the ein-word with non-structural case-assigning predicates) resulting in the by now well-known MIXED inflectional pattern. 44 42 Observe that jed -, classified as underlyingly an instance of the Q W based on the property to undergo AI, diverges in means of probing from the instances under analysis in previous chapters. Therefore, either those properties must be coded on each SLI separately or jed has to be recategorized as Q S ; since I take the complex SLI to end up in the latter head position at Spell-Out, nothing really hinges on this here. 43 One might object that a more economic derivation would propose ein to ‘tag along’ with jed in one application of IM with an intermediate landing-site at the EM-position of ein . However, observe that the complex head in this case would bear the label ein , rather than Q S , a property that will become important below (but cf. Pafel 1994: ch. 3). 44 An interesting observation in this context concerns the deviant forms of SLI and (homonymous) STRONG and WEAK adjectival inflection presented in chapter 3.3.4 c. above (cf. also Karnowski & Pafel 2004: 172f.): Observe that in the appropriate slots of the inflectional paradigm of the complex SLI, the right-hand terminal subpart ein inflects according to its status as an SLI (to a lesser degree also acceptable with homonymous adjectival inflection), while the quantifier jed only ever exhibits adjectival inflection. i. der Geruch [ein-es/ en jed-en Wein-es] Observe that in both the uniformly valued as well as the disrupted φ-configurations, the two sets of features are active as can be derived from their independent inflection for verbally assigned case (which moreover triggers 2 nd cycle Agree on the ein-word with non-structural case-assigning predicates) resulting in the by now well-known mixed inflectional pattern.* 26 With these basic processes in place, let us turn to the idiosyncratic properties of Adjectival Impoverishment on the one hand and the optionality of ein on the other. Concentrating on the former, recall that the derivational timing of evaluation has been taken to be post-syntactically above (i.e. at Morphology, cf. ch. 3.3.4) while precedence/ dominance was identified as the relevant structural trigger. Nominal domains derived along the path just outlined are headed by Q S P which, albeit split and thus complex (i.e. the ‘label of the head’ in intuitive terms, cf. fn. 35 above), bears an instance (in the technical sense implemented in chapter three) of the φ-feature structure of HN itself. Therefore, prenominal LIs will undergo AI along the lines laid out in the last chapter. The subparts of the split head-position, however, are either featurally disrupted or uniform (in derivations on the basis of [φ[m/ n]]or [φ[f]]-configurations respectively): In uniform domains, the head-adjoined ein-word is located in a sisterhood relation with the internally merged weak quantifier; Adjectival Impoverishment hence applies in parallel to cases of extended sisterhood. In disrupted domains, AI does not apply in the isolated derivation of the nominal domain. Parallel to other (ein-word) SLIs, probing for [φ[f]], 2 nd cycle Agree is triggered by the ‘free’ assignment of non-structural case. This does not change with the ein-word coordinated under Q S0 , which is still part of the c-command domain of the predicate. However, in this case, the complement of the probing item has indeed changed since the first probing-attempt, still resulting, however, in identical outcomes since the sister of ein , its probing domain, bears the same value as its initial complement at EM. Therefore, jed is once again located in the appropriate structural con- 35 One might object that a more economic derivation would propose ein to ‘tag along’ with jed in one application of IM with an intermediate landing-site at the EM-position of ein . However, observe that the complex head in this case would bear the label ein , rather than Q S , a property that will become important below (but cf. Pafel 1994: ch. 3). 4.3 Loose Ends & Complex Formations 347 figuration for AI to apply at Morphology once ein has successfully received a value. Therein the mixed inflectional pattern of SLI-internal agreement is derived (pace Fehlisch 1986: fn. 7, but cf. Pafel 1994: ch. 3). Turning to the absence of ein , then, observe that jed is still taken to internally merge to Q S P and to furthermore subsequently initialize Adjectival Impoverishment from said position. Parallel to the derivational path laid out above, said application is uniform with prenominal LIs while the quantifier itself bears strong inflection throughout. We have hence derived the idiosyncrasies of nominal SLI-internal and -external inflectional patterns in the identification of two independent applications of AI. After having sketched the default derivation incorporating the complex SLI in the system elaborated in earlier chapters and identified the interactions of its terminal subparts therein, let me return to the most puzzling agreement phenomenon exhibited with ( ein -) jed -: As noted above in the initial analysis of its properties as a strong quantifier, the sole grammatical concatenation with other SLIs in the extended nominal domain constitutes partitive constructions, conjoining the element under consideration with genitive case-marked nominal hierarchies on the basis of a plural HN. The resulting constructions fully parallel those under discussion in chapter 4.3.1 b. above; however, and deviating from these, ( ein -) jed surprisingly exhibits singular gender-agreement, while the initial weak quantifier above agreed with the overt, i.e. plural, form of HN. This state of affairs is obviously mysterious under the mechanisms underlying nominal concord carefully carved out in previous chapters. I believe, however, that this unexpected pattern of agreement allows us to further refine the structural set-up of simple partitive constructions. To begin, observe that simple concatenations of the complex SLI with a singular HN do not exhibit nominal domain-internal case-assignment but inflect according to the case frame of the dominating predicate and the patterns of AI just outlined. This can be deduced from both inflection on prenominal adjectives as well as on HN itself in the appropriate combinations of φand case-values (with regular exceptions, cf. fn. 26, ch. 2.3.3 b., ch. 4.3.1 b. above). Partitive case-assignment is hence not a property of the quantifier itself. As has been furthermore laid out above, the full range of nominal SLIs can follow the complex quantifier in genitive-marked extended nominal domains. Observe that this also includes the possessive pronoun: (49) (ein) jed-es mein-er gut-en Bücher (ein) every neut [ my pl good pl books ] GEN This co-occurrence constitutes an additional surprising property of partitives on the basis of ( ein -) jed -, given that, as I argued above, the root ein-word is part of 348 4 IMPLICATIONS - Optionality & Complex Formations both complex formations (cf. the discussion surrounding (46) above). That one can identify two separate (and furthermore differently φand case-specified) instances of the root ein seems to call into question the preceding categorization of the possessive pronoun and therefore the relatedness of the two instances of the morpheme ein as well as ultimately the decompositional approach in toto. This is, however, not necessarily so: Concentrating on the inflectional properties of the SLIs in (49) above, observe that, while the complex quantifier inflects for the appropriate singular gender-value of the plural HN, the genitive-marked range of the partitive uniformly inflects in accordance with the overt morphological form of the nominal, i.e. plurally. To recap, then, simple partitive constructions on the basis of the complex quantifier ( ein -) jed display a dual nature of agreement in which a verbally case-marked and singularly inflecting quantifier heads a genitive and plural-marked nominal domain which itself might incorporate another instance of an SLI already incorporated in the former. This state of affairs is problematic under the conclusions reached for the underlying structure of simple partitives on the basis of weak quantifiers in ch. 4.3.1 b. It should, however, readily call to mind the rationale of previous approaches to partitives following Dean (1966), also laid out above. To wit, varying φ-agreement between variable and range of a partitive construction as well as multiple occurrences of a single SLI both point to the direction of an underlyingly complex structure of simple partitives as two separate nominal domains with HN of the variable elided under lexical non-distinctness. The featural mismatch of ein jed and the partitive nominal is therein accounted for: These LIs are born in two separate domains. As has been further laid out above, the functional structure of the respective domains has been taken to differ with restrictions of the construction itself requiring the embedded partitive noun (henceforth PN) to exhibit plural marking in the lower nominal domain of the range, while the complex SLI surfaces with its idiosyncratic singular agreement in the higher domain of the variable, parallel to simple quantified nominal structures. Thus, PN incorporates Num[φ[f[pl]]], while HN does not. Therefore, elision of HN enforces the illusion of a number mismatch on the surface which has not been observable in partitive constructions on the basis of a morphologically simple Q W , inflecting for [φ[f]] parallel to the overt form of PN. 36 Likewise, multiple occurrences of ein , simultaneously part of the quantifier and the possessive pronoun in a single complex nominal domain, do not contradict the decompositional approach: 36 Furthermore taking into account the patterns of agreement in simple partitives on the basis of the non-concatenated subparts jed and ein - (cf. endnote *25) respectively as well as from QST (cf. endnote *24), one might speculate to have found the featural basis for the composition of this complex SLI in the parallel restrictions of number/ gender-selection of its subparts. 4.3 Loose Ends & Complex Formations 349 As has been observed by case-assignment properties, the former is part of the nominal domain of HN, while the latter heads the domain of PN. These findings, however, do not scrutinize the analysis of simple quantified and uniformly casemarked nominal structures conducted in the foregone chapters. Recall that the occurrence of ein is exclusive to the buildup of either SLI therein. As I hope to have shown, instances of number mismatches and patterns of (im-)possible co-occurrences of multiple instances of SLIs together make a strong case in favor of the underlyingly complex structural set-up of simple partitive constructions. Building on these findings, I will further elaborate on partitive hierarchies below based on another complex quantifier in German, i.e. d meist before turning to the implications this SLI bears concerning the syntactic means of restricting semantic application in German. b. d meist - The second object of investigation in this subchapter is the complex quantifier d meist - (lit. the most, ‘most’). As will be shown, much of what has been laid out above concerning the syntax of complex SLIs on the basis of ( ein -) jed readily carries over to this lexicalized concatenation 37 joining together the determiner (represented here by the root d -) with a quantificational element. As can be furthermore read off the written form employed above, the latter item is not optional and meist therefore does not occur in isolation, i.e. as an independent SLI in the extended projection of the noun. Note that one observes the same exceptional syntactic ‘closeness’ herein as has been demonstrated for the former complex SLI above, which can again be deduced from the insertion of degree particles (i.e. recht , ‘quite’ below) in between its subparts d and meist vs. the default concatenation of D and Q W . (50) a. die recht vielen Männer the quite many men b. * die recht meisten Männer 38 the quite most men 37 Cf. Pafel (1994: 242; 2005: 27, 45), Karnowski & Pafel (2004: 172). 38 As one of the reviewers points out, insertion of the modal particle wohl (‘probably’) between the subparts of the complex SLI seems to result in a grammatical construction for many speakers: Although at least degraded in grammaticality for me, an informal google search for this configuration returned more than 6000 hits. Interestingly, such a high number of occurrences is not found with other modal particles inserted in this position. I will have to leave the issue of particular particle placement without further comment 350 4 IMPLICATIONS - Optionality & Complex Formations Given the criteria on strong vs. weak quantifier status as employed in the last section (i.e. the series of grammatical environments laid out in (39) above as well as SLI-internal inflectional properties), the complex SLI can be categorized as a strong quantifier and hence an element of the category Q S on the surface (but see below), while the second terminal subpart, concatenated with D 0 , must be characterized as an instance of Q W0 since it uniformly inflects according to the weak pattern in this concatenation and is thus taken to undergo AI (cf. also C&G: 44). We can, however, not readily attribute inflectional properties directly to inter-SLI applications of AI in this obligatory co-occurrence (contrary to those incorporating the optional ein-word above) since no contrast can be established here.* 27 The null hypothesis should nevertheless proceed in parallel to the proposals made in section 4.3.2 a. Thus, meist externally merges in Q W0 and subsequently raises to Q S0 , thereby crossing the determiner which subsequently adjoins to it, resulting in SLI-internal AI on the quantifier. However, we find data that challenge these claims: Observe that, apart from the determiner, the possessive pronoun is likewise able to occur in concatenation with this quantifier. 39 Since the latter category is based on the root ein -, we observe probing for [φ[f]] rather than [φ] as in the case of D 0 . As is expected, SLI-internal inflection in these cases follows the mixed pattern. However, and deviating from the claims made above, coordinated adjectives likewise do so, resulting in sequential strong Q-Adj inflection with meist coordinated under the possessive pronoun in [φ[m/ n]] number/ gender in structural case. All these configurations are exemplified in (51) and (52) below for complex SLIs incorporating D and Poss PRO respectively, on the basis of accordingly specified HNs. (51) masc/ neut nom d-er/ -as meist-e gut-e [*[nom][φ*[[sg][m/ n]]]] [*[nom][φ*[[sg][m/ n]]]] [*[nom][φ*[[sg][m/ n]]]] gen d-es meist-en gut-en [*[gen][φ*[[sg][m/ n]]]] [*[gen][φ*[[sg][m/ n]]]] [*[gen][φ*[[sg][m/ n]]]] dat d-em meist-en gut-en [*[dat][φ*[[sg][m/ n]]]] [*[dat][φ*[[sg][m/ n]]]] [*[dat][φ*[[sg][m/ n]]]] acc d-en/ -as meist-en/ -e gut-en/ -e [*[acc][φ*[[sg][m/ n]]]] [*[acc][φ*[[sg][m/ n]]]] [*[acc][φ*[[sg][m/ n]]]] here and only note that a contrast to the default concatenation of D and Q W exists in the general accessibility of the position under consideration. 39 The third head of the highest contextual phasal cycle, however, is barred from appearing with meist -, both in isolation as well as preceding the complex SLI itself incorporating D/ Poss PRO , in accordance with the proposed structural position of the latter in Q S0 itself. 4.3 Loose Ends & Complex Formations 351 (52) masc/ neut nom mein-ø meist-er/ -es gut-er/ -es [φ] [*[nom][φ*[[sg][m/ n]]]] [*[nom][φ*[[sg][m/ n]]]] gen mein-es meist-en gut-en [*[gen][φ*[[sg][m/ n]]]] [*[gen][φ*[[sg][m/ n]]]] [*[gen][φ*[[sg][m/ n]]]] dat mein-em meist-en gut-en [*[dat][φ*[[sg][m/ n]]]] [*[dat][φ*[[sg][m/ n]]]] [*[dat][φ*[[sg][m/ n]]]] acc mein-en/ -ø meist-en/ -es gut-en/ -es [φ] [*[acc][φ*[[sg][m/ n]]]] [*[acc][φ*[[sg][m/ n]]]] One can therefore conclude that the quantifier itself does not induce AI. A parallel path of the derivations of complex SLIs can, I believe, be maintained in the switch of the category label from Q to D in the current case. Hence, the formation of this LI in NS would indeed follow the more economic single-movement proposal sketched above in fn. 35 (i.e. Q W to D, arguably coupled subsequent movement to Q S for the sake of uniformity, cf. also endnote *24). The categorical status and internal set-up of the complex construction are illustrated below: (53) Optionality & Complex Formations I MPLICATIONS 271 (53) D 0 3 D 0 meist - Note that therein we would again propose two independent applications of AI under (extended) sisterhood, viz. from the lower D 0 to the weak quantifier as well as from the higher D 0 to the extended nominal projection, parallel to the inner workings proposed above for ein jed -. Therefore, when the possessive head partakes in the formation of said complex item, the result is a ‘uniformly disrupted’ (i.e. SLI-internally as well as -externally MIXED ) pattern of adjectival inflection which can be traced back to two applications of AI (or the lack thereof) at the Poss 0 -labels. Focusing on the location of this complex SLI itself in the nominal domain, let me turn to combinations of the quantifier with the second construction under analysis in ch. 0 above, i.e. PDCs. This tripartite SO results in grammatical structures in principle, although the concatenation is unlikely to be observed in natural language use. Nevertheless it bears additional implications for the categorical status of the complex SLI on the surface: (54) (dem) Martin sein-e meist-en gut-en Bücher [( the ) Martin ] DAT his most good books Recall that I argued above for additional movement of the complex quantifier from Poss PRO P further up the nominal spine to Q S P for the sake of uniformity with the remaining complex as well as the simple strong quantifier. However, in the light of the data just presented, one would have to stipulate an additional independent movement of the lexical possessive to Spec,Q S P, left adjacent to the SLI under analysis here. Alternatively, terminating IM of both unrelated configurations, viz. the weak quantifier as well as the lexical possessive at the prenominal possessive phrase seems a more economic option. Generalizing from here to all Poss PRO meist as well as Dmeist co-occurrences, movement of this complex SLI should uniformly be taken to terminate in the head position it attaches to and thus in the lowest phrase of the highest nominal cycle. I will follow this proposal though nothing really hinges on it here. Focusing on the set-up of nominal domains incorporating this SLI, one clearly notices that it occurs in uniformly casemarked extended projections when no additional SLIs partake in the formation (cf. 0 a., b. below); again, once D or Poss PRO occur between the quantifier and the nominal head, diverging case-values indicate the existence of two separate nominal domains (viz. HN and PN, cf. 0 c., parallel to 0 e., f. on the basis of ein jed above). This is surprisingly also true with two diverging higher SLIs, one concatenated with meist -, the other one following the complex SLI (cf. 0 d.). Moreover, partitives incorporating D/ Poss PRO on the basis of the concatenation Poss PRO meist are barred (cf. 0 e.). (55) a. d-as meist-e gut-e Wasser 50 [ the most good water ] NOM b. mein meist-es gutes Wasser [ my most good water ] NOM c. d-as meist-e d-es guten Wasser-s [ the most ] NOM [ the good water ] GEN d. d-as meist-e mein-es guten Wasser-s [ the most ] NOM [ my good water ] GEN e. * mein meist-es des/ mein-es guten Wasser-s [ my most ] NOM [ the / my good water ] GEN The proposals concerning the structural set-up underlying partitive constructions as discussed in ch. 0 b. and revised in the course of ch. 0 a. above hence readily carry over to the structures presented here but might be modified to apply to full phasal cycles rather than single SLIs. I have to leave this observation for further research here. 50 The prenominal adjective is employed here to facilitate possible genitive inflection on the nominal head in accordance with Gallmann’s (1996: (37)) ‘Genitive Rule’. Observe that the nominal domain nevertheless uniformly displays nominative case. Note that therein we would again propose two independent applications of AI under (extended) sisterhood, viz. from the lower D 0 to the weak quantifier as well as from the higher D 0 to the extended nominal projection, parallel to the inner workings proposed above for ein jed -. Therefore, when the possessive head partakes in the formation of said complex item, the result is a ‘uniformly disrupted’ (i.e. SLI-internally as well as -externally mixed ) pattern of adjectival inflection which can be traced back to two applications of AI (or the lack thereof) at the Poss 0 -labels. Focusing on the location of this complex SLI itself in the nominal domain, let me turn to combinations of the quantifier with the second construction under analysis in ch. 4.3.1 above, i.e. PDCs. This tripartite SO results in grammatical structures in principle, although the concatenation is unlikely to be observed in natural language use. Nevertheless it bears additional implications for the categorical status of the complex SLI on the surface: 352 4 IMPLICATIONS - Optionality & Complex Formations (54) (dem) Martin sein-e meist-en gut-en Bücher [( the ) Martin ] DAT his most good books Recall that I argued above for additional movement of the complex quantifier from Poss PRO P further up the nominal spine to Q S P for the sake of uniformity with the remaining complex as well as the simple strong quantifier. However, in the light of the data just presented, one would have to stipulate an additional independent movement of the lexical possessive to Spec,Q S P, left adjacent to the SLI under analysis here. Alternatively, terminating IM of both unrelated configurations, viz. the weak quantifier as well as the lexical possessive at the prenominal possessive phrase seems a more economic option. Generalizing from here to all Poss PRO meist as well as Dmeist co-occurrences, movement of this complex SLI should uniformly be taken to terminate in the head position it attaches to and thus in the lowest phrase of the highest nominal cycle. I will follow this proposal though nothing really hinges on it here. Focusing on the set-up of nominal domains incorporating this SLI, one clearly notices that it occurs in uniformly case-marked extended projections when no additional SLIs partake in the formation (cf. (55) a., b. below); again, once D or Poss PRO occur between the quantifier and the nominal head, diverging casevalues indicate the existence of two separate nominal domains (viz. HN and PN, cf. (55) c., parallel to (39) e., f. on the basis of ein jed above). This is surprisingly also true with two diverging higher SLIs, one concatenated with meist -, the other one following the complex SLI (cf. (55) d.). Moreover, partitives incorporating D/ Poss PRO on the basis of the concatenation Poss PRO meist are barred (cf. (55) e.). (55) a. d-as meist-e gut-e Wasser 40 [ the most good water ] NOM b. mein meist-es gutes Wasser [ my most good water ] NOM c. d-as meist-e d-es guten Wasser-s [ the most ] NOM [ the good water ] GEN d. d-as meist-e mein-es guten Wasser-s [ the most ] NOM [ my good water ] GEN e. * mein meist-es d-es/ mein-es guten Wasser-s [ my most ] NOM [ the / my good water ] GEN 40 The prenominal adjective is employed here to facilitate possible genitive inflection on the nominal head in accordance with Gallmann’s (1996: (37)) ‘Genitive Rule’. Observe that the nominal domain nevertheless uniformly displays nominative case. 4.3 Loose Ends & Complex Formations 353 The proposals concerning the structural set-up underlying partitive constructions as discussed in ch. 4.3.1 b. and revised in the course of ch. 4.3.2 a. above hence readily carry over to the structures presented here but might be modified to apply to full phasal cycles rather than single SLIs. I have to leave this observation for further research here. An additional argument in favor of the dual nature of partitive nominal domains comes from agreement facts of the determiner. Observe that in (55) a. and c. above, the determiner concatenated with the quantifier seems to display neuter agreement with HN. However, this is only apparently so: Observe that while D in its concatenation with meist indeed agrees with the noun in simple quantified nominal domains, this is not true in partitives, in which the initial instance of the determiner displays neuter agreement irrespective of the gender of PN. (56) a. d-er meist-e Wein [ the masc most wine ] NOM a'. d-ie meist-e Milch [ the fem most milk ] NOM b. d-as meist-e d-es Wein-s [ the neut most ] NOM [ the masc wine ] GEN b'. d-as meist-e d-er Milch [ the neut most ] NOM [ the fem milk ] GEN Here, the lower D agrees with PN while the higher D 0 bears the unmarked/ default value in German φ-agreement in accordance with the lexical non-distinctness requirement from C&G laid out in ch. 4.3.1 a. above (cf. Wiltschko 2009: 43, based on H&R’s (p. 486) proposal of the universal default value Inanimate/ Neuter for the class -node in their feature geometry). Observe that multiple, differently specified occurrences of D are likewise observable with the Spanish partitive strong quantifier la mayoría de los N as well as the Greek i perissoteri i N as reported by Etxeberria (2009: fn. 26, 2012: fn. 19, cf. also Giannakidou 2004: 122). For him, these data are relevant in the context of semantic domain restriction of nominal arguments in the syntax (i.e. the ‘explicit strategy’ to restrict the semantic application of nominal LIs, cf. von Fintel 1998), a function arguably held by the determiner in introducing a context set C (cf. Westerståhl 1984). Etxeberria therein builds on Gian- 354 4 IMPLICATIONS - Optionality & Complex Formations nakidou (2004), 41 herself arguing against Matthewson’s (2001) deviation from Generalized Quantifier Theory in the analysis of Salish St’át’imcets quantified nominal domains. As Matthewson argues based on argumental nominal hierarchies in said language, quantification universally proceeds along the following derivational lines: First, N conjoins with D of type <<e,t>,e>, returning an entity, which subsequently combines with Q of type <e,<<e,t>,t>> to return a generalized quantifier <<e,t>,t>. The determiner moreover introduces a variable over choice functions, rather than an iota function, returning one of the entities (singular or plural) which satisfy the predicate. Therein, D under Q functions referentially. Giannakidou (2004), arguing against this deviation of types of D and Q, reinterprets the required co-occurrence of those SLIs as a language-specific reflex to overtly restrict the domain of the nominal, a function standardly carried out either by partitives or covert domain variables at LF. Without going into detail concerning the semantic set-up and interaction of these simple or complex variables with the nominal core as well as with the quantifying elements (cf. Giannakidou 2004: ch. 4.), observe that Giannakidou additionally postulates a type-shifting operation (to wit, <<e,t>,t> → <e,t>) in the case of Salish St’át’imcets generalized quantifiers, thereby making the DP type <<e,t>,t> compatible with standard Q (i.e. <<e,t>,<<e,t>,t>>; cf. also Etxeberria & Giannakidou 2009: ch. 2.4, 2010: ch. 2.3). Furthermore equalizing the type shifter with an empty prepositional head located between the overt SLIs, and therein with the aforementioned partitives, Giannakidou finally excludes parallel applications in English and Greek where the availability of overt partitives blocks covert shifts. Her system, then, tightly links the context set to the determiner in languages overtly restricting the nominal domain but moreover hints at the necessity to permit this strategy in the covert domain restricting languages since both “complex determiners” (Giannakidou 2004: 121) in Greek (i.e. D-Q S ) as well as numerals and weak quantifiers, embedded under D in English, are observable. Building on these proposals, Etxeberria & Giannakidou (henceforth E&G, 2009, 2010), Etxeberria (2012) set out to further elaborate the system mainly on the basis of the contrast of Basque strong and weak quantifiers with the former obligatorily concatenated with D and the latter ungrammatical in such co-occurrences, an opposition already mentioned in Giannakidou (2004, cf. also Etxeberria 2008, 2009). Various simplifications of Giannakidou’s original claims therein target the determiner in that this SLI is now structurally understood to 41 Cf. also Etxeberria (2005, 2008); moreover the contributions in Giannakidou & Rathert (2009); an early statement of this rationale can be found in Greenberg’s (1978: 78) discussion of nominal classification systems. 4.3 Loose Ends & Complex Formations 355 constitute a type-preserving modifier while semantically, contextual restriction via C is identified as one of the functions making up definiteness (qua familiarity/ salience, viz. existence presupposition, as opposed to reference, i.e. the function of iota). 42 The properties associated with this version of the definite determiner are subsumed under the label ‘D DR ’ (P the property denoted by HN): (57) (= E&G 2009: (39), 2010: (27), Etxeberria 2012: (28)) Optionality & Complex Formations I MPLICATIONS 273 (57) (= E&G 2009: (39), 2010: (27), Etxeberria 2012: (28)) However, in contrast to Salish St’át’imcets, the authors furthermore propose that in Basque as well as in the other languages under consideration D DR is not able to contextually restrict the nominal head. Instead, the combination of D and N results in a determiner that functions referentially in these languages (i.e. in the second function subsumed under the notion definiteness, cf. E&G 2010: 104, Etxeberria 2012: 100). For D DR to apply to such configurations, a type shift to the appropriate input of the determiner is needed which these languages provide with overt partitive prepositions. However, for a subpart of the languages under consideration (viz. Basque and Greek in contrast to English), another path arriving at a D DR -ed nominal domains is available in the concatenation of quantified nominal hierarchies with D: 53 (58) ([Greek]) a. oli i fitites (= E&G 2010: (20) a.) all D.pl students b. o kathe fititis (= Giannakidou 2004: (32) b., [modified]) D.sg each student (59) (= Etxeberria 2005: ch. 2, (37), [modified, Basque]) a. mutil guzti-ak boy all-D.pl b. mutil bakoitz-a boy each-D.sg (60) a. * the all boys b. * the each boy Observe that the determiner is taken to constitute the dominating element in all these configurations (Basque is classified as a head-final language) and that this SLI is furthermore obligatorily and exclusively combined syntactically with the quantifier in the Basque (cf. 0, Etxeberria 2012: ch. 2.1). As elaborated by the authors, the reason for the ungrammaticality of the structures depicted in 0 is found in the fatality of the type mismatch in these languages, whereas in those patterning with Greek and Basque (as well as Salish St’át’imcets, cf. E&G 2009: 19), the function of the determiner is able to shift from iota to D DR . 54 In these concatenations, D is once again taken to modify the (now complex) nominal domain (‘Q-det’ the term for Q S , cf. fn. 53): (61) (= E&G 2009: (44), 2010: (32), Etxeberria 2012: (31)) , where Z is the relation denoted by Q-det In line with these semantic claims, this SLI is taken structurally to be formed by adjoining the determiner to the quantifier in parallel fashion to the mechanisms proposed above on the basis of complex SLIs in German. Moreover, returning to said structures with the proposals of E&G in place, we have gained additional, tentative evidence to propose a dedicated nominal domain for every overt occurrence of D (cf. fn. 54 above) and thus for a complex underlying phrasal set-up of simple partitive constructions in German since domain restriction only applies once per domain. Turning to the Spanish and Greek data given in the introduction to the current discussion, then, E&G bundle these together with nominal quantificational heads under the term ‘fraction expressions’ (cf. also E&G 2009: ch. 3.5, 2010: ch. 3.4, cf. furthermore E&G 2009: fn. 17 on this rationale in the treatment of Basque genitive constructions) and likewise classify them as underlyingly of the form “[DP plus DP]” (E&G 2009: 21), accordingly. Semantically, though, the initial determiner in concatenation with the quantifying (fraction) element is not classified as an instance of D DR by E&G. Below, I want to challenge that claim while returning to German data. 53 The authors split quantifiers along the lines of strong and weak instantiations, parallel to the classification laid out in chapter 1.3 above, and link the resulting classes to varying structural positions with strong instances located in a dedicated phrase and weak quantificational items located elsewhere (i.e. classified as adjectival/ modificational inside NP in E&G 2009: ch.2.2, head of a deeply embedded Number Phrase in E&G 2010: ch.4, Etxeberria 2012: ch. 6). Therein, all claims concerning quantifiers actually target Q S in the system elaborated here. Observe furthermore that the concatenations of D and Q W , excluded in the languages under consideration by E&G, are grammatical in German (cf. ch. 1.4.1. above). 54 This is not possible with partitives, however. To this end, the authors establish a parallel with nominal hierarchies modified by the same adjectival LI twice resulting in a shift in meaning of one of these items away from its lexical content to a degree modifier such as ‘very’ to avoid redundancy (cf. E&G 2009: 27, Etxeberria 2012: 93). The equally modificational determiner, however, seems to have no further function to shift to in these configurations (cf. E&G 2010: 110). However, in contrast to Salish St’át’imcets, the authors furthermore propose that in Basque as well as in the other languages under consideration D DR is not able to contextually restrict the nominal head. Instead, the combination of D and N results in a determiner that functions referentially in these languages (i.e. in the second function subsumed under the notion definiteness, cf. E&G 2010: 104, Etxeberria 2012: 100). For D DR to apply to such configurations, a type shift to the appropriate input of the determiner is needed which these languages provide with overt partitive prepositions. However, for a subpart of the languages under consideration (viz. Basque and Greek in contrast to English), another path arriving at a D DR -ed nominal domains is available in the concatenation of quantified nominal hierarchies with D: 43 (58) ([Greek]) a. oli i fitites (= E&G 2010: (20) a.) all D.pl students b. o kathe fititis (= Giannakidou 2004: (32) b., [modified]) D.sg each student 42 The identification of contextual domain restriction with (parts of) the concept of definiteness has been called into question by Martí (2009); cf. also the discussion in E&G 2009: ch. 5. 43 The authors split quantifiers along the lines of strong and weak instantiations, parallel to the classification laid out in chapter 1.3 above, and link the resulting classes to varying structural positions with strong instances located in a dedicated phrase and weak quantificational items located elsewhere (i.e. classified as adjectival/ modificational inside NP in E&G 2009: ch.2.2, head of a deeply embedded Number Phrase in E&G 2010: ch.4, Etxeberria 2012: ch. 6). Therein, all claims concerning quantifiers actually target Q S in the system elaborated here. Observe furthermore that the concatenations of D and Q W , excluded in the languages under consideration by E&G, are grammatical in German (cf. ch. 1.4.1. above). 356 4 IMPLICATIONS - Optionality & Complex Formations (59) (= Etxeberria 2005: ch. 2, (37), [modified, Basque]) a. mutil guzti-ak boy all-D.pl b. mutil bakoitz-a boy each-D.sg (60) a. * the all boys b. * the each boy Observe that the determiner is taken to constitute the dominating element in all these configurations (Basque is classified as a head-final language) and that this SLI is furthermore obligatorily and exclusively combined syntactically with the quantifier in the Basque (cf. (59), Etxeberria 2012: ch. 2.1). As elaborated by the authors, the reason for the ungrammaticality of the structures depicted in (60) is found in the fatality of the type mismatch in these languages, whereas in those patterning with Greek and Basque (as well as Salish St’át’imcets, cf. E&G 2009: 19), the function of the determiner is able to shift from iota to D DR . 44 In these concatenations, D is once again taken to modify the (now complex) nominal domain (‘Q-det’ the term for Q S , cf. fn. 43): (61) (= E&G 2009: (44), 2010: (32), Etxeberria 2012: (31)) Optionality & Complex Formations I MPLICATIONS 273 (57) (= E&G 2009: (39), 2010: (27), Etxeberria 2012: (28)) However, in contrast to Salish St’át’imcets, the authors furthermore propose that in Basque as well as in the other languages under consideration D DR is not able to contextually restrict the nominal head. Instead, the combination of D and N results in a determiner that functions referentially in these languages (i.e. in the second function subsumed under the notion definiteness, cf. E&G 2010: 104, Etxeberria 2012: 100). For D DR to apply to such configurations, a type shift to the appropriate input of the determiner is needed which these languages provide with overt partitive prepositions. However, for a subpart of the languages under consideration (viz. Basque and Greek in contrast to English), another path arriving at a D DR -ed nominal domains is available in the concatenation of quantified nominal hierarchies with D: 53 (58) ([Greek]) a. oli i fitites (= E&G 2010: (20) a.) all D.pl students b. o kathe fititis (= Giannakidou 2004: (32) b., [modified]) D.sg each student (59) (= Etxeberria 2005: ch. 2, (37), [modified, Basque]) a. mutil guzti-ak boy all-D.pl b. mutil bakoitz-a boy each-D.sg (60) a. * the all boys b. * the each boy Observe that the determiner is taken to constitute the dominating element in all these configurations (Basque is classified as a head-final language) and that this SLI is furthermore obligatorily and exclusively combined syntactically with the quantifier in the Basque (cf. 0, Etxeberria 2012: ch. 2.1). As elaborated by the authors, the reason for the ungrammaticality of the structures depicted in 0 is found in the fatality of the type mismatch in these languages, whereas in those patterning with Greek and Basque (as well as Salish St’át’imcets, cf. E&G 2009: 19), the function of the determiner is able to shift from iota to D DR . 54 In these concatenations, D is once again taken to modify the (now complex) nominal domain (‘Q-det’ the term for Q S , cf. fn. 53): (61) (= E&G 2009: (44), 2010: (32), Etxeberria 2012: (31)) , where Z is the relation denoted by Q-det In line with these semantic claims, this SLI is taken structurally to be formed by adjoining the determiner to the quantifier in parallel fashion to the mechanisms proposed above on the basis of complex SLIs in German. Moreover, returning to said structures with the proposals of E&G in place, we have gained additional, tentative evidence to propose a dedicated nominal domain for every overt occurrence of D (cf. fn. 54 above) and thus for a complex underlying phrasal set-up of simple partitive constructions in German since domain restriction only applies once per domain. Turning to the Spanish and Greek data given in the introduction to the current discussion, then, E&G bundle these together with nominal quantificational heads under the term ‘fraction expressions’ (cf. also E&G 2009: ch. 3.5, 2010: ch. 3.4, cf. furthermore E&G 2009: fn. 17 on this rationale in the treatment of Basque genitive constructions) and likewise classify them as underlyingly of the form “[DP plus DP]” (E&G 2009: 21), accordingly. Semantically, though, the initial determiner in concatenation with the quantifying (fraction) element is not classified as an instance of D DR by E&G. Below, I want to challenge that claim while returning to German data. 53 The authors split quantifiers along the lines of strong and weak instantiations, parallel to the classification laid out in chapter 1.3 above, and link the resulting classes to varying structural positions with strong instances located in a dedicated phrase and weak quantificational items located elsewhere (i.e. classified as adjectival/ modificational inside NP in E&G 2009: ch.2.2, head of a deeply embedded Number Phrase in E&G 2010: ch.4, Etxeberria 2012: ch. 6). Therein, all claims concerning quantifiers actually target Q S in the system elaborated here. Observe furthermore that the concatenations of D and Q W , excluded in the languages under consideration by E&G, are grammatical in German (cf. ch. 1.4.1. above). 54 This is not possible with partitives, however. To this end, the authors establish a parallel with nominal hierarchies modified by the same adjectival LI twice resulting in a shift in meaning of one of these items away from its lexical content to a degree modifier such as ‘very’ to avoid redundancy (cf. E&G 2009: 27, Etxeberria 2012: 93). The equally modificational determiner, however, seems to have no further function to shift to in these configurations (cf. E&G 2010: 110). In line with these semantic claims, this SLI is taken structurally to be formed by adjoining the determiner to the quantifier in parallel fashion to the mechanisms proposed above on the basis of complex SLIs in German. Moreover, returning to said structures with the proposals of E&G in place, we have gained additional, tentative evidence to propose a dedicated nominal domain for every overt occurrence of D (cf. endnote *26) and thus for a complex underlying phrasal setup of simple partitive constructions in German since domain restriction only applies once per domain. Turning to the Spanish and Greek data given in the introduction to the current discussion, then, E&G bundle these together with 44 This is not possible with partitives, however. To this end, the authors establish a parallel with nominal hierarchies modified by the same adjectival LI twice resulting in a shift in meaning of one of these items away from its lexical content to a degree modifier such as ‘very’ to avoid redundancy (cf. E&G 2009: 27, Etxeberria 2012: 93). The equally modificational determiner, however, seems to have no further function to shift to in these configurations (cf. E&G 2010: 110). 4.3 Loose Ends & Complex Formations 357 nominal quantificational heads under the term ‘fraction expressions’ (cf. also E&G 2009: ch. 3.5, 2010: ch. 3.4, cf. furthermore E&G 2009: fn. 17 on this rationale in the treatment of Basque genitive constructions) and likewise classify them as underlyingly of the form “[DP plus DP]” (E&G 2009: 21), accordingly. Semantically, though, the initial determiner in concatenation with the quantifying (fraction) element is not classified as an instance of D DR by E&G. Below, I want to challenge that claim while returning to German data. Note that it is expected even for structurally simpler nominal SOs headed by determiners in their function as D DR to exhibit certain semantic and syntactic properties that sets them apart from regular (i.e. iota) DPs (as well as failing to display others associated with reference as the other property subsumed under ‘definiteness’). Indeed, E&G (2009: ch. 3.2) list various syntactic characteristics in which the determiner in concatenation with a strong quantifier diverges from the one in isolated use. The instance of D concatenated with meist in German should hence be testable w.r.t. its semantic function. However, the aforementioned test environments only target the categorical status of the complex SO (i.e. DP vs. QP) to highlight the modificational nature of the determiner therein. Moreover, most of these criteria are not transferable to the language under investigation here for reasons not immediately related to the phenomenon at hand. 45 The single environment from those presented by E&G which is applicable to German in a modified version, then, gives opposing results: As the authors note, since it is possible to conjoin two NPs or APs under a single D in Basque, coordinating two strong quantified nominal hierarchies under one D should turn out equally well with DP constituting the topmost phrase. The result is negative and hence points in the direction of a modifying D, D DR , in the Basque QP. In German, such coordinations turn out semantically odd. 46 However, it is possible to modify this environment slightly: As has been observed above, AI is not dependent on the quantifier but the SLI it is concatenated with, which initially led to the proposal of a dominating head D/ Poss PRO in the complexes under analysis here. These observations are consistent with those from adjectival inflection in multiple nominal hierarchies coordinated under a single instance of said item: 45 These include restrictions coordinating DP under demonstratives in Greek as well as the impossibility of D to reduplicate therein in said language. E&G (2009: fn. 9) note the impossibility to remodel the former environment in Basque in which D and Dem are categorically identical. This argument also extends to the second criterion when focusing on German which does not exhibit DP reduplication. 46 Recall that in contrast to the Basque restrictions only few complex quantifiers exist in German (cf. Pafel 1994: ch. 3, also Pafel 2005: (1.61)) with those of the subtype D-Q semantically incompatible to the item under analysis: i.e. d selb -, d gleich - (both translated ‘the same’) and d jen - (relative pronoun, ‘s/ he who’). 358 4 IMPLICATIONS - Optionality & Complex Formations (62) die meisten alten Männer und jungen Hunde the most [ old weak men ] and [ young weak dogs ] As can be deduced therefrom, both adjectives display weak , hence impoverished, inflection. I therefore maintain my previous proposal of a dominating head D/ Poss PRO in the buildup of the complex quantifier d meist -, pace E&G. However, I do believe one nevertheless finds evidence that it fulfills the function D DR in these concatenations. To begin with, recall that contextual domain restriction has been taken to constitute the subpart of definiteness presupposing existence to the exclusion of reference (i.e. the iota-function). Turning to the second instance of the category D in German, the demonstrative, the semantic import diverging from the determiner, i.e. ostensive/ deictic information, is straightforwardly subsumed under the latter. This is furthermore supported by the inability of such SLIs to be involved in the composition of generic statements (cf. E&G 2009: 18, also ch. 5.2.2). Likewise, these items are unable to partake in the establishment of the complex quantifier. 47 Further dwelling on the semantic criteria associated with D DR by the authors, the aforementioned presupposition is moreover taken to enforce a restriction of non-emptiness on the set referred to by the nominal (E&G 2009: ch. 3.3, 2010: ch. 3.2, Etxeberria 2012: ch. 2.3). Therefore, subsequently negating the existence of all members in a previously established set turns out to be contradictory and thus ungrammatical. Observe that presuppositionality is not a property tantamount with the strong/ weak quantifier distinction: 48 (63) a. Wenn du viele Fehler findest, gebe ich Dir fünf Euro, if you many mistakes find give I you five euros b. aber vielleicht gibt es keine Fehler. but maybe exist it no mistakes ‘If you find many mistakes I’ll give you five euros, but maybe there are none.’ 47 This is the intuition of my informants though a quick google search does return a significant number of hits for the string “diese meisten” (‘these most’); however, abstracting away from earlier stages of German, these seem to conform to the indefinite/ kind uses of the SLI laid out in chapter 1.3.3 above. 48 As Musan (1999: 629f.) notes, context presupposition is moreover a property of ‘definites’ as well as (weakly quantified) partitives inheriting it from the former structures which are coordinated therein. The latter have moreover been argued to function as D DR by E&G in cases where context restriction at N is not possible. 4.3 Loose Ends & Complex Formations 359 (64) a. Wenn du alle Fehler findest, gebe ich Dir fünf Euro, if you all mistakes find give I you five euros b. aber vielleicht gibt es keine Fehler. but maybe exist it no mistakes ‘If you find all mistakes I’ll give you five euros, but maybe there are none.’ In contrast, turning to the complex SLI under consideration here, we do observe ungrammaticality associated with presuppositionality, as is expected by now. What is more surprising at this point is the fact that it likewise holds for the other complex SLI under discussion earlier in this chapter (cf. (66)) as well as the default concatenation of the determiner with a weak quantifier in German (cf. fn. 43, cf. (67)): (65) a. Wenn du die meisten Fehler findest, gebe ich Dir fünf Euro, if you the most mistakes find give I you five euros b. * aber vielleicht gibt es keine Fehler. but maybe exist it no mistakes ‘If you find most mistakes I’ll give you five euros, but maybe there are none.’ (66) a. Wenn du (einen) jeden Fehler findest, gebe ich Dir fünf Euro, if you (ein) every mistake find give I you five euros b. * aber vielleicht gibt es keine Fehler. but maybe exist it no mistakes ‘If you find every mistake I’ll give you five euros, but maybe there is none.’ (67) a. Wenn du die vielen Fehler findest, gebe ich Dir fünf Euro, if you the many mistakes find give I you five euros b. * aber vielleicht gibt es keine Fehler. but maybe exist it no mistakes ‘If you find the many mistakes I’ll give you five euros, but maybe there are none.’ Concerning the former complex SLI, one might be inclined to make a case concerning its function to contextually restrict the domain of HN and indeed a 360 4 IMPLICATIONS - Optionality & Complex Formations parallel proposal has been put forth by the authors concerning one of its English equivalents, namely each , incorporating a null-D (cf. E&G 2009: fn. 15) which arguably constitutes the basis for its derivation from every (cf. E&G 2010: 15), viz. the second equivalent of jed -. Even more so, and once again employing a decompositional rationale, the seemingly simple quantifier jed might be viewed to be itself complex, comprised of a weak quantifier and the determiner d -root (cf. ch. 1.4.4 above, Klinge 2008). A tentative argument in this direction might be found in the contrast with another SLI in German, the distal demonstrative jen -. This item employs the initial morpheme jed to the exclusion of d and bears deictic semantics which the former quantifier does not convey. However, note that the quantifier under consideration constitutes the prime SLI for use in generic statements. I will not be able go into detail on these matters here. Returning to the paradigm above, the concatenation of uniformly casemarked D and Q W in the environment in (67) demonstrates the ability of the determiner to function as D DR in base position in German under current assumptions. This observation carries over to the complex SLI subsumed under D 0 . Diverging from the findings of E&G, then, syntactically regular (i.e. nonmodifying) heads must still be taken as able to fulfil the function of contextual domain restriction semantically. German therein nevertheless seems to conform to the bivalent typology of E&G (2009: ch. 3.4) allowing contextual restriction at N (via D in base position as in Salish St’át’imcets at least with weak quantifiers) as well as at Q (via adjunction/ head-movement as the highest head rather than a modifier). It is, however, surprising to also find a partitive construction in German, which has been argued by E&G to function as a domain restrictor in languages which are unable to apply D DR at N. This construction furthermore readily combines with the former option of D DR contrary to the overt concatenations available in Salish St’át’imcets. However, I will have to leave these deeper issues without further comment here. Howsoever, following from the variability in morphological form of the complex strong quantifier d meist - (and arguably also the underlying set-up of ( ein -) je d -), we can finally associate D DR in German with a subpart of the nominal hierarchy, to wit, DP > Poss PRO P. While certain underlyingly weak quantifiers are forced to attach to the SLIs hosted therein to acquire the intended reading, the former elements themselves are able to contextually restrict even in simple concatenations. To extend even further, and building on (64) above, this moreover covers concatenations with the native Q S0 all as demonstrated in (68) below. Moreover, definiteness in German, as analyzed in the previous chapter, does not constitute the relevant notion as is demonstrated below with the possessive pronoun and a mass nominal bearing [φ[n]]: 4.4 Conclusion 361 (68) a. Wenn du all(-e) die Fehler findest, gebe ich Dir fünf Euro, if you all the mistakes find give I you five euros b. * aber vielleicht gibt es keine Fehler. but maybe exist it no mistakes ‘If you find all mistakes I’ll give you five euros, but maybe there are none.’ (69) a. Wenn du all mein Laub auffegst, gebe ich Dir fünf Euro, if you all my leaves sweep give I you five euros b. * aber vielleicht gibt es kein Laub. but maybe exist it no leaves ‘If you sweep all my leaves I’ll give you five euro, but maybe there are none.’ We can hence dissociate D DR from definiteness even in a language that implements this feature morpho-syntactically (in a modification of E&G’s (2009: 18) own claims) and instead associate it categorically with a subpart of the nominal spine elaborated above. Therein, the semantic motivation for the buildup and the proposed derivation of multi-member quantifiers is uncovered. 4.4 Conclusion This chapter focused on a selection of topics either touched on or influenced by the derivational apparatus and claims made in foregone chapters. Chapter 4.2 laid out the implications carried by the system for the optionality of both movement as well as inflection on quantifying elements in the nominal domain in German. Concerning the former, it has been demonstrated in chapter 4.2.1 that this type of IM necessarily begins its derivational life in the Numeration as an optional EPP/ OCC-feature. In a parallel fashion, the prerequisite for the latter has been identified as the superfluous coding of features in chapter 4.2.2, while obligatory inflection was traced back to the retrieval of nominal φ-features for the ongoing verbal derivation. With obligatorily absent inflections attributed to unsuccessful applications of Match/ Value as well as Match at root nodes in chapter three above, the full paradigm of quantifier inflections has therefore been accounted for. 362 4 IMPLICATIONS - Optionality & Complex Formations Chapter 4.3 then took stock of the applicability of the system elaborated throughout this inquiry to a variety of complex syntactic as well as lexical configurations beginning with the revision of possessor doubling constructions and simple partitives, already touched on in the course of previous chapters, before presenting preliminary accounts of the narrow-syntactic compositionality of the complex, i.e. multi-SLI quantifiers ( ein -) jed and d meist in terms of the phrasal heads identified earlier. Moreover, these latter analyses uncovered additional mechanisms underlying the aforementioned partitives (in the case of the former) as well as the explicit strategy of contextual domain restriction (in the case of the latter). It is, then, due to this wide selection of topics that said subchapter necessarily poses only a first approximation to their respective treatments under these new derivational proposals and further research is obviously needed for all of them, as has been put forth numerous times above. Nevertheless, I am positive to have demonstrated the wider practicability of the system and therefore take the set-up of the nominal domain, elaborated throughout chapter one and two, as well as the derivational mechanisms holding therein, carved out throughout chapter two and three, to pose a stable basis for inquiry into syntactic phenomena in the nominal domain in German apart from the core phenomenon of nominal concord. In the final section of this book, viz. chapter five below, I will revisit all the major stepping stones that led to this wholesome approach, i.e. the theory of φ-syntax for nominal concord in German, in sketching once again the path from the initial discussion on the simple axioms of the current inquiry in chapter one to the elaboration of the technical premises carved out on the basis of languagespecific phenomena. Furthermore, the chapter will also compile the technical guidelines and derivational principles uncovered in the course of the preceding analysis. Therein, a comprehensive recapitulation is provided for the reader. 4.4 Conclusion 363 5 Conclusion The present analysis set out to derive nominal concord as a derivational phenomenon in German and to therein reanalyze morphological definiteness as the locus of exponence of the complex φ-feature category number/ gender in the extended nominal projection. To this end, the book firstly set the theoretic stage for the analysis in a presentation of the broad theoretical framework of the Minimalist Program for linguistic theory, the current instantiation of Generative Grammar, by a compiled overview of the derivational model of FL in chapter 1.2. Subsequently, the chapter went on to lay out two core notions of the ensuing investigation, viz. grammatical definiteness and the strong / weak adjectival inflection dichotomy in chapter 1.3.1 and 1.3.2 respectively. Building on this, the rigid causal relation between the status of prenominal semi-lexical items regarding the former on the choice of the latter was carved out in chapter 1.3.3. This was achieved by translating the testing environments for definiteness from analyses conducted in English (a language that does not display said inflectional dichotomy on adjectival LIs) and show that these are applicable in the language under consideration. Subsequently, the relation between adjectival forms and the semi-lexical items they concatenate with was elaborated. Finally, the overlay of the resulting classes of prenominal elements revealed adjectival inflection to be indeed sensitivity to grammatical definiteness. This state of affairs has been subsequently depicted in the schematic reprinted below as (1): (1) (= chapter 1, (29)) SLI definite ↔ Adj WEAK SLI indefinite ↔ Adj STRONG The remainder of the chapter then concentrated on the phrasal set-up of the nominal domain in German. On the basis of linearizational phenomena, and worked out against trends in the research tradition as well as a variety of structural proposals found in the current literature, two decided phrases hosting (i) strong and weak quantifiers (i.e. Q S P, Q W P) as well as (ii) pronominal and lexical possessives (i.e. Poss PRO P, Poss LEX P), encircling the core categories N and D, have been proposed, arriving at the rigid phrasal set-up depicted in (2) below: 364 5 Conclusion (2) (= chapter 1, (77) b.) 299 5 Conclusion The present analysis set out to derive nominal concord as a derivational phenomenon in German and to therein reanalyze morphological definiteness as the locus of exponence of the complex φ-feature category number/ gender in the extended nominal projection. To this end, the book firstly set the theoretic stage for the analysis in a presentation of the broad theoretical framework of the Minimalist Program for linguistic theory, the current instantiation of Generative Grammar, by a compiled overview of the derivational model of FL in chapter 1.2. Subsequently, the chapter went on to lay out two core notions of the ensuing investigation, viz. grammatical definiteness and the STRONG / WEAK adjectival inflection dichotomy in chapter 1.3.1 and 1.3.2 respectively. Building on this, the rigid causal relation between the status of prenominal semi-lexical items regarding the former on the choice of the latter was carved out in chapter 1.3.3. This was achieved by translating the testing environments for definiteness from analyses conducted in English (a language that does not display said inflectional dichotomy on adjectival LIs) and show that these are applicable in the language under consideration. Subsequently, the relation between adjectival forms and the semi-lexical items they concatenate with was elaborated. Finally, the overlay of the resulting classes of prenominal elements revealed adjectival inflection to be indeed sensitivity to grammatical definiteness. This state of affairs has been subsequently depicted in the schematic reprinted below as 0: (1) (= chapter 1, (29)) a. SLI definite ↔ Adj WEAK b. SLI indefinite ↔ Adj STRONG The remainder of the chapter then concentrated on the phrasal set-up of the nominal domain in German. On the basis of linearizational phenomena, and worked out against trends in the research tradition as well as a variety of structural proposals found in the current literature, two decided phrases hosting (i) strong and weak quantifiers (i.e. Q S P, Q W P) as well as (ii) pronominal and lexical possessives (i.e. Poss PRO P, Poss LEX P), encircling the core categories N and D, have been proposed, arriving at the rigid phrasal set-up depicted in 0 below: (2) (= chapter 1, (77) b.) Q S P 3 Q S D 3 D Poss PRO P 3 Poss PRO Q W P 3 Q W NP 3 N Poss LEX P The chapter concluded with the discussion of pronominal SLIs in chapter 1.4.4 and therein presented an overview on two additional structural concepts that surfaced prominently in the course of the succeeding analysis, to wit, the decompositional approach to semi-lexical items and the feature-geometrical approach to formal features. However carefully carved out, the reader has been asked to keep in mind the stipulative character of both the main topics of this chapter. These have to be equalized with the tenets of the general theory in that the dismissal of either one trivializes the findings made on their basis in subsequent chapters. It is this characteristic that led to the categorization of the chapter as the ‘axioms’ current investigation. The chapter concluded with the discussion of pronominal SLIs in chapter 1.4.4 and therein presented an overview on two additional structural concepts that surfaced prominently in the course of the succeeding analysis, to wit, the decompositional approach to semi-lexical items and the feature-geometrical approach to formal features. However carefully carved out, the reader has been asked to keep in mind the stipulative character of both the main topics of this chapter. These have to be equalized with the tenets of the general theory in that the dismissal of either one trivializes the findings made on their basis in subsequent chapters. It is this characteristic that led to the categorization of the chapter as the ‘axioms’ current investigation. The goal of the ensuing chapter was twofold: On the one hand, I focused on the phasal status of the phrasal, prenominal SLIs identified over the course of chapter one, on the other hand, I elaborated on the structural location and the interdependencies of φ-feature values of the categories number and gender therein. Concerning the former, and starting out with Richards’ (2007) consideration on phase-plus-non-phase-complement (P-N) concatenations as instances of good design, data from ellipsis as well as extraction in/ from nominal domains in German were presented in parallel to those laid out by Bošković (2014) to therein apply his generalizations on phasal properties from Serbo- Croatian to the language under consideration. As has been noted, one category does not adhere to the highest-phrase-as-phase rationale put forth by Bošković, viz. the quantificational instances Q S and Q W . These were identified as the highest phrases though least prominent phase heads of their respective cycles. The phasal properties of SLIs in German were subsequently captured by means of a hierarchy of assignment with regard to the remaining SLIs, concatenated in a given linearization. As has been argued, the considerations on good phasal design from Richards (2007) still apply, since the rationale of phase-plus-non- 5 Conclusion 5 Conclusion 365 phase-complement carried over to the object of investigation under (i) the proposal of two cycles coupled with (ii) the aforementioned hierarchy of assignment. These notions are combined in the schematic in (3) below with * marking rigid phasehood (if the respective lexical item is overtly present), ( * ) marking variable phasehood, numbers in the top row referring to the position in the hierarchy of assignment, numbers in the lower row depicting the preference for the assignment of phasehood concerning the two cycles w.r.t. each other. The dotted line separates said cycles. (3) C ONCLUSION 300 The goal of the ensuing chapter was twofold: On the one hand, I focused on the phasal status of the phrasal, prenominal SLIs identified over the course of chapter one, on the other hand, I elaborated on the structural location and the interdependencies of φ-feature values of the categories number and gender therein. Concerning the former, and starting out with Richards’ (2007) consideration on phase-plus-non-phase-complement (P-N) concatenations as instances of good design, data from ellipsis as well as extraction in/ from nominal domains in German were presented in parallel to those laid out by Bošković (2014) to therein apply his generalizations on phasal properties from Serbo-Croatian to the language under consideration. As has been noted, one category does not adhere to the highest-phrase-as-phase rationale put forth by Bošković, viz. the quantificational instances Q S and Q W . These were identified as the highest phrases though least prominent phase heads of their respective cycles. The phasal properties of SLIs in German were subsequently captured by means of a hierarchy of assignment with regard to the remaining SLIs, concatenated in a given linearization. As has been argued, the considerations on good phasal design from Richards (2007) still apply, since the rationale of phaseplus-non-phase-complement carried over to the object of investigation under (i) the proposal of two cycles coupled with (ii) the aforementioned hierarchy of assignment. These notions are combined in the schematic in 0 below with * marking rigid phasehood (if the respective lexical item is overtly present), ( * ) marking variable phasehood, numbers in the top row referring to the position in the hierarchy of assignment, numbers in the lower row depicting the preference for the assignment of phasehood concerning the two cycles w.r.t. each other. The dotted line separates said cycles. (3) Q S( * ) D* Poss PRO( * ) Q W( * ) Num ( * ) 3 1 2 2 1 1 2 Subsequently, the system was contrasted with the framework of Despić (2011) as well as with apparently simpler structural proposals thereafter in chapter 2.2.2, the latter incorporating only a single functional head but permitting multiple specifiers instead. There, it has been shown that neither approach was able to account for the full paradigm of grammatical configurations in German while at the same time excluding those that have been judged as ungrammatical by my informants. The schematic in 0 above additionally depicts the sole structural modification to the extended projection, proposed in chapter one, namely the addition of the functional head Num 0 . This projection was first proposed in the analysis of complex remnants in split-topicalization constructions and was therein taken to host plural morphology. Turning to the second topic of the chapter, I subsequently focused on the location of gender-features in the nominal domain in German, following Ritter (1993). It could be observed that the parallel patterning of masculine and neuter gender as located at the core lexical projection N 0 contrasted with the surprising parallelism of plural number and female gender as located on the head of NumP with regard to Ritter’s own testing grounds. The remainder of the chapter was then devoted to the exact formalization of this division, discussing various instantiations of narrow-syntactic feature coding (e.g. Harley 1994, Harley & Ritter 2002a) and their resulting interactions in post-syntactic Morphology (e.g. Halle & Marantz 1993, Noyer ([1992] 1997) to account for the full paradigm of (i) singular, (ii) gendered mass, (iii) plural as well as (iv) feminized singular and plural nominals in German. The feature-structure, ultimately proposed for the two phrasal heads N 0 and Num 0 are given in 0 a. and b. respectively, below: (4) a. (= chapter 2, (63)) N 3 singular gender (minimal) (class) 3 [m] [n] b. (= chapter 2, (64)) Num ! [f] ! [pl] Subsequently, the system was contrasted with the framework of Despić (2011) as well as with apparently simpler structural proposals thereafter in chapter 2.2.2, the latter incorporating only a single functional head but permitting multiple specifiers instead. There, it has been shown that neither approach was able to account for the full paradigm of grammatical configurations in German while at the same time excluding those that have been judged as ungrammatical by my informants. The schematic in (3) above additionally depicts the sole structural modification to the extended projection, proposed in chapter one, namely the addition of the functional head Num 0 . This projection was first proposed in the analysis of complex remnants in split-topicalization constructions and was therein taken to host plural morphology. Turning to the second topic of the chapter, I subsequently focused on the location of gender-features in the nominal domain in German, following Ritter (1993). It could be observed that the parallel patterning of masculine and neuter gender as located at the core lexical projection N 0 contrasted with the surprising parallelism of plural number and female gender as located on the head of NumP with regard to Ritter’s own testing grounds. The remainder of the chapter was then devoted to the exact formalization of this division, discussing various instantiations of narrow-syntactic feature coding (e.g. Harley 1994, Harley & Ritter 2002a) and their resulting interactions in post-syntactic Morphology (e.g. Halle & Marantz 1993, Noyer [1992] 1997) to account for the full paradigm of (i) singular, (ii) gendered mass, (iii) plural as well as (iv) feminized singular and plural nominals in German. The feature-structure, ultimately proposed for the two phrasal heads N 0 and Num 0 are given in (4) a. and b. respectively, below: 366 5 Conclusion (4) C ONCLUSION 300 The goal of the ensuing chapter was twofold: On the one hand, I focused on the phasal status of the phrasal, prenominal SLIs identified over the course of chapter one, on the other hand, I elaborated on the structural location and the interdependencies of φ-feature values of the categories number and gender therein. Concerning the former, and starting out with Richards’ (2007) consideration on phase-plus-non-phase-complement (P-N) concatenations as instances of good design, data from ellipsis as well as extraction in/ from nominal domains in German were presented in parallel to those laid out by Bošković (2014) to therein apply his generalizations on phasal properties from Serbo-Croatian to the language under consideration. As has been noted, one category does not adhere to the highest-phrase-as-phase rationale put forth by Bošković, viz. the quantificational instances Q S and Q W . These were identified as the highest phrases though least prominent phase heads of their respective cycles. The phasal properties of SLIs in German were subsequently captured by means of a hierarchy of assignment with regard to the remaining SLIs, concatenated in a given linearization. As has been argued, the considerations on good phasal design from Richards (2007) still apply, since the rationale of phaseplus-non-phase-complement carried over to the object of investigation under (i) the proposal of two cycles coupled with (ii) the aforementioned hierarchy of assignment. These notions are combined in the schematic in 0 below with * marking rigid phasehood (if the respective lexical item is overtly present), ( * ) marking variable phasehood, numbers in the top row referring to the position in the hierarchy of assignment, numbers in the lower row depicting the preference for the assignment of phasehood concerning the two cycles w.r.t. each other. The dotted line separates said cycles. (3) Q S( * ) D* Poss PRO( * ) Q W( * ) Num ( * ) 3 1 2 2 1 1 2 Subsequently, the system was contrasted with the framework of Despić (2011) as well as with apparently simpler structural proposals thereafter in chapter 2.2.2, the latter incorporating only a single functional head but permitting multiple specifiers instead. There, it has been shown that neither approach was able to account for the full paradigm of grammatical configurations in German while at the same time excluding those that have been judged as ungrammatical by my informants. The schematic in 0 above additionally depicts the sole structural modification to the extended projection, proposed in chapter one, namely the addition of the functional head Num 0 . This projection was first proposed in the analysis of complex remnants in split-topicalization constructions and was therein taken to host plural morphology. Turning to the second topic of the chapter, I subsequently focused on the location of gender-features in the nominal domain in German, following Ritter (1993). It could be observed that the parallel patterning of masculine and neuter gender as located at the core lexical projection N 0 contrasted with the surprising parallelism of plural number and female gender as located on the head of NumP with regard to Ritter’s own testing grounds. The remainder of the chapter was then devoted to the exact formalization of this division, discussing various instantiations of narrow-syntactic feature coding (e.g. Harley 1994, Harley & Ritter 2002a) and their resulting interactions in post-syntactic Morphology (e.g. Halle & Marantz 1993, Noyer ([1992] 1997) to account for the full paradigm of (i) singular, (ii) gendered mass, (iii) plural as well as (iv) feminized singular and plural nominals in German. The feature-structure, ultimately proposed for the two phrasal heads N 0 and Num 0 are given in 0 a. and b. respectively, below: (4) a. (= chapter 2, (63)) N 3 singular gender (minimal) (class) 3 [m] [n] b. (= chapter 2, (64)) Num ! [f] ! [pl] The features number and gender are hence conflated into one complex φ-category. As I argued there, feminine gender can be understood to constitute a semantically numberless syntactic plural marker in German. The chapter concluded with the presentation of various inflectional phenomena in German that might arguably receive a unified treatment in the realm of premises laid out thus far, to wit, parallelisms of plural and feminine nominal domains concerning (i) umlaut, (ii) genitive suffixation, (iii) adjective inflection with feminine and plural personal determiners as well as (iv) the mixed adjectival inflection pattern in general. A final formalization of the proposal was conducted within the notions of target and controller gender from Corbett (1991) and concluded in a last refinement in his terms of subgender/ dependent target gender. Both main topics of the chapter thus build on the proposals from the foregoing section while providing the notions in which the ensuing analysis of nominal concord and morphological definiteness would be conducted. It is this interposed nature that led to the designation of the chapter as the ‘premises’ of the analysis. Chapter three subsequently set out to incorporate the findings arrived at thus far into a derivational system that proves capable to account for the full range of inflectional idiosyncrasies subsumed under the term nominal concord. To this end, I devoted some space to the examination and careful development of two core notions for the analysis to follow, to wit, φ-feature dependencies and the operation Agree in the initial subchapters 3.2.1 and 3.2.2 respectively. Therein, I elaborated on Bejar’s (2003) application of Harley & Ritter’s (2002a) feature geometry, on the one hand, to argue for the narrow-syntactic relevance 5 Conclusion 367 of φ-features and to formalize my previous findings, accordingly. This resulted in the proposal of the geometry in (5) below: (5) (= chapter 3, (10)) C ONCLUSION 301 The features number and gender are hence conflated into one complex φ-category. As I argued there, feminine gender can be understood to constitute a semantically numberless syntactic plural marker in German. The chapter concluded with the presentation of various inflectional phenomena in German that might arguably receive a unified treatment in the realm of premises laid out thus far, to wit, parallelisms of plural and feminine nominal domains concerning (i) umlaut, (ii) genitive suffixation, (iii) adjective inflection with feminine and plural personal determiners as well as (iv) the MIXED adjectival inflection pattern in general. A final formalization of the proposal was conducted within the notions of target and controller gender from Corbett (1991) and concluded in a last refinement in his terms of subgender/ dependent target gender. Both main topics of the chapter thus build on the proposals from the foregoing section while providing the notions in which the ensuing analysis of nominal concord and morphological definiteness would be conducted. It is this interposed nature that led to the designation of the chapter as the ‘premises’ of the analysis. Chapter three subsequently set out to incorporate the findings arrived at thus far into a derivational system that proves capable to account for the full range of inflectional idiosyncrasies subsumed under the term nominal concord. To this end, I devoted some space to the examination and careful development of two core notions for the analysis to follow, to wit, φfeature dependencies and the operation Agree in the initial subchapters 3.2.1 and 3.2.2 respectively. Therein, I elaborated on Bejar’s (2003) application of Harley & Ritter’s (2002a) feature geometry, on the one hand, to argue for the narrowsyntactic relevance of φ-features and to formalize my previous findings, accordingly. This resulted in the proposal of the geometry in 0 below: (5) (= chapter 3, (10)) INDIVIDUATION 9 [ f ] [sg] class ! 3 [pl] [m] [n] The following subchapter, on the other hand, examined the full body of operations that are taken to constitute the sole motor of narrow-syntactic derivations. To begin, the notions employed in standard Minimalism were discussed and furthermore contrasted with various alternative proposals. The refined notions of features and their renewed narrowsyntactic locus of application were subsequently employed to flesh out a derivational apparatus whose components have each proven viable apart from the object under investigation in this book (namely in argument-predicate agreement in the clausal domain) and furthermore circumvent theory-internal shortcomings, identified over the course of the discussion. The core notions comprising said system are given below: (6) (= chapter 3, (41); Chomsky MI: (40) b., c.) a. D(P) is the sister of P. b. Locality reduces to ‘closest c-command.’ (7) (= chapter 3, (42); Bejar 2003: ch. 2, (35)-(38)) a. Entailment Condition Match is defined by entailment. b. Probe (F) and Goal (F’) match if Goal (F’) entails Probe (F). c. Match is evaluated at the root. where d. Probe (F) and Goal (F’) match if Goal (F’) entails root Probe (F). (8) (= chapter 3, (43); Bejar 2003: ch. 2, (58)) Condition on Value G(oal) values P(robe) iff f’(G) entails f(P) (9) (= chapter 3, (44)) Agree (Feature sharing version) Replace f(P) with f’(G), so that the same feature is present in both locations. Thereafter, the system was put to use in the initial discussion of mono-SLI nominal domains in chapter 3.3.1. Here, first mismatches of featural configurations and the subsequent reduction of the respective SLI’s feature set-up to partial default, viz. zero-morphology on the surface, were observed in accordance with the proposed workings carved out prior to that point. These latter notions were not only further sharpened in the course of the analysis, the selectional characteristics of the individual semi-lexical items as well as their ability to surface in the aforementioned reduced form were moreover recorded in the table of properties of SLIs. The following subchapter, on the other hand, examined the full body of operations that are taken to constitute the sole motor of narrow-syntactic derivations. To begin, the notions employed in standard Minimalism were discussed and furthermore contrasted with various alternative proposals. The refined notions of features and their renewed narrow-syntactic locus of application were subsequently employed to flesh out a derivational apparatus whose components have each proven viable apart from the object under investigation in this book (namely in argument-predicate agreement in the clausal domain) and furthermore circumvent theory-internal shortcomings, identified over the course of the discussion. The core notions comprising said system are given below: (6) (= chapter 3, (41); Chomsky MI: (40) b., c.) a. D(P) is the sister of P. b. Locality reduces to ‘closest c-command.’ (7) (= chapter 3, (42); Bejar 2003: ch. 2, (35)-(38)) a. Entailment Condition Match is defined by entailment. b. Probe (F) and Goal (F’) match if Goal (F’) entails Probe (F). c. Match is evaluated at the root. where d. Probe (F) and Goal (F’) match if Goal (F’) entails root Probe (F). (8) (= chapter 3, (43); Bejar 2003: ch. 2, (58)) Condition on Value G(oal) values P(robe) if f ’(G) entails f(P) (9) (= chapter 3, (44)) Agree (Feature sharing version) Replace f(P) with f ’(G), so that the same feature is present in both locations. 368 5 Conclusion Thereafter, the system was put to use in the initial discussion of mono-SLI nominal domains in chapter 3.3.1. Here, first mismatches of featural configurations and the subsequent reduction of the respective SLI’s feature set-up to partial default, viz. zero-morphology on the surface, were observed in accordance with the proposed workings carved out prior to that point. These latter notions were not only further sharpened in the course of the analysis, the selectional characteristics of the individual semi-lexical items as well as their ability to surface in the aforementioned reduced form were moreover recorded in the table of properties of SLIs. Building on these initial insights, chapter 3.3.2 set out to incorporate the assignment of case into the system. Focusing on the interactions between sentential and nominal domains therein, Danon’s (2011) DP Encapsulation Assumption constituted the basis for the ensuing discussion demonstrating that under neither version of the Phase Impenetrability Condition (PIC), featural interaction across said domains is fully accounted for. Subsequently, I proposed a simple principle to eradicate this defect and thus circumvent the critique put forth by Danon, the DP Inclusion Assumption, reprinted below: (10) (= chapter 3, (65)) DP Inclusion Assumption The highest nominal phase is part of the clausal spine for the means of TRANS- FER/ Spell-Out. Thereafter, structural and non-structural cases (viz. nominative/ accusative and genitive/ dative in German) were clustered with reference to Woolford (2006) and their respective means of assignment were discussed in turn under the premises of the current system: Structural case is dependent on φ-agreement between (parts of) the nominal domain and the case-‘bearer’, while non-structural case comes ‘free’ upon External Merge. The latter property was then made responsible for triggering the initial set of 2 nd cycle Agree phenomena, observed in the course of the analysis, viz. in-situ 2 nd cycle Agree of a partially reduced SLI into its complement. Building on this observation, additional effects of a second application of probing were unearthed in the discussion of stranding phenomena from below focus particles and strong quantifiers in the succeeding section. These configurations were incorporated into the system with the modification of Anti-Locality, viz. the relaxation of its application to Prolific Domains, the original proposal from Grohmann (2000, 2003a), pace Abels (2003). Therein, the availability of complex stranding configurations was traced back to the variation in membership of prolific domain regarding the SLIs involved, which in turn enables the nominal to move across multiple phrases in its left periphery. 5 Conclusion 369 The analysis of the properties of SLIs was taken up again in chapter 3.3.3 with the presentation of more complex, i.e. multi-SLI nominal domains. Therein, the derivational character of the phenomenon of nominal concord became most visible in the identification of further instances of mismatches in probing, 2 nd cycle Agree as well as the first occurrences of intervention effects. The system proved capable to account for the otherwise curious data involving e.g. (i) the skipping of zero-morphology, (ii) incompatibility of SLI concatenations restricted to a subset of number/ gender-paradigms and (iii) loss of inflection dependent on the structural environment. Appropriate examples are compiled below: (11) C ONCLUSION 302 Building on these initial insights, chapter 3.3.2 set out to incorporate the assignment of case into the system. Focusing on the interactions between sentential and nominal domains therein, Danon’s (2011) DP Encapsulation Assumption constituted the basis for the ensuing discussion demonstrating that under neither version of the Phase Impenetrability Condition (PIC), featural interaction across said domains is fully accounted for. Subsequently, I proposed a simple principle to eradicate this defect and thus circumvent the critique put forth by Danon, the DP Inclusion Assumption, reprinted below: (10) (= chapter 3, (65)) DP Inclusion Assumption The highest nominal phase is part of the clausal spine for the means of TRANSFER/ Spell-Out. Thereafter, structural and non-structural cases (viz. nominative/ accusative and genitive/ dative in German) were clustered with reference to Woolford (2006) and their respective means of assignment were discussed in turn under the premises of the current system: Structural case is dependent on φ-agreement between (parts of) the nominal domain and the case- ‘bearer’, while non-structural case comes ‘free’ upon External Merge. The latter property was then made responsible for triggering the initial set of 2 nd cycle Agree phenomena, observed in the course of the analysis, viz. in-situ 2 nd cycle Agree of a partially reduced SLI into its complement. Building on this observation, additional effects of a second application of probing were unearthed in the discussion of stranding phenomena from below focus particles and strong quantifiers in the succeeding section. These configurations were incorporated into the system with the modification of Anti-Locality, viz. the relaxation of its application to Prolific Domains, the original proposal from Grohmann (2000, 2003a), pace Abels (2003). Therein, the availability of complex stranding configurations was traced back to the variation in membership of prolific domain regarding the SLIs involved, which in turn enables the nominal to move across multiple phrases in its left periphery. The analysis of the properties of SLIs was taken up again in chapter 3.3.3 with the presentation of more complex, i.e. multi-SLI nominal domains. Therein, the derivational character of the phenomenon of nominal concord became most visible in the identification of further instances of mismatches in probing, 2 nd cycle Agree as well as the first occurrences of intervention effects. The system proved capable to account for the otherwise curious data involving e.g. (i) the skipping of zero-morphology, (ii) incompatibility of SLI concatenations restricted to a subset of number/ gender-paradigms and (iii) loss of inflection dependent on the structural environment. Appropriate examples are compiled below: (11) a. (= chapter 3, (51) a.) viel > Wein Q W N i. [φ[f]]  [φ[m]] ii. [φ] [φ[m]] Match:  Value:  b. (= chapter 3, (57) a.) mein > Wein Poss PRO N i. [φ[f]]  [φ[m]] ii. [φ] [φ[n]] Match:  Value:  c. (= chapter 3, (110)) mein viel-er Wein Poss PRO > Q W > N [φ] [φ[*m]] [φ[m]] (12) a. (= chapter 3, (113)) dies-e mein-e Milch D > Poss PRO > Num [φ[*f]] [φ[f]] [φ[f]] b. (= chapter 3, (115), [modified]) *dies mein Wein D > Poss PRO > N [φ] [φ] [φ[m]] (13) (12) C ONCLUSION 302 Building on these initial insights, chapter 3.3.2 set out to incorporate the assignment of case into the system. Focusing on the interactions between sentential and nominal domains therein, Danon’s (2011) DP Encapsulation Assumption constituted the basis for the ensuing discussion demonstrating that under neither version of the Phase Impenetrability Condition (PIC), featural interaction across said domains is fully accounted for. Subsequently, I proposed a simple principle to eradicate this defect and thus circumvent the critique put forth by Danon, the DP Inclusion Assumption, reprinted below: (10) (= chapter 3, (65)) DP Inclusion Assumption The highest nominal phase is part of the clausal spine for the means of TRANSFER/ Spell-Out. Thereafter, structural and non-structural cases (viz. nominative/ accusative and genitive/ dative in German) were clustered with reference to Woolford (2006) and their respective means of assignment were discussed in turn under the premises of the current system: Structural case is dependent on φ-agreement between (parts of) the nominal domain and the case- ‘bearer’, while non-structural case comes ‘free’ upon External Merge. The latter property was then made responsible for triggering the initial set of 2 nd cycle Agree phenomena, observed in the course of the analysis, viz. in-situ 2 nd cycle Agree of a partially reduced SLI into its complement. Building on this observation, additional effects of a second application of probing were unearthed in the discussion of stranding phenomena from below focus particles and strong quantifiers in the succeeding section. These configurations were incorporated into the system with the modification of Anti-Locality, viz. the relaxation of its application to Prolific Domains, the original proposal from Grohmann (2000, 2003a), pace Abels (2003). Therein, the availability of complex stranding configurations was traced back to the variation in membership of prolific domain regarding the SLIs involved, which in turn enables the nominal to move across multiple phrases in its left periphery. The analysis of the properties of SLIs was taken up again in chapter 3.3.3 with the presentation of more complex, i.e. multi-SLI nominal domains. Therein, the derivational character of the phenomenon of nominal concord became most visible in the identification of further instances of mismatches in probing, 2 nd cycle Agree as well as the first occurrences of intervention effects. The system proved capable to account for the otherwise curious data involving e.g. (i) the skipping of zero-morphology, (ii) incompatibility of SLI concatenations restricted to a subset of number/ gender-paradigms and (iii) loss of inflection dependent on the structural environment. Appropriate examples are compiled below: (11) a. (= chapter 3, (51) a.) viel > Wein Q W N i. [φ[f]]  [φ[m]] ii. [φ] [φ[m]] Match:  Value:  b. (= chapter 3, (57) a.) mein > Wein Poss N i. [φ[f]]  [φ[m]] ii. [φ] [φ[n]] Match:  Value:  c. (= chapter 3, (110)) mein viel-er Wein Poss PRO > Q W > N [φ] [φ[*m]] [φ[m]] (12) a. (= chapter 3, (113)) dies-e mein-e Milch D > Poss PRO > Num [φ[*f]] [φ[f]] [φ[f]] b. (= chapter 3, (115), [modified]) *dies mein Wein D > Poss PRO > N [φ] [φ] [φ[m]] (13) 370 5 Conclusion (13) C ONCLUSION 303 (13) a. (= chapter 3, (125)) all(-e) mein-e Milch Q S > Poss PRO > Num [φ[*f]] [φ[f]] [φ[f]] b. (= chapter 3, (126)) all mein Wein Q S > Poss PRO > N [φ] [φ] [φ[m]] Further notions have been refined and additional characteristics of the elements involved have been unearthed in the course of this subchapter, most notably w.r.t. the inflectional idiosyncrasies of the strong quantifier. To this effect, varying paradigms of licit inflection in concatenation with the two instantiations of category D were presented. With a shift in perspective from featural content to overt morphological realization, however, this apparently arbitrary state of affairs fell into place in its treatment as a morpho-phonological effect, to wit, dittology, evaluating phonological reduplication on adjacent LIs against a core pattern, viz. the fem/ pl-paradigms. Thus, the varying patterns were identified as the full range of logically possible applications of this rationale, termed the inclusive paradigm and the exclusive paradigm with Dem and Det respectively. The following subchapter 3.3.4 finally turned to adjectival inflection. It was argued with Sauerland (1996) that the WEAK pattern on said class of items results from impoverishment of its feature structure, triggered by precedence/ dominance of an appropriately specified SLI. The transposition of this rationale to the current system of number/ gender-values took either masculine, neuter for feminine together with its dependent plural (i.e. the sisters of the [sg]-node in 0 above) to constitute the target of deletion in post-syntactic Morphology. The process was moreover taken not to apply exclusively to adjectives and was therefore incorporated into the aforementioned table of properties of nominal SLIs. The subsequent analysis, then, found this process to apply to the structurally lowest semi-lexical item, Q W . Moreover, this element was uncovered to exhibit a probing as well as a non-probing instance, the latter uniformly surfacing with default agreement, consequently accompanied by STRONG adjectival inflection. The former, however, turned out to feature a multitude of grammatical inflectional options, all of which, nevertheless, conform to its probing properties carved out earlier in that chapter. It was proposed that all these idiosyncratic inflectional variations serve the sole purpose to cloud the categorical status of this lexical item, either adjectival or semi-lexical. Under this view, the full paradigm of inflection turned out ambiguous as either gradial SLI-Adj or sequential Adj-Adj agreement. The strategy was loosely referred to as syntactic agnosticism. Considerations on the motivation behind it were postponed until the following section on parallel effects concerning its strong counterpart, Q S . This SLI, in turn, was likewise identified as accompanying divergent adjectival inflection w.r.t. various number/ genderand case-values (however to a lesser extent). All the properties of quantifying elements listed above were argued to be historically motivated in their split from the class of adjectives and the establishment of decided categories of semi-lexical items. Therein, the degree of variation was taken to mirror the status of their reanalysis. The incorporation of the ability to undergo AI completed the aforementioned table of properties of semi-lexical items, which is given below. Observe that therein, every category is unambiguously identifiable by its specific setting of these characteristics. (14) (= chapter 3, (191)) Q S D Poss PRO Q W Adj Probing φ [φ] [φ] [φ[f]] [φ[f]] [φ] Default Agreement      Adjectival Impoverishment [φ[m/ n]]:  /  [φ[f]]:      Translating back from adjectival inflection to morphological definiteness (cf. the schematic in 0 above) in the conclusion to chapter three, the latter notion was finally accounted for under reference to (i) the phrasal set-up elaborated in chapter one (depicted in 0 above) and (ii) the phasal configurations argued for in chapter 2.2 (illustrated in 0 above) combined with (iii) the featural set-up, initially proposed in chapter 2.3 and elaborated throughout chapter three (laid out in 0 and 0 above), as well as (iv) the theory of Agree, also developed there (subsumed in 0-0) and applied to the items under consideration in 0. The resulting restriction is given below. As has been noted, these findings are moreover in line with the diachronic observations of Greenberg (1978) as well as the theoretical considerations from Gallmann (1996). Further notions have been refined and additional characteristics of the elements involved have been unearthed in the course of this subchapter, most notably w.r.t. the inflectional idiosyncrasies of the strong quantifier. To this effect, varying paradigms of licit inflection in concatenation with the two instantiations of category D were presented. With a shift in perspective from featural content to overt morphological realization, however, this apparently arbitrary state of affairs fell into place in its treatment as a morpho-phonological effect, to wit, dittology, evaluating phonological reduplication on adjacent LIs against a core pattern, viz. the fem/ pl-paradigms. Thus, the varying patterns were identified as the full range of logically possible applications of this rationale, termed the inclusive paradigm and the exclusive paradigm with Dem and Det respectively. The following subchapter 3.3.4 finally turned to adjectival inflection. It was argued with Sauerland (1996) that the weak pattern on said class of items results from impoverishment of its feature structure, triggered by precedence/ dominance of an appropriately specified SLI. The transposition of this rationale to the current system of number/ gender-values took either masculine, neuter or feminine together with its dependent plural (i.e. the sisters of the [sg]-node in (5) above) to constitute the target of deletion in post-syntactic Morphology. The process was moreover taken not to apply exclusively to adjectives and was therefore incorporated into the aforementioned table of properties of nominal SLIs. The subsequent analysis, then, found this process to apply to the structurally lowest semi-lexical item, Q W . Moreover, this element was uncovered to exhibit a probing as well as a non-probing instance, the latter uniformly surfacing with default agreement, consequently accompanied by strong adjectival inflection. The former, however, turned out to feature a multitude of grammatical inflectional options, all of which, nevertheless, conform to its probing properties carved out earlier in that chapter. It was proposed that all these idiosyncratic inflectional variations serve the sole purpose to cloud the categorical status of this lexical item, either adjectival or semi-lexical. Under this view, the full paradigm 5 Conclusion 371 of inflection turned out ambiguous as either gradial SLI-Adj or sequential Adj- Adj agreement. The strategy was loosely referred to as syntactic agnosticism. Considerations on the motivation behind it were postponed until the following section on parallel effects concerning its strong counterpart, Q S . This SLI, in turn, was likewise identified as accompanying divergent adjectival inflection w.r.t. various number/ genderand case-values (however to a lesser extent). All the properties of quantifying elements listed above were argued to be historically motivated in their split from the class of adjectives and the establishment of decided categories of semi-lexical items. Therein, the degree of variation was taken to mirror the status of their reanalysis. The incorporation of the ability to undergo AI completed the aforementioned table of properties of semi-lexical items, which is given below. Observe that therein, every category is unambiguously identifiable by its specific setting of these characteristics. (14) (= chapter 3, (191)) Q S D Poss PRO Q W Adj Probing φ [φ] [φ] [φ[f]] [φ[f]] [φ] Default Agreement      Adjectival Impoverishment [φ[m/ n]]:  /  [φ[f]]:      Translating back from adjectival inflection to morphological definiteness (cf. the schematic in (1) above) in the conclusion to chapter three, the latter notion was finally accounted for under reference to (i) the phrasal set-up elaborated in chapter one (depicted in (2) above) and (ii) the phasal configurations argued for in chapter 2.2 (illustrated in (3) above) combined with (iii) the featural setup, initially proposed in chapter 2.3 and elaborated throughout chapter three (laid out in (4) and (5) above), as well as (iv) the theory of Agree, also developed there (subsumed in (6)-(9)) and applied to the items under consideration in (14). The resulting restriction is given below. As has been noted, these findings are moreover in line with the diachronic observations of Greenberg (1978) as well as the theoretical considerations from Gallmann (1996). (15) (= chapter 3, (195), [modified]) Morphological Definiteness in German The highest phase head must bear a gender-value. The chapter thus derived the phenomenon of morphological definiteness from the notions elaborated in earlier sections. It is this property that led to its designation as the ‘inference’ of the current analysis. 372 5 Conclusion Finally, chapter four turned to a variety of phenomena that had been touched upon in the course of the foregone analysis, either openly or implicitly through the mechanisms proposed therein. These were bundled and discussed in two decided subchapters, to wit, chapter 4.2 on optionality of movement (ch. 4.2.1) and inflection (ch. 4.2.2) regarding quantifying SLIs as well as chapter 4.3 on the formation of complex syntactic objects (ch. 4.3.1) and complex lexical items (ch. 4.3.2). The former notions had already surfaced numerous times throughout the preceding chapters. Approaching these from the realm of the fully articulated theory in the current context, however, allowed for strong claims to be made concerning the underlying mechanisms. To this end, the subchapter started out by highlighting the non-cyclic implications of a theory of feature inheritance and reviewing various approaches to sentential derivations formalizing and incorporating so-called ‘tucking-in’ as a consequence thereof. Such strategies were subsequently excluded from the theory under elaboration. Therefore, maintaining the successive cyclic nature of the derivation of nominal concord while simultaneously retaining the rejection of feature inheritance enforced the unambiguous identification of the trigger of optional movement to be found in an optional EPP/ OCC-feature on the quantifying head, in line with standard assumptions. In the subsequent incorporation of complex stranding phenomena, this notion was furthermore revised into a property of the respective stranded phase head and coupled with its variability in membership of prolific domains. This, in turn, enables a shift in the resulting reading accompanied by the facilitation of movement through its edge, in contrast to the semantically vacuous applications of IM found with simple stranding. The second section of said subchapter then turned to optional vs. obligatorily present inflection on said semi-lexical items. As was demonstrated therein, this distinction turned out to be congruent with the proposed phasal status of the respective element, with the strong quantifier necessarily exhibiting inflection when bearing phasal status and its weak counterpart, on the contrary, only doing so in its role as a non-phase. Subsequently elaborating on phasal configurations in complex compositions of the nominal domain unearthed the motivation behind these idiosyncratic restrictions in the need to preserve nominal (i.e. φ-) features at the functional domain for the ongoing verbal derivation, irrespective of the Spell-Out/ TRANSFER of the lexical subpart, Num-N, which natively houses said values. Inflection qua φ-valuedness thus makes said features available under the DP Inclusion Assumption, while their occurrence beyond said structural configurations is conceived as redundant and turns optional: In the case of non-phasal Q S , either more prominent phase head of the higher cycle 5 Conclusion 373 necessarily bears said features while in the case of phasal Q W , the lexical subpart of the nominal domain is never spelled out prior to EM in the verbal domain. The second subchapter turned to a wider selection of idiosyncratic formations in the nominal domain in German and set out to account for these in terms of the theory elaborated above while simultaneously sharpening the relevant notions in the course of doing so. Commencing with complex syntactic objects, chapter 4.3.1 a. and b. revisited the formation of PDCs and simple partitive constructions respectively and attempted an approach based on movement between categorically related phrases for both of them (i.e. Poss-to-Poss and Q-to-Q). Various refinements followed from this for the system proposed previously in this book, inter alia concerning (i) the assignment of case from within nominal domains (viz. genitive case as possessive case assigned from N, partitive case assigned from Q S ), (ii) the structural set-up underlying partitive constructions (i.e. two coordinated nominal hierarchies for HN (variable) and PN (range), with the latter core deleted under identity with the former one) as well as (iii) the structural distance for the initiation of 2 nd cycle Agree, relativized w.r.t. the category of the initiating feature; the latter is reprinted below: (16) (= chapter 4, (38)) probing φ assigning case cycle-internally * DP-Poss PRO Q S -Poss PRO Q S -Poss PRO cycle-externally DP-Q W Q S -Q W V-Poss PRO V-Q W The final subchapter 4.3.2 shifted the focus to formations composed of two particular semi-lexical items, labeled ‘complex LIs’. Section 4.3.2 a. focused on the optional addition of the root ein-word to the weak quantifier jed while section 4.3.2 b. subsequently concentrated on the quantifier meist -, itself obligatorily conjoined with either D or Poss PRO . Over the course of the former analysis, inflectional interdependencies between the joint SLIs as well as between the resulting complex quantifier and coordinated adjectives uncovered two independent levels for the implementation of AI. These were interpreted as applying to the two copies of the head in a bare phrase structural model, already incorporated into the theory in previous chapters and were taken to depend on the individual subpart’s properties of probing, which, in turn, were categorized along the lines established in chapter three. These are given below along with the proposed structural set-up of the complex LI: 374 5 Conclusion (17) (= chapter 4, (47)) a. ein jed- Probing φ [φ[f]] [φ] Default Agreement   Adjectival Impoverishment   C ONCLUSION 305 of the former analysis, inflectional interdependencies between the joint SLIs as well as between the resulting complex quantifier and coordinated adjectives uncovered two independent levels for the implementation of AI. These were interpreted as applying to the two copies of the head in a bare phrase structural model, already incorporated into the theory in previous chapters and were taken to depend on the individual subpart’s properties of probing, which, in turn, were categorized along the lines established in chapter three. These are given below along with the proposed structural set-up of the complex LI: (17) a. (= chapter 4, (47)) ein jed - Probing φ [φ[f]] [φ] Default Agreement   Adjectival Impoverishment   b. (= chapter 4, (48)) Q S0 3 ein jed - Further refinements could be carved out by the subsequent insertion of this complex item in the simple partitive constructions, analyzed above. Therein, mismatches of φ-agreement on the surface pointed to the divergence in featural specification between the domain of the variable and the range, supporting the claims made earlier in the chapter. Turning to the final object of investigation, section 4.3.2 b. build on these insights and further modified the structural claims for the specific conjunction of D/ Poss PRO and meist before turning to the semantic motivation underlying it. The latter was found in parallel to E&G’s (2009, 2010, cf. also Etxeberria 2012) treatment of similar configurations in Greek and Basque (and ultimately in Giannakidous’s 2004 analysis of Salish St’át’imcets) in the role of the determiner to provide a context set and thus to semantically restrict the domain of HN, one of the functions taken by the authors to make up the concept of definiteness. Since the latter was split from any such categorical associations earlier, further inquiry successively revealed that this function of D, D DR , is not restricted to the particular concatenation under analysis and moreover neither a property of DP nor tantamount with the concept of definiteness in German, but instead a characteristic of a subpart of the extended projection of the noun, viz. DP > Poss PRO . The chapter thus revisited various aspects of the theory which have been either touched on in earlier stages of the preceding analysis or influenced by the proposals made therein. It is this property to explore the consequences of said findings for the system at large that led to its designation as the ‘implications’ of the current inquiry. The present inquiry, then, elaborated a holistic approach to inflectional interdependencies in and with the nominal domain in German. As has been laid out, a system build on standard assumptions about derivational progression in successive cyclic application of operations suffices to derive the vast heterogeneity of inflectional idiosyncrasies that make up the realm of nominal inflectional phenomena. Refinement, however, targeted the proposed ordering of φ-features by discarding decided categories in favor of interrelated values. Therein, it has been moreover demonstrated that certain features reduce to inflectional reflexes of the resulting hierarchy of the remaining values, the fate that befell morphological definiteness in chapter three. Thus, the current analysis was able to gain explanatory insight which remains buried under traditional terms of grammatical analysis. Nevertheless, more research is obviously needed to fully validate the claims made in the preceding chapters, as has been noted at various points throughout the book (regarding, inter alia, the relation of case-values to the proposed hierarchy, the coding of φ-features across languages as well as the broader applicability of the derivational system altogether). It is my hope to have pointed in a direction that appears fruitful to pursue. Further refinements could be carved out by the subsequent insertion of this complex item in the simple partitive constructions, analyzed above. Therein, mismatches of φ-agreement on the surface pointed to the divergence in featural specification between the domain of the variable and the range, supporting the claims made earlier in the chapter. Turning to the final object of investigation, section 4.3.2 b. build on these insights and further modified the structural claims for the specific conjunction of D/ Poss PRO and meistbefore turning to the semantic motivation underlying it. The latter was found in parallel to E&G’s (2009, 2010, cf. also Etxeberria 2012) treatment of similar configurations in Greek and Basque (and ultimately in Giannakidous’s 2004 analysis of Salish St’át’imcets) in the role of the determiner to provide a context set and thus to semantically restrict the domain of HN, one of the functions taken by the authors to make up the concept of definiteness. Since the latter was split from any such categorical associations earlier, further inquiry successively revealed that this function of D, D DR , is not restricted to the particular concatenation under analysis and moreover neither a property of DP nor tantamount with the concept of definiteness in German, but instead a characteristic of a subpart of the extended projection of the noun, viz. DP > Poss PRO . The chapter thus revisited various aspects of the theory which have been either touched on in earlier stages of the preceding analysis or influenced by the proposals made therein. It is this property to explore the consequences of said findings for the system at large that led to its designation as the ‘implications’ of the current inquiry. The present inquiry, then, elaborated a holistic approach to inflectional interdependencies in and with the nominal domain in German. As has been laid out, a system build on standard assumptions about derivational progression in successive cyclic application of operations suffices to derive the vast heterogeneity of inflectional idiosyncrasies that make up the realm of nominal inflectional phenomena. Refinement, however, targeted the proposed ordering of φ-features 5 Conclusion 375 by discarding decided categories in favor of interrelated values. Therein, it has been moreover demonstrated that certain features reduce to inflectional reflexes of the resulting hierarchy of the remaining values, the fate that befell morphological definiteness in chapter three. Thus, the current analysis was able to gain explanatory insight which remains buried under traditional terms of grammatical analysis. Nevertheless, more research is obviously needed to fully validate the claims made in the preceding chapters, as has been noted at various points throughout the book (regarding, inter alia, the relation of case-values to the proposed hierarchy, the coding of φ-features across languages as well as the broader applicability of the derivational system altogether). It is my hope to have pointed in a direction that appears fruitful to pursue. 376 Endnotes Endnotes *1 Roehrs (in prep.) actually does cite emphatic contexts in which a combination along these lines with pluralia tantum nouns are supposedly acceptable in certain dialects, not further specified: i. (= Roehrs in prep.: ch. 3, (3) b.) Was für eine {Ferien/ Typen/ Idioten/ Frauen}! what for a { holidays / guys / idiots / women } These are, however, fully ungrammatical for my informants as well as myself and will therefore not be considered in what follows. *2 These quantifiers may, however, optionally surface with overt inflection in lexical case configurations. Observe that accompanying adjectives therein follow the mixed inflectional pattern with masculine and neuter mass HNs: i. viel guter Reis [ many good weak rice ] NOM ii. viel-em guten Reis [ many good weak rice ] DAT I will not have to say much about this variation in the current introduction and only mention it here for the sake of completeness. The reader is referred to chapter 3.3.4 of which a substantial part is concerned with the phenomenon outlined above. *3 Notice that in order to adjust the arguments from Milsark to my prior line of reasoning, I reduced inter alia the ‘sm’-variant weak quantifier from the table in (19), to be deduced in what follows. This move left me with the indefinite article as the only overtly realized element from Milsark’s table (Milsark 1977: (13)) to demonstrate the ungrammaticality of weak elements in (20) d. (cf. Milsark 1977: (30) b.). I am aware of the shift in number from said datum to the current (21) b'. This change nevertheless does not interfere with the restrictions from the syntactic environments under consideration, as can be seen by the insertion-test, already employed in section 1.3.2, namely mass nouns: i. Some wine is red. (strong - P) ii. * Sm wine is red. (weak - P) Endnotes Endnotes 377 iii. * There is some wine in the kitchen. iv. There is sm wine in the kitchen. *4 The sentential constructions put to use throughout this sections do of course not constitute an exhaustive list of sensitive environments for morphological definiteness; cf. e.g. the sentence structure in i. below, taken from Pafel (1994: 272), in which differently modified temporal expressions w.r.t. the bipartition depicted in (22) likewise split along the lines identified therein, with weak nominal phrases resulting in grammatical, strong nominal hierarchies returning ungrammatical structures. i. später war er hier. (= Pafel 1994: (112) a., [gloss added]) later was he here *5 The above generalizations are reminiscent of those arrived at in Pafel (1994: 259): i. (= Pafel 1994: (82)) Wenn ein D + def ist, dann flektieren modifizierende Adjektive schwach. ‘if a D is + def , modifying adjectives inflect weak.’ [MB] ii. (= Pafel 1994: (83)) Wenn ein D def ist, dann flektieren modifizierende Adjektive stark. ‘if a D is def , modifying adjectives inflect strong.’ [MB] The interdependency can be generalized further by taking into account the parallel morphological forms of both SLIs and adjectives, the latter in cooccurrence with SLIs as well as in simple A-N concatenations. As discussed by Roehrs (2009: 135, cf. also references therein), this popular approach to inflection in the nominal domain is summed up in the principle in iii. below, whose underlying rationale will surface again in my analysis in chapter three. iii. Principle of Monoinflection (Roehrs 2009: ch. 4, (23)) The first element within a noun phrase carries the strong and the second one the weak ending. *6 Additionally, Zamparelli rules out further instances of recursively quantified nominal domains with reference to incompatibility of the respective values of the numberor definiteness-features of the items involved. He does not, however, specify a principled account for the application of either one of these mechanisms or interactions between them. For the lineari- 378 Endnotes zations in i., it is assumed without comment that some kind of semantosyntactic incompatibility bars these structures: i. * {The/ Some/ All} a man (= Zamparelli [1995] 2000: (332) a.) *7 C&S furthermore focus on reduced variants of third person singular nominative masculine and neuter pronouns. The contrast can be reproduced in colloquial German as well, with [a]/ [n], [sə] and [(ɐ)s] replacing weak er 3sg.NOM.masc / ihn 3sg.AKK.masc , sie 3sg.NOM.fem , es 3sg.NOM.neut : i. Ich hab` {n/ sə/ ɐs} gestern *{n/ sə/ ɐs} eingeladen. I have { n / sə / ɐs } yesterday *{ n / sə / ɐs } invited ii. (= C&S 1996: (51) b., [modified]) …, dass er/ *a und Maria intelligent sind that he / * a and Mary intelligent are *8 Alternatively, one might propose that lexical possessives themselves occupy the specifier of a null Poss-head in the complement of HN, thereby simplifying movement of the complex possessive when HN also heads an argument NP (see ch. 1, fn. 60). Nothing in my following analysis hinges on this, but as empty heads are suspicious theory-internally, I elaborate on the simpler alternative here. Observe that in any case, further backup for the phrasal configurations proposed thus far in this subchapter (i.e. (a) the pronominal possessive dominating the lexical possessive as well as (b) both of these dominating further complements of HN) comes from c-command relations, identified by anaphora-binding phenomena with picture-NPs (Condition A): Observe that (a) pronominal as well as lexical prenominal possessives are able to bind a reflexive in the complement of the noun, both in isolation and in co-occurrence with one another (i.-iii.); furthermore, (b) lexical possessives are also able to pose as the antecedent in postnominal position (iv). i. sein Bild von sich (selbst) his picture of him ( self ) ii. Martins Bild von sich (selbst) Martin GEN picture of him ( self ) iii. Martin sein Bild von sich (selbst) Martin DAT his picture of him ( self ) iv. das Bild Martins von sich (selbst) the picture Martin GEN of him ( self ) Endnotes 379 The structures above are modelled after Roehrs’ (2013: ch. 4.3) analysis of DP-internal binding in Norwegian. Concerning possessive-internal binding relations, the expected Condition C violations as well as the lack thereof in Northern Scandinavian structures (cf. Roehrs 2013: ch. 3.4) cannot be emulated for German since the only configuration in which the pronoun would be taken to c-command the R-expression (in the premises laid out here, i.e. prenominal pronoun coupled with postnominal lexical possessive) turns out ungrammatical (cf. v.). v. * sein Bild Martins (von sich (selbst)) his picture Martin GEN ( of him ( self )) *9 Salzmann (2011: 178f.) notes that cross-linguistically, the case of the lexical possessor in such constructions conforms to the most oblique one in a language w.r.t. the scale depicted in i.: i. genitive > dative > accusative > zero (= Salzmann 2011 (62)) However, the relevant LI/ SO bears dative in German, which also exhibits genitive case; Salzmann (2011: fn. 38) objects to this apparent counterevidence by arguing for the general ability to mark prenominal possessors with genitive in colloquial German. While his example is marginally acceptable (i.e. ii. below), isolated nominal domains constructed in parallel to data in the main text are ungrammatical: ii. (= Salzmann 2011: fn. 38, (ii)) Wer würde anderer Leute ihr Büro streichen? who would other. GEN people their office paint ‘Who would paint other people’s office? ’ iii. * sein sein Vater (based on (60) f.) his GEN his GEN father iv. * des Lehrstuhls sein Buch (based on (70)) [ the chair ] GEN his GEN book One might hence suspect the acceptability of ii. to relate to the sentential environment. I will not elaborate on this any further here but refer to ch. 4.3.1 a. below on case in possessive constructions. *10 We might have found a hint as to where the prenominal lexical possessive is located in the possibility to move said element together with HN from beneath the strong quantifier: 380 Endnotes i. Ich mag [alle Martins Bücher]. I like [ all Martin GEN books ] ii. [Martins Bücher] i mag ich [alle t i ]. [ Martin GEN books ] i like I [ all t i ] ii. parallels (14) b.’, if the position of the lexical possessive (Specifier of Poss PRO ) signals the presence of Poss PRO P and hence of the phase head below the quantifier. *11 It has to be noted that topicalization to sentence-initial position strongly depends on the context. Howsoever, establishing contrastive focus, the ungrammatical and marginal data in both (20) as well as (25) are much more degraded than parallel structures involving movement of the complete nominal domain, given it is the proposition of said complete domain and not the SLI that is contrasted: i. (= (20) a./ b.) Ich habe gute Bekannte, aber/ und [Freunde habe ich #diese/ *meine]. I have good acquaintances but / and [ friends have I # these / * mine ] ii. (= (25) a./ b.) Nudeln gab es nicht aber/ und [Reis habe ich *viel(-en) gekauft]. pasta exist it not but / and [ rice have I * many bought ] *12 Observe that topic left dislocation of plural and mass HNs out of the complement of the quantifiers is possible in contrastive constructions, see also ch. 2, fn. 13: i. Context: Speaker A reports about his participation in a tournament where one participant could in principle win several cups and several medals. Speaker A: Pokale habe ich wenige, [aber Medaillen i habe ich alle t i gewonnen]. cups have I few [ but medals i have I all t i won ] ‘I have won a few cups but all the medals.’ *13 Observe that the featural set-up, just proposed, makes additional claims about the recoverability of feminine gender (i.e. feminized masculine nominal domains) in ellipsis constructions like those employed in the discussion of phasehood throughout chapter 2.2, parallel to number: i. Ich habe Jans Freund kennen gelernt und du hast [die beiden <Freunde>] I have Jan’s friend sg met and you have [ the both < friends >] pl Endnotes 381 von Martin kennen gelernt. of Martin met . ii. * Ich habe Jans Freund kennen gelernt und du hast [die eine <Freundin>] I have Jan’s friend masc met and you have [ the one < friend >] fem von Martin kennen gelernt. of Martin met As can be deduced from i.-ii. above, this prediction is not borne out in German as is also the case in Spanish (i.e. Romance), which led Saab (2010: 69f.; cf. also Giurgea 2014: 52 and references therein) to uniformly locate gender below NumP (pace Ritter 1993) and inside the ellipsis-site. *14 Building on this observation, note that the ability of QST from below the strong and the weak quantifier (i.e. floating constructions and split-topicalization) is levelled to the structural configuration of non-phasal quantifier and phasal complement with regard to the assignment of phasehood depicted in (30) above for all cases reviewed in chapter 2.2.1. Furthermore focusing on the overt residues of nominal hierarchies in the case of split-topicalizations under the structural premises elaborated thus far, then, the application of said configuration as a testing environment for (in-)definiteness of the stranded SLI in Pafel (1994: 272) can now be identified as underlyingly sensitive to their external merging-site w.r.t. the two nominal phasal cycles: i. (= Pafel 1994: (113) a.-c., [gloss added]) Bücher hat Augustus _____ gelesen. books has Augustus _____ read  keine, viele, zwei, manche, einige, wenige, ein paar, … none, many, two, some, several, few, a couple * alle, beide, jedes, die, … all, both, every, the *15 The decision to present dative-, rather than genitive-marked strings, in which the pronouns themselves would surface as morphological ein-words, is not accidental but rooted in the fact that genitive-marked personal pronouns/ possessive pronouns in co-occurrence with (adjectives and) nouns result in ungrammatical strings (cf. ii. in opposition to the grammatical i.). 382 Endnotes i. Ich gedenke [dein-er]. I commemorate [ you ] GEN ii. * Ich gedenke [dein-er schön-en Mann-es/ Kind-es]. I commemorate [ you beautiful man masc / child neut ] GEN However, switching inflection to gender-specific es in accordance with HN induces an unambiguously possessive reading of the complete nominal domain: iii. Ich gedenke [dein-es schön-en Mann-es/ Kind-es]. I commemorate [ you beautiful man masc / child neut ] GEN ‘I commemorate your beautiful husband/ child.’ Recall that adjectival inflection is homonymous between the strong and weak ending in the genitive of the masculine and neuter paradigms. In the feminine paradigm, however, these are morphologically distinct while the pronominal forms overlap instead. Here, we find a parallel picture with strong inflected adjectives resulting in ungrammatical structures, while weak inflected adjectives as well as simple Poss PRO -N concatenations again enforce possessivity (cf. iv.-v.). iv. * Ich gedenke [dein-er schön-er Frau]. I commemorate [ you beautiful strong woman ] GEN v. Ich gedenke [dein-er schön-en Frau]. I commemorate [ you beautiful weak woman ] GEN ‘I commemorate your beautiful wife.’ Therein, a hint as to the relation of genitive personal and possessive pronouns (homonymy vs. identity), left open in the preceding chapter, might be uncovered. *16 Observe that the conjunction of these SLIs becomes acceptable in concatenation with masculine HNs in nominative case when the pronoun displays first or second person plural possessor agreement: i. dies-er unser Hund [ this our dog masc ] NOM ii. dies-er euer Hund [ this your pl dog masc ] NOM This effect, however, is rooted in the morphological form of the possessives as indicated by the omission of hyphens in the data above: The pronouns Endnotes 383 surface as bare, i.e. suffixless stems, parallel to the remaining forms in the prardigm; however, the stem-final er mirrors the appropriate inflection on the demonstrative which facilitates the reanalysis of the form, thus its acceptability. This approach gains support from parallel configurations based on neuter HNs in both structural cases, which are unacceptable for the full paradigm of D-Poss PRO concatenations. The only diverging factor on the surface is the morphological dissimilation of inflection on the demonstrative and the final syllable of the possessive stem: iii. * dies-es unser Pferd [ this our horse neut ] NOM/ ACC iv. * dies-es euer Pferd [ this your pl horse neut ] NOM/ ACC *17 Another instance of such syntactic agnosticism might be found w.r.t. the diverging morphological form of the ein-word Poss PRO in accusative masculine as mentioned above. Observe that in spoken colloquial German, the ein -root article can be reduced to the final consonant / n/ in cases where it surfaces with zero-morphology (i.e. structural cases in non-feminine gender). All other instances are only reducible to consonant plus inflectional suffix. This is demonstrated with the nominative and dative masculine, below: i. / n/ Mann steht vor der Tür. [ a man ] NOM stands in.front.of the door ii. / n/ *(-em) Mann vertraue ich. [ a man ] DAT trust I With the accusative masculine, surfacing with the suffix -en, however, both options become available: iii. Ich suche / n/ (-en) Mann. I am.looking.for [ a man ] ACC *18 This is true under the early definition of ‘minimal domain’ in the specification of equidistance from Chomsky (MP) as well as the reformulation in later works (Chomsky MI) and the phasal modification of equidistance itself before its final dismissal in favor of the dislocation of intervening SOs by phonological rules (Chomsky DbP: 28; cf. also ch. 4, fn. 2). All three formulations are given below for the reader to verify: 384 Endnotes i. (Chomsky MP) a. (ch. 4, (86) c.) The minimal domain Min(δ(CH)) of CH is the smallest subset K of δ(CH) such that for any γ ∈ δ(CH), some β ∈ K reflexively dominates γ. b. (ch. 4, (189)) γ and β are equidistant from α if γ and β are in the same minimal domain. ii. (Chomsky MI) a. (p. 123) The minimal domain of a head H is the set of terms immediately contained in the projections of H. b. ((41)) Terms of the same minimal domain are ‘equidistant’ to probes. iii. (Chomsky DbP: (44)) Terms of the edge of HP are equidistant from probe P. As Chomsky (DbP: 21) notes, iii. does not subsume the full set of terms captured by equidistance on the basis of minimal domains but is restricted to the relation of various specifier positions of a single head and hence does not concern us here. *19 Phase-based simultaneity (hence counter-cyclicity) of operations has already been hinted at without recourse to feature inheritance in earlier stages of the program (cf. DbP: 18 on Value and Spell-Out, also ibid.: 27f. on the application of MLC in evaluating multiple instances of IM; BEA on all operations, including IM and EM, the latter again liberated from phase-final application in recent versions, cf. PoP: 42; cf. also EKS 2012: ch. 11.4.1). Interestingly, later instantiations incorporating feature inheritance (i.e. AUGB, OP, PoP) do not discuss its tampering, non-extending consequences for IM, thought in part its counter-cyclic ones (cf. e.g. PoP: fn. 29, also AUGB: fn. 26 where any counter-cyclicity of feature-inheritance itself is brushed aside as intuitive without reference to its implications for IM). However, as observed by EKS (2012: 253), the earliest instantiation of the NTC indeed allowed for such counter-cyclic tucking-in at the specifier position (R the “relevant basic relation,” MI: 137): i. (Chomsky MI: 137, (59)) Given a choice of operations applying to α and projecting its label L, select one that preserves R(L, γ). Endnotes 385 *20 One additional scarce pattern of disrupted nominal domains with strong quantifiers in German not discussed here concerns cases of preposed quantifying SLI with a postverbal, arguably in-situ, (pro-)nominal remnant, referred to as ‘backward floating’ in Vater (1980) and exemplified below: i. Alle sind die Gäste gegangen. (= Vater 1980: (49) a., [gloss added]) all i are [t i the guests ] gone ii. Alle sind sie gegangen. (= Vater 1980: (50) a., [gloss added]) all i are [t i they ] gone This construction seems to implement additional restrictions (cf. Vater 1980: ch. 5, ch. 6.2), partially concerning the source of the proposed movement, and would hence take us too far afield. Observe, however, that these constructions are generally compatible with the proposals of the main text, arguably with the implementation of true optionality in the sense of variable size of the goal in the checking of obligatory EPP following Biberauer & Richards (2006). *21 This foremost concerns the dual nature of case on possessive pronouns. Observe that here as well as in chapter one above I specified agreement between the pronoun and HN to involve the category case along with φ even though the SLI has consistently been glossed as bearing genitive case whilst co-occurring with nominals in all lexical and structural cases in chapter one. While nominal concord via suffixation indeed involves case, possessive pronouns also consistently bear inherent genitive as can be deduced from the two licit interrogatives in ii. and iii. targeting various subparts of the possessive nominal domain in complement position of a non-structural case-assigning predicate in i. Once again, the dual nature of features on the possessive pronoun can be taken to reinforce a decompositional approach (cf. also ch. 3.3.3). i. Er hat sein-em Vater immer vertraut. he has [ his GEN dat father ] DAT always trusted ii. Wem hat er immer vertraut? who DAT has he always trusted iii. Wessen Vater hat er immer vertraut? [ whose father ] GEN has he always trusted *22 In what follows, I will have little to say about such concatenations; while the option of conjoining two pronominal (in contrast to nominal) LIs should 386 Endnotes in principle open the door to altering the person-value of the complete construction (i.e. concatenations of two first or second person dative and possessive case-marked pronouns respectively; cf. i., ii. below), observe that such structures are judged ungrammatical by my informants parallel to concatenations of homophonous dativeand possessive-marked feminine third person pronouns (cf. iii.). As should be expected at this point in the analysis, this state of affairs extends to any person-value in the plural paradigm (suffixial agreement with HN follows [φ[m/ n]] below): i. * mir mein N my DAT my GEN N ii. * dir dein N your DAT your GEN N iii. * ihr ihr N her DAT her GEN N *23 As noted in chapter one above with reference to Postal ([1966a] 1970), generic readings rescue preposed adjective constructions with indefinite nominal hierarchies in argument position. Furthermore, observe that unrestricted quantification by Q S supports a generic reading, a tendency that I tried to counter by simultaneously altering tense in both (30) and (31) b. Furthermore, as Postal ([1966a] 1970: fn. 9) himself notes, additional modification of the test environment in (31) b. by restrictive relatives is able to sharpen the predictions along the lines tested in the main text since these notions are taken to be generally incompatible. The test environment below is modelled after Postal’s (fn. 9) own example. Observe that domains incorporating Q S are predicted to turn out fine, while those incorporating nonpartitive Q W should turn out ungrammatical. Note, however, that the resulting structures are highly unusual and quite complex themselves, which is why these linearizations are generally judged as odd by my informants and marked accordingly. Nevertheless, a preference for i. over ii. was reported by some speakers, which can be traced back to the contrast under analysis. i. # Teuer wie alle Butter, die ich heute kaufte, war, … expensive as all butter that I today bought was ii. ## Teuer wie viele Butter, die ich heute kaufte, war, … expensive as some butter that I today bought was *24 A further argument for the occupation of this structural position can be found in the parallel behavior of the native strong quantifier all and the Endnotes 387 complex SLI concerning QST (cf. Fehlisch 1986, ch. 1.3 for the presentation of an extensive paradigm). Without going into too much detail, observe that stranding of ( ein -) jed is licit only from nominal hierarchies incorporating a head of the highest contextual phasal cycle (cf. also Fehlisch 1986: 97, Pafel 1994: 239) in contrast to split-topicalization structures on the basis of weak quantifiers (cf. ch. 2.2.1 above). Interestingly, however, another mismatch in number-agreement can be found in the resulting configurations of the form depicted in i., parallel to (39) f. below, in which the (now predicatively case-marked) nominal domain in plural moves from below the quantifier marked with the according singular gender-agreement: i. Die Affen haben (ein) jed-er eine Banane bekommen. (Affe[φ[[sg][m]]]) [ the apes ] have (ein) every masc a banana received ii. (Ein) Jed-er Affe hat eine Banane bekommen. [(ein) Every masc ape ] sg has a banana received For Fehlisch (1986: ch. 3), the additional deviation concerning this feature in QST vs. in-situ configurations (in i. vs. ii.) poses an argument against the transformational linkage of such constructions on the basis of the SLI under consideration. I will get back to these types of feature mismatches below. *25 An even more puzzling observation related to the data in (40) a.-c. can be found in partitives building solely on the ein-word. Parallel to the paradigm in the main text, the ein-word likewise displays uniform singular genderagreement (cf. i.-iii.). However, as has been noted before, ein standardly displays zero-agreement in concatenation with nominal domains bearing [φ[m/ n]] in structural case (cf. iv.-vi.). i. ein-es der Bücher ein neut [ the books ] GEN ii. ein-er der Männer ein masc [ the men ] GEN iii. ein-e der Frauen ein fem [ the women ] GEN iv. ein Buch ein-ø book [φ[[sg][n]]] v. ein Mann ein-ø man [φ[[sg][m]]] vi. ein-e Frau ein fem woman [φ[[sg][f]]] 388 Endnotes *26 An interesting observation in this context concerns the deviant forms of SLI and (homonymous) strong and weak adjectival inflection presented in chapter 3.3.4 c. (cf. also Karnowski & Pafel 2004: 172f.): Observe that in the appropriate slots of the inflectional paradigm of the complex SLI, the righthand terminal subpart ein inflects according to its status as an SLI (to a lesser degree also acceptable with homonymous adjectival inflection), while the quantifier jed only ever exhibits adjectival inflection. i. der Geruch [ein-es/ en jed-en Wein-es] the smell [ein-sli/ homonym every homonym wine masc ] GEN ii. der Geruch [ein-es/ en jed-en Bier-es] the smell [ein-sli/ homonym every homonym beer neut ] GEN Observe furthermore that jed in isolation exhibits the same variability between the two inflectional patterns; Karnowski & Pafel (2004: fn. 5) therefore trace the above restriction back to the varying categorical status of the right-hand terminal subpart in the aforementioned contexts. However, I believe that this inflectional property does not contradict the classification of jed -, advocated here, since parallel (albeit more variable) inflectional effects have been observed with another SLI of the same category Q W in the subchapter mentioned above. *27 Moreover, once again taking deviant SLIand homonymous adjectival inflection into account (cf. ch. 3.3.4 c., endnote *26), SLI-internal inflection is unambiguously adjectival in the case of the quantificational terminal subpart (and unambiguously semi-lexical in the case of the determiner): i. der Geruch [d-es meist-en Wein-es] the smell [ the sli most homonym wine masc ] GEN ii. der Geruch [d-es meist-en Bier-es] the smell [ the sli most homonym beer neut ] GEN Bibliography 389 Bibliography Abels, Klaus. 2003. Successive Cyclicity, Anti-Locality, and Adposition Stranding . Ph.D. Dissertation, University of Connecticut, Storrs. Abney, Steven P. 1987. The English Noun Phrase in its Sentential Aspect . Ph.D. Dissertation, MIT, Cambridge. Alexiadou, Artemis. 2001. Functional Structure in Nominals: Nominalization and Ergativity . (= Linguistik Aktuell/ Linguistics Today (LA) , 42.) Amsterdam, Philadelphia: John Benjamins. ---. 2004. “Inflection Class, Gender and DP Internal Structure”. Explorations in Nominal Inflection ed. by Gereon Müller, Lutz Gunkel & Gisela Zifonun, 21-50. Berlin: Mouton de Gruyter. ---, Liliane Haegeman & Melita Stavrou. 2007. Noun Phrase in the Generative Perspective . (= Studies in Generative Grammar [SGG] , 71.) Berlin, New York: Mouton de Gruyter. Anderson, Mona. 1983. “Prenominal Genitive NPs”. Linguistic Review 3: 1.1-24. Baker, Mark C. 1985. “The Mirror Principle and Morphosyntactic Explanation”. Linguistic Inquiry 16: 3.373-415. ---. 2008. The Syntax of Agreement and Concord . (= Cambridge Studies in Linguistics , 115.) Cambridge: Cambridge University Press. ---. 2015. Case: Its Principles and its Parameters . (= Cambridge Studies in Linguistics , 146.) Cambridge: Cambridge University Press. Balcik, Ines. 2007. Grammatik: Sichere Antworten auf knifflige Fragen . (= PONS Zweifelsfrei Deutsch , 2.) Stuttgart: Ernst Klett. Barbiers, Sjef. 2005. “Variation in the Morphosyntax of One”. The Journal of Comparative Germanic Linguistics 8: 3.159-183. Bat-El, Outi. 1986. Extraction in Mordern Hebrew Morphology . Master's Thesis, UCLA, Los Angeles. Bayer, Josef & Jaklin Kornfilt. 1994. “Against Scrambling as an Instance of Move- Alpha”. Studies on Scrambling: Movement and Non-Movement Approaches to Free Word-Order Phenomena . (= Studies in Generative Grammar [SGG] , 41.) ed. by Norbert Corver & Henk van Riemsdijk, 17-60. Berlin, New York: Mouton de Gruyter. Bejar, Susana. 2000. “Structural Markedness in Formal Features: Deriving Interpretability”. Revue québécoise de linguistique 28: 1.47-72. ---. 2003. Phi-Syntax: A Theory of Agreement . Ph.D Dissertation, University of Toronto, Toronto. --- & Milan Rezac. 2009. “Cyclic Agree”. Linguistic Inquiry 40: 1.35-73. Belletti, Adriana & Luigi Rizzi. 1981. “The Syntax of Ne: Some Theoretical Implications”. Linguistic Review 1: 2.117-154. Benincà, Paola & John Haiman. 1992. The Rhaeto-Romance Languages . London: Routledge. Bibliography 390 Bibliography Benincasa, Marco. 2014. “On the Evaluation of Morphological Form of the German Universal Quantifier”. Facets of Linguistics: Proceedings of the 14th Norddeutsches Linguistisches Kolloquium 2013 in Halle an der Saale . (= Hallesche Sprach- und Textforschung , 12.) (Norddeutsches Linguistisches Kolloquium (NLK) 14), Halle an der Saale, 15-17.03.2013 ed. by Anne Ammermann, Alexander Brock, Jana Pflaeging & Peter Schildhauer, 11-20. Frankfurt (a.M.), Berlin, Bern, Bruxelles, New York, Oxford, Wien: Peter Lang. Bernstein, Judy B. 1991. “DPs in Walloon: Evidence for Parametric Variation in Nominal Head Movement”. Probus 3.101-126. ---. 1993. Topics in the Syntax of Nominal Structure across Romance . Ph.D. Dissertation, CUNY, New York. ---. 2008. “English th-forms”. Essays on Nominal Determination: From Morphology to Discourse Management . (= Studies on Language Companion Series (SLCS) , 99.) ed. by Henrik Høeg Müller & Alex Klinge, 213-232. Amsterdam, Philadelphia: John Benjamins. Biberauer, Theresa, Anders Holmberg, Ian Roberts & Michelle Sheehan. 2010. Parametric Variation: Null Subjects in Minimalist Theory . Cambridge, et al. Cambridge University Press. --- & Marc Richards. 2006. “True Optionality: When the Grammar doesn't Mind”. Minimalist Essays . (= Linguistik Aktuell/ Linguistics Today (LA) , 91.) ed. by Cedric Boeckx, 35-67. Amsterdam, Philadelphia: John Benjamins. Bobaljik, Jonathan D. 2003. “Floating Quantifiers: Handle with Care”. The Second Glot International State-of-the-Article Book: The Latest in Linguistics . (= Studies in Generative Grammar [SGG] , 61.) ed. by Lisa L.-S. Cheng & Rint Sybesma, 107-148. Berlin, Boston: De Gruyter Mouton. Boeckx, Cedric. 2008. Understanding Minimalist Syntax: Lessons from Locality in Long- Distance Dependencies . (= Generative Syntax , 9.) Malden, Oxford, Carlton: Blackwell Publishing. ---. 2009. “On the Locus of Asymmetry in UG”. Catalan Journal of Linguistics 8.41-53. Bonet i Alsina, M. Eulélia. 1991. Morphology after Syntax: Pronominal Clitics in Romance . Ph.D. Dissertation, MIT, Cambridge. Borer, Hagit. 2005. In Name Only: Structuring Sense . Volume 1. Oxford, New York: Oxford Univ Press. Bošković, Željko. 1997. The Syntax of Nonfinte Complementation: An Economy Approach . (= Linguistic Inquiry Monographs , 32.) Cambridge, London: MIT Press. ---. 2005. “On the Locality of Left Branch Extraction and the Structure of NP”. Studia Linguistica 59: 1.1-45. ---. 2007. “On the Locality and Motivation of Move and Agree: An Even More Minimal Theory” Linguistic Inquiry 38: 4.589-644. ---. 2008. “What will you have, DP or NP? ”. Proceedings of the Northeast Linguistic Society 37 . (= NELS , 37.) ed. by Emily Elfner & Martin Walkow, 101-114. Almherst: GLSA (Graduate Linguistic Student Association). ---. 2009. “Unifying First and Last Conjunct Agreement”. Natural Language and Linguistic Theory 27: 3.455-496. Bibliography 391 ---. 2011. “On Unvalued Uninterpretable Features”. NELS 39: Proceedings of the 39th Annual Meeting of the North East Linguistic Society . (1.) (39th Annual Meeting of the North East Linguistic Society), Cornell University, 07-09.11.2008 ed. by Suzi Lima, Kevin Mullin & Brian Smith, vol. 1, 109-120. Almherst: GLSA (Graduate Linguistic Student Association). ---. 2012. “On NPs and Clauses”. Discourse and Grammar: From Sentence Types to Lexical Categories . (= Studies in Generative Grammar [SGG] , 112.) ed. by Günther Grewendorf & Thomas E. Zimmermann, 179-245. Boston, Berlin: Walter de Gruyter. ---. 2013. Getting Really Edgy: On the Edge of the Edge . Ms: University of Connecticut. ---. 2014. “Now I’m a Phase, Now I’m Not a Phase: On the Variability of Phases with Extraction and Ellipsis”. Linguistic Inquiry 45: 1.27-89. Bouchard, Denis. 2002. Adjectives, Number and Interfaces: Why Languages Vary . Amsterdam, Boston: North Holland/ Elsevier. Brame, Michael. 1981. “The General Theory of Binding and Fusion”. Linguistic Analysis 7: 3.277-326. ---. 1982. “The Head-Selector Theory of Lexical Specifications and the Nonexistence of Coarse Categories”. Linguistic Analysis 10: 4.321-325. Branigan, Phil. 2013. Cyclicity and the Approach the Probe Principle . Ms: Memorial University. ---. 2014. Approaching the Edge in Syntactic Derivation . Ms: Memorial University. Brisson, Christine M. 1998. Distributivity, Maximality, and Floating Quantifiers . Ph.D. Dissertation, Rutgers University, Brunswick. Brody, Michael. 1997. “Perfect Chains”. Elements of Grammar: A Handbook in Generative Syntax ed. by Liliane Haegeman, 139-167. Dordrecht: Kluwer. Bussmann, Hadumod. 1996. Routledge Dictionary of Language and Linguistics . London, New York: Routledge. Butler, Jonny. 2004. “On Having Arguments and Agreeing: Semantic EPP”. York Papers in Linguistics [series 2]: 1.1-27. Cardinaletti, Anna. 1998. “On the Deficient/ Strong Opposition in Possessive Systems”. Possessors, Predicates and Movement in the Determiner Phrase . (= Linguistik Aktuell/ Linguistics Today (LA) , 22.) ed. by Artemis Alexiadou & Chris Wilder, 17-54. Amsterdam, Philadelphia: John Benjamins. --- & Giuliana Giusti. 1992. “Partitive Ne and the QP Hypothesis”. Proceedings of the XVII Meeting of Generative Grammar ed. by Elisabetta Fava, 121-141. Turin: Rosenberg and Sellier. --- & ---. 2006. “The Syntax of Quantified Phrases and Quantitative Clitics”. The Blackwell Companion to Syntax . (= Blackwell handbooks in linguistics , 19.) ed. by Martin Everaert & Henk van Riemsdijk, vol. 5, 23-93. London: Blackwell. --- & Michal Starke. 1994. “The Typology of Structural Deficiency: On the Three Grammatical Classes”. University of Venice WPL 4: 2.41-109. --- & ---. 1996. “Deficient Pronouns: A View from Germanic”. Studies in Comparative Germanic Syntax: Volume 2 . (= Studies in Natural Language and Linguistic Theory , 38.) ed. by Höskuldur Thráinsson, Samuel D. Epstein & Steve Peter, 21-65. Dordrecht: Kluwer Academic. 392 Bibliography --- & ---. 1999. “The Typology of Structural Deficiency: A Case Study of the Three Classes of Pronouns”. Clitics in the Languages of Europe: EUROTYP 5 . (= Empirical Approaches to Language Typology [EALT] , 20.) ed. by Henk van Riemsdijk, 145-234. Berlin, New York: Mouton de Gruyter. Chakroun, Slaiem. 1991. Stufung und Reihung als Verbindung attributiver Adjektive in der Nominalklammer . (= Tuduv-Studien .) München: Tuduv-Verlagsgesellschaft. Chierchia, Gennaro. 1998. “Plurality of Mass Nouns and the Notion of ‘Semantic Parameter’”. Events and Grammar . (= Studies in Linguistics and Philosophy , 70.) ed. by Susan D. Rothstein, 53-103. Dordrecht: Kluwer Academic. Chomsky, Noam. 1975. “Questions on Form and Interpretation”. Linguistic Analysis 1.75-110. ---. 1986. Barriers . (= Linguistic Inquiry Monographs , 13.) Cambridge, London: MIT Press. ---. 1991. “Some Notes on Economy of Derivation and Representation”. Principles and Parameters in Comparative Grammar . (= Current Studies in Linguistics , 20.) ed. by Robert Freidin, 417-454. Cambridge, London: MIT Press. ---. 1993. “A Minimalist Program for Linguistic Theory”. The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger . (= Current Studies in Linguistics , 24.) ed. by Kenneth L. Hale & Samuel J. Keyser, 1-52. Cambridge: MIT Press. ---. 1995a. The Minimalist Program . (= Current Studies in Linguistics , 28.) Cambridge, London: MIT Press. ---. 1995b. “Bare Phrase Structure”. Evolution and Revolution in Linguistic Theory: Studies in Honor of Carlos P. Otero . (= Georgetown Studies in Romance Linguistics .) ed. by Héctor Campos, 51-109. Washington, D.C. Georgetown University Press. ---. 2000. “Minimalist Inquiries: The Framework”. Step by Step: Essays on Minimalist Syntax in Honor of Howard Lasnik ed. by Roger Martin, David Michaels & Juan Uriagereka, 89-155. Cambridge: MIT Press. ---. 2001. “Derivation by Phase”. Ken Hale: A Life in Language . (= Current Studies in Linguistics , 36.) ed. by Michael J. Kenstowicz, 1-52. Cambridge, London: MIT Press. ---. 2004. “Beyond Explanatory Adequacy”. Structures and Beyond: The Cartography of Syntactic Structures . Volume 3. (= Oxford Studies in Comparative Syntax .) ed. by Adriana Belletti, 104-131. New York, et al. Oxford Univ Press. ---. 2007. “Approaching UG from Below”. Interfaces + Recursion = Language? Chomsky's Minimalism and the View from Syntax-Semantics . (= Studies in Generative Grammar [SGG] , 89.) ed. by Uli Sauerland & Hans M. Gärtner, 1-29. New York: Mouton de Gruyter. ---. 2008. “On Phases”. Foundational Issues in Linguistic Theory: Essays in Honor of Jean- Roger Vergnaud . (= Current Studies in Linguistics , 45.) ed. by Robert Freidin, Carlos P. Otero & Maria L. Zubizarreta, 133-166. Cambridge, London: MIT Press. ---. 2013. “Problems of Projection”. Lingua 130.33-49. Cinque, Guglielmo. 1990. Types of A-bar-Dependencies . Cambridge: MIT Press. ---. 1994. “On the Evidence for Partial N-Movement in the Romance DP”. Paths Towards Universal Grammar: Studies in Honor of Richard S. Kayne . (= Georgetown Studies in Bibliography 393 Romance Linguistics .) ed. by Guglielmo Cinque, Jan Koster, Jean-Yves Pollock, Luigi Rizzi & Raffaella Zanuttini, 85-110. Washington, D.C. Georgetown University Press. ---. 1999. Adverbs and Functional Heads . Oxford, NY et al. Oxford Univ Press. --- & Luigi Rizzi. 2008. “The Cartography of Syntactic Structures”. CISCL Working Papers on Language and Cognition 2.43-59. Corbett, Greville G. 1991. Gender . (= Cambridge Textbooks in Linguistics .) Cambridge: Cambridge University Press. ---. 1994. “Gender and Gender Systems”. The Encyclopedia of Language and Linguistics ed. by R. E. Asher & Simpson, J. M. Y., vol. 3, 1347-1353. Oxford, New York: Pergamon Press. Corver, Norbert. 1990. The Syntax of Left Branch Extractions . Ph.D. Dissertation, Tilburg University, Tilburg. --- & Henk van Riemsdijk, eds. 2001a. Semi-Lexical Categories: The Function of Content Words and the Content of Function Words . (= Studies in Generative Grammar [SGG] , 41.) Berlin, Boston: De Gruyter Mouton. --- & ---. 2001b. “Semi-Lexical Categories”. Semi-Lexical Categories: The Function of Content Words and the Content of Function Words . (= Studies in Generative Grammar [SGG] , 41.) ed. by Norbert Corver & Henk van Riemsdijk, 1-22. Berlin, Boston: De Gruyter Mouton. Cowper, Elisabeth & Daniel Currie Hall. 2009. “Argumenthood, Pronouns, and Nominal Feature Geometry”. Determiners: Universals and Variation . (= Linguistik Aktuell/ Linguistics Today (LA) , 147.) ed. by Jila Ghomeshi, Ileana Paul & Martina Wiltschko, 97-120. Amsterdam, Philadelphia: John Benjamins. Crisma, Paola. 1991. Functional Categories inside the Noun Phrase: A Study on the Distribution of Nominal Modifiers . Tesi di Laurea, University of Venice, Venice. ---. 1996. “On the Configurational Nature of Adjectival Modification”. Grammatical Theory and Romance Languages: Selected papers from the 25th Linguistic Symposium on Romance Languages (LSRL XXV) . (= Current Issues in Linguistic Theory , 133.) ed. by Karen Zagona, 59-72. Amsterdam, Philadelphia: John Benjamins. Croft, William. 1994. “Semantic Universals in Classifier Systems”. Word 45.145-171. Cruse, Alan. 2004. Meaning in Language: An Introduction to Semantics and Pragmatics . (= Oxford Textbooks in Linguistics .) Oxford, New York: Oxford Univ Press. Danon, Gabi. 2011. “Agreement and DP-Internal Feature Distribution”. Syntax 14: 4.297-317. Dean, Janet. 1966. Determiners and Relative Clauses . Ms: MIT. Déchaine, Rose-Mary & Martina Wiltschko. 2002. “Decomposing Pronouns”. Linguistic Inquiry 33: 3.409-442. Delsing, Lars-Olof. 1993. The Internal Structure of Noun Phrases in the Scandinavian Languages . Ph.D. Dissertation, University of Lund, Lund. ---. 1998. “Possession in Germanic”. Possessors, Predicates and Movement in the Determiner Phrase . (= Linguistik Aktuell/ Linguistics Today (LA) , 22.) ed. by Artemis Alexiadou & Chris Wilder, 87-108. Amsterdam, Philadelphia: John Benjamins. Despić, Miloje. 2010. “Markedness and Marked Features in Serbian”. Proceedings of the Formal Approaches to Slavic Linguistics 18 (FASL 18): The Second Cornell Meeting (For- 394 Bibliography mal Approaches to Slavic Linguistics (FASL) 18), Cornell University,NY, 15-17.05.2009 ed. by Wayles Browne, Adam Cooper, Alison Fisher, Esra Kesici, Nikola Predolac & Draga Zec, 1-20. Ann Arbor: Michigan Slavic Publications. ---. 2011. Syntax in the Absence of Determiner Phrase . Ph.D. Dissertation, University of Connecticut, Storrs. ---. 2013. “Binding and the Structure of NP in Serbo-Croatian”. Linguistic Inquiry 44.239-270. Di Domenico, Elisa. 1997. Per una Teoria del Genere Grammaticale . Padova: Unipress. Dudenredaktion, ed. 1984. Duden: Grammatik der deutschen Gegenwartssprache . (= Duden , 4.) Mannheim, et al. Dudenverlag. ---, ed. 2007. Duden: Richtiges und gutes Deutsch . Wörterbuch der sprachlichen Zweifelsfälle. (= Duden , 9.) Mannheim, et al. Dudenverlag. ---, ed. 2010. Duden: Das Stilwörterbuch . Grundlegend für gutes Deutsch. (= Duden , 2.) Mannheim, et al. Dudenverlag. ---, ed. 2016. Duden: Die Grammatik . Unentbehrlich für richtiges Deutsch. (= Duden , 4.) Mannheim, et al. Dudenverlag. Emonds, Joseph E. 1985. A Unified Theory of Syntactic Categories . (= Studies in Generative Grammar [SGG] , 19.) Berlin, Boston: De Gruyter Mouton. Epstein, Samuel D., Hisatsugu Kitahara & T. D. Seely. 2012. “Structure Building that can't Be”. Ways of Structure Building . (= Oxford Studies in Theoretical Linguistics , 40.) ed. by Myriam Uribe-Etxebarria & Vidal Valmala, 253-270. Oxford, New York, et al. Oxford Univ Press. ---, --- & ---. 2013. “Simplest Merge Generates Set Intersection: Implications for Complementizer ‘Trace’ Explanation”. Proceedings of GLOW in Asia IX 2012: The Main Session (GLOW in Asia), Mie University, 04-06.09.2012 ed. by Nobu Goto, Koichi Otaki, Atsushi Sato & Kensuke Takita, 77-92. --- & T. D. Seely. 2002. “Rule Applications as Cycles in a Level-Free Syntax”. Derivation and Explanation in the Minimalist Program ed. by Samuel D. Epstein & T. D. Seely, 65-89. Oxford: Blackwell. Etxeberria, Urtzi. 2005. Quantification and Domain Restriction in Basque . Ph.D. Dissertation, University of the Basque Country (UPV/ EHU), Vitoria-Gasteiz. ---. 2008. “On Quantification in Basque and on How Some Languages Restrict their Quantificational Domain Overtly”. Quantification : Universals and Variation . (= North-Holland Linguistic Series: Linguistic Variations , 64.) ed. by Lisa Matthewson, 225-276. Bingley: Emerald. ---. 2009. “Contextually Restricted Quantification in Basque”. Quantification, Definiteness, and Nominalization . (= Oxford Studies in Theoretical Linguistics , 24.) ed. by Anastasia Giannakidou & Monika Rathert, 76-107. Oxford, New York, et al. Oxford Univ Press. ---. 2012. “The Way the Definite Determiner Affects Quantifiers in Basque (and Beyond)”. Noun Phrases and Nominalization in Basque: Syntax and Semantics . (= Linguistik Aktuell/ Linguistics Today (LA) , 187.) ed. by Urtzi Etxeberria, Ricardo Etxepare & Myriam Uribe-Etxebarria, 79-110. Amsterdam, Philadelphia: John Benjamins. Bibliography 395 --- & Anastasia Giannakidou. 2009. Contextual Domain Restriction across Languages: Definiteness, Indefiniteness and the Structure of QP . Ms: IKER/ CNRS, University of Chicago. --- & ---. 2010. “Contextual Domain Restriction and the Definite Determiner”. Context- Dependence, Perspective and Relativity . (= Mouton Series in Pragmatics (MSP) , 6.) ed. by Francois Recanati, Isidora Stojanovic & Neftali Villanueva, 93-126. Berlin, Boston: De Gruyter Mouton. Fanselow, Gisbert. 1988. „Aufspaltung von NPn und das Problem der ‘freien’ Wortstellung“. Linguistische Berichte 114.91-113. ---. 2002. “Quirky Subjects and other Specifiers”. More than Words: A Festschrift for Dieter Wunderlich . (= Studia Grammatica , 53.) ed. by Ingrid Kaufmann & Barbara Stiebels, 227-250. Berlin: Akademie Verlag. ---. 2013. “Morphological Mismatches in Discontinous Noun Phrases”. Interfaces of Morphology: A Festschrift for Susan Olsen . (= Studia Grammatica , 74.) ed. by Holden Härtl, 11-26. Berlin: Akademie Verlag. --- & Damir Ćavar. 2002. “Distributed Deletion”. Theoretical Approaches to Universals . (= Linguistik Aktuell/ Linguistics Today (LA), 49.) ed. by Artemis Alexiadou, 65-107. Amsterdam, Philadelphia: John Benjamins. ---, Denisa Lenertovà & Thomas Weskott. 2008. “Studies on the Acceptability of Object Movement to Spec,CP”. The Discourse Potential of Underspecified Structures . (= Language, Context and Cognition , 8.) ed. by Anita Steube, 413-438. Berlin, New York: De Gruyter. Fehlisch, Ulrike. 1986. “Jeder: Stellungs- und Referenzeigenschaften”. Zur Syntax der Determinantien . (= Studien zur deutschen Grammatik , 31.) ed. by Heinz Vater, 83-122. Tübingen: Narr. Fox, Danny. 1995. “Economy and Scope”. Natural Language Semantics 3.283-341. ---. 1999. Economy and Semantic Interpretation . Cambridge: MIT Press. Frampton, John & Sam Gutmann. 2000. Agreement is Feature Sharing . Ms: Northeastern University. ---, ---, Julie Legate & Charles Yang. 2000. Remarks on ‘Derivation by Phase’: Feature Valuation, Agreement, and Intervention . Ms: Northeastern University, MIT. Frey, Werner. 2005. “Zur Syntax der linken Peripherie im Deutschen”. Deutsche Syntax: Empirie und Theorie . (= Göteborger Germanistische Forschungen , 46.) (Symposium in Göteborg), 13.-15. Mai 2004 ed. by Franz-Josef d' Avis, 147-171. Göteborg: Acta Universitatis Gothoburgensis. Gallmann, Peter. 1996. “Die Steuerung der Flexion in der DP”. Linguistische Berichte 164.283-314. ---. 2004. “Feature Sharing in DPs”. Explorations in Nominal Inflection ed. by Gereon Müller, Lutz Gunkel & Gisela Zifonun, 121-160. Berlin: Mouton de Gruyter. Gengel, Kirsten. 2009. “Phases and Ellipsis”. Linguistic Analysis 35: 1-4 Special Issue on Phase Edge Investigations.21-42. Giannakidou, Anastasia. 2004. “Domain Restriction and the Arguments of Quantificational Determiners”. Proceedings of SALT 14 (14th Semantics and Linguistic Theory 396 Bibliography Conference), Northwestern University, May 14-16, 2004 ed. by Robert B. Young, 110-126. Ithaka, NY: Department of Modern Languages, Cornell University. --- & Monika Rathert, eds. 2009. Quantification, Definiteness, and Nominalization . (= Oxford Studies in Theoretical Linguistics , 24.) Oxford, New York, et al. Oxford Univ Press. Giorgi, Alessandra & Giuseppe Longobardi. 1991. The Syntax of Noun Phrases . (= Cambridge Studies in Linguistics , 57.) Cambridge: Cambridge University Press. Giurgea, Ion. 2014. “Possible Syntactic Implementations of the Controller vs. Target Gender Distinction: The View from Ambigenerics”. Language Sciences 43.47-61. Giusti, Giuliana. 1990. “Floating Quantifiers, Scrambling and Configurationality”. Linguistic Inquiry 21: 4.633-641. ---. 1991. “The Categorial Status of Quantified Nominals”. Linguistische Berichte 136.438-454. ---. 1992. La Sintassi dei Sintagmi Nominali Quantificati: Uno Studio Comparativo . Ph.D. Dissertation, University of Venice, Venice. ---. 1997. “The Categorial Status of Determiners”. The New Comparative Syntax . (= Longman Linguistics Library .) ed. by Liliane Haegeman, 95-123. London: Pearson Longman. --- & Nedžad Leko. 1996. “Definite and Indefinite Quantity Expressions in Bosnian”. Determinatezza e indeterminatezza nelle lingue slave 5: atti del convegno svoltosi a Firenze, 26-28 ottobre 1995 ed. by Rosanna Benacchio, Francesca Fici & Lucyna Gebert, 127-145. Padova: Unipress. --- & ---. 2005. “The Categorial Status of Quantity Expressions”. Lingvistički Vidici ed. by Nedžad Leko, 121-184. Sarajevo: Forum Bosniae. Greenberg, Joseph H. 1963. “Some Universals of Grammar with Particular Reference to the Order of Meaningful Elements”. Universals of Human Language ed. by Joseph H. Greenberg, 73-113. Cambridge: MIT Press. ---. 1978. “Gender Markers”. Universals of Human Language: Volume 3 . Word Structure ed. by Joseph H. Greenberg, 48-82. Stanford, CA: Stanford University Press. Grewendorf, Günther. 1989. Ergativity in German . (= Studies in Generative Grammar [SGG], 35.) Dordrecht, Providence: Foris. Grimm, Jakob. 1870. Deutsche Grammatik . Berlin: Dümmler. Grimshaw, Jane B. 1991. Extended Projection . Ms: Brandeis University. ---. 2005. “Extended Projection”. Words and Structure ed. by Jane B. Grimshaw, 1-73. Stanford: CSLI Publications. Grohmann, Kleanthes K. 2000. Prolific Peripheries: A Radical View from the Left . Ph.D. Dissertation, University of Maryland, College Park. ---. 2003a. Prolific Domains: On the Anti-Locality of Movement Dependencies . (= Linguistik Aktuell/ Linguistics Today (LA) , 6.) Amsterdam, Philadelphia: John Benjamins. ---. 2003b. “Successive Cyclicity under (Anti-)Local Considerations”. Syntax 6: 3.260- 312. --- & Liliane Haegeman. 2003. “Resuming Reflexives”. Nordlyd 31.46-62. --- & Phoevos Panagiotidis. 2005. “An Anti-Locality Approach to Greek Demonstratives”. Contributions to the Thirtieth Incontro di Grammatica Generativa: Venice, Bibliography 397 February 26-28, 2004 (Incontro di Grammatica Generativa), Venice, IT, 26-28.02.2004 ed. by Laura Brugè, Giuliana Giusti, Nicola Munaro, Walter Schweikert & Giuseppina Turano, 243-263. Venice, IT: Università Ca’ Foscari, Venezia, Department of Language Sciences/ ATTI 2. Haider, Hubert. 1988. “Die Struktur der deutschen Nominalphrase”. Zeitschrift für Sprachwissenschaft 7: 1.32-59. ---. 1990. “Topicalization and Other Puzzles of German Syntax”. Scrambling and Barriers , ed. by Günther Grewendorf and Wolfgang Sternefeld, (= Linguistik Aktuell/ Linguistics Today 5.), 93-112. Amsterdam: John Benjamins. ---. 1992. “Die Struktur der Nominalphrase: Lexikalische und funktionale Strukturen”. Deutsche Syntax: Ansichten und Aussichten . (= Jahrbuch des Instituts für deutsche Sprache , 1991.) ed. by Ludger Hoffmann, 304-333. Berlin, New York: Walter de Gruyter. Hall, Barbara. 1962. All about Predeterminers . Ms: MIT. Halle, Morris. 1997. “Distributed Morphology: Impoverishment and Fission”. MITWPL 30.425-449. --- & Alec Marantz. 1993. “Distributed Morphology and the Pieces of Inflection”. The View from Building 20: Essays in Linguistics in Honor of Sylvain Bromberger . (= Current Studies in Linguistics , 24.) ed. by Kenneth L. Hale & Samuel J. Keyser, 111-176. Cambridge: MIT Press. Hanson, Rebecca, Heidi Harley & Elisabeth Ritter. 2000. “Underspecification and Universal Defaults for Person and Number Features”. Proceedings of the 2000 Annual Conference of the Canadian Linguistic Association ed. by John T. Jensen & Gerard van Herk, 111-122. Ottawa: Cahiers Linguistiques d'Ottawa/ Ottawa Papers in Linguistics. Harley, Heidi. 1994. “Hug a Tree: Deriving the Morphosyntactic Feature Hierarchy”. MITWPL 21: Papers on Phonology and Morphology . (= MIT Working Papers in Linguistics , 21.) ed. by Andrew Carnie & Heidi Harley, 275-288. Cambridge: MITWPL. --- & Elisabeth Ritter. 2002a. “Person and Number in Pronouns: A Feature-Geometric Analysis”. Language 78: 3.482-526. --- & ---. 2002b. “Structuring the Bundle: A Universal Morphosyntactic Feature Geometry”. Pronouns: Grammar and Representation . (= Linguistik Aktuell/ Linguistics Today (LA) , 52.) ed. by Horst J. Simon & Heike Wiese, 23-39. Amsterdam, Philadelphia: John Benjamins. Harris, James W. 1991. “The Exponence of Gender in Spanish”. Linguistic Inquiry 22: 1.27-62. ---. 1994. “The Syntax-Phonology Mapping in Catalan and Spanish Clitics”. MITWPL 21: Papers on Phonology and Morphology . (= MIT Working Papers in Linguistics , 21.) ed. by Andrew Carnie & Heidi Harley, 321-353. Cambridge: MITWPL. Heine, Bernd. 1982. “African Noun Class Systems”. Apprehension: Das sprachliche Erfassen von Gegenständen I . Bereiche und Ordnung der Gegenstände ed. by Hansjakob Seiler & Christian Lehmann, 189-216. Tübingen: Narr. Hellan, Lars. 1986. “The Headedness of NPs in Norwegian”. Features and Projections ed. by Pieter Muysken & Henk van Riemsdijk, 89-122. Dordrecht: Foris. 398 Bibliography Heycock, Caroline & Roberto Zamparelli. 2005. “Friends and Colleagues: Coordination, Plurality, and the Structure of DP”. Natural Language Semantics 13.201-270. Hiraiwa, Ken. 2005. Dimensions of Symmetry in Syntax: Agreement and Clausal Architecture . Ph.D. Dissertation, MIT, Cambridge. Hornstein, Norbert. 2001. Move! : A Minimalist Theory of Construal . Oxford: Blackwell. Jackendoff, Ray S. 1968. “Quantifiers in English”. Foundations of Language 4: 4.422-442. ---. 1977. X̄ Syntax: A Study of Phrase Structure . (= Linguistic Inquiry Monographs , 2.) Cambridge, London: MIT Press. Julien, Marit. 2005. Nominal Phrases from a Scandinavian Perspective . (= Linguistik Aktuell/ Linguistics Today (LA) , 87.) Amsterdam, Philadelphia: John Benjamins. Karnowski, Paweł & Jürgen Pafel. 2004. “A Topological Schema for Noun Phrases in German”. Explorations in Nominal Inflection ed. by Gereon Müller, Lutz Gunkel & Gisela Zifonun, 161-188. Berlin: Mouton de Gruyter. Kayne, Richard S. 1993. “Toward a Modular Theory of Auxiliary Selection”. Studia Linguistica 47: 3-31. ---. 1994. The Asymmetry of Syntax . (= Linguistic Inquiry Monographs , 25.) Cambridge, London: MIT Press. ---. 2000. Parameters and Universals . Oxford, et al. Oxford Univ Press. ---. 2005. “Some Notes on Comparative Syntax, with Special Reference to English and French”. The Oxford Handbook of Comparative Syntax . (= Oxford Handbooks in Linguistics .) ed. by Guglielmo Cinque & Richard S. Kayne, 3-69. Oxford, New York: Oxford Univ Press. Khim, Alain. 2005. “Noun Class, Gender, and the Lexicon-Syntax-Morphology Interfaces: A Comparative Study of Niger-Congo and Romance Languages”. The Oxford Handbook of Comparative Syntax . (= Oxford Handbooks in Linguistics .) ed. by Guglielmo Cinque & Richard S. Kayne, 459-512. Oxford, New York: Oxford Univ Press. Kibort, Anna. 2008. “Definiteness: Grammatical Features”. http: / / www.features.surrey. ac.uk/ features/ definiteness.html. Klinge, Alex. 2008. “Stating the Case for þ-root and hw-root Determiners”. Essays on Nominal Determination: From Morphology to Discourse Management . (= Studies on Language Companion Series (SLCS) , 99.) ed. by Henrik Høeg Müller & Alex Klinge, 233-264. Amsterdam, Philadelphia: John Benjamins. Kniffka, Gabriele. 1986. “Zur Distanzstellung von Quantoren und Quantifikatoren”. Zur Syntax der Determinantien . (= Studien zur deutschen Grammatik , 31.) ed. by Heinz Vater, 57-82. Tübingen: Narr. Koptjevskaja-Tamm, Maria. 2003a. “Action Nominal Constructions in the Languages of Europe”. Noun Phrase Structure in the Languages of Europe: EUROTYP 7 . (= Empirical Approaches to Language Typology [EALT] , 20.) ed. by Frans Plank, 723-759. Berlin, New York: Mouton de Gruyter. ---. 2003b. “Possessive Noun Phrases in the Languages of Europe”. Noun Phrase Structure in the Languages of Europe: EUROTYP 7 . (= Empirical Approaches to Language Typology [EALT] , 20.) ed. by Frans Plank, 621-722. Berlin, New York: Mouton de Gruyter. Bibliography 399 Lamontagne, Greg & Travis, Lisa de Mena. 1988. “The Syntax of Adjacency”. The Proceedings of the Sixth West Coast Conference on Formal Linguistics . (= WCCFL , 6.) ed. by Megan Crowhurst, 173-185. Stanford: Stanford Linguistics Association. Larson, Richard K. 1988. “On the Double Object Construction”. Linguistic Inquiry 19: 3.335-391. Lehmann, Christian. 1988. “On the Function of Agreement”. Agreement in Natural Language: Approaches, Theories, Descriptions ed. by Michael Barlow & Charles A. Ferguson, 55-66. Stanford, CA: Center for the Study of Language and Information. Lenerz, Jürgen. 1977. Zur Abfolge nominaler Satzglieder im Deutschen . (= Studien zur deutschen Grammatik , 5.) Tübingen: TBL Verlag Gunter Narr. Link, Godehard. 1974. “Quantoren-Floating im Deutschen”. Syntax und generative Grammatik . (= Linguistische Forschungen , 12.) ed. by Ferenc Kiefer & David M. Perlmutter, vol. 2, 105-127. Frankfurt (a.M.): Akademische Verlagsgesellschaft Athenaion. Littlefield, Heather. 2006. Syntax and Acquisition in the Prepositional Domain: Evidence from English for Fine-Grained Syntactic Categories . Ph.D. Dissertation, Boston University, Boston. Löbel, Elisabeth. 1990a. “D und Q als funktionale Kategorien in der Nominalphrase”. Linguistische Berichte 127.232-264. ---. 1990b. “Q as a Functional Category”. Syntactic Phrase Structure Phenomena in Noun Phrases and Sentences . (= Linguistik Aktuell/ Linguistics Today (LA) , 6.) ed. by Christa Bhatt, Elisabeth Löbel & Claudia M. Schmidt, 133-158. Amsterdam: John Benjamins. ---. 1994. “KP/ DP-Syntax: Interaction of Case-Marking with Referential and Nominal Features”. Theoretical Linguistics 20: 1.37-70. Longobardi, Giuseppe. 1994. “Proper Names and the Theory of N-Movement in Syntax and Logical Form”. Linguistic Inquiry 25: 4.609-665. ---. 2008. “Reference to Individuals, Person, and the Variety of Mapping Parameters”. Essays on Nominal Determination: From Morphology to Discourse Management . (= Studies on Language Companion Series (SLCS) , 99.) ed. by Henrik Høeg Müller & Alex Klinge, 189-212. Amsterdam, Philadelphia: John Benjamins. Lumsden, John S. 1992. “Underspecification in Grammatical and Natural Gender”. Linguistic Inquiry 23: 3.469-486. Lyons, Christopher. 1995. Proper Nouns, Generics, and the Count-Mass Distinction . (= Paper presented at the ESRI Research Seminar .) University of Salford. ---. 1999. Definiteness . (= Cambridge Textbooks in Linguistics .) Cambridge: Cambridge University Press. Marantz, Alec. 2001. Words . Ms: MIT. ---. 2006. Phases and Words . Ms: NYU. Martí, Luisa. 2009. “Contextual Restrictions on Indefinites: Spanish 'algunos' vs. 'unos'”. Quantification, Definiteness, and Nominalization . (= Oxford Studies in Theoretical Linguistics , 24.) ed. by Anastasia Giannakidou & Monika Rathert, 108-132. Oxford, New York, et al. Oxford Univ Press. Matthewson, Lisa. 2001. “Quantification and the Nature of Crosslinguistic Variation”. Natural Language Semantics 9: 2.145-189. 400 Bibliography McGinnis, Martha J. 1998. Locality in A-movement . Ph.D. Dissertation, MIT, Cambridge. Merchant, Jason. 1996. “Object Scrambling and Quantifier Float in German”. Proceedings of the North East Linguistic Society (NELS) 26 (North East Linguistic Society (NELS) 26), Cambridge, Mass, 27-30.10.1995 ed. by Kiyomi Kusumoto, 179-193. Amherst: GLSA (Graduate Linguistic Student Association). Milsark, Gary L. 1974. Existential Sentences in English . Ph.D. Dissertation, MIT, Cambridge. Milwark [sic! ], Gary. 1977. “Toward an Explanation of Certain Peculiarities of the Existential Construction in English”. Linguistic Analysis 3.1-29. Miyagawa, Shigeru. 2011. “Optionality”. The Oxford Handbook of Linguistic Minimalism . (= Oxford Handbooks in Linguistics .) ed. by Cedric Boeckx, 354-376. Oxford: Oxford Univ Press. Motsch, Wolfgang. 2004. Deutsche Wortbildung in Grundzügen . (= Schriften des Instituts für Deutsche Sprache , 8.) Berlin, New York: Walter de Gruyter. Müller, Michael. 1986. “Zur Verbindbarkeit der Determinantien und Quantoren”. Zur Syntax der Determinantien . (= Studien zur deutschen Grammatik , 31.) ed. by Heinz Vater, 33-56. Tübingen: Narr. Musan, Renate. 1999. “Temporal Interpretation and Information-Status of Noun Phrases”. Linguistics and Philosophy 22: 6.621-661. Neeleman, Ad & Van De Koot, Hans. 2006. “Syntactic Haplology”. The Blackwell Companion to Syntax . (= Blackwell handbooks in linguistics , 19.) ed. by Martin Everaert & Henk van Riemsdijk, vol. 4, 685-710. London: Blackwell. Norde, Muriel. 1997. The History of the Genitive in Swedish: A Case Study in Degrammaticalization . Ph.D. Dissertation, Univertity of Amsterdam, Amsterdam. Norris, Mark. 2012. “Towards an Analysis of Concord (in Icelandic)”. Proceedings of the 29th West Coast Conference on Formal Linguistics: WCCFL 29 ed. by Jaehoon Choi, E. A. Hogue, Jeffrey Punske, Deniz Tat, Jessamyn Schertz & Alex Trueman, 205-213. Somerville: Cascadilla Proceedings Project. ---. 2014. A Theory of Nominal Concord . Ph.D Dissertation, University of California, Santa Cruz. Noyer, Robert R. 1992. Features, Positions and Affixes in Autonomous Morphological Structure . Ph.D. Dissertation, MIT, Cambridge. ---. 1997. Features, Positions and Affixes in Autonomous Morphological Structure . New York: Garland Publishing. Nunes, Jairo. 2004. Linearization of Chains and Sideward Movement . (= Linguistic Inquiry Monographs , 43.) Cambridge, London: MIT Press. Olsen, Susan. 1990. “AGR(eement) in the German Noun Phrase”. Syntactic Phrase Structure Phenomena in Noun Phrases and Sentences . (= Linguistik Aktuell/ Linguistics Today (LA) , 6.) ed. by Christa Bhatt, Elisabeth Löbel & Claudia M. Schmidt, 39-49. Amsterdam: John Benjamins. Pafel, Jürgen. 1994. “Zur syntaktischen Struktur nominaler Quantoren”. Zeitschrift für Sprachwissenschaft 13.236-275. ---. 2005. Quantifier Scope in German . (= Linguistik Aktuell/ Linguistics Today (LA) , 84.) Amsterdam, NL, Philadelphia: John Benjamins. Bibliography 401 Panagiotidis, Phoevos. 2002. Pronouns, Clitics and Empty Nouns: 'Pronominality' and Licensing in Syntax . Amsterdam, Philadelphia: John Benjamins. Payne, John. 1993. “The Headedness of Noun Phrases: Slaying the Nominal Hydra”. Heads in Grammatical Theory ed. by Greville G. Corbett, Norman M. Fraser & Scott McGlashan, 114-139. Cambridge: Cambridge University Press. Pesetsky, David. 1982. Paths and Categories . Ph.D. Dissertation, MIT, Cambridge. ---. 1989. Language-Particular Processes and the Earliness Principle . Ms: MIT. --- & Esther Torrego. 2001. “T-to-C Movement: Causes and Consequences”. Ken Hale: A Life in Language . (= Current Studies in Linguistics , 36.) ed. by Michael J. Kenstowicz, 355-426. Cambridge, London: MIT Press. --- & ---. 2007. “The Syntax of Valuation and the Interpretability of Features”. Phrasal and Clausal Architecture: Syntactic Derivation and Interpretation . In Honor of Joseph E. Emonds. (= Linguistik Aktuell/ Linguistics Today (LA) .) ed. by Simin Karimi, Vida Samiian & Wendy K. Wilkins, 262-294. Amsterdam, Philadelphia: John Benjamins. Picallo, M. C. 1991. “Nominals and Nominalization in Catalan”. Probus 3: 3.279-316. Pollard, Carl J. & Ivan A. Sag. 1994. Head-Driven Phrase Structure Grammar . Stanford, Chicago: Center for the Study of Language and Information; University of Chicago Press. Pollock, Jean-Yves. 1989. “Verb Movement, Universal Grammar and the Structure of IP”. Linguistic Inquiry 20: 3.365-424. Postal, Paul M. 1966a. “On So-Called 'Pronouns' in English”. Report of the Seventeenth Annual Round Table Meeting on Linguistics and Language Studies . (= Monograph Series , 19.) (17th Annual Round Table) ed. by Francis P. Dinneen, S. J., 177-126. Washington, D.C. Georgetown University Press. ---. 1966b. “A Note on 'Understood Transitively'”. International Journal of American Linguistics 32: 1.90-93. ---. 1970. “On So-Called Pronouns in English”. Readings in English Transformational Grammar ed. by Roderick A. Jacobs & Peter S. Rosenbaum. Waltham: Ginn & Company. Preminger, Omer. 2011. Agreement as a Fallible Operation . Ph.D Dissertation, MIT, Cambridge. ---. 2013. “That’s not How You Agree: A Reply to Zeijlstra”. Linguistic Review 30: 3.491- 500. Quirk, Randolph & Jan Svartvik. 1966. Investigating Linguistic Acceptability . Berlin, Boston: De Gruyter Mouton. Radford, Andrew. 1977. “Counter-Filtering Rules”. York Papers in Linguistics 7.7-45. ---. 1979. “The Functional Basis of Transformations”. Transactions of the Philological Society 77: 1.1-42. ---. 1993. “Head-Hunting: On the Trail of the Nominal Janus”. Heads in Grammatical Theory ed. by Greville G. Corbett, Norman M. Fraser & Scott McGlashan, 73-113. Cambridge: Cambridge University Press. Reis, Marga & Heinz Vater. 1980. “Beide”. Wege zur Universalienforschung: Sprachwissenschaftliche Beiträge zum 60. Geburtstag von Hansjakob Seiler ed. by Gunter Brettschneider & Christian Lehmann, 360-386. Tübingen: Narr. 402 Bibliography Rezac, Milan. 2003. “The Fine Structure of Cyclic Agree”. Syntax 6: 2.156-182. Richards, Marc. 2007. “On Feature-Inheritance: An Argument from the Phase Impenetrability Condition”. Linguistic Inquiry 38: 3.563-572. Ritter, Elisabeth. 1988. “A Head-Movement Approach to Construct-State Noun Phrases”. Linguistics and Philosophy 26: 6.909-929. ---. 1991a. Evidence for Number as a Nominal Head . (= Paper Presented at GLOW 26 .) Leiden. ---. 1991b. “Two Functional categories in Noun Phrases: Evidence from Modern Hebrew”. Perspectives on Phrase Structure: Heads and Licensing . (= Syntax and Semantics , 25.) ed. by Susan D. Rothstein, 37-62. San Diego: Emerald Group Publishing. ---. 1992. “Cross-Linguistic Evidence for Number Phrase”. Canadian Journal of Linguistics 37.197-218. ---. 1993. “Where's Gender? ”. Linguistic Inquiry 24: 4.795-803. Rizzi, Luigi. 1997. “The Fine Structure of the Left Periphery”. Elements of Grammar: A Handbook in Generative Syntax ed. by Liliane Haegeman, 281-337. Dordrecht: Kluwer. Roehrs, Dorian. in preparation. Semantically Vacuous Elements: Adjectival Inflections and the Article ein in German . (Provisional Title). http: / / ling.auf.net/ lingbuzz/ 002457. ---. 2005. Possessives Consist of Heads and Complements: Some Notes . http: / / ling.auf.net/ lingbuzz/ 000302. ---. 2009. Demonstratives and Definite Articles as Nominal Auxiliaries . (= Linguistik Aktuell/ Linguistics Today (LA) , 140.) Amsterdam, Philadelphia: John Benjamins. ---. 2012. “Complex Determiners: A Case Study of German ein jeder”. Interdisciplinary Journal for Germanic Linguistics and Semiotic Analysis 17: 1.1-56. ---. 2013. “Possessives as Extended Projections”. Working Papers in Scandinavian Syntax 91.37-112. Ross, John R. 1972. “The Category Squish: Endstation Hauptwort”. Papers from the 8th Regional Meeting of the Chicago Linguistic Society, University of Chicago, Linguistic Department, April 14-16, 1972 ed. by Paul M. Peranteau, Judith N. Levi & Gloria C. Phares, 316-328. Chicago: CLS. ---. 1973. “Nouniness”. Three Dimensions of Linguistic Research ed. by Osamu Fujimura, 137-257. Tokyo: TEC Company Ltd. ---. 2004. “Nouniness”. Fuzzy Grammar: A Reader ed. by Bas Aarts, David Denison, Evelien Keizer & Gergana Popova, 351-422. Oxford: Oxford Univ Press. Saab, Andrés L. 2010. “(Im)possible Deletions in the Spanish DP”. Iberia: An International Journal of Theoretical Linguistics 2: 2.45-83. Saito, Mamoru. 2001. “Movement and Theta-Roles: A Case Study with Resultatives”. The Proceedings of the Second Tokyo Conference on Psycholinguistics ed. by Yukio Otsu, 35-60. Tokyo: Hituzi Shobo. ---, T.-H. J. Lin & Keiko Murasugi. 2008. “N'-ellipsis and the Structure of Noun Phrase in Chinese and Japanese”. Journal of East Asian Linguistics 17.247-271. Salzmann, Martin. 2011. “Silent Resumptives in Zurich German Possessor Relativization”. Null Pronouns . (= Studies in Generative Grammar [SGG] , 106.) ed. by Melani Wratil & Peter Gallmann, 141-221. Berlin, Boston: De Gruyter Mouton. Bibliography 403 Sauerland, Uli. 1996. The Late Insertion of Germanic Inflection . Ms: MIT. Schoorlemmer, Erik. 2009. Agreement, Dominance and Doubling: The Morphosyntax of DP . Ph.D. Dissertation, LOT, Utrecht. Schütze, Carson T. 2016. The Empirical Base of Linguistics: Grammaticality Judgments and Linguistic Methodology . (=Classics in Linguistics, 2.) Berlin: Language Science Press. Schwarzschild, Roger. 1996. Pluralities . Dordrecht, Boston: Kluwer Academic. Shlonsky, Ur. 1991. “Quantifiers as Functional Heads: A Study of Quantifier Float in Hebrew”. Lingua 84.159-180. ---. 2010. “The Cartographic Enterprise in Syntax”. Language and Linguistics Compass 4: 6.417-229. Sigur∂sson, Halldór Á. 1996. “Icelandic Finite Verb Agreement”. Working Papers in Scandinavian Syntax 57.1-46. Sportiche, Dominique. 1988. “A Theory of Floating Quantifiers and its Corollaries for Constituent Structure”. Linguistic Inquiry 19: 2.425-451. Starke, Michal. 2001. Move Dissolves into Merge: A Theory of Locality . Ph.D Dissertation, University of Geneva, Geneva. Sternefeld, Wolfgang. 2006. Syntax: Eine morphologisch motivierte generative Beschreibung des Deutschen . Band 1. Tübingen: Stauffenburg. Svenonius, Peter. 2004. “On the Edge”. Peripheries: Syntactic Edges and their Effects ed. by David Adger, Cécile de Cat & George Tsoulas, 261-287. Dordrecht: Kluwer. Szabolcsi, Anna. 1981. “The Possessive Construction in Hungarian: A Configurational Category in a Non-Configurational Language”. Acta Linguistica Academiae Scientiarum Hungaricae 31.161-189. ---. 1983. “The Possessor that ran away from Home”. The Linguistic Review 3: 1.89-102. ---. 1987. “Functional Categories in the Noun Phrase”. Approaches to Hungarian 2: Theories and Analyses . (= Approaches to Hungarian , 2.) ed. by István Kenesei, 167-189. Szeged: Jate. ---. 1992. “Subordination: Articles and Complementizers”. Approaches to Hungarian 4: The Structure of Hungarian ed. by István Kenesei & Csaba Pléh, 123-137. Szeged: Jate. ---. 1994. “The Noun Phrase”. The Syntactic Structure of Hungarian . (= Syntax and Semantics , 27.) ed. by Stephen R. Anderson & Katalin F. Kiss, 179-275. San Diego: Emerald Group Publishing. Takahashi, Masahiko. 2011. Some Consequences of Case-Marking in Japanese . Ph.D Dissertation, University of Connecticut, Storrs. Tănase-Dogaru, Mihaela. 2011. “Diagnostic Criteria for Semi-Lexicality”. Bucharest Working Papers in Linguistics 13: 1.151-165. Tappe, Hans-Thilo. 1989. “A Note on Split Topicalization in German”. Syntactic Phrasestructure Phenomena in Noun Phrases and Sentences ed. by Christa Bhatt, Elisabeth Löbel, and Claudia Maria Schmidt (= Linguistik Aktuell/ Linguistics Today (LA) , 6.), 159-179. Amsterdam, Philadelphia: John Benjamins. Thomason, Sarah G. & Daniel L. Everett. 2001. “Pronoun Borrowing”. Proceedings of the Annual Meeting of the Berkeley Linguistics Society 27: 1.301-315. 404 Bibliography Ticio, Emma M. 2010. Locality Domains in the Spanish Determiner Phrase . (= Studies in Natural Language and Linguistic Theory , 79.) Heidelberg, London, New York: Springer Dordrecht. Valois, Daniel. 1991. The internal Syntax of DP . Ph.D Dissertation, UCLA, Los Angeles. van Riemsdijk, Henk. 1989. “Movement and Regeneration”. Dialect Variation and the Theory of Grammar ed. by Paola Benincà, 105-136. Dordrecht: Foris. ---. 1990. “Functional Prepositions”. Unity in Diversity: Papers Presented to Simon C. Dik on his 50th Birthday ed. by Harm Pinkster & Inge Genee, 229-242. Berlin, Boston: De Gruyter Mouton. ---. 1998. “Categorial Feature Magnetism: The Endocentricity and Distribution of Projections”. Journal of Comparative Germanic Linguistics 2.1-48. Vater, Heinz. 1979. Determinantien: Teil I . Abgrenzung, Syntax. (= KLAGE (Kölner Linguistische Arbeiten Germanistik) , 6.) Trier: L.A.U.T. ---. 1980. “Quantifier Floating in German”. The Semantics of Determiners . (= Croom Helm Linguistic Series .) ed. by Van der Auwera, Johan, 232-249. London, Baltimore: Croom Helm; University Park Press. ---. 1984. “Determinantien und Quantoren im Deutschen”. Zeitschrift für Sprachwissenschaft 3: 1.19-42. ---. 1986. “Zur Abgrenzung der Determinantien und Quantoren”. Zur Syntax der Determinantien . (= Studien zur deutschen Grammatik , 31.) ed. by Heinz Vater, 13-31. Tübingen: Narr. Vinka, Mikael. 2001. “Impoverishment as Feature Deletion: Dual and Plural Agreement in Sámi”. Lund Working Papers in Linguistics 48.183-191. von Fintel, Kai. 1998. The Semantics and Pragmatics of Quantifier Domains . Ms: MIT. Watanabe, Akira. 2010. “Notes on Nominal Ellipsis and the Nature of no and Classifiers in Japanese”. Journal of East Asian Linguistics 19. Westerståhl, Dag. 1984. “Determiners and Context Sets”. Generalized Quantifiers in Natural Language . (= Groningen-Amsterdam Studies in Semantics (GRASS) , 4.) ed. by Johan van Benthem & ter Meulen, Alice G. B., 45-72. Dordrecht: Foris. Williams, Edwin. 1981. “Argument Structure and Morphology”. Linguistic Review 1.81-114. Wiltschko, Martina. 2009. “What’s in a determiner and how did it get there? ”. Determiners: Universals and Variation . (= Linguistik Aktuell/ Linguistics Today (LA) , 147.) ed. by Jila Ghomeshi, Ileana Paul & Martina Wiltschko, 25-66. Amsterdam, Philadelphia: John Benjamins. Witkoś, Jacek. 2004. “Reviewing Phases: On some Concepts and Consequences of Phase-Based Minimalism”. Poznań Studies in Contemporary Linguistics 39.197-251. Woolford, Ellen. 2006. “Lexical Case, Inherent Case, and Argument Structure”. Linguistic Inquiry 37: 1.111-130. Yang, Dong-Whee. 2008. On Edge Feature Movement . Ms: MIT. Zaliznjak, Andrej A. 1964. “K voprosu o grammatičeskix kategorijax roda i oduševlennosti v sovremennon russkom jazyke”. Voprosy jazykoznanija 4.25-40. Zamparelli, Roberto. 1995. Layers in the Determiner Phrase . Ph.D. Dissertation, (Revised Version), University of Rochester, Rochester. Bibliography 405 ---. 2000. Layers in the Determiner Phrase . New York: Garland Publishing. Zeijlstra, Hedde. 2012. “There is only One Way to Agree”. Linguistic Review 29: 3.491- 539. Zwarts, Joost. 2011. “Lexical and Functional Properties of Prepositions”. Lexikalische und grammatische Eigenschaften präpositionaler Elemente ed. by Dagmar Haumann & Stefan J. Schierholz, 1-18. Berlin, Boston: De Gruyter. 406 List of Short Form Citations 406 List of Short Form Citations C&G Cardinaletti & Giusti (2006) C&S Cardinaletti & Starke E&G Etxeberria & Giannakidou EKS Epstein, Kitahara & Seely H&R Harley & Ritter (2002a) P&T Pesetsky & Torrego (2007) AUGB Chomsky (2007) BEA --- (2004) BPS --- (1995b) DbP --- (2001) MI --- (2000) MP --- (1995a) MPLT --- (1993) OP --- (2008) PoP --- (2013) The book focusses on the grammatical feature definiteness in German, visible in the inflection of adjectives (“ein schönes Kind” vs. “das schön-e Kind”). It argues for an analysis of this effect that draws a connection to the visible categories of number and gender on nouns and related words rather than an abstract property. This conclusion rests on the conflation of the established grammatical categories into a single one, number-gender, which explains a vast body of grammatical phenomena in German and principles of language in general. ISBN 978-3-8233-8133-4 Benincasa A Phi-Syntax for Nominal Concord Marco Benincasa A Phi-Syntax for Nominal Concord Morphological Definiteness as Gender-Agreement in the German DP