<?page no="0"?> Gunter Narr Verlag Tübingen Silke Höche Cognate Object Constructions in English A Cognitive-Linguistic Account Language in Performance LiP <?page no="1"?> Cognate Object Constructions in English <?page no="2"?> 41 Edited by Werner Hüllen† and Rainer Schulze Advisory Board: Thomas Herbst (Erlangen), Andreas Jucker (Zürich), Manfred Krug (Bamberg), Christian Mair (Freiburg i. Br.), Ute Römer (Hannover), Andrea Sand (Trier), Hans-Jörg Schmid (München), Josef Schmied (Chemnitz) and Edgar W. Schneider (Regensburg) <?page no="3"?> Silke Höche Cognate Object Constructions in English A Cognitive-Linguistic Account Gunter Narr Verlag Tübingen <?page no="4"?> Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über <http: / / dnb.d-nb.de> abrufbar. Dissertation an der Ruhr-Universität Bochum (D 294) © 2009 · 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. Gedruckt auf säurefreiem und alterungsbeständigem Werkdruckpapier. Internet: www.narr.de E-Mail: info@narr.de Druck und Bindung: Ilmprint, Langewiesen Printed in Germany ISSN 0939-9399 ISBN 978-3-8233-6489-4 <?page no="5"?> Acknowledgements When I seemed to die a thousand small deaths during my final examinations as a student of English Literature and Linguistics, I wasn't aware that dying deaths, dreaming dreams and telling tales would one day be the subject of my doctoral thesis on grammatical constructions, let alone a complete book. Now, that the book is finished, it is time to thank a thousand thanks to all those people who have never stopped supporting me and believing that this enterprise would come to a successful end. My greatest thanks are owed to my mom, my sister, Fritz and my grandma for their love and constant moral, emotional and financial support. Thank you for always being there for me and providing me with a safe and warm haven during this adventurous journey! I would also like to acknowledge friends and colleagues who reduced the stress of writing a doctoral thesis, especially Darja, Ingo, Martina and Andrea, who lent me their ears, opinions and (technical) support. Special thanks are due to Prof. Dr. Rainer Schulze, the editor of this book series, and Stefan Thomas Gries for sharing with me his expertise and knowledge on statistics and the interpretation of data. Last but most importantly I would like to express my deep gratitude to my research supervisor Prof. Dr. Doris Schönefeld for her sustained encouragement, guidance, patience and advice. Several of the ideas and models presented and discussed in this book owe their final shape to stimulating suggestions by and insightful discussions with her. Her enthusiasm for linguistic research has always been an inspiration for me and kept me going in phases of doubt and uncertainty. <?page no="7"?> VII Table of Contents 1 Introduction .............................................................................................. 1 1.1 Aims of the book ............................................................................. 1 1.2 Methodological issues .................................................................... 3 1.3 Overview of the book ..................................................................... 5 2 Cognate Object Constructions: Past and Present .............................. 8 2.1 Introduction ..................................................................................... 8 2.2 COCs in Descriptive / Reference Grammars .............................. 9 2.2.1 Henry Sweet (1891) ............................................................... 9 2.2.2 Otto Jespersen (1927) ............................................................ 10 2.2.3 Quirk et al. (1985) .................................................................. 12 2.2.4 Huddleston & Pullum (2002) .............................................. 15 2.2.5 Summary: Descriptive Grammar accounts of COCs ....... 16 2.3 Generative Grammar approaches to COCs ................................. 17 2.3.1 Generative Grammar / Government and Binding: Chomsky (1965, 1981) ........................................................... 17 2.3.2 Cognate Objects as adjuncts: Jones (1988) ......................... 20 2.3.3 Cognate Objects as arguments: Massam (1990) ................ 23 2.3.4 Summary and discussion: COCs in Generative Grammar 28 2.4 Functional Grammar accounts of COCs ...................................... 29 2.4.1 COCs as instances of valency extension: Dik (1997) ........ 30 2.4.2 Cognate Objects as Participants: Halliday (1985, 2004) ... 32 2.4.3 A Functional Constraint on the COC: Kuno & Takami (2004) ....................................................................................... 34 2.4.4 Summary and discussion: COCs in Functional Grammar 37 2.5 Macfarland (1995): Cognate Objects and the adjunct / argument distinction ........................................................................................ 38 2.5.1 Constraints on the form of COCs ....................................... 38 2.5.2 COCs compared to Light Verb Constructions .................. 41 2.5.3 Evidence for COs as arguments .......................................... 42 2.5.4 Summary and discussion ..................................................... 45 2.6 Conclusion ....................................................................................... 46 3 Basic Tenets of Cognitive Linguistics ................................................. 49 3.1 Introduction: Cognitive Linguistics and Cognitive Grammar 49 3.2 The non-autonomy of language ................................................... 50 3.3 The symbolic nature of grammar ................................................. 52 <?page no="8"?> VIII 3.4 Meaning as conceptualization ...................................................... 54 3.4.1 Subjectivist and encyclopaedic semantics ......................... 54 3.4.2 Construal ................................................................................ 55 3.5 Linguistic categorization ............................................................... 60 3.5.1 Linguistic categories as prototype categories ................... 60 3.5.2 Language usage as categorization ...................................... 62 3.6 The usage-based account of language ......................................... 63 3.7 Constructions and construction grammars ................................ 66 3.7.1 Construction grammars: Common grounds ..................... 66 3.7.2 Predictability and compositionality of constructions ...... 69 3.7.3 Autonomous syntax vs. reductionism? ............................. 70 3.7.4 Lexical polysemy and constructional meaning ................ 73 3.8 Conclusion ......................................................................................... 74 4 A Cognitive Linguistic description of COCs: General issues ........ 76 4.1 Introduction ..................................................................................... 76 4.2 The meaning of COCs ..................................................................... 77 4.3 Events, results, landmarks: Different types of Cognate Objects 79 4.3.1 Eventive Object vs. Object of Result ................................... 79 4.3.2 Landmarks and products ..................................................... 84 4.3.3 Derivational directions ......................................................... 86 4.3.4 Types of COs - A summarizing overview ........................ 88 4.3.5 Cognate Subjects? .................................................................. 89 4.4 The composite structure [V + CO], or: How to integrate an object with an intransitive verb ........................................................ 92 4.4.1 Langacker's approach: Where does the e-site come from? 92 4.4.2 COC S : Arguments or adverbials? ....................................... 95 4.4.3 COCs as argument structure constructions - Testing Goldberg's model .................................................................. 100 4.4.4 A 'hybrid - model' for COCs ............................................... 106 4.5 Repeated content: Redundancy vs. Iconicity .............................. 108 4.5.1 Iconicity in language ............................................................ 108 4.5.2 The "more of form" ............................................................... 110 4.5.3 The iconicity of repetition .................................................... 112 4.6 Summary and conclusion ................................................................ 117 5 Consulting the corpus: Towards a usage-based network of COCs 118 5.1 Introduction ..................................................................................... 118 5.2 CL and corpus linguistics: The usage-based approach to language .......................................................................................... 119 5.3 General methodological considerations ...................................... 122 <?page no="9"?> IX 5.4 Semantic classes and COC types .................................................. 124 5. 5 A collexeme analysis of COCs - Measuring the association between verb and construction .................................................... 132 5.6 A usage-based network of argument structure constructions 137 5.7 Summary and conclusion .............................................................. 143 6 The Transitivity of COCs ...................................................................... 144 6.1. Introduction ..................................................................................... 144 6.2. Transitivity in Cognitive Grammar ............................................. 144 6.3 Hopper & Thompson (1980): Features of transitivity ............... 148 6.4 The transitivity prototype .............................................................. 150 6.5 Other event scenarios ..................................................................... 153 6.6 COCs — Means of elaborating event structure .......................... 159 6.6.1 Unaccusative verbs ............................................................... 159 6.6.2 Unergative verbs ................................................................... 163 6.6.3 Transitive verbs ..................................................................... 165 6.7 The transitivity continuum ............................................................ 167 6.8 To passivize or not to passivize? .................................................. 171 6.9 Summary and conclusion .............................................................. 179 7 ‘Packaged’ events: Aspectual characteristics of COCs ..................... 180 7.1 Introduction ..................................................................................... 180 7.2 Aspect, aktionsart, aspectuality ..................................................... 181 7.3 Boundedness: Concepts and implications .................................. 184 7.4 Iterative processes ........................................................................... 187 7.5 Continuous processes ..................................................................... 189 7.6 Peripheral cases ............................................................................... 195 7.7 Conclusion ....................................................................................... 198 8 Mid-level schemas: The make-up of the CO-nominal ..................... 200 8.1 Introduction ..................................................................................... 200 8.2 Determiners in the CO-phrase ...................................................... 200 8.3 Modification patterns ..................................................................... 209 8.4 Summary and conclusion .............................................................. 213 9 COCs vs. VACs: Putting alternations to the test ............................... 215 9.1 Preliminary considerations ........................................................... 215 9.2 Live happily or live a happy life ? ...................................................... 220 9.3 Die naturally or die a natural death? ................................................ 223 <?page no="10"?> X 9.4 Smile slowly or smile a slow smile? .................................................. 226 9.5 Summary and discussion ............................................................... 228 10 Don’t have a run, run a run! — LVCs and COCs compared ............ 231 10.1 Introduction ..................................................................................... 231 10.2 A brief portrayal of LVCs .............................................................. 233 10.2.1 General observations ....................................................... 233 10.2.2 Formal characteristics of LVCs ...................................... 234 10.2.3 Syntactic and semantic issues in the study of LVCs ... 235 10.3 Parallels between COCs and LVCs ? ........................................... 237 10.4 The semantics of selected LVCs in comparison to COCs ......... 240 10.4.1 Have-LVCs: Have a dance vs. dance a dance .................... 240 10.4.2 Give-LVCs: Give a smile vs. smile a smile ........................ 243 10.4.3 Take-LVCs: Take a breath vs. breathe a breath .................. 246 10.4.4 Do-LVCs: Do a dance vs. dance a dance .......................... 249 10.5 Summary and conclusion ............................................................. 250 11 Lower-level schemas: Collocations, idioms, register ....................... 254 11.1 Preliminary considerations ........................................................... 254 11.2 Variable low-level schemas ........................................................... 256 11.3 Fixed units ........................................................................................ 259 11.4 Register ............................................................................................. 263 11.5 Summary and conclusion: A usage-based network of constructions ................................................................................... 269 12 I have a tale to tell: Conclusion and outlook ..................................... 272 12.1 Summary of findings ...................................................................... 272 12.2 Future perspectives ........................................................................ 277 13 References ................................................................................................. 282 14 Appendix ................................................................................................... 298 <?page no="11"?> XI List of abbreviations 1. Theoretical Frameworks CL Cognitive Linguistics CG Cognitive Grammar FG Functional Grammar GG Generative Grammar 2. Construction types and constituents COC Cognate Object Construction EV/ R 1 COCs of the type 'Event/ Result' R 2 COCs of the type 'Result' A COCs of the type 'Affected' LVC Light Verb Construction TLV True Light Verbs TO Transitivizing Object VAC Verb Adverb Construction VAV Vague Action Verb <?page no="13"?> 1 1 Introduction 1.1 Aims of the book Sentences like (1) a. Tessa stopped crying and sighed a deep, uncontrollable sigh like a yawn. (G12: 1551) 1 b. Delighted to get some response at last, Lord Cumbermound grinned his evil grin. (H8B: 1221) c. Her secret love for Robert must stay hidden, and perhaps eventually it would die a slow and painful death. (JYE: 3864) display a construction which has become known as 'Cognate Object Construction' (COC) in linguistic descriptions. 'Cognate', from Latin cognatus, meaning 'related by blood', however, has turned out to be a problematic notion when applied to the portrayal of such verb-object combinations, especially as regards descriptive and/ or definitional criteria for the specification of the types of relations that hold between verb and object, be they of a semantic, morphological or syntactical nature. While the description and analysis of the construction played some role in Generative Grammar research, where it was discussed as a challenge to certain established principles and theoretical constructs (e.g. subcategorization frames or case-assignment), not much attention has been paid to this phenomenon in more recent approaches to and models of language, such as Functional Grammar or Cognitive Linguistics. This apparent 'gap' has been one major motivation for the research project to be presented here: a profound description of COCs 2 as meaningful patterns within the framework of Cognitive Linguistics (CL), exploiting its methodological and descriptive inventory and its assumptions about mental representation of language in speakers. The notion of 'construction', in most general and neutral terms defined as a pairing of form and meaning (i.e. a symbolic unit), has in recent times 1 If not indicated otherwise, all the examples given are quoted from the British National Corpus (BNC). 2 The use of the plural form "constructions" is deliberate and will come to be explained in the course of this investigation. (One general note on footnotes: Numbering of footnotes starts anew from 1 in each chapter for the sake of clarity and readerfriendliness). <?page no="14"?> 2 become a core concept in linguistic description, being part and parcel of descriptive models of so-called construction grammar approaches to language (see esp. Fillmore, Kay & O'Connor 1988). Proponents of construction grammar, a family of related grammatical theories, view language as consisting of a taxonomic inventory of constructions as elementary units, instead of as atomic syntactic units and rules of their combination into larger structures. Assumptions and principles of construction grammar are inherent to models of linguistic representation developed within the framework of Cognitive Linguistics (see especially Goldberg 1995, 2006; Lakoff 1987; Langacker 1987a, 1991a). Being couched into the CL paradigm, the present investigation has as one of its main objectives to depict COCs as symbolic units, viz. unique pairings of form (syntactic and phonological aspects) and meaning (semantic and pragmatic aspects), and to spell out their lexical, semantic and formal properties. As noted initially, the notion of 'cognate' has been the source of controversies about the types of defining relations that hold between verb and object, and hence the types of object that can be recognized as cognate forms. This shows in the various attempts by scholars to single out syntactic, morphological and semantic criteria for the definition and delimitation of the category of 'cognate objects'. The present approach differs from previous accounts in that it broadens the category and incorporates into the analysis forms which have so far not been considered as cognate objects. Extending the scope of possible candidates of cognate objects, I will abandon the idea of one single COC-type and instead depict a family of different, but related types of COCs. My attempt to demonstrate the relationship between the various constructional patterns and their more or less schematic representations within a constructional network takes up and builds on the notion of a 'taxonomic inventory' of linguistic units, eventually resulting in the integration of COCs into a larger network of constructions. The postulation of constructional schemas will be tenable only if it is accompanied by an assessment of their plausibility with respect to actual usage and exploitation by speakers when processing frequently used phrases such as I just wanna live my life or possibly creating novel or unusual forms, such as I am so damn flam, I slam a slam 3 . The present project thus constitutes and promotes a usage-based approach to COCs in that it draws on and abstracts away from concrete, naturally occurring language samples as documented in the British National Corpus (BNC), currently being the largest available corpus of present-day English. This approach paves the way for one further objective to be realised in this survey: the development of a network of COCs which comprises schemas which not 3 Line of the song Soul by the Pound by the band "Common"; found on http: / / www.lyricsfind.com/ c / common / 5915.html (online access: 02/ 07/ 07). <?page no="15"?> 3 only differ with respect to their degree of abstractness, but which, moreover, are weighted on the basis of their frequency of usage. 1.2 Methodological issues As spelt out in the preceding section, one major goal of my investigation of COCs is a close analysis of their semantics, i.e. their conceptual content. There exists a broad spectrum of methodologies for the exploration of conceptual structure (which surfaces in linguistic structure), such as (the traditional) introspection, cross-linguistic comparison of linguistic phenomena of typologically related and unrelated languages, computer-aided investigation of large language corpora, inferential statistical analyses and comparison of the data obtained from such corpus investigation, psycholinguistic experiments such as eye-movement measures or lexical decision tasks, to name just a few (see Gonzalez-Marques et al. 2007, Janssen & Redeker 1999). Although there is a wide recognition for the striving for converging evidence among the cognitive linguists community, i.e. evidence combining findings from introspection, corpus analyses and psycholinguistic and neurolinguistic experimentation, a large part of the studies done in the field lacks such a comprehensive foundation, presenting results arrived at by 'isolated' methodological procedures (i.e. by introspection only, or by corpus investigation only etc.). My investigation of COCs attempts at the integration of at least two sources of evidence: introspection in conjunction with theoretical analysis and corpus investigation in conjunction with inferential statistical analysis. Although the application of purely introspective methods has been criticized heavily and vehemently by some cognitive linguists (see, among others, Cuyckens et al. 1997; Geeraerts 1999, 2006; Gibbs 2007), one can, however, not neglect or even deny the importance of introspection for the development of Cognitive Linguistics in general, and for the analysis of conceptual/ semantic structure in particular. Talmy (2007: xii) assesses the role of introspective procedures in linguistic research as follows: The methodology of introspection […] has been central in the development of cognitive linguistics and continues as its main methodology, […] and its particular profile of limitations has in part led to the pattern in the use of other methodologies. Gibbs (2007: 3), who pleads for a stronger integration of empirical methods into cognitive linguistic research, does, nonetheless, also grant some importance to introspections in that they form a "valuable sources for constructing hypotheses" and in that they have led to findings about structures of the human conceptual system which may "indeed be correct and thus psychologically plausible". <?page no="16"?> 4 According to Talmy's (2007) account, which evaluates the accessibility of linguistic categories and structures to introspection, it is the aspect of meaning (i.e. conceptual content) which is most easily accessible to introspective analysis, and here in particular the meaning of concrete words, phrases and sentences (cf. ibid.: xiii) 4 . My objective is to explore and describe the meaning of a more schematic, abstract meaning of a construction, which, due to its less concrete nature, offers only moderate accessibility to introspection. Yet, while constructional meaning itself might not be as readily accessible, generalizations over concrete instances of the construction (i.e. sentences/ utterances), whose meaning, following Talmy, can be retrieved through introspection, will support the postulation of a more abstract meaning of constructional templates. The collection of such concrete instances, i.e. naturally occurring tokens of COCs, however, falls to a great extend under the realm of corpus linguistics, where huge language corpora can provide the linguistic researcher with a contemporary and authentic cross-section of conventionalized speech phenomena. Complementing introspective investigation with corpus-linguistic procedures has the advantage of making observations and results measurable 5 : Not only can the recorded speech samples illustrate the particular linguistic forms under investigation, but also can careful statistical analyses of these data bring to light subtle patterns of usage, e.g. preferences of a specific construction for particular forms of modification, determination, tense etc., or even reveal the existence of ready-made, lexicalized phrases such as idioms. For my investigation of the particular constructional patterns, I compiled a large collection of instances of COCs by means of an extensive (yet, not exhaustive) search of the British National Corpus (BNC), a 100 million word collection of samples of contemporary texts of written and spoken language from a wide range of sources 6 . The obtained data, more than 3000 examples of COCs, form the complete basis of my description and analysis - be it introspective, theoretical or statistical. Making use of introspective and theoretical description and employing the descriptive apparatus of CL, I will explore semantic and syntactic idiosyncrasies of COCs, which have partly been taken up already by researchers, who, however, were working in different theoretical paradigms. As for the statistical approach, phenomena and patterns of concrete usage are to be identified, described and ex- 4 As Talmy (2007: xiii) explains: "Meaning is a consciousness phenomenon and, if it is to be taken on as a target of research, introspection - itself a process occurring in consciousness is the relevant instrumentality able to reach its venue." 5 The fact that results gained through introspection cannot be operationalized and measured objectively often led and leads to the rejection of this method in scientific research (as was especially the case in Behaviourism). 6 A detailed description of my compilation of the data and all related procedures is provided in Chapter 5. <?page no="17"?> 5 amined critically, especially against the background of findings of previously conducted studies on COCs. 1.3 Overview of the book With respect to the purposes pursued and the methods applied, the study presented in this book basically follows a tripartite structure: Chapters 2 and 3 have a summarizing function: Chapter 2 is to provide an overview of the treatment of COCs in different phases in the history of linguistic description, taking into consideration the domineering linguistic models of the respective stages. Traditional-descriptive approaches, analyses couched into the Generative Grammar paradigm and studies with a Functional Grammar background will be summarized, discussed and evaluated. In Chapter 3 I will briefly delineate basic tenets and common goals of Cognitive Linguistics and introduce those concepts, assumptions and descriptive devices of the paradigm which are to be deployed for the depiction of the construction as an idiosyncratic assembly of form, meaning and function in the remaining chapters of the study. Chapter 4 is basically of a descriptive nature, focusing on selected aspects of the construction. It will investigate the semantics of COCs, describing different types of the construction and the respective conceptual content verbalized by these various forms. Moreover, the question in how far principles of iconicity can be drawn upon as a motivation of the particular pairing of form and meaning will be addressed. Chapters 5-11, being empirically driven, will present the results of a corpus analysis of the actual usage of COCs in contemporary English, based on the British National Corpus (BNC). These chapters exemplify the fruitful relationship between CL and corpus-based approaches to language in that data of natural usage are statistically evaluated and integrated into a network model of constructions. Furthermore, the validity of observations and claims made in foregoing chapters is checked against facts and figures elicited from the corpus data, allowing for the depiction of more concrete and fine-grained types of COCs. All of the aspects investigated in those chapters incorporate observations which are related to and build on concrete samples of speech found in the BNC. At times, examples found on the World Wide Web (itself a 'mega-corpus' of language) will be quoted to complement instances taken from the BNC and to illustrate forms which are not attested in the corpus analysed 7 . 7 As the Czech linguist Karel Oliva noted only recently on the 2 nd international conference on Grammar and Corpora (Liblice, Sept. 25-27, 2007), currently available corpora <?page no="18"?> 6 Chapter 5 presents initial steps towards a usage-based network of constructions, discussing methods of corpus compilation, describing procedures of statistical analysis, and offering a first description and specification of possible abstract (i.e. higher-level) schemas of different types of COCs. Chapter 6 centres on the notion of transitivity as one crucial component of COCs. Here, the constructional types are scrutinized as a means to provide event scenarios with a particular force-dynamic frame. The leading question will be whether COCs can be claimed to profile energetic interactions, which might be assumed on the basis of their form as monotransitive constructions. Concepts of verbal aspect are examined in Chapter 7, where COCs are depicted as "packaging" devices, imposing temporal boundaries on the events which are verbalized via these forms. While the discussion provided in Chapters 5-7 focuses on form and functions of so-called high-level schemas of constructional representation, Chapter 8 moves on to an investigation of mid-level schemas, studying the make-up of the CO-NP in greater detail. Issues of determination and modification are approached from a usage-based perspective. Remaining in the realm of mid-level schemas, Chapters 9 and 10 will offer a comparison of COCs with seemingly related constructions. Firstly, COCs will be contrasted with intransitive verb-adverb constructions (VACs), i.e. the focus will be put on alternations of the type laugh a hearty laugh (COC) vs. laugh heartily (VAC). Selected verbs occurring in both constructions will be examined with respect to the modifying material they are preferably used with, which concerns adjectival modifiers in the CO-NP in the case of COCs and adverbs in the case of VACs. This procedure aims at the identification of subtle differences in meaning between the constructional variants on the one hand, and possibly pre-fabricated chunks of language (i.e. entrenched units) on the other. Secondly, in many cases, contents that are expressed by the use of COCs can also be verbalized by so-called light verb constructions (LVCs), i.e. constructions with verbs that are semantically light or bleached and where the primary sense of the construction is carried by the verb's complement, as in to give a cry or to have a dance. The two constructions are compared with respect to the functions they fulfil and the subtle shades of meaning they possibly convey. Moving down to the lowest levels of representation in a network of constructions, Chapter 11 will then provide an investigation of patterns which allow for little or no flexibility as regards their formal/ lexical makeup: In discussing strong collocations and idiomatic expressions, the stepby-step development of the network of constructions will be comple- "cover only subsets of positives examples" (so stated in his plenary talk "Phenomena- Oriented Corpora: A Manifesto", delivered Sept. 27 th , 2007). <?page no="19"?> 7 mented and completed. Moreover, aspects of register variation in the exploitation of all kinds of schemas are to refine the portrayal of COCs pursued in this comprehensive investigation of constructional form and meaning. Finally, Chapter 12 will summarize the main findings of the research project and comment on future perspectives. <?page no="20"?> 8 2 Cognate Object Constructions: Past and Present 2.1 Introduction The following chapter will present an investigation into how the case of COCs has been treated in different linguistic frameworks of modern linguistics. For the purpose of such a survey I shall examine how the construction is approached in descriptive/ reference grammars of English (2.2), Generative-Transformational Grammar (2.3), and Functional Grammar (2.4) as representatives of the many models of linguistic description that have emerged in the course of (recent) time. Section 2.5 will offer a detailed discussion of a dissertation thesis by Talke Macfarland (1995) 1 , the most comprehensive and extensive study on COCs carried out so far. In order to ease comprehension, a short description of the relevant frameworks will precede the actual discussion of the analysis of and findings on COCs put forward by particular studies. Crucial issues that are necessary for grasping the seemingly 'problematic' status of COs will be considered in the light of the respective linguistic paradigms and related to their handling of COs. These issues embrace concepts of transitivity, syntactic categories such as object and adverbial, the distinction between argument and adjunct, and the nature of thematic roles, all of which are indispensable to capture the complexity of the construction investigated in this study. As will become obvious, a straightforward handling of the status of COs seems to pose a problem for all of the models dealt with here. An evaluation of the plausibility of the descriptions of COCs and the ways how the different models solve to integrate the pattern into the theorems they work with will be part of each description. Some of the points will be taken up again in later chapters, where they will be contrasted with suggestions and solutions offered by models of Cognitive Linguistics. 1 Due to its comprehensiveness, Macfarland's analysis and a discussion thereof will be presented in a separate section. <?page no="21"?> 9 2.2 COCs in Descriptive / Reference Grammars 2.2.1 Henry Sweet (1891) I shall start my survey on the treatment of COCs with Henry Sweet's work A New English Grammar, published in 1891. I have chosen this book as a first reference since it marks the abandonment of purely prescriptive grammar in favour of descriptive approaches to language 2 , of structuralism even 3 . Many of the linguists who have done research on COCs picked as their basis of discussion Otto Jespersen's account of these structures in The Philosophy of Grammar or A Modern English Grammar (Vol. III), published in 1924 and 1927, respectively (see below). Yet, Sweet, who published his work several years earlier, already dedicates a passage to the description of COCs: Sometimes an intransitive verb is followed by a noun in the common form which repeats the meaning of the verb, as in sleep the sleep of the just, fight a good fight, where the noun is simply the verb converted into a noun, and in fight a battle, run a race, where the noun repeats the meaning but not the form, of the verb. Such object nouns are called cognate objects. A cognate object must essentially be an abstract noun. (Sweet 1891: 91; emphasis added, SH) Sweet's definition already includes most of the key notions that are hotly debated in more recent studies on COCs and which will occur over and over again in the present discussion: Firstly, as regards the types of verbs occurring in the construction, he confines them to intransitives. Transitivity in Sweet's understanding refers to complementation patterns of verbs transitive verbs being those that require a "noun word or noun-equivalent in the direct object relation to serve as a complement" (ibid.: 89), while intransitives occur without such a complement. Unfortunately, he explains the notion of 'direct object' in terms of "complet[ing] the meaning of a transitive verb" (ibid.: 50) which leaves the reader with a circular definition of the concepts of 'direct object' and 'transitive verb'. Sweet also addresses those intransitive verbs that are followed by a noun phrase, as in to run a mile. Observing a similarity between such NPs and adverbs, he gives the label 'adverbial objects' to these types of constituents (ibid.: 91). Although Sweet does not make it explicit, one might assume that COs fall into this 2 As Sweet claims in the foreword of his work: "As my exposition claims to be scientific, I confine myself to the statement and explanation of facts, without attempting to settle the relative correctness of divergent usages" (ibid. 1891: xi). 3 Aarts (1986: 365) praises Sweet for "inaugurating a new era in the history of English linguistics" due to his introduction of several new concepts for the analysis of language, among them the distinction between scientific grammar and practical grammar and that between formal syntax (grammatical forms) and logical syntax (grammatical categories). <?page no="22"?> 10 category, if one takes into consideration his defining criterion of intransitivity for the verbs that can be combined with cognate nouns. The label 'adverbial object' obviously merges two different sentence functions into one and reflects the difficulties that arise with the attempt to categorize COs as either direct object or adverbial, which is a central field of discussion in the CO literature, as will become obvious in the further course of this survey. The second noticeable feature Sweet extracts for his description of COs is the strong semantic relationship between the verb and the CO, with the meaning of the latter 'merely' being a repetition of the former. Thirdly, as regards aspects of form, both nouns which are morphologically derived from or related to the verb, and nouns that show no morphological relation to the verb can be considered COs, as long as there are 'semantic bonds' between verb and noun 4 . This makes his description a little loose since it lacks a characterization of the strength of these semantic ties. Following from Sweet's considerations, all phrases made up of a combination of a verb and a noun that is synonymous or hyponymous to the 'original' cognate object would fulfil the criteria listed by the author. To sleep a slumber or to fight a combat could then easily be added to the examples he offers. Also, it remains open whether expressions like to run a marathon, in which marathon denotes a subcategory of run or race and clearly repeats aspects of the meaning of the verb, would then need to be discussed as COCs. A fourth criterion listed by Sweet, relating to the abstract meaning of a CO, is debateable. It might be true that the majority of COs denote rather abstract, intangible entities, yet, examples like to sow a seed or to weave a web (which are cited as instances of COCs in other studies) run counter to Sweet's proposal. Considering that Sweet treats COCs in only a few lines, it is remarkable that he already touches upon most of the features that will play a major role in studies on this construction of later periods in 20 th century and present day linguistics. 2.2.2 Otto Jespersen (1927) Otto Jespersen's A Modern English Grammar, being another milestone in descriptivist approaches to the structure of English, will be consulted for its explanation of COs in the following. This seven-volume work, published between 1909 and 1949 5 , is written in the same tradition as Henry Sweet's 4 Such considerations are reflected in the distinction between morphological cognates and semantic cognates. This distinction is applied to the description of the kind of verbs which enter COCs by Visser (1963) and Baron (1971), who allow morphological cognates for transitive as well as for intransitive verbs but restrict transitive verbs to occur with semantic cognates only. 5 Part III, which contains the section on COs, was published in 1927. <?page no="23"?> 11 grammar in that, besides being descriptive rather than prescriptive, it focuses on spoken language (cf. Aarts 1986: 367). Jespersen elaborates on the nature of COs in more detail than Sweet does and introduces some interesting observations: First of all, he mentions a frequently noticed parallelism between a class labelled 'object of result' 6 and that of cognate objects. He, however, rejects a self-contained category of COs and describes it as a "subdivision under the object of result", for which he gives no distinct criteria of definition 7 except for listing some characteristic verbs that take an object of result (such as produce, pronounce, construct etc. [1954(1927): 232]). He justifies this 'fusion' of classes by equating the NP following the verb in the king lived the life of an exile with "something what resulted from his (manner of) living" (ibid.: 234) and thus with an object of result. With respect to the general notion of object, Jespersen describes as crucial characteristics its intimate connection with the verb of a sentence and its identification "[…] by asking Whom, or What with the subject and the verb of the sentence" (ibid.: 229), in respect to which COs as a subcategory of objects of result would doubtlessly qualify as objects. As for COs as members of the special class 'object of result', Jespersen does not - in contrast to Sweet restrict their occurrence to combinations with intransitive verbs, but extends their usage to "verbs that are otherwise transitive"( ibid.: 234). Jespersen also works with the concept of transitivity in the sense of patterns of verb complementation, but not without pointing out the difficulties of a strict division between classes of intransitive and transitive verbs: It is customary to divide verbs into transitive and intransitive. But in English at any rate, it is impossible to make a sharp distinction between two classes, and we should rather speak of a transitive and an intransitive use of verbs, for many verbs which are generally transitive, i.e. take an object (or two objects), are very often used without any object, and other verbs, which are as a rule intransitive, may at times be connected with an object. (ibid.: 319f.) It is surely the last remark of the quote which refers to and is to account for cases like to sigh a sigh or to die a death. Jespersen completes his description of COs with some remarks on the complementation of the CO-NP, doubting the occurrence of bare, i.e. unmodified COs (as in the previous examples) in natural speech, since such bare objects do not add any new information to the verbal notion. It is ex- 6 The underlined NPs in the following examples, found in Jespersen (1954[1927]: 232), belong to the class of object of results: the dressmaker made the dress, the girl paints flowers, the child wrote a letter. 7 He does, however, emphasize the special nature of objects of results: "All objects other than those of result we shall here term ordinary objects" (Jespersen 1954[1927]: 230, italics in original). <?page no="24"?> 12 actly the expression of new information that Jespersen considers as a speaker's motivation for the use of COCs: In real speech this kind of object is used for one purpose only, namely to add a descriptive or qualifying trait, which could not be conveniently joined to the verb in the usual way as a subjunct 8 . (ibid.: 235) It seems then that Jespersen reduces the use of a COC to a means to make up for a lexical gap a particular language might have in not offering an appropriate adverb to describe an action expressed by the verb 9 . He compares to fight the good fight with to fight well and concludes that the former is surely "not the same thing" as the latter (ibid.). The above quote indicates that Jespersen, like Sweet and like many others to follow, recognizes parallels between COs and adverbials. Moroever, he includes into his discussion of COCs sentences such as (1) a. Then he smiled, a shy nervous smile. b. Kitty laughed - a laugh musical but malicious. (cf. ibid.: 236), in which the respective NPs are extracted out of the VP ( or clause, even) and rather constitute something like an apposition to the verb. If the term 'object' is defined as an entity that fulfils the function of complementing the verb, then cases like those in (1) cannot be labelled 'objects'. What becomes obvious in Jespersen's treatment of COs once more is the difficulty to define absolute and definite boundaries between the functional categories 'object' and 'adverbial' and - even more so to use these for the definition of COs. The author merges the two functions for his delineation of COs although he unmistakably assigns them to the class of objects at the beginning of his discussion. What is striking about Jespersen's analysis of COs is the absence of any remarks on semantic or morphological relationships between verb and CO. He neither pins down criteria that define a repetition of aspects of meaning of the verb by the CO phrase nor criteria that allow for an identification of COs on the basis of formal aspects (i.e. issues of derivation). 2.2.3 Quirk et al. (1985) It might appear as a big jump from Jespersen's A Modern English Grammar to Quirk et al.'s (1985) A Comprehensive Grammar of the English language (CGEL); however, the choice of the latter as a third source of insight into how COs are treated by Traditional/ Reference Grammars can be easily 8 i.e. adverbial-like phrases 9 This is seen as the primary function of COs by Dixon (1991: 118), who claims that "English grammar has much more restricted possibilities for adverbial modification of verbs than for adjectival modification of nouns - hence the usefulness of cognate NPs". <?page no="25"?> 13 explained: Other works that are listed as important representatives of that approach (see Aarts 1986, 1991; Robins 1986; Leitner 1986) either do not acknowledge the issue at all or draw on the analyses suggested by Sweet and Jespersen (e.g. Zandvoort 1945). Moreover, the publication of Chomsky's Syntactic Structures (1957) and Aspects of the Theory of Syntax (1965) led to a shift of emphasis from the description of linguistic data to an exploration of the relation between these data and a person's linguistic knowledge, so that only few grammars following the traditional style were written during this period 10 . The discussion of COs by Quirk et al. can be found in a subchapter titled "Semantic roles of clause elements", where they are set up as a semantic class within a variety of roles that can be fulfilled by object phrases. The reader finds, besides 'cognate object', the following semantic roles of direct objects: 'affected participant', 'locative object', 'resultant object', 'eventive object', and 'instrumental object' (Quirk et al. 1985: 741, 749ff.). Whereas the names of these roles actually reveal something about the semantic contribution the objects make to the verb phrase, 'cognate' does not tell much about such a contribution it merely hints at an etymological relation between object and verb. Quirk et al.'s description of this particular structure is done in less detail than in Jespersen's account and runs as follows: A cognate object is similar to a resultant object in that it refers to an event indicated by the verb. […] In this type of object, the noun head is semantically and often morphologically related to the verb. The object can therefore not be considered a participant. Its semantic function is to repeat, wholly or partially, the meaning of the verb. Most cognate objects tend to convey a rather orotund style. The noun is generally modified. The verb and the object are then equivalent to the verb and a corresponding adverbial. (Quirk et al. 1985: 750) This definition mostly combines criteria that had already been laid down by Sweet and Jespersen: Semantic and morphological relations between verb and object are spelled out (cf. Sweet) and aspects of the make-up of the NP described (cf. Jespersen). Unlike Jespersen, Quirk et al. figure COs as a category on their own and not as a subcategory of objects of results. The difference between the former and the latter, however, remains unexplained. 10 As Aarts (1986: 376) nicely summarizes: "All remained quiet on the traditional grammar front" until 1972, which saw the publication of Quirk et al's A Grammar of Contemporary English, the predecessor of the work to be discussed in the following sections. CGEL is a "modern" traditional grammar in that it has benefited from the work of structuralist and generative-transformational grammarians, integrating their findings into the traditional framework without applying the formalizations and technical apparatus applied in the more recent works (cf. Algeo 1991: 133). <?page no="26"?> 14 Nonetheless, the definition also brings into play some new aspects as regards the characteristics of COs: First of all, Quirk et al. identify them as a form of stylistic devices. Orotund, meaning 'inflated', 'bombastic', 'pompous', suggests that COs are to be found in more elaborate pieces of writing or public speeches rather than in informal conversation. Secondly, Quirk et al. exclude COs from a class of sentence elements referred to as 'participants', a class that the authors understand as "entities realized by noun phrases, whether such entities are concrete or abstract" (ibid.: 740). Although they explicitly list agentive, affective and recipient roles as belonging to the class of participants, the reader is neither given any semantic description thereof nor an indication of why COs are categorized differently 11 . Parallels between combinations of verb + CO and adverbials were observed by Jespersen and were implicit in Sweet's analysis; Quirk et al. continue this 'tradition' when they remark on the equal status of COs and adverbials, illustrating their considerations with the following example: (2) a. They fought a clean fight. b. They fought cleanly. (ibid.: 750) This is similar to the example found in Jespersen (see p.12), however, he offered the alternating sentences to make clear a difference in meaning, while Quirk et al. emphasize correspondences. Interestingly enough, no observations about the verb's (in)transitive status are included. Quirk et al. make use of the intransitive-transitive distinction, with transitivity defined as the verb's complementation by one or two objects. (cf. ibid.: 54ff.). A verb can be transferred from one category to the other, a process the authors refer to as a particular form of 'conversion', which leads to a 'change of secondary word class' (ibid.: 1565). Instances of shifts from 'intransitive' to 'transitive', from 'transitive' to 'intransitive', or from 'intransitive' to 'copular' are listed. Although the authors consider COs as direct objects and list expressions of COCs that contain verbs 12 which they explicitly put into the category of intransitive verbs 13 (die [ibid.: 1169], live [ibid.: 746; 1175]), they do not comment on them as instantiations of 'partial conversion'. Drawing on the examples Quirk et al. use for the illustration of such a type of conversion, one is tempted to regard COCs as cases thereof. 11 It needs to be noticed, however, that Quirk et al. point at their description of participants as being "tentative" (ibid.: 741). 12 The following verbs are contained in the example sentences: sing, fight, die, live, breathe (ibid.: 750). 13 Quirk et al. note that one should rather speak of a verb's in-, monoor ditransitive use or complementation (ibid.: 1168). <?page no="27"?> 15 2.2.4 Huddleston & Pullum (2002) Huddleston and Pullum's Cambridge Grammar of the English Language (2002) will serve as the last source of information about the handling of COs in the framework of Reference Grammars. Similar to Quirk et al.'s work, this grammar might be called a "modern traditional" one, for it takes account of the progress that has been made in the description of English grammar in more recent theories 14 . The way Huddleston & Pullum treat COs deviates in some respect from the accounts described so far: A cognate object is one where the head noun is a nominalization of the verb […]. In some cases the selection of a cognate object is of no syntactic significance: They built a hideous building and I can smell an appalling smell belong to the same construction as They built a mansion and I can smell rotting meat. Sing is arguably basically intransitive, but it allows many objects beside the cognate song […]. But there are also verbs where the cognate object is not freely replaceable by a noncognate one. […]. The semantic role might again be said to be factitive. (Huddleston & Pullum 2002: 305) The authors take up transitivity as an important concept to explain the class of COs, yet, allow verbs from both the transitive and the intransitive category to enter a combination with this particular object phrase 15 . Apparently, as I take it, they implicitly suggest a tripartite division of this construction, depending on the verb occurring therein: 1. Transitive verb + cognate object / other direct object from a wide semantic range (build a building, smell a smell [ibid.]) 2. Basically intransitive verb + cognate object/ other direct object from a more restricted range (run, walk, sing [ibid.]) 3. Intransitive verb+ cognate object only (e.g. cough, grin, laugh, snore, die, dream [ibid.]) The semantic function that is assigned to COs by Huddleston & Pullum reflects Jespersen's suggestions to treat them as 'object of result'. 'Factitive' as employed in their terminology denotes particular kinds of themes that "come into existence by virtue of the process expressed" (ibid.: 233). 14 In their preface, the authors express their pretension to "bridge the large gap between traditional grammar and the partial description of English grammar proposed by those working in the field of linguistics" (Huddleston & Pullum 2002: xv). Since they do not explicitly ascribe their accounts to any theoretical framework, I chose to include them in this section. 15 Similar to Quirk et al. (1985), Huddleston & Pullum (2002) apply the concept of 'transitivity' to the use of verbs (ibid.: 216). <?page no="28"?> 16 The authors complete their description of COCs with clear statements about the modification of the NP, which they define as "obligatory" 16 (ibid.: 305), and the function COs fulfil: As the head noun itself is readily implied by the verb it does not normally occur on its own […]. Rather, the noun is modified in some way, as by the adjectives in [the] examples. And these adjectives typically describe the process expressed in the clause and thus have the same kind of function as a manner adverb. (ibid.: 673; emphasis added, SH) As in the accounts discussed before, parallels between COs and adverbs of manner are observed, which to some degree defines the contribution COs make to the clause, yet, Huddleston & Pullum do not point at the stylistic value of or differences in meaning between alternating sentence pairs. The examples (3) a. She fought a heroic fight. b. She fought heroically. are commented on as meaning "essentially the same" (ibid.). For (4) He died a long and agonising death the inclusion of a CO phrase is explained as a means to make up for the absence of an adverb *longly to express the same meaning. This parallels Jespersen's proposal of COs as a solution for lexical gaps in a language. 2.2.5 Summary: Descriptive Grammar accounts of COCs Although there lie some 110 years between the publication of Sweet's A New English Grammar (1891) and Huddleston & Pullum's Cambridge Grammar of the English Language (2002), the major issues in giving a logical presentation of COs in conformity with traditional syntactic and semantic categories have not changed essentially. Basic questions such as morphological form, syntactic and semantic function within the sentence and the nature of verbs that can enter such a construction are taken up by all grammars discussed in the preceding section, although the different works emphasize different criteria. Despite these differences one particular line of thought is present with all authors: They explicitly give the syntactic label "object" to COs, at the same time observing a similarity of CO phrases with adverbs of manner. Readers might have noted that the preceding survey of the treatment of COCs in traditional / reference grammars of English does not contain a reference to Biber et al.'s Longman Grammar of Spoken and Written English 16 In an earlier work, Huddleston (1984: 193) notes: "Having a cognate noun […] enables us to add an adjectival modifier - without such a modifier, the object would be pointless and the sentence hardly acceptable" (emphasis added, SH). <?page no="29"?> 17 (1999), which is based on a 40 million word corpus of text representing different registers and thus comes closest to the approach pursued in the present account, i.e. a usage-based approach to language. The reason for its absence in my overview is quite simple: Biber et al. did not include a discussion of COCs in their grammar, a fact which fairly clearly underlines the need for a description of the construction based on real language usage. 2.3 Generative Grammar approaches to COCs While the works described in the preceding sections aim at a description of pure "surface phenomena" of a particular language (i.e. English), Generative Grammar is a highly formalized approach to language and concerned with the theory of a type of grammar that is claimed to be an innate component of the human mind / brain. Noam Chomsky, the father of the generative grammar model of language, sees it as the main task of a linguistic theory to develop an account of linguistic universals that, on the one hand, will not be falsified by the actual diversity of languages and on the other hand, will be sufficiently rich and explicit to account for the rapidity and uniformity of language learning, and the remarkable complexity and range of the generative grammars that are the product of language learning. (Chomsky 1965: 27-28) In order to accomplish this vision of language, generative linguists have postulated sets of general principles of grammar which are assumed to be part of the innate knowledge a language learner employs when faced with the enormous task of language acquisition. These sets of principles are formulated in terms of fundamental notions of grammatical analysis, e.g. case (focusing on the distribution of phonetically realized NPs), argument structure (mediating the distribution of arguments with respect to predicates) or binding (capturing the distribution of bound elements such as anaphors, pronouns etc.) (cf. Freidin 1996: 119ff.). 2.3.1 Generative Grammar / Government and Binding: Chomsky (1965, 1981) Regarding an adequate account of COCs in Generative Grammar (GG), the principles and concepts which are set up by Chomsky (1965, 1981) 17 to capture the syntactic and semantic relations between a verb and the NPs complementing it are of interest here, and these are subcategorization features of a lexeme, -theory and case theory. 17 I will not include Chomsky's most recent model, proposed and developed in his book The Minimalist Program (1995), since the studies discussed in the following section primarily draw on Chomsky's Lectures on Government and Binding (1981). <?page no="30"?> 18 A verb's subcategorization frame basically spells out its syntactic environment, more specifically, the range of complements the verb permits. This information is contained in the lexicon 18 , a set of lexical entries listing all the properties of an item that are idiosyncratic to it, such as morphophonological structure and syntactic features (cf. Chomsky, 1965: 94, 1981: 5) As regards complementation, the entry of the verb love, for example, will be specified with the feature 'V + [ __ NP], indicating that the verb immediately precedes an NP within the same VP (a representation which corresponds to the notion 'transitive'), while the verb die, on the other hand, will be specified with 'V + [__ #], representing that die does not permit a complement within the same VP (corresponding to 'intransitive'). -theory relates to the assignment of thematic roles to arguments by the lexical head of a phrase. With respect to verbs, this can be explained in the following way: As the lexical head of a VP, a transitive verb not only 'calls for' complements satisfying its subcategorization features, but also ascribes to them a thematic status (i.e. its semantic function). Chomsky assumes a role for each NP and clause, a claim which he formulates in his " criterion", serving as a formal device which is to implement the one-to-one match between argument and -role: Each argument bears one and only one -role, and each -role is assigned to one and only one argument. (Chomsky 1981: 36) Both of these principles (i.e. subcategorization and -criterion) need to be satisfied for COs if they are considered as arguments (i.e. direct objects) of the verb. Thus, for the phrase to die a gruesome death, it must be explained how an NP which obviously appears in the direct object position can follow a verb that has the subcategorization frame 'V+ [__ #] ' and therefore can actually neither take an NP nor assign a thematic role to this NP. With respect to the assignment of -roles, Chomsky offers a 'loop-hole' for NPs that are part of idiomatic phrases. In take advantage of Bill, for example, the verb selects a particular idiom chunk (i.e. advantage) which is not assigned a -role. For cases in which a so-called non-argument is subcategorized by the head, Chomsky invents the new -role "#" to ensure that each subcategorized position is a position (ibid.: 37). One might argue then that, since several of the COCs have an idiom-like status, gruesome death in the above example can be assigned the "#"- -role. However, the requirement that -marking presupposes subcategorization (ibid.: 38) excludes this solution of the problem because, as stated above, die - in con- 18 Chomsky sets up his grammar as a rule system with several subcomponents: (i) lexicon, (ii) syntax: (a) categorical component, (b) transformational component; (iii) Phonological Form component, (iv) Logical Form component. The lexicon and the categorical component of syntax constitute the so-called base which generates deep structures (Chomsky 1981: 5). <?page no="31"?> 19 trast to take has a subcategorization frame that does not offer a second position (besides the one for the "victim" of the process), unless die is 'granted' a second entry in the lexicon, spelling out additional syntactic and semantic features. In order to handle further problematic cases, Chomsky introduces socalled quasi-arguments (1981: 325), contrasting them with true arguments that have "potentially referential function". Quasi-arguments, on the other hand, "lack any such function as a matter of grammatical principle" 19 . Adger (1994), for example, transfers Chomsky's notion of quasi-arguments to measurement phrases, a class of phrases which COs are often described to belong to (see Tenny 1994). Considering CO phrases as instances of quasi-arguments at least solves the problem of their thematic status, since quasi-arguments are not assigned a -role. A third principle that needs to be considered to illustrate the problematic nature of COs in GG frameworks is Chomsky's Case Theory, a basic theory according to which certain lexical categories can assign case to mark syntactic or thematic relations between certain constituents. The theory states that "every noun with a phonetic matrix must have Case", which is assigned to NPs "by virtue of the configurations in which they appear" (Chomsky 1981: 49). Only arguments that have been assigned case can be assigned a -role. In English, Chomsky claims, the lexical categories 'verb' and 'preposition' are case-assigners, with transitive verbs assigning objective case to the NPs they govern. With respect to COs it is an open question whether they appear with case-marking at all 20 and if so, which element is responsible for the assignment of case, for intransitive verbs do not assign case. For several linguists working with the Generative paradigm, COCs have been of some interest in order to test for the general principles addressed above (i.e. case assignment or the distribution of theta-roles). Two 19 Compare the following sentences in which it functions as a true argument in the first version, but as a quasi-argument in the second: a) It is on the table. vs. b) It is raining. 20 Examples from German, having a richer inflectional system than English, give evidence that COs are marked for case in some languages. In German, COs appear with the accusative case: a) Er starb einen qualvollen Tod 'He died a-ACC painful-ACC death'. b) Sie lief einen phänomenalen Lauf 'She ran a-ACC phenomenal-ACC run.' Note, however that in the German equivalent of die a … death, the NP also frequently occurs with genitive case, primarily expressing the cause of dying (much more than the manner of dying): c) Er starb eines natürlichen Todes. 'He died a-GEN natural-GEN death-GEN'. <?page no="32"?> 20 of them will be discussed in more detail here. Jones (1988), opting for a treatment of COs as adjuncts, draws on the characteristics of COs in order to demonstrate the need to revise Chomsky's Case theory, whereas Massam (1990) advances considerations in favour of COs as direct objects, sanctioned by the assignment of a thematic role. 2.3.2 Cognate Objects as adjuncts: Jones (1988) The main question Jones raises in his study is whether to treat COs as arguments or adjuncts of the verbs they occur with, basing his line of argument on Chomsky's concepts of Theta Theory and Case Filter (1981). Jones' concern lies with constructions "in which a normally intransitive verb occurs with what appears to be a direct-object NP whose head noun is the event or state nominalization 21 of the verb" (1988: 89). With respect to the classes of verbs that appear with COs, Jones distinguishes between two types of CO constructions: Some expressions contain non-ergative verbs (dance a dance, dream a dream) and can undergo passivization (A merry dance was danced by Sam.) a characteristic which leads Jones to assign to them the status of "genuine[ly] transitive" (ibid.: 91) 22,23 . Others involve verbs that are ergative, i.e. the NP appearing as the subject in S-structure is the object of the VP in D-structure (cf. ibid.: 90), such as the NP the old lady in The old lady eventually died 24 . In his discussion and his considerations, Jones exclusively focuses on the latter type, exemplified by sentences such as (5) a. John died a gruesome death. b. Bill sighed a weary sigh. Trying to find evidence which supports the argument status of COs, Jones at first suggests a marginal transitive use of ergative verbs such as die, live, sigh, assuming a process "which allows some intrinsic semantic component of an intransitive verb to be reanalysed as a selection restriction on a direct object NP" (ibid.). According to this approach, COs should behave like 21 Unfortunately, Jones does not illustrate the distinction between event and state nominalization in COCs with examples. This would have been of interest especially since Massam (1990, see below) extracts "event nominal" as a distinctive feature for the identification of COs. 22 As will be discussed at length in later sections, it is commonly assumed that the ability of an expression to occur in a passive construction presupposes the transitive status of the expression and thereby the object status of the NP which is to appear in the subject position of the passive form. 23 Having labelled them as transitive, Jones does not take up again CO expressions which contain unergative verbs in the further course of his analysis, thereby implying that he excludes them from the class of COCs. 24 Note that Jones makes use of the term 'ergative' (=unaccusative) in a seemingly broad sense, classifying live a life as belonging to this category. For a more detailed discussion of the unergative-unaccusative distinction, see Ch. 6.5. <?page no="33"?> 21 normal direct objects, being 'equipped' with objective case and -role. Direct objects allow for passivization (see ftn. 22), i.e. the NP can be moved to the subject position as a consequence of the fact that the -role and objective case are absorbed by the passive participle in the passive construction (ibid.: 91). However, the assumed ungrammaticality of the sentences (6) a. * A gruesome death was died by John b. * A weary sigh was sighed by Bill. leads Jones to the conclusion that either the requirements which case theory imposes on the movement of NPs differ for some types of objects or that the interpretation of COs as arguments needs to be abandoned in favour of their classification as adjunct predicates. He opts for the second alternative, observing that the difference between the sentences (7) a. John died gruesomely. b. Bill sighed wearily. , which contain manner adverbs, and those in (5) "is more a matter of style than of meaning" (ibid.: 93), and hence suggests to treat COs "as modifiers of the VP, on a par with […] manner adverbs" (ibid.) 25 . Following Zubizaretta (1987), who proposes a class of adjunct-predicates (including manner adverbs) which are not assigned a -role by the head element but rather function as predicates modifying the constituent in which they occur, Jones considers COs as such adjunct predicates and thus can at least solve the problem of the -criterion. Yet, things are more difficult for the assignment of case, since Chomsky's case theory explicitly requires that "every noun" be marked for case (see above). Of the three possible solutions Jones discusses, he considers the modification of Chomsky's Case filter as the most plausible one. The theory is altered in such a way that only NPs that are assigned a -role can also carry case (in Chomsky's original version, casemarking precedes the assignment of -roles). Thus, since COs are claimed to not carry -roles, case cannot be assigned to them. Jones' revised version 25 Although Jones does not make it explicit, his approach parallels to some extent the model of "semantically internal" and "semantically external" modification, as discussed by Ernst (1981) with respect to adjectival modification of idioms (see also Langlotz 2007). While internal modification truly aims at the specification of a nominal head, with external modification the premodifier has the semantic status of an external adjective with an adverbial interpretation, as exemplified in the following sentence pair, which is claimed to be synonymous and where quick in (a) functions as an external modifier: a) The lady carried out a quick attack on the tramp b) The lady quickly carried out an attack on the tramp. (quoted in Langlotz 2007: 32). Jones' suggestion to treat modified COs as adverbials thus renders the modifying constituent similar to external modifiers. <?page no="34"?> 22 of Case theory can now account for the fact that COs cannot surface in the subject position of a passivized sentence: The subject position is casemarked with structural case (nominative), and as such is incompatible with an element that has the features [- -role, -Case]. This combination of features is only ascribed to elements that are non-arguments, a fact which lends evidence to Jones' interpretation of COs as non-arguments. The weak point in Jones' line of argument, however, is his starting out from the claim that COCs (in his more restricted sense, i.e. those forms containing ergative verbs) cannot be passivized, a fact which can easily be disproved by an analysis of language data as found, for example, in the BNC. It is an interesting fact that an earlier attempt by Burzio (1986) to 'remedy' Chomsky's Case theory, known as Burzio's Generalisation, obviously predicts the opposite analysis for COCs: All and only the verbs that can assign -roles to the subject can assign accusative Case to an object. (Burzio 1986: 165, quoted in Kuno & Takami, 2004: 108) This generalization, which permits unergative verbs to assign accusative case to an object which is not subcategorized for (due to their 'abiltiy' to assign -roles to their subjects) was taken as the basis of Levin & Rappaport Hovav's (1995) account of COs as direct objects of their respective verbs. Jones' analysis, favouring the discussion of COs as adjuncts, finds support in approaches by Zubizaretta (1987) and Moltmann (1990) 26 . His claim that COCs are merely stylistic variants of simple verb expressions which include an adverbial that corresponds to a (possible) adjective modifier preceding the CO reminds one very much of the views of those proponents of transformational approaches who equate meaning with the fulfilment of truth conditions: In accordance with such an understanding of meaning, generative-transformational grammarians postulated relations of synonymy between formally different expressions if they had the same truthconditional values. Such truth-conditional synonymy, in turn, served as a basis for the postulation of a shared underlying D-structure, permitting the identification and description of transformational relationships between the different surface forms. In view of that, the COC variant and the simple verb variant might be linked by a transformational rule which derives the former from the latter. In fact, Jones hints at such a derivational relation between COCs and their verb +adverbial counterparts, stating that in COCs […]the head noun acts as a surrogate for the verb, with the result that the modification relation assigned to the NP by the adjective or other modifiers is trans- 26 Zubizaretta (1987: 10) claims that the postverbal element is an adverbial which modifies an incorporated constant, whereas Moltmann (1990: 10) argues that COs are optional predicates over the event argument of the verb. <?page no="35"?> 23 ferred to the VP at the level of LF (or semantic representation). (ibid.: 93, emphasis added, SH) Although one cannot deny that in some cases there might exist a quasiparaphrase relation between simple verb + adverbial structures and COCs, this is not necessarily the case for every pair, as had already been noticed by Jespersen (cf. p. 12): (8) They fought the good fight They fought well. Horita, quoting Iwakura (1976), points at the below sentence pair which - reportedly - cannot be judged as equivalent: (9) a. Mary dreamed a strange dream. b. *Mary dreamed strangely. (Horita 1996: 224) Moreover, what Jones' suggestion cannot explain are the attested cases of COCs in which the object phrase occurs unmodified (10a), or COs that have modifiers that obviously lack an adverbial counterpart (10b): (10) a. Happily, at the time of going to print the craze is dying a death. (AS3: 1339) b. In the dark, I dreamed the unpopulated dreams of the amnesiac. (GVL: 711) Clearly, it must be asked which modification relation can be transferred from NP to VP in the case of absent modifiers, as in (10a), or in the case of non-existing corresponding adverbs, as in (10b). To sum up, Jones' analysis, which remedies Chomsky's Case Filter to make it compatible with COCs involving ergative verbs (in Jones' understanding), does not offer clear solutions that can account for the wide variety of COs found in natural language usage. 2.3.3 Cognate Objects as arguments: Massam (1990) Reacting to those studies that supposedly present evidence for the status of COs as adjuncts, Massam raises a discussion full of counterarguments 27 . She presents facts which speak in favour of COs as direct objects (i.e. arguments) that receive a patient -role from the verb. In the course of her argument she tries to answer a whole complex of questions concerning 27 Massam's (1990) and Macfarland's (1995; see section 2.5) approaches are based on and to some extent integrate claims of those scholars among the generative grammarians who operate with so-called lexicosemantic rules (e.g. Levin & Rappaport 1988, Pinker 1989). They basically claim that a verb's subcategorization frame can be predicted from the semantics of the verb, formulating lexical rules which are intended to operate on the semantic structure of a lexeme. Different verb senses (output) are arrived at through the application of generative lexical rules to a verb stem (input). <?page no="36"?> 24 characteristics of COs that have been observed in other studies and will be discussed in the following. Massam, like Jones, distinguishes between two different types of COCs: so-called transitivizing object (TO) constructions, in which otherwise intransitive verbs occur with a limited range of objects (Tosca sang a song/ an aria), and true COCs. Objects of the former type behave like "regular" objects in allowing for a variety of structures in which they can appear, and for different kinds of modifications, which is not the case for the latter type. These structures are the 'target' of test procedures, which are applied as a method to distinguish between true objects and cognate objects, and have been used throughout the literature on (cognate) objects. For reading convenience they are summarized in Table 2.1. : Features True / Transitivizing Objects True Cognate Objects Passivization YES The Irish jig was danced by Bernadette Dooley. NO *A silly smile was smiled by Ethel. Topicalization YES The Irish jig, nobody danced. NO * A silly smile, nobody smiled. Free Pronominalization YES I sang the aria, then Tosca sang it. NO *Maggie smiled a silly smile, then her brother smiled it, too. Definite Object Possible YES Fred danced the slow number. NO ? He lived the quiet life. Can be questioned YES What did Tosca sing? NO *What did he die? Restrictions on Form of NP NO She sang a song/ a beautiful song/ the song you like. YES ? He died a death. He died a gruesome death. Object Necessarily Cognate NO Bernadette danced the Irish jig. YES *? He smiled a silly grin Table 2.1 Tests for the distinction between "regular" objects and true cognate objects (cf. Massam 1990: 163.f) The characteristics displayed by COs seem to support the view that true COs fall out of the category of direct objects. However, as Massam argues, COs also have properties which are incompatible with general features of <?page no="37"?> 25 adverbial modifiers. She notes that, firstly, they cannot co-occur with a direct object unlike normal adverbials, and secondly, that they are casemarked in languages with overt case-marking and as a consequence of being assigned case need to be adjacent to the verb 28 . On the basis of these observations she concludes that COs are structural objects (ibid.: 166). Attempting to explain the transitive status of 'formerly' intransitive verbs, Massam considers the procedure of lexical subordination, advanced as a process that extends the meaning of a verb: It operates on the Lexical Conceptual Structure (LCS) 29 of a verb, adding a level of meaning so that the new verb has an argument which undergoes a change of state/ location/ existence (Levin & Rappaport 1988: 281f.) This process is illustrated in the following clauses, with the relationship between versions (11a) and (b) expressed by the presentations in square brackets: (11) a. Tosca sang. [x verb] b. Tosca sang an aria. [x CAUSE [y BECOME EXPRESSED ]] BY [x VERB] (Massam 1990: 171) The new LCS is mapped to the predicate argument structure and realized syntactically in that y is mapped to an internal argument position, yielding: [x VERB y] (cf. ibid.: 171). In Massam's view, lexical subordination can account for both TO and CO constructions, which makes her set up the following generalizations: 1. Lexcial subordination of unergative verbs to CAUSE- EXIST/ EXPRESS predicates creates TO constructions. (ibid.: 172) 2. Lexical subordination of unergative verbs to CAUSE/ EXIST predi- 28 These restrictions reflect the assumed ungrammaticality of the following sentences: a) *He fought the enemy a heroic fight. (V+ DO+ DO) (but: He fought the enemy heroically). b) *? He lives peacefully a full life. (CO-NP not adjacent to the verb) 29 As introduced by Levin & Rappaport (1988), LCS describes the lexical semantic representation of a lexeme in contrast to 'Predicate-Argument-Structure', which captures its lexical syntactic representation. In their paper they show that lexical subordination is at work for several other types of 'transformations', as demonstrated by the following examples: a) Dennis hammered the metal. Dennis hammered the metal flat. (causative construction) b) Carla squeezed the juice. Carla squeezed the juice out. (verb-particle constr.) c) Pauline smiled. Pauline smiled her thanks. (gesture-expression constr.) d) Jack moaned. Jack moaned his way out the door. ("one's-way" construction) (Levin & Rappaport 1988: 275ff.) <?page no="38"?> 26 cates where y = subordinated event creates CO constructions. (ibid.: 173, emphasis added, SH). Accordingly, the CO can be interpreted as a patient which comes into existence through an action undertaken by the agent (or, as Massam formulates it: "The patient is the event itself" [ibid.]). A similar approach is suggested by Larson (1988), who argues that unergative verbs are allowed to be "'thematically reconstrued' as transitives"(ibid.: 386). This process, which Larson labels "Cognate Object Formation", results in the integration of COs into the verb phrase, modelled by the author as shown in Fig.2.1. Fig.2.1 Unergatives: Cognate Object Formation (Larson 1988: 387) According to Larson, the unergative verb die in (a), which is considered as thematically intransitive, is thematically reconstructed as a transitive verb (b), occurring with a CO which is treated as a genuine direct object. Unlike Massam, Larson does, however, not make explicit which -role is assigned to the cognate NP. Apart from validating the status of COs as direct objects, Massam's findings have further consequences for the description of COCs. First of all, her definition of COs as denoting the event itself undoubtedly departs from traditional approaches of labelling COs as objects of result. Secondly, she restricts COCs to unergative verbs that can be represented by CAUSE/ EXIST predicates. Concerning their thematic role, Massam argues for an "affected patient"-reading of COs. Evidence she draws from findings that point at di- VP NP V' John V YP die VP NP V' John V NP die a painful death a) b) Thematic intransitive Categorial transitive Thematic transitive Categorial intransitive <?page no="39"?> 27 rect objects being involved in aspectual properties of their verbs 30 . For verbs of change, "it [i.e. the direct/ affected object, SH] measures out over time the event described by the verb, through the change (of state or location) which it undergoes" (ibid.: 177). Thus in (12) Marcy knitted a sweater, the action described by the verb is finished when the sweater is finished. Similar sentences including COs in combination with a temporal adverbial delimiting the event are shown to be grammatical: (13) a. ? Casey laughed a hearty laugh in 20 seconds but: b. *Casey laughed heartily in 20 seconds. (quoted ibid.) This test is indicative of the telicity of the event verbalized in an utterance, because only temporally bounded events can be measured out (the 'measuring out' being expressed by the adverbial in 20 seconds in the above examples). Massam's comparison of properties of COs with further properties of typical Patient arguments 31 seemingly support her claim that COs can be seen as direct objects (arguments) carrying the patient -role. This of course leads back to the question why COs behave so differently from other objects (see Table 2.1). As regards the inability of COs to passivize, Massam suggests that this is characteristic of all direct objects which involve a necessarily bound element, i.e. the lexical item from which they are projected involves a bound variable inside the object variable (ibid.: 180). The fact that COs are created patients and no object can be created twice plausibly explains why COs cannot freely be pronominalized by definite pronouns, as exemplified in (14a): (14) a. *John sighed a weary sigh, then Harry sighed it. b. John sighed a weary sigh, then Harry sighed one. c. John sighed a sigh, and it was a weary one. ( ibid.) 30 Cf. Tenny (1987, 1994) who observes that "[d]irect internal arguments are the only overt arguments which can 'measure out' the event" (ibid. 1994: 11). More details will be given in Ch. 6, which provides an account on the influence COs can have on the aspectual behaviour of verbal expressions. 31 These features are suggested by Dowty (1991) for the description of a so-called Proto- Patient: 1. change of state (including coming-to-being) 2. incremental theme (determinant of aspect) 3. causally affected by theme 4. stationary (relative to movement of Proto-Agent) 5. referent may not exist independent of action of verb, or may not exist at all. Massam notes that (4) is difficult to apply to any event and thus problematic for COs as eventive objects. <?page no="40"?> 28 However, created objects can be referred to by indefinite pronouns, and so can COs, as shown in (14b). Also, reference by an indefinite pronoun as in (14c) is possible (cf. ibid.: 181). The ungrammaticality of questions eliciting the CO as an answer is explained by Massam as follows: "[It] is clear that to question the CO is to question what is lexically given". This explanation is doubtful, especially if one takes into account the claim that COs usually occur with some kind of modification, adding new information. Thus, questioning the CO cannot be reduced to an "act" of asking for the obvious. The criterion of new information added to the CO by modification also permits their topicalization: (15) It was a weary sigh that John sighed. (ibid.) In sum, besides 'adducing evidence' for COs as arguments realizing the Patient role as their special -role, Massam demonstrates that the constraints on the kind of constructions COs can occur in is not only a syntactic issue, but is also a matter of pragmatics. This is true for sentences containing an unmodified CO, too: Instead of describing such expressions as ungrammatical one should view them as cases of redundancy (cf. ibid: 182), a view which is debatable and which will be examined in Chapter 4 in regard to principles of iconicity in language. 2.3.4 Summary and discussion: COCs in Generative Grammar Government and Binding theory and other generative-transformational approaches to syntactic structure comprise models in which syntactic arrangements are projected from the specifications included in the lexical entries of lexemes. Therefore, of main concern for an analysis of COCs within the GG paradigm are lexical specifications of the verbs occurring in the construction with respect to their subcategorization frames, argument structure arrangements and the (non-)assignment of case, and transformational processes which may result in the change of such specifications (as illustrated by Larson [1988] and Massam [1988]). If COs are arguments of their verbs, the only solution to the problem of how to represent intransitive verbs occurring with an internal argument is the postulation of a second entry in the lexicon: Die, for example, would need to be listed with a subcategorization frame that spells out its complementation by either a true argument or a non-argument or a quasi-argument (depending on how the CO is categorized within such an approach). Since due to the very nature of COs these types of arguments are restricted to one particular item for each verb (i.e. there is only one CO), no generalisations can be drawn, and thus for each (conventionally) intransitive verb which is able to occur with a CO a separate entry in the lexicon is necessary. This vividly illustrates what Goldberg (1995) describes (and criticizes) as GG's treatment of <?page no="41"?> 29 lexical entries as the "last refuge of the idiosyncratic", in that all idiosyncratic properties are ascribed to individual lexical items 32 . Being a highly formalized approach to the description of speakers' linguistic competence rather than performance, GG leaves hardly any room for an evaluation of the function of COs or their usage in actual speech. Of all the studies reviewed in the preceding sections, only Massam gives some thought to the question of the functional motivation of such a construction when she comments on a shift in telicity in the verbalized event which is caused by the inclusion of a CO into the clause. Issues of usage and function of linguistic elements take centre stage in models of Functional Grammar, which will be dealt with in the next section. 2.4 Functional Grammar accounts of COCs Functional Grammar (FG) is an approach to language which focuses on the social and communicative functions of the linguistic system. Contrary to the formal paradigm of Generative-Transformational Grammar, in which language is described in a strictly algorithmic fashion and independent of matters of actual usage, the functional paradigm considers language as "an instrument of social interaction among human beings" ( (Dik 1997a : 2) in the first place. Linguistic categories, functions and structures are analysed and described with respect to the communicative and interactional functions they fulfil. Among the guiding principles of this approach one finds that a theory of language […] should not be content to display the rules and principles underlying the construction of linguistic expression for their own sake, but should try, wherever this is possible at all, to explain these rules and principles in terms of their functionality with respect to the ways in which these expressions are used. (Dik 1997a: 4) In envisaging grammar as a "sub-component of a model of a natural language using (NLU) system in which the human linguistic capacity is linked to epistemic, logical, perceptual and social capacities" (Siwierska 1991: 1), the Functional Grammar paradigm shares important characteristics with the Cognitive Linguistics approach to language (see Ch. 3). Several models have been developed within the functional framework, with Dik's and Halliday's being the most elaborate ones, both offering comprehensive and subtle models of grammar which aim to provide a "complete account of sentence structure from the underlying semantic representation to the surface phonetic form" (Siewierska 1991: 2). Of these 32 Note that in Chomsky's most recent model of generative grammar, his Minimalist theory (1995), the concept of the lexicon as a 'refuge of the idiosyncratic' is maintained (cf. ibid. 1995: 235f.). <?page no="42"?> 30 two approaches, only Halliday's work explicitly mentions COCs (see 2.4.2). Dik does not take up the phenomenon of COCs in his description and none of the examples he gives as illustrations for his claims include such a form, so that I can only speculate on their status in his model (2.4.1). The most specific and most detailed account of COCs within the FG paradigm is offered by Kuno & Takami) (2004) and is examined in 2.4.3. 2.4.1 COCs as instances of valency extension: Dik (1997) In Dik's model, clause structure is represented as a system of abstract underlying structures that are expanded on different levels or layers of formal and semantic organization. Starting off on the lowest level with predicates as basic elements, the full clause structure is arrived at in a "step-by-step fashion", proceeding through the different stages 33 . As for COs, one would expect information on their inclusion either at the level of nuclear predication, if they are considered as arguments of the verb (which would mean that the verb is 'endowed' with such an argument specification) or at the level of core predication, where lexical satellites, specifying, for instance, the manner of the action designated by the predicate, are added. Therefore, if the CO counts as an element describing manner, as described in some of the discussed literature, it would be integrated into the clause at this layer. Yet, as noted before, Dik does not take up a discussion of COCs. For a description of COCs within this framework it might be of interest, though, that in his treatment of predicate formation, Dik describes the process of valency extension, which is to explain complementation patterns in which, for example, an intransitive predicate is 'turned into' a transitive one (Dik 1997b: 8f.). This form of extension is exemplified by the following sentence pair, in which the 'original' agent-subject (the soldiers) in (16a) appears as goal-object (in Dik's terminology) in (16b), and a new agent is inserted and 'takes over' the subject position: (16) a. The soldiers marched to the camp. b. The corporal marched the soldiers to the camp. (ibid.) 33 These steps are: nuclear predication (level 1, inclusion of arguments) => core predication (level 2, where information about the internal dynamics [aspect] and Manner, Speed, Instrument, Direction, Beneficiary are spelt out) => extended predication (level 3, where temporal, spatial and cognitive specifications are added) => proposition (level 4, extended by information about the speaker's evaluation or attitude) => clause (level 5, with illocutionary force markers added). The extension of the predicate on the respective levels is realized by both grammatical operators and lexical satellites. (cf. Dik 1997a: 58 ff., emphasis added, SH). <?page no="43"?> 31 In order to make certain that the newly formed predicate will be sanctioned by the expression rules 34 , which might detect "one argument too many" (ibid.: 9) in the predicate, Dik formulates the 'Principle of formal adjustment' and the 'Principle of semantic adjustment 35 , in accordance with which the NP the soldiers in (b) will be sanctioned by the expressions rules as a second argument with the thematic role Goal. In this light, one could consider COCs with verbs conventionally used in an intransitive construction (i.e. dream, live, laugh etc.) as instances of valency extension, where the derived construction then conforms formally and semantically to the non-derived, i.e. transitive predicates. This implies that the added element in, e.g. He lived a sad life, extends the formerly oneplace predicate live, and fulfils both the syntactic and semantic functions of the second argument of an 'original' transitive construction. With this approach, the CO could be considered a 'true' argument, realizing the semantic function of Goal, the regular function of second arguments. Goossens (1994), who builds on Dik's model, criticizes functionalists for not having paid much attention to transitivity, a term which he regards as a key notion of both traditional and more theoretically oriented grammars (Goossens 1994: 65). He follows Rice (1987) in distinguishing between prototypical and non-prototypical instantiations of transitivity, with the prototypical transitive event being one […] in which the two asymmetrically related entities are involved in some unilateral activity. The activity requires forceful movement or some energized transfer instigated by one entity resulting in either contact with or some observable effect in the other. (Rice 1987: xii, so quoted in Goossens 1994: 66) Goossens identifies clauses containing an element with the semantic function Goal as defined in FG as equivalents of the transitive prototype, but extends the category of transitive predicates in order to include Effected Goals, i.e. entities "which depend for [their] existence on the successful realization of the [state of affairs] to which [they] belong" (ibid.: 71). While I would understand (prototypical) COs as belonging to the group of Effected Goals since they owe their existence to a successfully completed activity referred to by the verb, Goossens introduces a separate class labelled 'second arguments/ terms other than Goal', offering examples like John died the 34 Expression rules determine the form, order, and the prosodic constituents of the underlying clause structure (Dik 1997b: 49). 35 1. Principle of formal adjustment (PFA): "Derived construction types of type T tend to mould their expressions after the typical expression model of non-derived constructions of type T. 2. Principle of Semantic adjustment (PSA): "To the extent that the PFA is yielded to, the derived construction will also tend to adjust to the semantics of non-derived constructions of type T" (ibid.: 20). <?page no="44"?> 32 death of a saint and John ran two miles. He characterizes them as more problematic cases and argues: "I take both of these to contain a satellite rather than a second argument, but especially the characterization of the so-called cognate object […] is tentative" (ibid.: 73, emphasis added, SH). In the course of his argument, Goossens also attempts at an account of passivized sentences in which a second argument with a semantic function other than Goal, or a third argument or a level-1 satellite, is turned into the subject of the sentence as in (17) The five miles was run by the entire team. (ibid.: 75f.) For an adequate and plausible explanation of these structures, Goossens relies on cognitive processes such as the speaker's ability to view the entities as effected or affected (either literally or metaphorically), i.e. to construe a transfer of energy between two or more entities. Surprisingly, the author does not consider construal as a possible explanation for COs as Effected Goals. 2.4.2 Cognate Objects as Participants: Halliday (1985, 2004) Halliday (1985, 2004) assumes a tripartite functional 'segmentation' of a clause, distinguishing the categories "process", typically realized by a verbal group, "participant", typically encoded by a nominal group, and "circumstance", usually expressed by an adverbial group or a prepositional phrase (cf. Halliday 1985: 102, 2004: 175). The 'participant group' consists of two subcategories, based on the participant's function or involvement in the process. On the one hand, there are those that are directly involved in the process, including Actor, Goal 36 , Behaver, Senser, Phenomenon etc. (Halliday 1985: 132). On the other hand, he describes a group of participants which consists of Beneficiary and Range, being 'oblique' or 'indirect' participants. It is the Range category that is of interest here, since the cognate object is one of its 'main representatives'. As Halliday formulates: In the 'logical' terminology' […], where Actor is 'logical subject' and Goal is 'logical object', the Beneficiary is 'logical indirect object' and the Range would be 'logical cognate object'. (1985: 132, emphasis added, SH) 37 Range is defined as that element that "specifies the range or scope of the process" and described as "the meaning which is behind the classical category of cognate objects" (ibid.: 134). Halliday weakens the claim expressed in the above quote by not making (morphological) cognateness a necessary feature for elements of this category; he notices, however, that many Range 36 In Halliday’s system, the role Goal also includes what is referred to as 'Theme' by other approaches. 37 In an earlier paper, Halliday (1967: 58) notes that this terminology would be overspecific not only in regard to 'object' but also in regard to 'cognate'. <?page no="45"?> 33 elements are close in meaning to the verb. The following examples are used as illustrations by the author: sing a song of sixpence, with a CO as range element, or do you play croquet with the Queen today? , where the range element is semantically related to the verb, but also Mary climbs the mountain and John plays the piano. In the latter two examples, the Range is an entity which exists independently of the process but which specifies the domain over which the process takes place (ibid.), whereas in the first two examples, the Range is "rather another name for the process" (ibid.: 135). These differences are reflected in the labels Halliday attributes to them: 'Range: entity' and 'Range: process', respectively, with COs belonging to the latter type. As for a speaker's motivation to restate "the process itself in a nominal form" (ibid.: 149), he suggests that this construction enables the speaker to further specify the quantity [sing five songs], class [sing gospels] or quality [sing adorable songs] of the processes described. In an earlier paper, published 1967, Halliday had pointed at another function of the range element, describing its contribution to the extension of a sentence in that it allows for the inclusion of other elements: The introduction of a 'range' nominal […] allows the presence of a beneficiary: while she sang John and she danced Jim are impossible, she sang John a song and she did Jim a dance are acceptable. Range specifies the extension of a process, or in some case specifies the process itself (where the verb is neutral); but it also objectifies it so that it becomes not merely specific, as quantity or quality, but also as it were a 'participant' in the clause, and thus relatable directly to a beneficiary […]. (ibid.: 61, emphasis added, SH) In other words, similar to the observations by Jespersen and Huddleston & Pullum discussed in previous sections, Halliday describes COCs as devices for the lexical elaboration of sentences and, moreover, for the expression of a more complex participant constellation. With regard to his participant categories, Halliday (1985) includes some remarks on the problems which might arise for the distinction between Range and Goal: He observes that they are particularly difficult to distinguish in clauses describing material processes with only one direct participant (They played five games before tea), in the case of which the Range is syntactically treated as if it were a Goal, allowing the Range to become Subject in a passivized clause (Five games were played before tea., cf. ibid.: 136) 38 . What is more, in his description of a third group of clausal 38 See Halliday (1985: 136) for a compilation of grammatical idiosyncrasies distinguishing between Range and Goal. Among these features are the restrictions that the Range "cannot be a personal pronoun, and it cannot normally be modified with a possessive". However, counterexamples of COCs (expressing Range) are recorded in the BNC: a) Personal Pronoun: If you tell him to jump a jump twice, he does it; but when you tell him to jump it a third time […] (AR5: 1536). <?page no="46"?> 34 elements, the so-called 'circumstantial elements', which specify Extent and Location in space and time, Manner, or Cause (among others), Halliday (ibid.: 137) comments: "There is no sharp line separating (circumstantial) expression of Extent from (Participant) expressions of Range[…]". This overlap of categories of clausal elements clearly parallels the difficulties of a clear categorization of COs as either verbal arguments or adverbials/ adjuncts in the approaches described in foregoing sections. In the latest edition of his Introduction to Functional Grammar (2004), Halliday comes up with a more specific categorization of the clausal elements he had grouped together as 'range' participants in earlier editions. Depending on the semantic category of the predicate they occur with, these elements are relabelled 'Scope' (in 'material' clauses, e.g. die a death 39 ), 'Behaviour' (in 'behavioural clauses, e.g. sing a song) or 'Verbiage' (in 'verbal' clauses, e.g. tell a tale) (cf. ibid.: 193; 251; 255 f.). They still fall under the category 'clause participants', 'Range' being reintroduced as the general term for the described elements. It is noteworthy that Halliday no longer mentions the special status of COs in being prone to appear in the role of Range elements. 2.4.3 A Functional Constraint on the COC: Kuno & Takami (2004) Kuno & Takami solely couch their analysis of COCs into a discussion of the unergative-unaccusative distinction of intransitive verbs, reacting to the claim of several scholars (e.g. Larson 1988, Levin & Rappaport Hovav 1995, Macfarland 1995, Massam 1990) that the acceptability of the construction can be accounted for by such a distinction in that only unergative verbs 40 are selected by the construction (what Kuno & Takami call the "unergative restriction"). It is the authors' aim to show that the acceptability of COCs is bound to the construction's semantic function and discourse factors rather than to the type of intransitive category the verbs belong to. The starting point for their analysis is examples of COCs which involve verbs that have been shown to be unaccusative verbs in a wide range of studies (see Kuno & b) Possessive modifier: He smiled his smile at them. (H7A: 1090) 39 Some of the following examples are slightly adapted in order to cover CO phenomena, however, the verbs used in the examples are listed by Halliday under the respective categories. 40 The authors define unergative and unaccusative verbs as follows: "Unergative verbs are (i) those describing willed or volitional acts, taking agents as their subjects (e.g., play, speak, shout, work) ; and (ii) those describing certain involuntary bodily processes, taking experiencers as their subjects (e.g., cough, belch, hiccup, vomit). […] unaccusative verbs are (i) those whose subjects are semantically themes or patients (e.g., burn, drop, tremble, float), (ii) those of existing and happening (e.g. hang, remain, happen, occur); and (iii) aspectual verbs (e.g. begin, start, end)" (Kuno & Takami 2004: 106). <?page no="47"?> 35 Takami 2004: 111ff.), running counter to the predictions of the unergative restriction: (18) a. The general died the death of a hero. b. Mary blushed a deep/ sudden blush. c. The apples fell just a short fall to the lower deck, and so were not too badly bruised. d. The tree grew a century's growth within only ten years. (ibid.: 111-116) Moreover, the authors invalidate the claim that the COC is restricted to intransitive verbs that take as their objects only cognate nouns, a claim which captures the assumed unacceptability of utterances such as *He dreamt a nightmare (Baron 1971: 81) or *He laughed an ironical smile (Horita 1996: 241). Counterexamples are presented by Kuno & Takami: (19) a. He slept a fitful slumber. b. Van Aldin laughed a quiet little cackle of amusement. (ibid.: 118) Building on studies of COCs in which the CO-NP is portrayed as designating the result of an action, Kuno & Takami advance an explanation which can account for the acceptability of the examples in (19) in contrast to the ones judged as unacceptable above: Whereas, for example, a fitful slumber can readily be considered as a natural result of the act of sleeping (slumber is seen as a synonym of sleep), an ironical smile does not belong to the subset of possible outcomes of an act of laughing (the latter involving sounds, which the former does not) (ibid.: 120). These observations lead the authors to postulate a preliminary "Functional Constraint on the Cognate Object Construction": In the cognate object construction, the cognate object (the whole NP) must represent a specific state or event that is a subset of the possible states/ events resulting from the action represented by the verb. (ibid.: 121; emphasis added, SH) This constraint is also taken to explain why a CO-NP must necessarily be accompanied by a modifying element, since otherwise the NP does not refer to a subset (as in * Mary laughed a laugh). The "Functional Constraint" quoted above is extended by one further restriction, claiming that "the speaker's specific reference to the state or event represented by the cognate object must be either conventionally or contextually justifiable" (ibid.: 125). This is to explain Kuno & Takami's observation that speakers readily accept COCs which include additional information on why a speaker has chosen to verbalize the result of the event, whereas sentences without such information are considered to be marginal: <?page no="48"?> 36 (20) a. ? ? The apples fell a short fall. b. The apples fell a short fall to the lower deck, and so were not to badly bruised. (ibid.: 124) The additional restriction is taken to account for the unacceptability of COCs in which the whole NP represents manner rather than result (ibid.: 126): (21) a. *The snow melted a slow melt. b. *We approached a strange approach. Unfortunately, the authors do not include information on acceptability judgements on sentences with differently modified CO-NPs, i.e. not expressing manner (e.g. a short melt, a different approach). Furthermore, it does not become quite clear by which criteria CO-NPs denoting results are distinguished from CO- NPs describing manner, especially as the similar example Bill sighed a weary sigh (ibid.: 105) is acknowledged as an acceptable instance of a result-denoting COC (yet, weary denotes manner! ). What is more, the adjective slow is not uncommonly used as a modifying element within others CO-NP (smile a slow smile, die a slow death, dance a slow dance). The authors leave it open what exactly causes the manner-denoting quality of the whole CO-NP and thus the unacceptability of the examples in (21). A final point taken up by Kuno & Takami concerns passivized COCs, whose occurrence is negated in some studies but is affirmed by the authors, as in: (22) a. It is a smile that could be smiled by the whole country. b. Pictures were taken, laughs were laughed, food was eaten. (ibid.: 1127f.) The authors suggest that a passive form of the COC is most acceptable and most frequently used with an unmodified CO. In such cases, the COs serves the rearrangement of information structure, enabling passivization, which would otherwise (i.e. in the absence of a direct object) not be possible. Thus, a last 'amendment' is added to the "Functional Constraint on the Cognate Object Construction": […] in active sentences, if a cognate object does not have a modifier, as in laugh laughs, it is tautological and therefore, there is no justification for its use. On the other hand, in the passive construction, a cognate object without a modifier is acceptable, as long as Passivization is justifiable. (ibid.: 130) Kuno & Takami have compiled a considerable collection of COC examples which are partly extracted from newspaper articles, novels, or earlier studies on the construction, and partly invented by the authors and/ or elicited through acceptability judgements by native speakers. The data seem to provide ample evidence for their hypothesis that the form and usage of COCs is not so much constrained by the verb's syntactic class (i.e. ergative <?page no="49"?> 37 vs. unergative), but that, in fact, such constraints arise from the semantics of CO-NPs (type vs. subtype; result vs. manner), the nature of coand context, and information structure. However, a few weak points in Kuno & Takami's study must be noted (in addition to the fact that the example sentences are not always based on authentic language data): Besides the implausible result-manner distinction already pointed at, the question why unmodified COs, which are claimed to be unacceptable by the authors, do occur in natural language usage is not accounted for. Also, they define the COC as a construction involving intransitive verbs which take a resultant object, either designated by a cognate noun or a possible subtype. A similar pattern can be observed for constructions that contain transitive verbs such as weave a web, tell a tale or build a building, which also can be complemented by non-cognate nouns referring to subtypes. Yet, these are not considered as instances of COCs due to their verbs' status as 'transitive'. Evidently, like other authors discussed before, Kuno & Takami view transitivity in terms of the syntactic surroundings of a verb, which, in turn, is taken to result in a clear distinction between intransitive and transitive verbs. Note also that the issue of transitivity is not taken up by the authors as a motivating factor for the use of a COC with an 'originally' intransitive verb. Such an alternative view will be presented in Chapter 6. 2.4.4 Summary and discussion: COCs in Functional Grammar Also in the paradigm of Functional Grammar, linguists are faced with the problem of ascribing COs to any of the categories they have established within their theories. The fact that COs appear in several subgroups suggested by Halliday, his fuzzy distinction between Range and Circumstance, and Goossens' emphasis of the tentative nature of COs as satellites are obvious indications for the class of COs being anything but homogeneous. Goossens' approach to non-prototypical transitive clauses, which is to account for the passivization of clause elements that deviate from prototypical objects of a transitive verb in terms of construal, clearly opens up possibilities for the recognition of COs as clause participants rather than as phrases designating circumstances. His view on transitivity as means of expressing energetic transfer between entities makes such a classification more plausible than Halliday's grouping, which also yields an interpretation of COs as clause participants. Kuno & Takami background the discussion of the syntactic status of COs, focusing instead on functional constraints on the occurrence of these objects. In their approach, the use of the construction is described as depending to a great extent on the expressive needs of speakers, rather than on syntactic characteristics of the verb. With the authors integrating information structure and pragmatic aspects as explanatory factors, their ac- <?page no="50"?> 38 count of COCs most clearly reflects the overall aim of FG to explain linguistic structures in terms of communicative and interactional demands of language users. 2.5. Macfarland (1995): Cognate Objects and the adjunct / argument distinction The most comprehensive analysis of COCs so far has been carried out by Macfarland (1995) in a dissertation thesis in which she discusses on the basis of this construction procedures, methods and techniques that have been developed to distinguish between adjuncts and arguments. In the following sections I will give an account of and discuss the most important observations and arguments listed by Macfarland, at the same time consulting and considering works and analyses that have contributed with their ideas to her analysis or which have advanced similar ideas. 2.5.1 Constraints on the form of COCs For her investigation, in the course of which she argues for the argument interpretation of COs, Macfarland compiled a corpus of naturally occurring tokens 41 , explicating a clear set of definitional criteria for those instances of COs which she incorporated into her collection. These criteria are stated in the following: Building on previous treatments of the construction, I have defined the cognate object construction as involving verb-noun pairs which are either zero-related or which share a root morpheme and are not derived by means of affixation […]. Close examination of a corpus of constructions that satisfy the definitional criteria of section 1.2 [i.e. morphological relations, SH], however reveals that the verbs in the cognate object construction have a creation verb reading and the nouns have a result object interpretation. Therefore I incorporate into the definition of the cognate object construction the constraint that the cognate object must be a result object. (Macfarland 1995: 48; emphasis added, SH) Before laying down these criteria, Macfarland discusses several other characteristics of COs that had been suggested by other authors (some of which were dealt with in the preceding sections), such as transitivity patterns of verbs or issues of modification and type of determination of the CO phrase. Macfarland includes in her analysis both verbs that are generally dealt with as intransitive (e.g. die, live, smile) and those that are generally considered 41 These are tokens cited in grammars or linguistics papers or obtained through computer searches of files from the Gutenberg project, the Wall Street Journal files and Nexis (cf. Macfarland 1995: 10). <?page no="51"?> 39 transitive (e.g. do, sing, tell), noticing that if COs are true objects the intransitive status of the verbs in the first group is questionable (ibid.: 12). As regards the "modification constraint", Macfarland opposes those views that claim the occurrence of a bare, unmodified CO-NP to be unacceptable or at least a rare phenomenon. She argues along the lines of Baron (1971: 90f.), who takes up this point of discussion when she tries to explain a semantic difference between transitive and intransitive verbs occurring with a CO. For this purpose, Baron works with the concepts 'Genericity', describing "a general class which can be subdivided into types", and 'Specificity', denoting "a member of a general class" (ibid.). Transitive verbs such as sing are associated with Genericity since they can occur with objects that denote types of songs, e.g. sing a ballad/ aria/ anthem, while intransitive verbs such as dream are associated with Specificity. With this class of verbs, Baron argues, an unmodified CO does not add any semantic content and thus would be redundant and unacceptable, since, for example, the only thing one can dream is a dream. With transitive verbs, on the other hand, generic (semantic) COs are not redundant but "rather specify genus" (ibid.), since one can fight a fight as well as a battle, duel or skirmish. In this respect, the occurrence of unmodified COs can be expected with transitive verbs. The question whether COs can be accompanied by definite determiners is one further point of Macfarland's discussion. The use of definite (strong) determiners 42 usually presupposes the existence of the entity denoted by the noun. Thus, if the CO is interpreted as something which is only created by the event described by the verb (i.e. object of result) or part of the event, it should not occur with these types of determiners. Macfarland, on the other hand, claims that COs are inherently definite, independent of the type of determiner preceding the cognate noun: Even if definiteness is understood in the pragmatic sense of invoking a previously mentioned discourse entity […], cognate objects must be considered to be definite. In the cognate object construction, the verb evokes the following object. There is no smile without the action of smiling, no dance without the action of dancing, no thought without the action of thinking. Thus, whether cognate objects occur with definite/ strong determiners or with indefinite/ weak determiners, they must be considered definite. (ibid.: 22) This alternative view on definiteness serves Macfarland as an explanation for the occurrences of COs with definite determiners attested in her corpus. Besides constraints on the form of COs, Macfarland also puts forward constraints on their meaning. She emphasizes the resultative nature of COs as an essential definitional criterion, relying on Fillmore (1968), who had pointed out that a distinction between object of result (effected object) and 42 This class of determiners includes: definite articles, possessive pronouns, demonstrative pronouns, and quantifiers such as all and every. <?page no="52"?> 40 regular object (affected object) is syntactically relevant 43 . Moreover, as claimed by Macfarland, the interpretation of COs as result objects has crucial impact on the interpretation of the verb it is combined with. Whereas verbs like smile have an activity interpretation outside the COC, they result in a creation verb interpretation when a CO is added (ibid.: 26). Further evidence for such a result interpretation is provided by the fact that the majority of COCs are instances of verb-noun combinations in which the noun is derived from the corresponding verb, i.e. a deverbal noun. Macfarland even speculates about this derivational direction being a necessary feature, although 16% of the verbs in her corpus are denominal (e.g. fish, nail, file, name etc.). These verbs she actually excludes from her discussion of COs, as she does with object NPs that are derived from their verbs by means of affixation. Thus, although paint a painting, draw a drawing, or build a building obviously fulfil all other criteria listed by Macfarland, they deviate formally from zero-derived COs or those that share their root with the corresponding verb. The decision to neglect them for further analysis is questionable, especially since morphemes such as -ing or -ment do not have an inherent meaning that precludes a result-reading of derived nouns with such affixes. Mittwoch (1998), who discusses COs in Hebrew, approaches them from a different angle, both as regards their 'result' reading and the morphological criteria for their identification as COs. She makes the following observation in her general introduction to the matter: Morphological cognateness between the verb and the object noun is neither a sufficient nor a necessary condition for the phenomenon to be discussed here. It is not sufficient, since some morphologically cognate objects do not denote events: […] build a tall building, sing a song, dance a dance. (Mittwoch 1998: 312; emphasis added, SH) A result interpretation of the given examples, including an instance with the derived form building, is her favoured alternative. Therefore, she draws a distinction between COCs that have a result reading and those with an event reading, independent of the respective noun in the CO phrase. That is, even a CO noun that qualifies as result object in some instances might be read as an event argument in other occurrences. The result-event distinction is taken up by Macfarland, contrasting examples such as 43 Fillmore (1968) notes that only sentences with affected objects but not with effected objects allow a paraphrase with do to. Compare the following pairs: a) John killed the dog. (affected) What John did to the dog was kill it. b) John dreamed a dream. (effected) *What John did to the dream was dream it. Another difference arises in case assignment, with affected objects being assigned Objective Case and Effected Objects being assigned Factitive Case, the latter being defined as "the case of the object or being resulting from the action or state identified by the verb, or understood as part of the meaning of the verb" (Fillmore 1968: 25). <?page no="53"?> 41 (23) a. Life here had been lived on a scale and in a style she knew nothing about. b. She was only twenty-seven, and the life [which] she had lived in Hotel Manhattan had seriously abraded her […]. (ibid. 1995: 61.) The example in (23a) is recognized as an instance of the group of COs with an event reading, whereas in the second example the COs adopts a result reading. However, Macfarland views such a distinction as problematic since it needs to be done on a case by case basis (ibid.: 56f.), where co-text and modifying elements which have an influence on the respective interpretation of the CO as either result or event must be considered. 2.5.2 COCs compared to Light Verb Constructions In order to clarify the status of COs as arguments, Macfarland carries out a detailed comparison with a construction that is often described as similar to COCs in the literature, the so-called light verb construction (LVC). This construction type is characterized by a combination of verbs such as have, take, do or give with an NP denoting an event or an action (which often occur in COCs as well), as displayed in the following examples: (24) a. There was a pause, and I heard him take a deep breath. (FEE: 1847) b. He did a short angry dance by the bed. (FSP: 822) c. Burney gave a bashful grin. (ANL: 1331) In her comparison of COCs with LVCs Macfarland examines the different semantic roles ascribed to the respective NPs in each construction, arguing that whereas the nominals in LVC are event arguments, i.e. arguments not referring to participants of the event but the event itself, such an analysis is not appropriate for COs. Similar to the tests that are utilized to distinguish true objects from (more adjunct-like) COs (see p. 25), several test procedures are supposed to demonstrate the difference between LVCs and COCs. For reasons of convenience, these procedures are summarized in Table 2.2 44 . It is an interesting, though confusing fact that some of the test procedures applied here are identical to those used by Massam in order to distinguish between true objects and COs, yet with the opposite result. In Massam's approach, COCs display those characteristics which Macfarland postulates for LVCs in order to distinguish them from COCs. In this light, Massam's analysis renders COCs more similar to LVCs than the account proposed by Macfarland. 44 Cf. Macfarland (1995: 60ff.), who makes use of data and results from Kearn's (1988) study of light verbs. COC examples in the table are mine, extracted from the BNC. <?page no="54"?> 42 Test Procedure Light Verb Construction Cognate Object Construction Passivization *A groan was given by the man on the right.(Kearns) For twenty years my life has been lived within the Left. (EG0: 71) Wh-Movement 45 ? ? The groan which he gave startled me.(Kearns) Some young women might find the life [which] we live here unexciting. (JY7: 3149) Pronominalization ? ? The deceased gave a groan around midnight and gave another one just after two.(Kearns) If you tell him to jump a jump twice, he does it; but when you tell him to jump it a third time, he says, I've had it […]. (AR5: 1536) Definiteness *Who gave the groan just now? (Kearns) She could hardly bear to think the thought […]. (EFP: 33) Table 2.2 Test procedures for the distinction between LVCs and COCs With regard to differences in the semantic properties of LVCs and COCs, Macfarland notes that besides the fact that the verbs in LVCs are semantically empty or insignificant they also denote actions that are briefer and more casual. This difference seems to be one result of the different type of argument associated with the two constructions, an overt event argument for LVCs and a thematic object for COCs (ibid.: 65). 2.5.3 Evidence for COs as arguments It is mainly their role of an object of result which is responsible for the inconsistencies in classifying COs as either arguments or adjuncts since effected objects are generally considered as adjuncts in the literature. Differences in the behaviour of arguments and adjuncts have been captured in diagnostic tests which serve Macfarland as a means to demonstrate argument characteristics of COs. The majority of these diagnostic tests target at the constituency of the verb phrase, since one of the main differences between arguments and adjuncts is that the former are integrated into the VP and consequently cannot be separated from it. Table 2.3 presents an overview of these tests and their application to COCs (cf. Macfarland 1995: 102ff.). The behaviour exhibited by COCs in the respective constructions support an interpretation of COs as arguments. Like Massam (1988; see 45 This criterion is defined by Kearns (1988: 54) as follows: "The nominal complement of a [LVC] cannot be the focus of a WH-question, or be modified by a relative clause". <?page no="55"?> 43 Test Procedure Arguments Adjuncts Though -Movement (preposes entire VP) Chris smiled a happy smile. Smile a happy smile though Chris did,(everyone could see…) Chris smiled that day. => *Smile that day though Chris did, (everyone could see…) VP-Preposing (preposes entire VP) I wanted Chris to smile a happy smile, and smile a happy smile she did that day. *I wanted Chris to smile that day, and smile that day she did. Do-So Copying (do so = proform of V', must substitute all arguments, can but need not substitute adjuncts, compare a) and b)). Chris smiled a happy smile, and Mary did so, too. *Chris smiled a happy smile, and Mary did so a sarcastic smile. I saw Chris that day. a) I saw Chris that day, and John did so, too. b) I saw Chris that day, and John did so another day. Long wh-movement Chris wondered [whether Lee smiled a happy smile] ? [What kind of smile] i did Chris wonder [whether Lee smiled t i ]? Chris wondered [whether Lee read that day]. *What day i did Chris wonder [whether Lee read t i ]? Table 2.3 Diagnostic tests for an adjunct-argument distinction above) and Tenny (1994), Macfarland seeks support for the assumed argument status of COs in aspectual properties of the construction in which they appear 46 . Following Olsen's (1994, 1997) classification of aspectual classes (State, Activity, Accomplishment, Semelfactive, Stage-Level-State), Macfarland shows that bare CO verbs differ from their COC counterparts with respect to the class they belong to. Olsen's aspectual classes are distinguished through a combination of absent or present features such as "telic", "dynamic" and "durative". Verbs can change their aspectual class, but only through the addition of features, original features cannot be lost. Macfarland assigns COCs to the class of accomplishments, which carry the aspectual features [+telic] 47 , [+dynamic], [+durative] (ibid.: 131). 46 Tenny (1994: 38f.) notices a shift in aspectual behaviour when intransitive unergative verbs occur with COs. While they typically describe non-delimited events in their 'bare' form, the inclusion of a CO results in a delimited reading of the event (cf. ftn. 30). 47 As will be discussed in Ch. 7, the term "telic" is ambiguous to some degree: While traditional descriptions take "telic" to refer to processes which inherently involve a terminal point (cf. Östen 1981), more recent approaches use the term in the sense of 'boundedness'. Macfarland quotes a definition of "telic" by Olsen (1994: 61), denoting "the existence of an end result to which a situation will lead, not necessarily the actual attainment of such an end", she does, however, not make explicit in which sense she applies the term (so quoted in Macfarland 1995: 132). <?page no="56"?> 44 Macfarland's analysis of the corresponding bare CO verbs brings to light a rather heterogeneous behaviour with respect to the distribution of the aspectual features and thus their membership in any of the aspectual classes. As she demonstrates, all the verbs compiled in her corpus are uniform concerning the presence of the feature [+dynamic], which leads her to the conclusion that stative verbs, which have the feature [-dynamic], are unlikely to appear in a COC. Most other feature combinations are attested for the verbs (ibid.: 136ff.): Verbs such as bark, bow or kiss are marked for the feature [+dynamic] only, belonging to the class of so-called semelfactives, i.e. verbs which express the momentary and single occurrence of an action. Here, the COC is said to add a notion of durativity. Compare the following: (25) a. He bowed. b. He bowed his correct little bow. (ibid.) Live, smile and sing are classified as activities with the features [+dynamic, +durative]. The addition of a cognate object to these verbs is assumed to express telicity of the event. While die belongs to the category of achievements, carrying the features [+telic, +dynamic], tell or sing display the features [+telic, +dynamic, +durative] and are thus categorized as accomplishments. Both sing and tell occur with both cognate and non-cognate objects, which contribute to the telic nature of the verb phrase. Macfarland sees the main difference between these two types of objects in their emphasis of telicity: Whereas non-cognate objects add new semantic information, COs are semantically non-informative and draw "the hearer's attention on the aspectual information provided by the cognate object" (ibid.: 147). However, she does not comment on the difference between modified and unmodified CO-NPs, with the former definitely adding new semantic information. According to her argument, only unmodified CO-NPs would then emphasise the telic reading of the event described, yet, as frequently noticed in relevant studies on the construction, even though unmodified COs do occur, their frequency is comparatively low. Macfarland does, however, not comment on the number of attested unmodified CO phrases in her collection of COCs to underscore her argument. Felser & Wanner (2001) also consider shifts in aspect as a major function of the inclusion of a CO into a VP. The authors aim at a description of what they label 'unselected objects', such as COs and reaction objects (as in Warm thanks were smiled at the audience). Focusing on intransitive unergative verbs, they discuss the role of COs as theme arguments. Here, they observe parallels with 'ordinary' transitives that occur without an overt theme, i.e. cases in which the theme does not surface since it can be understood as part of the verb. Thus, <?page no="57"?> 45 (26) a. Mary smiled. entails b. Mary smiled some kind of smile. just as (27) a. They were eating. entails b. They were eating something. (ibid.: 115). Through this and other observations the two authors arrive at the conclusion that in objectless transitive and unergative structures, the understood Theme argument is actually part of the sentence's semantic representation. However, the Theme argument of unergatives is not an event participant in the same way as the Theme selected by eat or drink is. Rather, it usually serves to delimit the event described aspectually. (Felser & Wanner 2001: 117; emphasis added, SH) As regards the acceptability of COCs, they also adhere to the modification constraint since - in their view a bare object does not make any contribution to the semantics of the clause. The realization of an unmodified CO "will normally be precluded for reasons of economy" (ibid.: 118). The overall conclusion they draw from their analysis is that unergatives are always transitive but permit their internal arguments to be an empty category (what Felser & Wanner call "null pronominal"). Is the Theme spelled out, it adds features of telicity to the verbal representation of an event. Accordingly, the only difference between (28) a. He fought heroically. and b. He fought a heroic fight. is that the latter sentence has a telic reading, i.e. is temporally bounded and taken to include the terminal point of the action . With their discussion of the contribution of COs to aspectual characteristics of an event description Tenny (1994), Macfarland (1995) and Felser & Wanner (2001) support recent linguistic studies on verbal aspect which stress that "aspectual distinctions are distinctions between linguistic expressions and are not properties of events in themselves" (Rothstein 2004: 2). 2.5.4 Summary and discussion In the preceding sections, I gave an overview of the rich evidence that Macfarland collected to prove the argument status of COs versus their classification as adjuncts. Although she successfully applies a whole range of tests to corroborate the argument interpretation, some inconsistencies, especially as regards the semantic nature of CO phrases, remain. This holds in particular for the distinction between an event and result interpretation of the CO-NP, for which she cannot provide clear criteria. Asserting the existence of COs describing events, she does, however, not extend her defi- <?page no="58"?> 46 nition to include such an interpretation and exclusively discusses them as objects of result. One of the major conclusions Macfarland draws from her analysis of COs as arguments is that their inclusion into the verbal expression describing an event leads to a shift in its aspectual characteristics. It is, however, debatable whether such aspectual characteristics are the only motivation for the integration of a CO. Such a claim leaves open the question why COs need to be modified or what kinds of semantic contributions the obligatory modifier adds. If the object phrase is included to temporally delimit the event, one would, above all, expect terms of measurement or quantifiers as modifying items. 2.6 Conclusion The purpose of the preceding chapter was to give an overview of the multifarious approaches undertaken to describe and define COs and to solve their somewhat problematic status with respect to their semantic, morphological and syntactic idiosyncrasies. The criteria applied to distinguish COs from other kinds of objects and the different labels given to their assumed semantic and syntactic roles are summarized in Table 2.4. One of the main problems of the discussions and descriptions analysed and evaluated in this chapter is the attempt of their authors to work with clearly delineated, self-contained categories, which is reflected in inconsistencies and - in the extreme case - circularity in these accounts. Circularity is nicely illustrated in a comment by Macfarland (1995): If a cognate object is a true object, then the verb with which it occurs is transitive. And if the verb occurring with a cognate object is a purely transitive verb (i.e. it cannot generally appear without an object), then the cognate object must be a true object. (ibid.: 12) As this statement clearly demonstrates, the distinction between transitive and intransitive verbs and the attempt to pin down the definition of COCs to such syntactic categories is one of the factors that create confusion in some of the descriptions. If the CO is a true object (and at least the label 'object' is given to COs by the majority of researchers, see Table 2.4), its occurrence cannot be restricted to verbs that are categorized as 'intransitive'. More promising is the approach taken up by Goossens (1994), who discusses transitivity in the light of prototype-theory and thus by means of categories which exhibit gradience in the goodness of membership and allow for overlaps with other categories. Although he does not first and foremost focus on COCs, Goossens at least acknowledges the possibility to treat them as peripheral, non-prototypical cases of transitive constructions. <?page no="59"?> 47 The endeavour to classify COs as either result objects or event objects also fails on the basis of assuming rigid categories and restricting an item's membership to one category only. In Chapter 4 I will argue that COs can be of both types, i.e. resultative and eventive, resulting in the polysemous nature of constructions involving COs. In Cognitive Linguistics, the theory to be drawn on in the remaining chapters of this book, polysemy is taken to be a naturally occurring phenomenon of both lexical items and grammatical structures which is reflected in the complexity of linguistic categories as networks of related meanings. Concepts and models such as prototype theory, polysemy and mechanisms of human cognition such as construal, to be discussed in the following chapter, are useful means to describe and present COCs from a different, cognitive-linguistic perspective. Author Verb type Syntactic function Semantic role of CO Relation between verb and object Modification of NP Sweet (1891) intr. object / conversion or semantic rel. / Jespersen (1927) intr. & trans. object object of result / + Zandvoort (1945) trans. object / semantic relation / Fries (1952) / object undergoer of action / / intr. adjunct / morphological relation Visser (1963) trans. object object of result semantic/ morpholog. relation / Baron (1971) intr. trans. object object of result semantic relation generic: specific: + Halliday (1985) / participant range semantic relation / Quirk et al. (1985) / object cognate object semantic (morphological) + Zubizaretta (1987) / adjunct / / + <?page no="60"?> 48 Jones (1988) intr.ergative adjunct / nominalization / Massam (1990) intr. object patient (eventive theme) / - Macfarland (1995) intr. & transitive argument object of result zeroderivation, ablaut -/ + Mittwoch (1998) intr.unerg. / object of result, object of event morphological/ semantic relation + Wanner & Felser (2001) intr.unerg. object Aspectual theme morphological relation + Huddleston& Pullum (2002) intr. & trans. object Factitive theme nominalization + Kuno & Takami (2004) intr. (unacc. & unerg.) / transitive object object of result semantic relation + Table 2.4 Overview of definitional criteria of selected approaches to COCs <?page no="61"?> 49 3 Basic Tenets of Cognitive Linguistics 3.1 Introduction: Cognitive Linguistics and Cognitive Grammar Cognitive Linguistics 1 (CL) is closely related to Functional Grammar (FG) approaches to language (see Ch. 2.4.) in that its description of linguistic structure is founded on and geared towards the functions language serves, most importantly the representation of knowledge and the communication thereof (which Givón [1998: 41] labels "megafunctions" of language). Both approaches are part of the movement known as "Cognitive-Functional Linguistics", developing and establishing itself as an alternative to the Generative Grammar paradigm, and denying the autonomy of language. Compared to FG, CL stands out through its emphasis on the semiological, i.e. symbolizing function of language and the crucial role of conceptualization in social interaction (cf. Langacker 1998: 1, 2008: 7f.). As such, it forms a stark contrast to the mathematical-formal approach utilized in generative linguistics. Cognitive linguists aim at the investigation of language via cognition, having recognized that this route offers a promising and prolific approach to the analysis of linguistic structures. Needless to say that, in turn, this approach to the study of language is essentially an attempt to understand human cognition, which is assumed to shape the structures of language. This is present in and demonstrated by all studies in Cognitive Linguistics. The purpose of the following sections is to outline those basic tenets, notions and descriptive tools of CL which I consider necessary and fundamental for the description and exploration of COCs within this theoretical framework. One of the most fully elaborated and most innovative models within CL is Langacker's Cognitive Grammar (CG), a theory of linguistic structure which has been developed since the late 1970s, centring on the conception of language as an essential facet of cognition. Since this model will serve as a guiding line for my description of characteristics of COCs, the main focus will be put on assumptions of Langacker's descriptive framework. However, key ideas of other, related models, notably Goldberg's (1995, 2006) account of argument structure constructions, and approaches to linguistic categorization (esp. Taylor 1995), will be incorpo- 1 Cognitive Linguistics (with capital letters) needs to be distinguished from cognitive linguistics, the latter generally referring to all approaches to language as a mental phenomenon (which, among others, also includes Generative Grammar). <?page no="62"?> 50 rated into the discussion. The chapter has a summarizing character; many of the basic notions will be examined more closely in subsequent chapters when applied to grammatical description. The key tenets of cognitive linguistics are laid out, examined and discussed in great detail in Geeraerts & Cuyckens (2007), Langacker (1987, 1991, 2008), Lakoff (1987), Cruse & Croft (2004), and Evans & Green (2006) 2 . 3.2 The non-autonomy of language One of the fundamental achievements of Cognitive Linguistics is its renunciation of viewing language as an autonomous mental ability, a view which was prevailing under Chomskyan linguistics. From the perspective of CL, language is part and parcel of cognition and is inherently tied to all other cognitive abilities of human beings, involving and reflecting basic forms of conceptual operations such as perception, conception 3 , attention etc. Violi (2004, quoted in Dirven 2005: 17) goes so far as to argue that "[p]erception, action, language cannot any more be considered as totally autonomous and independent modules, they must become functional specifications in a common unitary configuration" (emphasis added, SH). Such a position does not completely give up the assumption of a language faculty with innate structures which underlie the acquisition and 'execution' of language, but it does question the uniqueness of these structures in an autonomous language module (cf. Langacker 1987: 13). The way linguistic knowledge is represented is assumed to be essentially the same as the representation of other conceptual structures. This strongly implies that linguistic knowledge is basically conceptual structure (cf. Croft & Cruse 2004: 2) and that linguistic categories and structures can be captured via conceptually based descriptions. A demonstration of such a type of grammatical description is provided, for example, by Langacker (e.g. 1987a, b; 2005), who offers a systematic, conceptually based (notional) definition of verbs and nouns, when he describes nouns as designating THINGS (i.e. bounded regions in some conceptual domain) and verbs as designating PROCESSES (i.e. as a relation between components which evolves along a temporal scale). In generative linguistic paradigms, these classes are considered as purely syntactic categories, rendering them as autonomous from cognition. Furthermore, as hinted at in Violi's quote cited above, to cognitive linguists, cognition (and consequently language) is crucially intertwined with bodily experience: 2 For various summaries, see, for example, Dirven (2005), Gibbs (1996), Janda (2006), Langacker (1988a, 1988b) and Ungerer & Schmid (2006). 3 Talmy (1996) speaks of "ception" as the general process subsuming both perception and conception. <?page no="63"?> 51 [C]ognitive linguistics (and its allies in philosophy, anthropology, and psychology) views knowledge as arising out of people's bodily interactions with the world. Knowledge is seen not as static, propositional, and sentential, but as grounded in patterns of bodily experience. These patterns emerge throughout sensorimotor activity as we manipulate objects, orient ourselves spatially or temporally, and direct our perceptual focus for various purposes. (Gibbs 1996: 41) Cognitive linguists are guided by the "embodiment of meaning" premise, considering human experience of bodily existence as a fundamental source of meaning. Our earliest bodily (i.e. sensorimotor) experiences are said to form the basis for the emergence of so-called "image schemas". This term was simultaneously introduced by Johnson (1987) and Lakoff (1987), and defined as an abstract, possibly pre-conceptual 4 representation of a "recurring, dynamic pattern of our conceptual interactions and motor programs that give coherence and structure to our experience" (Johnson 1987: xiv, cf. also Johnson 2005: 19f.). The list of such schemas includes, among many others, CONTAINMENT , BALANCE , UP - DOWN , BLOCKAGE , PROXIMITY etc. As Gibbs (2005: 113) quite nicely comments, image schemas "represent the essential glue that binds embodied experience, thought and language". Thus, they constitute a central component of a model of language as a nonautonomous cognitive ability which is inextricably linked with bodily experience 5 : Goldberg (1998), for example, describes argument structure constructions such as the caused-motion construction or the transitive construction as linguistic patterns that emerge through patterns of experience, hypothesising that "constructions that correspond to basic simple sentence types encode as their central senses, event types that are basic to human experience" (ibid.: 205). As regards language processing, Gibbs & Perlman (2006) impressively demonstrate in one of numerous studies how bodily experience (both actual and simulated) and bodily posture of speakers influence and facilitate the comprehension of metaphorical and nonmetaphorical language involving spatial concepts. Viewed from a more general perspective, the close association between language, cognition and bodily (but also social and cultural) experience springs from the fact that knowledge of the world is a fundamental source of meaning (if meaning is taken to be of an encyclopaedic nature, see below): With knowledge being grounded in all kinds of experience, linguistic meaning cannot but be deemed to be experientially grounded as well. 4 There is still some controversy as to whether image schemas are preconceptual or only preverbal structures. While Lakoff and Johnson claim that they are preconceptual in nature, Radden & Panther (2004: 27) note that it is "intuitively more plausible to assume that, only after experiencing a number of image-schematic situations, a child can form abstract image-schematic concepts." 5 See Hampe (ed.) (2005) for a collection of most recent studies on the significance of image schemas in thought and language. <?page no="64"?> 52 3.3 The symbolic nature of grammar At the core of the CL paradigm one finds the conception of language being inherently symbolic in nature. That is, language is understood as an openended assembly of form-meaning pairings. What at 'first sight' reminds one of the Saussurean description of linguistic signs soon emerges as an original and elaborate account on the meaningfulness of all linguistic forms, be they morphemes, lexemes, phrases or sentences, and the structures involved in the arrangement of these elements. Hence, not only (lexical) morphemes and lexemes are considered as pairings of form and meaning, but also syntax, i.e. the patterns which combine morphemes and lexemes into larger configurations. The role of syntax within an inherently symbolic linguistic system is posited by Langacker (1988a: 5) as follows: [G]rammar is intrinsically symbolic, having no independent existence apart from semantic and phonological structure. Grammar is describable by means of symbolic units alone, with lexicon, morphology, and syntax forming a continuum of symbolic structures 6 . In order to fully comprehend this way of describing linguistic expression, it is necessary to view in more detail the notion of 'symbolic unit', i.e. the symbolic relationship between semantic and phonological structures. The term 'unit' comprises all and only those cognitive structures which are automatised by the speaker in such a way that they can be used as integrated wholes, without the speaker having to pay attention to their individual parts or their particular arrangements (cf. Langacker 1987a: 494). These units vary to a great extent with respect to, for example, specificity and complexity. To give an example concerning varying specificity, the symbolic unit dog comprises the semantic pole DOG ("canine") with the phonological pole [d g], resulting in the bipolar unit [[ DOG ]/ [d g]]. The conceptual class of nouns can be described in a similar fashion, yet on a more abstract level, i.e. as the pairing of a highly schematic phonological pole " X " and a schematic semantic pole "thing" (the notional definition of nouns), yielding the symbolic unit [[ THING ]/ [ X ]]. As for their varying complexity, symbolic units range from simple morphemes, e.g.[[ PAST ]/ [d]] and lexemes [[ GUM ]/ [g m]] to more complex forms such as morphologically complex words (admiration), phrases (on the hook, highly appreciated) or sentences (How is it going? She fought the good fight.). These specific, complex symbolic units, too, can be captured by more abstract, schematic representations, e.g. [ THING / NP PROCESS / VP THING / NP ] for a pattern that is traditionally known as the monotransitive construction. The assembly of such a 6 Wierzbicka (1988: 1) argues along similar lines when she writes: "If semantics is to be defined as the study of meaning encoded in natural languages, then syntax is part of semantics". <?page no="65"?> 53 Morphemes Complex Lexcial items/ Set Phrases Symbolic Complexity Schematicity Category Schemas Constructional Schemas (Grammar) (Lexicon) wide range of units eventually results in a continuum of symbolic structures. Fig.3.1 and Table 3.1 give an overview of the variety of symbolic units postulated in CL, which form the so-called syntax-lexicon continuum. The acceptance of such a continuum corresponds to and correlates with the negation of a completely autonomous syntax, i.e. the notion of grammar being distinct from both the lexicon and a semantic module. Fig. 3.1 Schematicity and complexity of symbolic units (slightly adapted from Langacker 2005: 108) Type of unit Traditionally known as Example complex and mostly schematic syntax determiner noun (i.e. noun phrase), noun verb noun noun (i.e. ditransitive construction) complex and (mostly) specific idioms, set phrases, proverbs in the long run, head over heels, Home is where the heart is. complex but bound morphology verb-ed simple and schematic word class, part of speech determiner, noun, adjective simple and specific lexeme/ lexicon the, good, fight Table 3.1 The syntax-lexicon continuum (adapted from Croft & Cruse 2004: 255) As will be explicated in section 3.7, complex schematic symbolic units are used in CG (and in other comprehensive, often cognitive theories of grammar, subsumed under the cover term construction grammar) to encapsulate grammatical knowledge and serve the same function as rules set up in tra- <?page no="66"?> 54 ditional and generative approaches to grammar, in that they account for a speaker's ability to combine linguistic elements in regular ways. In CG, linguistic regularities (patterns) are represented in the form of complex units which are abstracted away from and are schematic relative to instantiating expressions. The term "schema" is used to refer to such an abstract representation which captures commonalities of a range of expression, emerging through comparison and categorization, while the term "instantiation" holds for a subtype of such a schema. Schemas are postulated for all types of linguistic units, thus, one can propose a phonological schema [CVC] for units such as [p t], [s k], [mæn] etc., or the schema [NP V NP] of a monotransitive pattern for expressions such as John loves Mary, They shot the sheriff or The argument needs elaboration. 3.4 Meaning as conceptualization 3.4.1 Subjectivist and encyclopaedic semantics The CL approach to meaning embraces a subjectivist view in that the meaning of an expression involves the way a speaker chooses to think and reflect about an entity, situation or scene. Janda (2006: 11) captures CL's goals in describing meaning in a very straightforward way: It is necessary to remember that all experience is filtered by perception, and that as a consequence language is not a description of the real world (nor any possible world), but rather a description of human perception of reality. Therefore, when we examine meaning, our goal is not to find a correspondence between utterances and a world (real or otherwise), but rather explore the ways in which meaning is motivated by perceptual and conceptual capacities. (emphasis added, SH) Thus, one of the prime principles endorsed by a cognitive semantics is that meaning resides in the speaker's mind. The notions of 'concept' and 'conceptualisation' play a crucial role in this account of semantic structure, as meaning is basically identified with conceptualisation. Concepts, simply defined, are basic units of mental representation, mirroring existing properties of the world as experienced by the individual speaker. Conceptualisation one form of mental experience refers to the cognitive ability to form such concepts. It comprises novel and established conceptions, facets of sensorimotor experience, and the perception and processing of the social, linguistic and cultural context (cf. Langacker 1998: 3). With these different types of experience contributing to the formation of concepts and thereby eventually to meaning, the latter must be seen as emerging through a process of active and dynamic construction which draws on these different types of experience and resources of knowledge. For this <?page no="67"?> 55 reason, it is more appropriate to speak of linguistic expressions as evoking instead of encoding meaning. This goes hand in hand with an encyclopaedic view on linguistic semantics, which approaches lexical items merely as 'door openers' or "points of access" (Evans & Green 2006: 215) into enormous conceptual networks of stored (world) knowledge. The strict separation between linguistic and pragmatic specifications of meaning is not maintained in CL as for the full comprehension of concepts speakers draw on several different knowledge domains involving even presumably peripheral facets of knowledge associated with a concept. Different contexts of usage of a lexeme (i.e. a lexicalized concept) might call for the activation of different aspects of its meaning. Therefore, lexemes are no longer approached as dictionary-type entries with clearly delineated specifications of meaning. The difference between the 'dictionary view' of word meaning and the 'encyclopaedic view' is succinctly summarized by Evans & Green (2006: 215): […] while dictionaries represent a model of knowledge of linguistic meaning, the encyclopaedic view represents a model of the system of conceptual knowledge that underlies linguistic meaning. As has been stated above, in CL all linguistic structures of all types of complexity and schematicity are considered as meaningful, and thus both lexical and grammatical elements are exploited by speakers for their structuring of conceptual content. Langacker sums up his position with respect to the semantic value of all linguistic symbols as follows: The key to recognizing the meaningfulness of grammar lies in adopting a conceptualist semantics that properly accommodates construal […]. As an inherent aspect of its conceptual-semantic value, every lexical and grammatical element incorporates a particular way of construing conceptual content - either its own content or that evoked by other elements. (Langacker 2001: 6, italics in original) Langacker introduces the notion of construal, a set of cognitive operations which are basic for the shaping and structuring of entities in cognitive domains. These cognitive processes will be explained in greater detail in the following section. 3.4.2 Construal Construal (alternatively labelled mental imagery) as described by Langacker (e.g. 1987a, 1991a) combines a set of basic human cognitive abilities with which speakers operate when they verbally convey their specific conceptualization of an event. Construal rests on mental operations by which speakers structure a given situation, comprising their ability to "construe a situation in alternate ways for purpose of thought or expression" (Langacker 1991a: 549). As situations and events can typically be arranged in alternate ways, the choice of particular lexemes and particular grammatical struc- <?page no="68"?> 56 tures mirrors only one of different possible construals. In light of this theoretical background, the grammar of a language can be understood as constituting a stock of conventionalized imagery, with alternating structures being associates of alternative construals of a given situation. Langacker's position that "meaning is a function of both conceptual content and the 'image' imposed on it" (ibid.) undoubtedly manifests a conception of semantic structure which is opposite to assumptions of truthconditional semantics, the latter analysing concepts/ conceptual content in relation (i.e. their application) to "real world" situations as basic to and sufficient for semantic analysis. The expressions (1) a. A car hit Cathy. b. Cathy was hit by a car. might be considered as truth-functionally equivalent in referring to the same "real world" situation, while in fact they reflect different conceptualizations of the situation, expressed via different grammatical arrangements which are offered by the structural inventory of English. That is, two expressions may evoke the same conceptual content, yet they differ semantically due to the different construals they verbalize. In the examples offered in (1), this difference lies in the prominence which is given to the participants of the event, i.e. whether the attention is directed at the car (as in [1a]), describing what the car 'did', or at Cathy (as in [1b]), focusing on what happened to her. Quite naturally, it comes as one corollary that in CL alternating grammatical constructions (e.g. active and passive variants of a verbal expression) are not seen as being related by transformational procedures, but rather as conceptually motivated forms which emerge independently of each other. That is, CL negates any of these alternating forms the status of being more basic or serving as an underlying/ deep structure for transformations, which is one of the core statements of the "content requirement" or "substance principle" of CG (Langacker 1987a: 23ff., 1998: 13f.): [T]he only structures validly posited are those that figure directly in the data (i.e. actual expression and their parts) or that emerge from the data via the basic cognitive processes of abstraction […] and categorization. (ibid. 1998: 13f.; emphasis added, SH) In order to construe a given situation in alternate ways, speakers make use of so-called construal operations. A wide range of different construal operations are suggested and discussed in the CL literature, among them such procedures as 'profiling' and 'fictive motion' (as means to adjust attention or salience), 'categorization' and 'figure-ground alignment' (reflecting comparison), and 'individuation' (relating to the construal of gestalt-like structures) (see Croft & Cruse 2004: 46ff. for a summary and classification of construal operations). Some of these operations - 'profiling', 'figure- <?page no="69"?> 57 ground organization', 'metaphor' and 'categorization' which have attracted considerable attention in CL-based studies and which work on a more general level, will be portrayed below. Other, more specific procedures, such as 'summary scanning', 'end-point focus', 'reification' and 'force-dynamics', which I think to be crucial for my account of COCs, will be introduced in more detail in the relevant sections of my analysis of the construction. 3.4.2.1 Prominence A ubiquitous aspect of construal is the prominence which is given to entities and/ or substructures of a conception. One special form of assigning prominence is called profiling, i.e. the imposition of a profile onto a base. Base 7 refers to the conceptual domain which is presupposed for the comprehension of particular expressions / concepts, whereas profile refers to an inherent substructure of the base, which is 'singled out' by a particular semantic unit. The concept KNUCKLE , for example, can only be sufficiently defined against its base FINGER , and HANDLE only unfolds its full meaning in the context of its base CUP / MUG . Two linguistic expressions can have the same base but profile different aspects of it and thus differ in meaning: Trunk and twig "pick out" different regions of the larger cognitive structure TREE , and buy, sell and pay highlight differing relations within a COMMER- CIAL EVENT scenario (as described in Fillmore 1977). As Langacker (1987a: 186) notes, to consider two linguistic expressions as different profiles of the same base automatically entails the recognition of their semantic relationship. Another dimension of prominence, which is essential for the apprehension of relations between entities, is the trajector / landmark organisation. Whereas profiling is a matter of what an expression designates, the trajector / landmark alignment mirrors the focal prominence given to the participants of a profiled relation. Relations, by definition, hold between at least two participants (entities), and when we perceive such a relation, the involved participants are arranged in a particular constellation, one participant usually being more prominent than the other(s). This most salient participant is labelled trajector (tr) in CG, and the other (less) salient entities are referred to as landmarks (lm). Trajectors, being prominent in conceptualization, usually surface in a prominent position in a verbal expression, i.e. the subject position, whereas landmarks are candidates for object slots or adverbial expressions. In a FIGHT scenario, for example, involving the entities JERRY , RON and STONE , the participants can be arranged differently in a 7 The term base used by Langacker is similar to Fillmore's (1978) notion frame or Lakoff's (1987) Idealized Cognitive Model (ICM). <?page no="70"?> 58 verbal account of the scene, depending on the prominence speakers allot to these: (2) a. Jerry (tr) hit Ron (lm 1 ) with a stone (lm 2 ). b. Ron (tr) was hit by Jerry (lm 1 ) with a stone (lm 2 ). c. A stone (tr) hit Ron (lm). d. Jerry (tr) used a stone (lm 1 ) to hit Ron (lm 2 ). The examples clearly illustrate that the trajector/ landmark constellation significantly contributes to different construals of the same situation. They present only some of even more possible verbalizations, all of which differ in the arrangement of and prominence given to the entities being part of the scenario. Conceptual prominence of one entity (and thus, its status of a trajector) can arise for different reasons: As the term 'trajector' already suggests, moving entities are more likely candidates for this role than stationary objects, and smaller objects are chosen as the primary focus of attention rather than large objects (see Ungerer & Schmid 2006: ch. 4). Moreover, speakers can place salience onto entities through their deliberate choice of one entity functioning as the trajector, possibly overriding expectations and thereby creating humorous or alienating effects, as in (3) a. The desk stood under the computer. instead of the expected b. The computer stood on the desk. The account of phrasal / sentential organization in terms of trajector/ landmark alignment reflects CG's endeavour of defining linguistic categories on the basis of their conceptual content rather than on formal features. Here, the traditional notions of subject and object are captured by schematic conceptual characterizations based on focal prominence (pertaining to a basic mental ability) instead of morphological and syntactic idiosyncrasies (subject-verb agreement or word order in English). The latter are, as Langacker (2005: 130) points out, "symptomatic of the conceptual import of subject and object, not definitional". 3.4.2.2 Conceptual metaphor Conceptual metaphor, which has attracted great attention in CL over the last decades, is also discussed as a construal operation by Langacker (1987a), Croft & Cruse (2004) and Taylor (2002). Within this paradigm, the concept of metaphor has become 'elevated' from merely being a figure of speech to constituting a mental operation which reflects a speaker's ability to understand or construe one concept in terms of another. In their seminal work Metaphors we live by, Lakoff & Johnson (1980) introduce the notion of metaphor as a mechanism of the mind, recognizing conceptual metaphor as a "cross-domain mapping", by which entities or structures of one conceptual domain (target domain) are construed in terms of another (source <?page no="71"?> 59 domain). Metaphorical expressions, i.e. linguistic expressions that verbalize these metaphors, amount to 'surface phenomena' of such mappings 8 and offer insights into the conceptual apparatus which is structured by a rich system of metaphors, which in turn is to a great extent grounded in and shaped by bodily experience (especially by the sensorimotor system) 9 . To give a very basic example: In the expression He is in love, the description of an emotional state is expressed via a metaphor which maps a spatial concept to the domain of emotions (the mapping being EMOTIONS ARE CON- TAINERS ). That is, LOVE is understood as a bounded region (a container), and to experience this feeling is understood as being inside this container (likewise, one can also be out of love). The basic human experience in which this metaphor is grounded is our permanent experience of being either inside or outside such containers of the most different nature: the womb, a crib, the playpen, clothes, rooms, buildings etc. Moreover, we manipulate objects, putting them into or taking them out of boxes, bottles, cans etc. Last but not least, we are aware of our bodies as three-dimensional containers affected by 'inand outputs'. The concept of containment is based on one of the most fundamental of our bodily experiences and as such is the source of many metaphorical mappings, e.g. TIME IS A CONTAINER (He finished the task in five minutes), ABSTRACT ENTITIES ARE CONTAINERS (He was in mortal danger), ACTIVITIES ARE CONTAINERS (I am not so much into skating), to name just a few. In CL, conceptual metaphor has been investigated as a pervasive 'tool' for the extension of all types of linguistic categories (see below). The polysemy of lexemes and syntactic structures could be shown to rest to an important extent on the metaphorical extension of the core meaning of the respective lexeme or construction: Lakoff (1987), for example, gives a coherent account of the highly polysemous preposition / particle over, demonstrating how non-spatial senses (He got over this failure) form metaphorical extensions of spatial senses (He jumped over the fence); Newman (1996) offers a similarly systematic description of the meanings of the verb give. 8 In Lakoff & Turner's system, the names given to these mappings (graphically represented in small capitals) reflect the domains which are associated by the particular metaphor. The commonly cited metaphorical mapping ARGUMENT IS WAR , for example, indicates that the concept of ARGUMENT is understood in terms of WAR , as in She was bombed with questions and accusations. 9 Since the publication of Lakoff & Johnson's Metaphors we live by in 1980, innumerable studies and projects have analysed linguistic phenomena for the identification of conceptual metaphors, and investigated the relationship between human (bodily) experience and conceptual structure. The impact Lakoff & Johnson's work had on studies of metaphors and studies of the mind is impossible to summarize at this place. See, among many others, Barcelona (2000), Johnson (1987), Kövecses (1986, 2005), Lakoff & Turner (1989), Stefanowitsch (2006), Sweetser (1990) or Turner (1987). Lakoff & Johnson (1999) reveal how major tenets of Western philosophy build on and are structured by basic conceptual metaphors. <?page no="72"?> 60 Goldberg (1995) depicts how the use of argument structure constructions (in particular the ditransitive construction) can be extended via conceptual metaphor (Sasha gave them candy. Sasha kissed them a warm welcome.) In the present study, conceptual metaphor will be drawn on as an explanatory device for COCs as monotransitive constructions, being extensions of a core form of monotransitives. Accounting for the polysemy of linguistic forms in terms of a core form, extensions thereof etc. anticipates another crucial assumption of CL: the discussion of linguistic elements as categories which are structured around a prototype. Treating them as categories with the same characteristics as other conceptual categories quite naturally follows from the fundamental principle of the framework to consider linguistic knowledge as being represented in the same form as other conceptual structures. The next section will elaborate on this assumption. 3.5 Linguistic categorization 3.5.1 Linguistic categories as prototype categories Another fundamental (almost primordial) cognitive 'strategy' which underlies human conceptual organization is that of assembling concepts and knowledge structures in form of prototype categories. Prototype theory has developed as an alternative to the 'classical theory' of categories which is ultimately based on Aristotle's ideas of categorization. According to the Greek philosopher, a category is defined by a set of necessary and sufficient features which often come in form of binary attributes. Depending on the presence or absence of the defining features, an item belongs to a category if and only if it exhibits all these features. That is, membership in a category is an all-or-nothing issue, and, consequently and logically, all members of a category must be equally good members. These are assumptions, however, which run against results of psychological studies revealing that (1) subjects are flexible in extending categories, (2) given different contexts, subjects ascribe items to different categories, i.e. categories are context-dependent and may overlap, and (3) subjects judge some members of a category to be better examples, i.e. membership is graded (cf. Taylor 1995: 40 ff.) 10 . As Janda (2006: 13) wittily remarks: "Rather than having a 10 The 'famous' BIRD category shall illustrate (in a very simplified way) these basic ideas: According to the classical theory, if the feature [+ FLY ] is necessary for an animal to be categorized as a member of the category BIRD , a swallow clearly belongs to the category, whereas an ostrich, lacking this feature, does not. In prototype theory, however, an ostrich is recognized as a more peripheral member of the BIRD category, sharing <?page no="73"?> 61 defining boundary and no internal structure, human categories tend to have a defining internal structure and no boundary". The idea of prototypes originated in cognitive psychology in the 1970s and was formulated by Rosch and her colleagues (e.g. 1978), who, in turn, drew on insights of the studies by cognitive anthropologists such as Berlin and Kay (cf. Taylor 1995: 8ff.). The theory basically maintains that a category, especially a complex category, is built around a prototype. The nature of such a prototype, however, is still open to dispute; two main approaches concerning its representation can be distinguished: Originally, Rosch understood as a prototype a relatively abstract mental representation that assembles the key attributes or features that best represent instances of a given category, i.e. a schematic representation of the most salient or central characteristics associated with the category in question. In this model, the centrality of a member of a category hinges on the number of relevant features possessed by it. Put simply: The more features an item possesses, the better an example it is of the category in question. However, further developments in prototype theory led to a change in the conception of 'prototype', with scholars arguing for a prototype being represented in form of an exemplar, i.e. an ideal or best example. In such a model, the centrality of any member of the category is a function of its similarity to the ideal example, i.e. the more similar an item is to the exemplar, the more central a member it is of the category (cf. Croft & Cruse 2004: 81f; Evans & Green 2006: 249f.). As noted by Croft & Cruse (2004: 82, quoting Hampton 1997), both types of prototype representation may come 'into service': While simple concepts such as colour or shape might be more easily handled by the similarity approach, drawing on an exemplar, more complex concepts like BIRD or VEHICLE are dealt with much better by the 'key-attribute' version. On the basis of its prototype (leaving aside the question about its nature of representation), a category is extended by means of various cognitive processes (e.g. comparison, prominence, conceptual metaphor and metonymy) to gradually less prototypical members. Due to such a gradual extension, members of a category usually display varying degrees of membership and representativeness 11 . Lakoff (1987) introduced the notion 'radial category' as a label for categories in which category members are extended from a central case, i.e. all of the non-central cases are described as radiating from the central case. Originally applied to the designation of lexical items such as colour terms or names for plants and animals (i.e. nominal categories), prototype some features with the prototypical bird (which was shown to be the robin in the USA), e.g. having a beak and feathers, and lacking others, e.g. the ability to fly. 11 Recent studies have shown that categories are by no means stable constructs, but can be created and extended on-line. See Croft & Cruse (2004: 92f.), who advocate a 'dynamic construal approach' to conceptual categories. <?page no="74"?> 62 categories have come to be investigated and tested for verb categories, speech acts and eventually for syntactic categories such as nouns and verbs, and grammatical constructions. Probing for prototype effects, Fillmore (1978) examined the semantics of the verb climb, Coleman & Kay (1981) analysed the speech act verb lie, and Bates & MacWhinney (1982) studied the grammatical notion of subject (see Taylor 1995 for an overview of prototype studies). Syntactic constructions as prototype categories are discussed in Lakoff's (1987) analysis of there-constructions, Rice's (1987) account of transitive constructions, Kemmer's (1992) investigation of reflexives, Taylor's (1995) discussion of possessive and transitive constructions, and Takahashi's (1999) work on imperatives and passives. All of these studies present the respective constructions as radial categories or at least assume that they are. Taylor (1998: 198) emphasizes the necessity to capture syntactic constructions in terms of prototype categories: More complex categories, i.e. syntactic constructions, also turn out to be prototypically structured. A construction may be defined, very generally, as any established pairing of a form with a meaning […]. Given this definition, syntactic constructions need to be characterized in both their formal and their semantic aspects. Precisely because of their complexity, syntactic constructions are especially likely to give rise to prototype effects, in that examples of a construction may match some, but not all of the formal and/ or semantic aspects. And, just as with word classes and real-world categories, syntactic constructions can merge into each other at their fuzzy boundaries. (emphasis added, SH) In Chapter 5 it will be shown that COCs can be discussed in the light of prototype theory from a twofold perspective. Firstly, COCs may be considered as a family of constructions having prototypical and peripheral instances. Secondly, COCs will be described as subtypes of transitive constructions, which, if compared to a transitive prototype, display characteristics which render them as more or less peripheral forms. 3.5.2 Language usage as categorization Grouping items into larger categories on the basis of shared features or recognizing objects as instances of a particular class is a natural human impulse and elementary to cognition. Langacker (1987a: 101ff.) describes the ability to compare one entity against the background provided by another as a fundamental aspect of construal, with categorization being one of its basic manifestations (cf. also Croft & Cruse 2004, who include categorization in their list of construal operations). CL describes language usage itself as an act of categorization in two ways. First, if speakers opt to encode a particular experience by means of a particular expression, they do so on the basis of prior experiences to which the expression has been applicable (cf. Croft & Cruse 2004: 54 ) which is undoubtedly an act of categorization. Second, as described in section 3.2, CG proposes a wide array of <?page no="75"?> 63 symbolic units constituting the linguistic knowledge of speakers. These do not function as structure-generating devices, but the symbolic units on all levels of schematicity and complexity are considered as templates for the sanctioning of linguistic expressions. That is, each actual usage of a linguistic form in a particular context (i.e. a usage event 12 ) be it in language production to simply refer to an object or to verbalize a complex scenario, or in comprehension to interpret the utterance as intended by the speaker proceeds as a kind of alignment against the inventory of conventionalized structures the language user has acquired. If, for example, a speaker makes use of the symbolic unit [[ KNIVE ] / [na f]] (viz. a conventionalized structure, = (A)) to name a piece of cutlery he is holding in his hands (i.e. the usage event; = [B]), (A) is used to categorize components of [B] for the purpose of assuring the well-formedness of the utterance. Linguistic well-formedness can thus be described as a case of full sanctioning of [B] by (A), that is, [B] is fully categorized as an instance of (A). In the case of unconventional usage or ill-formedness of an utterance, [B] is not or only partially sanctioned by the conventionalized form (A) (e.g. in the case of using [[ KNIVE ] / [na f]] to refer to a letter opener) (cf. Langacker 1999a: 99ff., 2008: 224ff.). Crucial for the recognition of structures being wellor ill-formed instances of a particular symbolic representation is our ability to conceive the same content on different levels of schematicity. Therefore, in order to make judgements on the well-formedness of the utterance The ostrich swallowed an apple, for example, the language user has to recognize it as an instance of the more abstract, complex unit of monotransitive constructions, i.e. a template which sanctions its use. Issues on the emergence and representation of such schema - instance relations are taken up in the subsequent section in which I attend to the usage-based approach to language inherent to CL. 3.6 The usage-based account of language Cognitive linguists pursue an approach to language which describes linguistic structure as arising from and interacting with actual language use. The model of linguistic representation that has been developed and promoted within this framework is a usage-based one in which substantial importance is given to the actual use of the linguistic system and a speaker's knowledge of the full range of linguistic conventions, regardless of 12 Usage event is to be understood as the concrete performance of a 'linguistic act' in a particular context. It is "[t]he pairing of vocalization, in all its specificity with a conceptualization representing its full understanding. A usage event is thus an utterance characterized in all the phonetic and conceptual detail a language user is capable of apprehending" (Langacker 1999a: 99). <?page no="76"?> 64 whether these conventions can be subsumed under more general statements. [It is a] nonreductive approach to linguistic structure that employs fully articulated schematic networks and emphasizes the importance of low-level schemas. (Langacker 1987a: 494, emphasis added, SH) Goldberg acknowledges the advantage of a usage-based model in its combining "facts about the actual use of linguistic expressions such as frequencies and individual patterns that are fully compositional" with "more traditional linguistic generalizations" (2006: 25). A usage-based model maximally emphasizes the role of usage events for the representation of a speaker's linguistic knowledge, which, as was pointed out in 3.3., is assumed to consist of a structured inventory of symbolic units of different degrees of complexity and schematicity. These units can be thought of as generalizations over concrete usage events, their presentation and availability in the minds of speakers being strongly determined by the frequency a speaker is 'exposed' to the usage (i.e. production or perception) of the relevant structures: Every time a speaker activates a particular linguistic structure, its representation as a unit is simultaneously strengthened and might ultimately become fully entrenched 13 . That is, units which are activated frequently can progressively be accessed more easily, whereas units which are hardly used gradually fade and 'perish'. This way of capturing linguistic knowledge undoubtedly assumes a 'dynamic grammar' representation in the speaker's mind, where the knowledge structures are constantly under construction and subject to reorganization due to their more or less frequent activation. It also illustrates the interactive nature of such a system: Not only does a speaker activate and thereby strengthen structures of his own linguistic system when producing utterances, but he also provides input for his interlocutor(s) which contributes to the activation and strengthening of patterns in the hearer's system. CL approaches language with a "bottom-up"-orientation and thus gives a certain priority to low-level-schemas, i.e. lexically specific units whose representation in the system is itself grounded in repeated usage events. These form the source for the schematization of more and more abstract higher-level schemas and thus must be seen as the starting point for the growth of a network of structures which range from "the fully idiosyncratic to the maximally general" (Langacker 1999a: 92). The frequent usage of different, concrete instances of a particular pattern may lead to the abstraction of a template which comprises all the commonalities that are inherent to these expressions. Thus, a speaker's encounter with utterances such as 13 As Langacker explains, there is a continuum ranging from novel forms to deeply entrenched units. Even the first usage event of a form leaves some trace in the speaker's mental representation, i.e. it "constitutes an initial step along the path of progressive entrenchment and conventionalization" (Langacker 2000: 10f.). <?page no="77"?> 65 She likes bananas, They like fast cars, Jerry likes surprises results in the more general schema [[ LIKER ] [ LIKE ] [ LIKED ]], and an even more abstract schema [[ THING ] [ PROCESS ] [ THING ] as the possible result of an abstraction away from other similar, more general schemas such as [[ WANTER ] [ WANT ] [ WANTED ]] (as in She wanted a Porsche), [[ SEER ] [ SEE ] [ SEEN ]] (Alec saw the planes) or [[ EATER ] [ EAT ] [ EATEN ]] (Claire ate a whole bag of popcorn) (cf. Dabrowska 2004: 208f.). The integration of units and schemas at the different levels of abstractions is realized on the basis of categorization, and the links that hold between schemas and subschemas are established through particular categorizing relationships, which basically come in two types: First, an abstract schema (A) and instantiations (B, C; fully sanctioned by A) are linked in pairwise manner by elaboration (A B/ C) or schematization (B/ C A) links respectively. Instances B and C are related by similarity. Second, an instance D which does not completely specify a higher-level schema (but is still similar enough to B and C to be recognized as a somewhat unconventional form of A) is integrated into the network by means of an extension link 14 , which, in turn, might eventually lead to the modification of schema A to A' so that the new abstract schema then fully captures B, C and D (cf. Langacker 2000: 12f.). Through the described acts of categorization the network is constantly reshaped and/ or extended. The representation of linguistic knowledge in a usage-based network is maximalist and non-reductionist in that it permits the co-existence of concrete units alongside schematized units on several levels of abstraction. The resulting construct is a highly redundant network which allows speakers to access information represented in it in various ways. As noted above, with respect to the weight and significance of the various concrete and more or less abstract schemas in a network, priority is given to low-level forms. Langacker advocates their importance, explicating that low level schemas, expressing regularities of only limited scope, may on balance be more essential to language-structure than high-level schemas representing the broadest generalizations. A higher-level schema implicitly defines a large "space" of potential instantiations. Often, however, its actual instantiations cluster in certain regions of that space, leaving other regions sparsely inhabited or uninhabited altogether. An adequate description of linguistic convention must therefore provide the details of how the space has been colonized. (Langacker 2000: 29) Speakers seem to preferably deploy schemas which are close to actual usage events, whereas high-level schemas function as organizing entities and are available for the sanctioning of novel units which might come to occupy 'uninhabited' regions if used only frequently enough. 14 As noted above, one primary source for the extension of schemas is conceptual metaphor. <?page no="78"?> 66 Recently, the CL framework has come under criticism for not living up to exactly this basic manifesto of being a usage-based linguistics. One of the main points of criticism is aimed at an imbalance between the framework's theoretical background which clearly requests empirical research (e.g. corpus investigations, statistical surveys of language usage, psycholinguistic experiments, etc.), on the one hand, and the actual shortcoming of wideranging applications of empirical methods in studies and description offered by many of the scholars working with this paradigm, on the other (see Geeraerts 2006, Gries 2006): Grammatical models like Cognitive Linguistics and Construction Grammar in particular, which accept the existence of networks in which elements or levels may have different degrees of salience, should be particularly sensitive to differences in frequency of alternative expressions. That is to say, a usage-based linguistics needs quantification and statistical analysis. (Geeraerts 2006: 37, italics in original) Geeraerts, analysing and evaluating programs of recent conferences with a CL background with respect to the prominence given to empirical methods in the papers presented, arrives at the conclusion that although a keen and vested interest in empirical methods can be observed, one is still confronted with a dominance of traditional methods (i.e. conceptual analysis) of linguistic description (cf. ibid.: 34f.). The investigation of COCs presented in this study depicts an attempt to bridge this gap between theory and implementation by the development of a network of structural patterns which is based on the careful analysis and statistical evaluation of language data obtained from actual usage as represented in the BNC. 3.7 Constructions and construction grammars 3.7.1 Construction grammars: Common grounds As has become obvious from the discussion of linguistic units and schemas amounting to constructional templates which are arranged in schematic hierarchies, CG and other related theories known as construction grammars place constructions at the heart of the linguistic inventory. This comes as a major correlate of describing language within a usage-based theory: Since linguistic units arise by abstraction from usage events, a usage-based approach is necessarily a construction-based approach. That is, constructions are basic rather than epiphenomenal, and rules are nothing more than schematic constructions […]. A construction is merely an assembly of symbolic structure. (Langacker 2001: 4, italics in original) <?page no="79"?> 67 The role of constructions as ascribed to them in the CL paradigm starkly differs from the notion thereof pursued in other frameworks 15 . Taylor (1998: 186) summarizes the 'importance' given to constructions in Generative Grammar: Constructions are merely "taxonomic epiphenomena" (Chomsky 1991: 417), of little interest in themselves. This approach entails that speakers of a language do not need to learn the constructions in their language; they need only know the general principles that determine their grammaticality. Hence, in GG constructions are treated as the outcome of an application of general rules and principles which grasp all the regularities of a particular language and determine syntactic well-formedness. The assumption of constructions as stored patterns as part of the grammar is rejected, which is inherent in the very make-up of GG models (e.g. Government & Binding models, Minimalist Syntax model, see, for example, Chomsky 1981,1995; Hornstein et al. 2005, Radford 2004): They include as their basic components a lexicon, specifying lexical items and their idiosyncratic properties, but "excluding whatever is predictable by principles of UG or properties of the language in question" (Hornstein et al. 2005: 15), and a computational system, arranging these items in a particular way. Thus, regular complex patterns, i.e. forms which conform to the rules of a language, are claimed to be generated through the application of algorithms and built 'from scratch', with little or no room left for 'pre-packaged chunks', having gained such unit-status due to highly frequent usage. GG models of language are clearly driven by the striving for a most economic representation of linguistic knowledge, a goal which lies at the core of the Generative Grammar paradigm and which figures prominently in Chomsky's most recent Minimalist Program (1995). As was laid out in section 3.3, in the framework adopted here the concept of construction comprises each (complex) unit which can be described as a pairing of form and meaning. This is one of the ground assumptions of the several models that have been developed in the last two decades which view constructions as the basic units of language. Although the construction grammar models with a CL background 16 encompassing Langacker's 15 See Schönefeld (2006) for a concise overview and discussion of the term construction as used in frameworks outside the construction grammar models. 16 Construction Grammar, as originally proposed by Fillmore et al. (Fillmore, Kay & O'Connor 1988, Kay & Fillmore 1999) and adopted, among others, by Lambrecht (1994) and Michaelis (2004), will not be included in the discussion, since the theoretical background of this model differs considerably from the position promoted in this study. The model has its roots in generative accounts of language and aims at capturing linguistic competence rather than being usage-oriented. Striving for maximal generalization in language, its proponents eschew redundant specifications of constructions. <?page no="80"?> 68 Cognitive Grammar 17 (esp. 1987a, 1991a), Croft's Radical Construction Grammar (2001) and Goldberg's Cognitive Construction Grammar (1995, 2006) highlight different facets of the nature of constructions and differ, among other things, in their treatment of word classes and grammatical relations (see section 3.7.3), they have a common core of guiding principles, most of which have already been spelt out in preceding sections 18 : 1. constructions (i.e. any entrenched symbolic unit ) are the primary objects of description 2. the frameworks are mono-stratal / non-derivational 3. lexicon and grammar form a continuum of constructions 4. constructions are linked in networks of inheritance 5. regularities of a language are captured in the form of constructional schemas, which are abstractions of their instantiating expressions 6. a large proportion of the vast number of constructions that make up the linguistic inventory are idiosyncratic (e.g. construction idioms) in relation to regular productive / core patterns as a special case 7. a descriptive apparatus which can capture idiosyncratic forms can easily be extended to core constructions 8. providing a motivational explanation for the posited constructions is a central aspect of description 9. emphasis is put on the psychological plausibility of the postulated constructs However, despite this common ground, important differences between the models must be recognized. The most prominent distinctions between Langacker's Cognitive Grammar and Goldberg's Construction Grammar will briefly be outlined in the following 19 . The latter approach is of special 17 Broccias (2006: 108) views Langacker's theory as "by far the most comprehensive theory of grammar available in the cognitive linguistic camp", Goldberg's and Croft's models being to some extent "notational variants of Cognitive Grammar". 18 The overview is mainly adapted from Langacker (2005: 102) and Goldberg (2006: esp. Ch.10). See also Broccias (2006) and Schönefeld (2006). 19 I neglect Croft's model here since it will not be deployed for my description of COCs in the remaining chapters of the present study. For the same reason, I only mention here that one of the basic differences between Langacker's and Goldberg's model is the complexity of the units which are viewed as constructions. While Langacker reserves the term construction for complex units which "consists of two or more component structures that are integrated to form a composite structure" (ibid. 1991: 5), Goldberg incorporates morphemes in her definition of constructions, when she characterizes constructions as "learned pairings of form with semantic or discourse function, including morphemes or words, idioms, partially lexically filled and fully general phrasal patterns" (Goldberg 2006: 5, emphasis added, SH). <?page no="81"?> 69 interest here since Goldberg focuses on argument structure constructions and as such might provide a suitable framework for a discussion of COCs as monotransitive patterns. 3.7.2 Predictability and compositionality of constructions Especially the notion of predictability originally separated Goldberg's (1995) model of constructions from Langacker's (and Croft's) variants. Whereas Goldberg in an earlier version of her model considered particular patterns as constructions if and only if their meaning was "not strictly predictable from [the construction's] component parts or from other previously established constructions" (Goldberg 1995: 4), Langacker and Croft work with a more comprehensive definition of construction including form-meaning pairings that are fully predictable on the basis of their components and the principles combining them. That is, while in Langacker's account, the 'fully regular' expression Close the door is considered as a construction, since due to its frequent usage it has possibly achieved the status of an entrenched expression, in Goldberg's model, the phrase was denied a unit status since it lacks any unpredictable features and can be constructed in accordance with higher-level schemas (e.g. those sanctioning an imperative, monotransitive construction). Only expressions such as the now famous sentence Sam sneezed the napkin off the table, whose meaning cannot be arrived at via its component parts 20 , counted as constructions. In her most recent work, however, Goldberg (2006) broadens her definition of construction when she adds to the above quoted formulation that "patterns are stored as constructions even if they are fully predictable as long as they occur with sufficient frequency" (Goldberg 2006: 5; emphasis added, SH). This altered concept of construction now parallels Langacker's notion of "entrenched units" as constructions. A closely related issue is that of the degree of compositionality of those structures which are entrenched as units as compared to that of novel expressions which are arrived at through the sanction of higher-level schemas. According to the compositionality principle (postulated by Frege and promoted by formal theories of semantics), the meaning of a complex expression derives from the meaning of its component parts and the manner in which they are combined. CL discards the conception of strict compositionality since a strictly compositional meaning of an expression is the exception rather than the norm. Although it cannot be denied that in large part, component parts contribute to the composite meaning of an expres- 20 The sentence expresses 'caused motion', an aspect of meaning which is not inherent in the meaning of sneeze but is added to it by the meaning ascribed to the argument structure construction [Subj V Obj Obl path/ loc ] , which is: X causes Y to move Z path/ loc (cf. Goldberg 1995: esp. ch.7). <?page no="82"?> 70 sion, its full semantics is not predictable on the basis of these components alone 21 . Especially due to the polysemy of the majority of lexical items in a language, full predictability of composite meaning is highly questionable. What is more, the structural ambiguity of sentences and polysemous syntactic structures, idiomatic and metaphoric expressions add to the implausibility of full compositionality (cf. Taylor 2002: 97ff.). At best, the notion of partial compositionality can be advocated, which in turn, is a matter of degree. A common view is that entrenchment (i.e. unit status) correlates with a higher degree of opacity and non-predictability of an expression's meaning, i.e. the semantics of fixed expressions are claimed to be less predictable than those of novel expressions (cf. Langacker 2005: 140). With a definition of "fixed expression" as posited in CG, this claim must be questioned, as seemingly highly frequent, complex forms such I love you, I'm alright or brown hair are not less compositional than less common expressions as I love cockatoos, I'm schizophrenic or brown socks. In this respect, Langacker asserts that "fixed and novel expressions are roughly comparable in their degree of compositionality. What differs instead is merely the extent to which noncompositional aspects of their meaning are entrenched and conventionalized" (ibid.: 140). In other words, although I am alright, as a fixed form of answering the speech act formula How are you? , has become conventionalized with respect to this pragmatic function (i.e. its entrenched noncompositional component), it does not differ in degree of compositionality from the phrase I am schizophrenic, which lacks such a conventionalized pragmatic aspect. 3.7.3 Autonomous syntax vs. reductionism? There is an ongoing controversy between Langacker and Goldberg with regard to the levels of organization of constructions. In her now classic work Constructions: A Construction Grammar Approach to Argument Structure, Goldberg (1995) develops a systematic account of argument-structure constructions, most elaborately the ditransitive construction, as (unpredictable) form-meaning pairings. Argument structures are described as meaningful syntactic assemblies which specify and even add to the semantics of verbs occurring in the particular construction. Table 3.2 provides an overview on those English argument structure constructions that have been described as meaningful patterns by Goldberg so far. 21 See Langacker (1999a: 15f.) and Taylor (2002: 105f.) who demonstrate that even expressions such as pencil sharpener and the football under the table, which might appear as good candidates for examples of strict compositionality, in fact comprise facets of meaning which cannot be derived from their component parts alone. <?page no="83"?> 71 Construction / Example Meaning Form Double Object Construction Kate wrote Jim a letter. X causes X to receive Z Subj V Obj Obj 2 Caused- Motion Sam sneezed the napkin off the table X causes Y to move Z Subj V Obj Obl Resultative He bored the audience stiff. X causes Y to become Z Subj V Obj XCOMP Intr. Motion The boat sailed into the harbour. X moves Y Subj V Obl Transitive Thelma hit Louise. X acts on Y Subj V Obj Possessive Sam landed a good job. X acquires / possesses Y Subj V Obj Table 3.2 Argument structure constructions (cf. Goldberg 1998: 206) For the schematic representation of the constructions, Goldberg opts for a type of notation as shown in Fig. 3.2 for the ditransitive / double object construction, which is said to (prototypically) communicate the transfer of an object from a volitional agent to a willing recipient (or, more technically, X CAUSES Y to RECEIVE Z). The schema represents the semantic part of constructions ( CAUSE - RECEIVE , AGENT , RECIPIENT , PATIENT ) in the upper tier which is expressed through a particular syntactic arrangement (the formal part) captured in terms of syntactic functions (V, Sub, Obj, Obj 2 ) in the lower tier. By way of the middle tier ( PRED ) the construction specifies that, Fig. 3.2 The Ditransitive Construction (Goldberg 1995: 50) R: instance, PRED < > means Syn V < SUB OBJ OBJ 2 > R Sem CAUSE - RECEIVE < agt rec pat > <?page no="84"?> 72 for each verb used in the pattern, the verb and its own array of participants 22 are to be merged with (integrated into) the predicate and arguments provided by the construction. The major point of criticism by Langacker is Goldberg's postulation of a separate syntactic level expressed by syntactic categories such as 'Subj', 'Obj' etc., which - in Langacker's view - marks one step back towards the autonomy of syntax. In his model, no separate level of grammatical form is claimed to exist, but grammar is said to reside precisely in the particular symbolic pairing of form and meaning (Langacker 2005: 105). That is, what symbolizes meaning is phonological form (including the arrangement of words in a particular order), not syntax. The difference between his model of constructions and that of Goldberg is illustrated by Langacker as shown in Fig. 3.3. Goldberg, however, argues that the use of the labels 'Subj', 'Obj', 'N', 'V' is merely a means to capture the form of particular constructions in "easily recognizable" terms, the labels being "metageneralizations over construction-specific categories" (Goldberg 2006: 221). Moreover, she rejects Langacker's maximally reductionist accounts of notions such as 'object' or 'subject' (i.e. his completely reducing it to symbolic structure), and adopts a non-reductionist view, claiming that "there are often interactions between parts that lead to emergent properties that can only be described Fig. 3.3 The different layers of constructions (Langacker 2005: 105) 22 Goldberg distinguishes between participant roles of the verb, which are associated with a verb's frame semantics, and argument roles, which are part of the construction (cf. ibid. 1995: 43). Semantic Structure Phonological Structure Semantic Structure Grammatical Form Phonological Structure Symbolic Structure Symbolic Structure a) Cognitive Grammar b) Construction Grammar <?page no="85"?> 73 at the level of the whole" (Goldberg 2006: 222) 23 . To support her argument, she points to construction-specific properties of semantic roles, which cannot be discussed independently of the particular construction, and Croft's claim of syntactic roles being construction-specific (ibid.). With respect to a description of COCs, Goldberg's view seems plausible, since a cognate object can only be accounted for in relation to the verb it is a cognate of, and thus cannot be described independently of the construction in which it occurs (even if it might be possible to describe the notion of 'object' in a reductionist fashion, i.e. as the secondary focal point of a given scenario). In relation to her criticism of Langacker's reductionist account of syntactic roles and word classes, Goldberg also explicitly rejects Langacker's 'essentialist definitions' of notions such "object", "subject", "verb", "thing", i.e. his claim that these categories can fully be defined conceptually. Neither she nor Croft have adopted such an 'essentialist view', for, as they remark, no independent evidence (i.e. reliable speaker intuitions or psycholinguistic experimentation) could be adduced so far (Goldberg 2006: 223). 3.7.4 Lexical polysemy and constructional meaning One crucial corollary of Goldberg's account of argument structure constructions is the reduction of semantic content of verbs to one or a few basic senses, with the full semantic specifications added by the respective constructions in which the verb appears. Thus, for kick, which can be used in several different argument structure constructions (some of which are illustrated below, cited in Goldberg 1995: 11), only a basic sense, e.g. "strike with the foot", is necessary for a sufficient semantic representation of the verb, coming along with two participant roles, an AGENT and a PATIENT . Additional aspects of meaning (as described below in brackets) and additional argument arrangements are claimed to be inherited from the construction as a meaningful schematic unit: (4) a. Pat kicked the wall. b. Pat kicked Bob black and blue. (resultative) c. Pat kicked the football into the stadium. (caused motion) d. Pat kicked Bob the football. (ditransitive / transfer) Under an opposite view, as taken, for example, by scholars with a lexicalist 23 The discussion whether grammar must be described as a reducible set of entities and structures is metaphorically illustrated by the composition of water. Whereas Langacker (2005: 105) states that in order to recognize the nature and structure of water, one must acknowledge the existence of its elements hydrogen and oxygen, Goldberg (2006: 222) counters that no such elementary, reductionist account of water is going explain "why water is wet, nor why it is used the way it is". <?page no="86"?> 74 background 24 , kick would be considered as a highly polysemous verb. Several different entries for the verb would need to be posited, each entry being associated with a different lexical representation. Langacker (2005), too, objects to Goldberg's semantic minimalism, for several reasons. Firstly, he regards Goldberg's description as an attempt to minimize lexical polysemy, and more generally, as an endorsement of a parsimonious view on grammar (i.e. in the sense of "the best grammar is the shortest grammar", Langacker 2005: 150). Within such a grammar, the postulation of different senses of kick would be highly 'un-parsimonious' since they could be retrieved from the respective construction in which the verb is used. As has been outlined above, CG assumes a highly redundant representation of linguistic structure and discards notions of representational parsimony since a maximally economic grammar may not be psychologically realistic. Secondly, Langacker questions the plausibility of strictly distinguishing between lexical and constructional meaning. Rather, he suggests, the relation between word meaning and constructional meaning should be viewed as a continuum: Whereas in some cases, the meaning of construction and verb is completely congruent (as is the case with send and the causedmotion construction, in I send a letter to the editor), in other cases the construction clearly adds meaningful components to a verb's meaning (as in the famous sneeze-example). In between these two poles are verbs that are on their way of developing new senses as they become more and more entrenched in association with a particular construction. According to Langacker, this might be the case for the caused-motion meaning of kick, a verb which is not too unfamiliar in the caused-motion construction. As will be shown in the chapters to follow, verbs occurring in the COC reflect such a continuum. For some CO-verbs, the use in a monotransitive construction is fully congruent with their inherent semantics, whereas others acquire additional semantic specifications through their use in such a construction. In sum, Langacker does not deny the significance and validity of the meaningfulness of argument structure constructions as posited by Goldberg, but he warns of a minimalist lexical semantics, reducing verb meaning to a minimum and maximizing constructional meaning. 3.8 Conclusion The purpose of the preceding chapter was to present Cognitive Linguistics as an outstanding model of linguistic description and to give an overview 24 A more recent tendency in generative grammar is to focus on lexical items and their projection into syntactic structures, and to study those principles whereby the semantic representation of the lexical items determines the syntactic environment in which they can occur (see Levin & Rappaport 1995, 2005; Pinker 1989). <?page no="87"?> 75 of those principles, concepts and tools which are crucial for a description of COCs as unique pairings of form and meaning. I have pointed out that as one of the core principles of the CL framework one finds the commitment to a psychologically plausible description of linguistic phenomena, a subscription to the study of language as a facet of cognition which cannot be explored in isolation from other cognitive abilities and human experience, be it bodily, social or cultural. Linguistic structure is assumed to be grounded in and shaped by these abilities and experiences. Moreover, it has been shown that instead of strictly separating between syntax and semantics, linguists working with the CL paradigm advocate a model of language in which all linguistic elements and structures are meaningful, the semantic content being specific in some cases and highly schematic in others. This view amounts to a model of linguistic structure which builds on constructions, i.e. a repertoire of form-meaning pairings at all levels of complexity and specificity which speakers rely on for the organization of conceptual content and the sanction of (novel) linguistic expressions. This repertoire of templates originates in and grows through concrete usage of forms by speakers and is subject to constant modification and/ or extension In the following chapters, COCs will be explored as inherently meaningful patterns, grounded in basic human experience. Hence, they will be described as complex symbolic units which allow speakers to construe an event in a particular fashion, rendering certain facets of a processual scenario as particularly prominent. They will be presented as constructions which display the full scope of the specific-schematic scale, ranging from the fully idiosyncratic to the maximally schematic. <?page no="88"?> 76 4 A Cognitive Linguistic description of COCs: General issues 4.1 Introduction This chapter presents a general discussion of issues which are related to the conceptual content of COCs, starting out from a brief description of the construction as given by Langacker (1991a). In the first part of this chapter the focus will be put on the cognitive abilities which speakers possibly exploit for the understanding of a scenario verbally reflected in a COC. This part of the analysis also tackles the problem of a systematic classification of possible different types of cognate objects and the question whether there is such a thing as a cognate subject. Furthermore, the issue of how to treat the CO with respect to its grammatical purpose will be attended to from the perspective of CG. As summed up in Chapter 2, the syntactic status of the CO has been a controversial issue in the history of studies on COCs, with some scholars considering it a true object, whereas others describe it as an adjunct or adverbial of the verb phrase. In the approach presented here, COCs are dealt with as monotransitive constructions, i.e. the CO functions as the direct object (a primary landmark). The question which naturally comes along with this conception is how the adopted theoretical paradigm can explain the occurrence of a direct object with verbs which are most often associated with one participant only, i.e. 'classical intransitive' verbs such as die, smile or dream. In order to answer this question, the possibilities which Langacker's CG framework offers for such an explanation will be investigated and compared with the potential solutions provided by Goldberg's model of argument structure constructions. The latter allows for a description of verbs used in an argument structure arrangement which seems unusual for them - a characteristic which obviously holds for most of the verbs described as occurring in a COC in the respective literature. Finally, the phenomenon of COCs will be described against the background of iconicity in language. I will discuss the specific form of COCs as being motivated, i.e. as mirroring a particular load of conceptual content and/ or a particular relation or constellation of the conceptualized entities (objects, events 1 ). 1 Here and henceforth, I will understand 'event' as a sub-type of Langacker's category 'process', with 'event' referring to all non-stative forms of complex temporal relations (cf. ibid. 1987a: 244ff.). <?page no="89"?> 77 4.2 The meaning of COCs In Foundations of Cognitive Grammar, Langacker briefly comments on the COC, citing it as evidence for a speaker's flexibility of coding and construal (1991a: 364f., see also Langacker 2003). He views the structure of the construction (e.g. Alice yawned a big yawn) as deviating to some extent from the expected grammatical structure (Alice yawned) and as such as a less typical, marginal instantiation of a transitive construction. According to Langacker, two factors enable such a deviation from the more 'central', expected pattern, viz. […] our capacity for conceptual reification, which allow an event to be coded in nominal form; and the tolerance of redundancy, up to and including full overlap between the components of a complex expression. (ibid., emphasis added, SH) It is the combination of these two factors that make the COC such an interesting construction type. Conceptual reification is a particular mode of construal, based on a speaker's capability of construing a complex event, which is built up of several component states, as a gestalt-like, (temporally) bounded entity. Finite verbs (verbalizing events) usually encode a complex temporal relation, in which a process is presented as unfolding through time, i.e. its components states are depicted as successively transforming from one state into another. Such a manner of construal is defined as 'sequential scanning'. The second form of scanning, labelled 'summary scanning', renders the relation of the event's single component states as an atemporal constellation. Here, the components are viewed collectively, forming a single gestalt (cf. ibid.1987a: 144f., 248f., 2008: 119f.). A combination of summary scanning and shift in profile is at work in nominalization (i.e. the conceptual reification of a verbal process 2 ), giving rise to a particular type of noun - the episodic noun, which designates one episode of the process expressed by the verb. Thus, a smile, for example, is the result of reifying the complex temporal relation designated by the verb smile. As a consequence, the verb and its corresponding episodic noun evoke the same conceptual base; yet, the profiles which are imposed on this base by verbal and nominal form in each case differ: While for verbs, the temporal succession of the component states is prominent, indicated by the bold bar along 2 Construing actions and events as concrete objects is basic to human conceptualization, and verbal expressions reflecting this form of construal can be found easily: a) But on the run to the post Bassetja was passed by Velma. (K2D: 2733) b) [T]he height of his jumps coupled with the careless speed of his preparation was breathtaking. (A1D : 80) c) [T]hey are not going to stay up unless they wait until the shower has passed. (A0H: 902) <?page no="90"?> 78 the time arrow in Langacker's diagrammatic representation of processes, for nominalizations we find the characteristic shift of the profile to a thing (in the case presented here the process itself is reified), the focus now being on a bounded entity, containing the abstract region which is occupied by a set of component states. In this scenario, all temporal relations are 'cut out'. The conceptual relationship between verb and its nominalised form is depicted in Fig. 4.1 3 . Fig. 4.1 Reification (Source: Langacker 1991a: 24) The fact that verb and noun share the same conceptual base explains the redundant nature of COCs, which Langacker mentions as the second factor characteristic of the pattern. The question whether "redundant" must be equated with "void of meaning" is one of the issues of this study and will be a motif throughout the discussion on the functional load of COs in later sections of this chapter (see 4.5). As has been stated above, reification as a cognitive operation renders the referent as an object or a mass, and as such can be construed as the participant of an event. Note that if a speaker wanted to focus attention on a process as a complex, reified event alone, another construction type would perhaps express such construal much more naturally, viz. the light verb construction (LVC) 4 . Consider the following examples: 3 Following Langacker's mode of notation, temporal relations are depicted as a sequence of component states, which in turn involve one or more participants: These are indicated by a small circle (for the trajector) and a small square (for the landmark), and the relation which holds between them by a solid line. The single component states evolve along a temporal axis. Structures of a base which are profiled are represented in bold print. For further information on how to interpret the diagrams depicting conceptual content, see especially Langacker 1987a: 244ff. 4 It needs to be noted, however, that not every COC can be roughly paraphrased by a LVC, and that the particular light verb that is chosen also adds some nuance of meaning (see Ch. 7). Verb Cognate-Object Nominal <?page no="91"?> 79 (1) a. He smiled a forced smile. b. He gave a forced smile. (2) a. They fought a furious fight. b. They had a furious fight. The comparison of these constructions undoubtedly shows that one of the specific features of COCs is their double encoding of conceptual content, i.e. verb and object profile different facets of the same conceptual base. It may be claimed that COCs and LVCs are at two opposite poles as regards the semantic relationship between verb and object. Whereas in LVCs the main semantic content is not carried by the light verb but by the noun functioning as the head of the direct object, in COCs the object head repeats aspects of the semantic content of the verb. Both constructions have in common that to some extent they can be expressed with a syntactically simpler alternant, e.g. to fight furiously (to have a furious fight > to fight a furious fight), which, however, exclusively presents the event as a temporal sequence of its components states; the concept of a holistic presentation of the event is absent. With these rather general observations and introductory remarks on the semantic characteristics of COCs I will now turn to a close-up analysis of the features of different types of COs. 4.3 Events, results, landmarks: Different types of Cognate Objects The Cognitive Linguistic description of COCs given so far has remained unspecific about the nature of the reified object. What exactly is the outcome of the reification process? Or, to put it differently: Are all COs reifications of the process designated by the verb? In the following, I will explore the complex character of the semantic relationships that can hold between verb and its CO, for it allows us to set up a general, CL definition of "cognate object". 4.3.1 Eventive Object vs. Object of Result The literature on COCs offers two different viewpoints on the semantics of COs. One camp argues for its interpretation as an object of result (e.g. Macfarland 1995, Mittwoch 1998), that is the outcome of the action specified by the verb, while other researchers consider it as an eventive object (Massam 1988, Moltman 1989), i.e. the noun presents the event itself. Such a controversy can be readily solved within CL in describing the COC as a polysemic construction in which the different readings of the CO as either event or result of the event are linked through construal transformations, <?page no="92"?> 80 such as end-point focus, metaphor or metonymy (as in the case of a result standing for an action). Let us first (re-)consider a possible CL description of eventive COnominals. Following Langacker's description, it has been argued above that nominalization can be described as the result of summary-scanning the single components of a process and shifting the profile from a process to a thing by 'removing' any kind of temporal relations. A depiction of the CO as illustrated in Fig 4.1 is based on these modes of conceptual processing and surely comes close to an eventive interpretation of such nominals, comprising the single component states which add up to the event. Also, the notion of a succession of the single components of the event is still present to some extent in such a conceptualization, which, however, is not to mean that the single states are scanned sequentially (which would contradict the process of summary scanning). Rather, it suggests that a certain degree of dynamicity is still inherent in the conceptualization of an event as a thing. Therefore, to better capture the nature of an eventive CO, the view on event-nominals (i.e. reified processes) presented so far must be slightly adapted: Instead of Langacker's model of event nominals, which is characterized by a complete suspension of the temporal features of the related process, I propose a description in which these are still preserved in a weaker form, yet no longer figure prominently (as is the case for processes) 5 it is still the profiled bounded region occupied by the component states which is most characteristic of this model of event nominals. This altered representation of eventive COs will serve as the basis of any further discussion in this study. The result reading is closely related to the event reading and becomes available through a conceptual 're-profiling' of the reified event: A procedure which can be described as a non-spatial equivalent of the imageschematic transformation of end-point focus (Lakoff 1987, Johnson 1987) enables speakers to interpret the CO as a result of the process. In the cognitive domain of SPACE , one of the most basic image schemas is that of PATH , which emerges out of our daily experience of moving from one point to 5 The difficulty of properly depicting eventive COs by means of summary scanning mirrors recent discussions on the status and usefulness of the two modes of scanning. Broccias & Hollmann (2007) see a theory-internal relevance for the postulation of these mental operations (in allowing for the distinction between processes and atemporal relations) yet doubt their psychological plausibility, especially due to the lack of supporting external evidence. The "weakened" form of summary scanning hypothesized here which preserves temporal relations in event nominals (i.e. things) somewhat corresponds to the suggestion by Matlock (2004) that even nominals might involve sequential scanning. See also Egan (2008) for a discussion of a third mode of scanning, viz. recursive scanning, which he introduces for an analysis of -ing-forms which apparently cannot be adequately described by either of the "original" scanning modes. <?page no="93"?> 81 another. The basic elements of the PATH schema are: a) a starting point, b) an endpoint, and c) a sequence of connected, intermediate locations (cf. Johnson 1987: 113). Fig. 4.2 The PATH image schema (Source: Johnson 1987: 114, adapted) Many spatial prepositions have this PATH schema as an integral part of their meanings: over, through, into, around 6 etc. all designate a specific type of path. However, in some usages, these prepositions only designate the end-point of the path, as displayed in the following examples: (3) a. She came from 'over the hill', from the higher part of Littondale. (CB5: 69) b. There's a café that stays open a Greek place, just around the corner… (H0F: 3477) c. My case is at a friend's house down the street. (CN3: 2090) The end-point sense of the prepositions is linked to the more central sense (describing the path from source to goal) by an image-schematic transformation called end-point focus, a cognitive operation which enables us to "follow […] the path of a moving object, and then focus on the point where it comes to rest" (Johnson 1987: 26). A B Fig. 4.3 End-point focus ( SOURCE and PATH "faded out") 6 A whole canon of literature has been written on the polysemy of prepositions since the early days of Cognitive Semantics. These studies explain the polysemous character of preposition on the basis of image schemas, image-schematic transformations, metaphoric and metonymic links etc., resulting in networks of interrelated senses of the respective preposition. For case studies, see, for example, Boers (1996), Brugman (1988), Herskovits (1986), Lakoff (1988), and Tyler & Evans (2004). A B PATH PATH <?page no="94"?> 82 Drawing on the insights of cognitive metaphor research which revealed that human beings map experience from the domain of SPACE to the domain of TIME and as one consequence of such a mapping perceive of AC- TION (temporal) as MOTION (spatial), we can plausibly suggest a transfer of the spatial PATH schema to events, which are conceptualized as unfolding along a TEMPORAL PATH 7 . All the component states of an event are thus equivalents of what Johnson describes as "contagious locations" in a spatial scenario. By means of the image schematic transformation "path-focus to end-point-focus" the event receives a result reading. Accordingly, the previous component states of the event expressed by the verb are left schematic, only the final, resultant state (the state coinciding with the end-point of the temporal path) is profiled, as illustrated in the below diagram 8 : Fig. 4.4 End-point focus transformation: object of result The two subschemas of event reading and result reading thus can be said to stand in a metonymic relationship, more specifically, a part-whole relationship, with one facet of an event guiding the user to the whole event. A sentence such as Anna smiled her smile therefore activates two possible interpretations: The CO either presents a reification of the whole event (based on a somewhat weakened form of summary scanning) with all component states being evoked simultaneously, or the final outcome of the described event is profiled, i.e. the focus is put on the final component state, with the preceding states being only schematically present. A CO associated with the event interpretation will be referred to as Event CO (EV), a CO with a result reading as Event-Result-CO (EV/ R 1 ). Both forms, however, have in common that they allow for the construal of an active 7 This metaphor has been implicitly employed already in all those figures which make use of a time line. 8 In order to display the thing-like nature of the result-nominal and to indicate the contribution of conceptual reification an ellipsis encircles all of the event's component states - both schematic and focused on. The prominent focus lying on the final component is represented by bold print. <?page no="95"?> 83 involvement of the subject (most likely an AGENT or EXPERIENCER , see Ch.6) in an action by focusing on either the 'creation' of the whole event or its final result. As regards the availability of the two interpretations of the CO, it should be noted that, in many cases, the respective reading of the CO as either Event or Event-Result form is determined by a modifying element within the CO-NP: According to Iwasaka (2007: 10, ftn. 16; 20) so-called 'property adjectives' (happy, small, merry, sad) facilitate a result reading, while adjectives denoting processual characteristics such as temporal extension (sudden, fast, brief, quick, slow) cause the construal of a CO as an event. Thus, (4a) is likely to induce a result interpretation, b), however, will most probably evoke an event interpretation: (4) a. Guido smiled a small smile devoid of humour. (JXT: 154) b. 'Think about it, anyway.' He smiled a quick smile. (H97: 3387) Yet, for COC expressions which combine property adjectives and adjectives denoting processual characteristics (5a), the ambiguity of the CO may remain unresolved, as is the case for expressions including unmodified cognate nouns where the absence of modifying element yields an 'equibiased' (and thus ambiguous) reading (5b): (5) a. Vitor smiled the slow crooked smile which had once had his female fans attempting to push their telephone numbers sometimes accompanied by intimate items of underwear into his hand. (JY9: 305) b. He grinned a grin, then went back in and pulled out number three. (www.poetry4kids.com/ poem; accessed 09/ 09/ 2008) Moreover, aspectual characteristics of the whole construction may determine one particular interpretation. Apparently, COCs used in a progressive form facilitate an event reading (even if the CO-NP contains a property adjective), while those in a simple tense form allow for both readings: (6) a. I saw Signe standing among the red-roofed airport huts even before we landed, and while we taxied in she was running and waving and smiling a gigantic smile. (HR7: 14) b. Ronni smiled a light smile, seeking to pour oil on troubled waters. (JXT: 223) The latter observation on the interrelation between CO construal and aspectual characteristics anticipates points of discussion which will be addressed at length in Chapter 7. <?page no="96"?> 84 4.3.2 Landmarks and products Besides COs verbalizing the event or the result of an action, the literature offers descriptions of variants of COs whose conceptual content conforms to neither of the described forms: (weave) a web, (build) a building, (sow) a seed denote concrete entities instead of abstract, reified concepts. In section 4.2.2 I described the specific nature of COCs, one aspect of it being the combination of two different profiles of the same conceptual base - one related to the verbal and one related to the nominal constituent. Event reification, as demonstrated above, is just one particular form of nominalization. There are several other, more basic types of nominalization which merely "shift the profile of a verb to some nominal entity evoked as part of its inherent structure", as Langacker (1991a: 23) observes. The parts on which the profile can be shifted stand in various relationships to the base: They can originate as the verbal event's trajector (paint - painter, host host), its landmark (internal object) (choose - choice, feed food), its product (mark - mark, paintpainting) or its location (bake-bakery, dine diner) (cf. ibid). (Although Langacker does not make it explicit, the distinction between landmark [internal object] and product seems to correspond to the distinction between affected and effected object.) As in all of these relations verb and noun share semantic structure 9 , I will include in my discussion of COCs not only instances in which the CO denotes an event or the outcome of such (i.e. the object of result), but also such constructions as illustrated in the following examples: (7) a. Real plants should be planted with warmed water in the tank. (C95: 539) b. Transfer embryos or cells to acetone-cleaned slides, fix by dropping a single drop of ethanol. (EV6: 563) c. ABU, Mitchell, Ryobi and Shimano all produce excellent products. (HJE: 365) d. Don't draw such good drawings. (KD0: 10990) (7a) and (b) offer examples in which the CO is an internal landmark of the verb, whereas in (c) and (d) the COs designate products. The difference between objects profiling events and results on the hand, and objects denoting landmarks and products on the other, is that the former require some type of scanning of the component states of the verb, whereas the latter profile an inherent substructure of the event denoted by the verb, 9 The criterion "sharing of semantic structure" holds for any kind of verb-noun derivation, yet, as will be spelt out below, the particular type of semantic relationship between verb and noun further constrains the group of possible COs. <?page no="97"?> 85 Fig. 4.5 Nominalization (Source: Langacker 1991a: 24; see also Taylor 2002: 271) with any notion of a temporal sequence of component states being absent. Fig. 4.5 depicts the process of nominalization by means of elaboration of one of the inherent substructures of the verbal event. In CL, morphemes are considered units of minimal symbolic complexity. In the case of nominalization, a so-called nominalizer ( NR , any kind of morpheme deriving deverbal nominals), is 'at work'. The standard 10 of a nominalizing morpheme is a schematic process and its target profiles some substructure (e.g. tr or lm) of the process (cf. Langacker 1987a: 473). When NR combines with the verb, the verb elaborates 11 the schematic process in the profile of NR . The affix -ee, for instance, is a NR whose target prototypically profiles the PATIENT of a process. If -ee, serving as the profile determinant (traditionally 'head'), is integrated with a verb stem, the resulting composite structure designates the lm, most likely the PATIENT , of the activity. In the case of possible product and landmark COs, it is the lmsubstructure of the verbal event which is profiled as a result of nominalization, i.e. NR 's schematic target is the landmark of the process. Note that since NR s with different phonological content have a verb's lm as their target, I do not make restrictions to the form of NR contrary to e.g. Macfarland, who only 10 Langacker (1991a: 553f.). uses the following terminology: In a conceptually dependent structure, such as derivational morphemes, if seen as a function, 'standard' refers to the function's input, serving as an e-site for the autonomous component (in the case of deverbal nominals, the verb stem), whereas 'target' describes the function's output, corresponding to the composite structure and determining its organization. 11 See section 4.4.1 for a more detailed description of terms such as elaboration or e-site. tr lm N tr lm V NR (-ee) <?page no="98"?> 86 considered zero-derived nouns or nouns derived by stem alternation as possible CO candidates. Therefore, in the present study record a recordØ, pray a prayer, paint a painting, arrange an arrangement or decide a decision 12 will be discussed as instances of COCs. Note furthermore that in cases in which the CO is a nominalized form of those conceptual structures that relate to a verb's inherent landmark or product, the verb already provides an argument slot for the role specified by the CO. These forms present cases in which the number of the participants inherently associated with the event designated by the verb corresponds to the number of argument slots provided by the monotransitive construction (see section 4.4.3). 4.3.3 Derivational directions Another issue, concerning the morphological relation between verb and CO, must be taken up here. So far, the description of possible COs has focused on items which are nominalisations of facets of the verb's 'conceptual inventory' arrived at via a combination of (weak) summary scanning of its component states and profile shifting (event object), an additional endpoint focus transformation (object of result), or a shift of the profile to the verb's primary landmark (product, inherent lm). That is I have discussed deverbal nouns only, the direction of derivation being V N. However, how are we to treat combinations of verb and object in which the verb is derived from the noun? The following sentences present such instances: (8) a. Can I borrow this cup now to fish these fish? (KBW: 105) b. Then he nailed two six-inch nails up his nose. (CK5: 527) c. Nineteen farms were visited of which three were farmed full-time. (ALC: 72) d. […] they all counted the steps Ellen stepped. (CFY: 120) Should the direction of derivation serve as a definitional criterion for COCs, as suggested, for example, by Macfarland (1995) or Huddleston & Pullum (2002)? Of course, in treating only nouns designating objects of result as true COs, Macfarland predetermines the direction of derivation as V N, as this derivational relationship is correlated to the fact that results quite naturally presuppose actions and processes. The narrow constraints on possible candidates for COs with respect to the semantic relation between verb and object (as postulated by Macfarland) have been abandoned in this study in favour of the less restrictive specifications of COs spelt out in the above section, which provides a sys- 12 Instances of such forms are attested in the BNC; see Ch. 5. <?page no="99"?> 87 tematic account of conceptual relationships between verb and CO. However, the semantic relation between nouns and their denominal verbs is not as straightforward as with verbs and their deverbal nouns. The latter can be captured more easily, since, as Langacker claims, the nominalization of conceptual structure of verbs does not alter their conceptual content to a great extent: It is normal for a language to display patterns of nominalization whereby virtually any verb can be nominalized with no essential modification of its conceptual content.[…] [T]he semantic contribution of nominalization is limited to profiling -an aspect of construal. (ibid. 1991a: 25; emphasis added, SH) This is a logical consequence of the fact that verbs, profiling processes, are conceptually dependent on the participants of the process. Nouns, on the other hand, are conceptually autonomous; we can conceive of a THING without necessarily construing it as a participant in an event. Therefore, the derivation of verbs from nouns always brings with it an addition of conceptual content. The basic concept of [ FISH / THING ], for example, does not include any aspects of 'catch'. Notions such as 'spawn', 'ocean', 'seafood' or 'angle' might be evoked by [ FISH ] as part of a speaker's encyclopaedic knowledge, but are not as intricately linked to it as is, for example, the notion of 'triangle' to the concept of [ HYPOTENUSE ]. Nevertheless, the noun from which the verb is derived is an integral part of the verb’s semantic content: fish / [ PROCESS ] can be paraphrased as 'catch fish [ THING ] ' 13 , just like farm / [ PROCESS ] means 'to operate a farm [ THING ]'. In these cases the noun is incorporated into the verb as one of its basic landmarks, resulting in a conceptual overlap if the two are combined. Therefore, I see no reason for excluding combinations of denominal verb + noun from my analysis of COCs and do not make the direction of derivation, i.e. V N, a defining feature of the construction. Note, however, that 'cognate combinations' of verb and noun in which the verb is derived from the noun are less common than combinations of verb and deverbal noun. In the corpus of COCs which I have compiled for further analysis by means of a lexical search (see Ch. 5), only 26 % (i.e. 28 of 109) of the different types of verb-noun combinations involve pairs with the derivational direction N V (as quoted in the OED). With the exception of the pair breathe-breath, all of the denominal verbs in this corpus are materially unmarked, converted forms, yet, since the examples were retrieved on the basis of a lexical search (which is far 13 Clark & Clark categorize fish both as LOCATUM verb and CROP verb. (cf. Clark & Clark 1979: 771; 780).The former is derived from a noun (the parent locatum) which designates an item to be "moved" from a certain location, as in to fish the pond, which can be paraphrased as 'cause fish to be not in a specified location [i.e. the pond]'. The second meaning is equivalent to the representation offered here, i.e. the noun fish, from which the verb is derived, is the primary lm). <?page no="100"?> 88 from exhaustive), it can only be speculated that conversion is the defaultcase for derived verbs occurring in a COC. 4.3.4 Types of COs - A summarizing overview To sum up what has been said about the form and content of COs, I put forward the following specifications, which are at the same time criteria for the COs to be included in the further course of my study: In combination with verbs that lack an inherent landmark e-site, i.e. those denoting events which are inherently associated with one participant only, the CO designates a reification of the process profiled by the verb or the result of this process. With verbs that do have an inherent landmark e-site, i.e. verbs that prototypically occur in transitive constructions, the CO constitutes a nominalised form of this landmark. This also holds true for those combinations of cognate forms in which the verb is derived from the noun, most likely converted forms. No restrictions are placed on COs with respect to the phonological form of the nominalising element NR , as long as its schematic target is the lm of a process. From all this it can be concluded that the label 'COC' is given to all those patterns in which a verb is combined with a noun with which it shares form and a certain conceptual content, as specified above. Since the nature of the structure which is shared by the two components (i.e. the semantic relation between verb and CO) is of various types, we must expect that the emerging category of COs (and thus the whole construction) presents itself as complex and heterogeneous. In accordance with the conventions of CL to depict complex categories as networks in which category members are related through categorizing relationships such as instantiation or metaphoric and metonymic extension, the category COGNATE OBJECT comprising the different types of COs might be captured as shown in Fig. 4.6. The two major subgroups of COs that I have identified, are, firstly, those that denote a reified form of the event expressed by the verb, and secondly, those that designate a more or less inherent lm of the event referred to by the verb. These two groups are lower-level forms of the most abstract and general schema 'CO', each comprising further lower-level schemas that are instantiations of the more general level. The two subcategories of 'Reified CO' are linked via another type of relationship: The type 'Event-Result-Objects' can be rendered as a metonymic extension of the type 'Event Objects', drawing on the image-schema transformation 'end-point focus' (as described in 4.3.1). Note that for the different types of COs there is a decline in the degree of conceptual overlap between verb and object: While Event-COs are close to fully congruent with their verbs, Event-Result-COs only share the final <?page no="101"?> 89 Categorizing relationships: Instantiation : Extension: Fig. 4.6 Network of COs types component state with their respective verbs. Lm-COs display the lowest degree of conceptual overlap with their verbs, fully coinciding only with the landmark of their verbs. Since it is the members of the class labelled 'Reified COs' which are generally discussed as COs and which seem to be most problematic for a description of argument structure, the main focus of further points of discussions will be on this group (which does not mean that the other groups will be completely excluded from the analysis). 4.3.5 Cognate Subjects? The description of the semantic and morphophonological relationships between verb and CO worked out in the preceding sections calls for further considerations. First of all, it should be obvious now why, for example, to cook a cook cannot count as an instance of a COC. In this case, the object of cook is a nominalisation of the verb’s inherent trajector, and not, as defined here for COCs, its landmark. This, in turn, leads us to a brief excursus on the 'counterpart' of COCs - the cognate subject construction - and the question whether it would be justified to analyse the below example as an instance thereof: COGNATE OBJECT Lm-COs Inherent lm (affected) Product (effected) Reified COs Event-Result (focus on final outcome/ state) Event (holistic) Drop a drop Feed some food Fish some fish Weave a web Paint a painting Smile a smile Yawn a yawn Smile a smile Yawn a yawn <?page no="102"?> 90 (9) The cook cooked a delicious Mediterranean dish. Only few grammarians acknowledge the phenomenon of cognate subjects at all, and the definitions they offer are rather vague. Quirk et al. (1985: 750) describe this construction type as a case "in which it is the meaning of the subject that is presupposed by the verb", without any further explication of what they understand by 'presupposition'. Langacker only offers one example (see below), commenting on it as a construction "in which the subject and the verb redundantly convey the same conceptual content" (ibid. 1991a: 366). That is in a cognate subject construction, the trajector designates an entity which - unlike the AGENT subject in (9) depicts a conceptual reification of the process expressed by the verb: (10) a. The frost froze hard. (Quirk et al. 1985: 750) 14 b. Rain rains. (Langacker 1991a: 366) c. Still the thunder thundered and burned and shook them. (CJA: 2036) d. […] and the winds blew and the hail hailed and the rains poured down causing floods. (community2.webtv.net/ pantheralex/ LATESTNEWSV/ page2. html; accessed 09/ 05/ 07) Note that Langacker explicitly comments on the processual character of the noun rain in the cognate subject construction and not its profiling of the physical substance. This makes it different from rain occurring in a COC: (11) It's raining rain from the skies, and it's raining real tears from my eyes 15 . Here, in contrast to the cognate subject, the cognate object refers to the physical substance rather than to the process. The clausal subjects in the examples offered by Quirk et al. (cf. ftn.14) are also of an eventive character. If we compare the instances of Cognate Subject Constructions drawn on in the literature and analogous examples, it seems that the most likely candidates for such expressions are those nouns and verbs that encode weather and other natural phenomena. These 14 The other examples of cognate subject constructions offered by Quirk et al. include subjects that are only semantically but not morphologically related to the verb: a) The day dawned. b) The wind is blowing. Note that parallel to dawn, English had the now obsolete verb dayn, in which the relationship to its possible cognate subject day is both of a semantic and morphological nature. 15 Source: Cindy Bullen's song It's raining on prom night from the musical Grease (Lyrics: Jim Jacobs/ Warren Casey). <?page no="103"?> 91 verbs belong to the group of verbs known as avalent (zero-valent) verbs 16 that have no true subject (e.g. rain, drizzle, sleet, freeze, thaw) and are thus most frequently expressed with an expletive-it-construction, where it functions as an empty / dummy subject (cf. Allerton 1982: 94f.) However, I do not go as far as to claim that it, occurring in the same slot as the cognate subject, should be regarded as a 'dummy' for the cognate subject because the former construction represents the more conventional form of expressing the occurrence of natural events. The way speakers construe these types of events reflects to a certain extent common or folk beliefs about the causation or the coming about of natural phenomena such as thunder, wind or hail. The expletive-itconstruction verbalizes a construal in which the components of the event are represented as independent of any other participants (labelled 'absolute construal' by Langacker 1991a: 389), i.e. the cause/ source of such events remains unspecified, which is due to the virtual absence of an obvious causative entity. In a cognate subject construction, it is the event itself that is depicted as being responsible for the occurrence of the phenomenon expressed by the verb. Thus, instead of expressing the (very complex and specialist) knowledge about the origin of such natural phenomena we have gained in modern times, speakers still seem to prefer to present them in a rather 'naïve' way: They either just happen (as expressed in the expletive-itconstruction) or they cause themselves (cognate subject construction). Returning to the initial question whether cook (n) can be considered the cognate subject of cook (v), it must be weighed in how much the meaning of the subject is presupposed by the verb (see Quirk et al.'s definition above). Undoubtedly, the verb cook has two schematic e-sites, one of which is associated with an AGENT -trajector carrying out the activity. Yet, cook is only one of the many possible entities which can elaborate the schematic trajector e-site and is not as intricately related to the verb as is frost to freeze, where the trajector designates the event expressed by the verb or, what is also likely, the result thereof. It might be argued that the closer relationship between frost and freeze is precisely due to this ambiguity of frost, i.e. its denoting an event or a result the latter coming into existence through the process denoted by the verb (an observation which holds for all of the examples in [10]). In general, the relation between verb and object is closer than that between subject and verb, which is, for example, reflected in Constituent Analysis where the object-NP is analysed as part of the VP, whereas the subject-NP is not 17 . Moreover, in contrast to the examples of cognate subject constructions discussed so far, cook (n) designates an agent, 16 See section 4.4.1 for a discussion of the concept of 'valency'. 17 Moreover, the more or less close relationship between a verb and its arguments is reflected in the terminology of theta-theory: Direct objects are labelled "internal arguments", whereas the subject is described as "external argument". <?page no="104"?> 92 and not an event. Thus, the example in (9) should at best be considered as a peripheral instance of a cognate subject construction, as there is a close relation between subject and verb only on the morpho-phonological level. On the semantic level, the relationship is much 'looser', as cook is only one of many possible agents 18 . 4.4 The composite structure [V + CO], or: How to integrate an object with an intransitive verb In order to fully grasp the complex character of COCs, it is crucial to not only consider the single components of a COC as depicted in Fig. 4.1., but to look at the composite constituent [V + CO]. This, of course, entails the consideration of possible mechanisms of the components' integration into a more elaborate structure, i.e. the verb + CO pairing. The purpose of the subsequent sections is to discuss and compare linguistic models with respect to the solutions they offer as to how the more complex structure is arrived at. 4.4.1 Langacker's approach: Where does the e-site come from? One essential function of grammatical organization is the sanction of combining two or more symbolic structures to form a more complex structure, i.e. a composite structure. Component structures, their mode of combining, and the resulting composite structure all contribute to what has been discussed in Chapter 3 as a grammatical construction, i.e. a complex linguistic unit. CG's account of the underlying 'mechanism' for the integration of component structures into a more elaborate form centres around the notion of valence relations (Langacker 1987a: 277ff., 1988c). The concept of valency has its origins in chemistry, where it is understood as a measure of the maximal number of bonds a particular atom can form with other atoms. Its use in linguistics can be traced back to the French linguist Lucien Tesnière, who included 'valency theory' as a subtheory of his model of dependency grammar. In Tesnière's account, valency refers to the ability of a verb to predetermine its syntactic environment, its capacity for "combining with particular patterns of other sentence constituents" (Allerton 1982: 2). Depending on the number of elements verbs can bind they can be classed as avalent (rain, drizzle; see section 4.3.5) monovalent (sleep, bark), divalent 18 Similarly, such a looser semantic relationship holds between verb and lm-COs (see p. 84f.), which also denote one of many other possible PATIENTS . Thus, lm-COs would need to be placed at the periphery of the category 'Cognate Object'. See Ch. 5 for a network model of COCs, including prototypes and more peripheral instances. <?page no="105"?> 93 (answer, cost), trivalent (give, teach, wish), and, possibly, tetravalent (charge, pay) (cf. ibid.: 94ff.; Herbst 2004). According to Langacker's description of valence relation, the combination of elements into a composite form is only permitted if the two component structures share certain substructures, i.e. if there is an "overlap between two conceptions that permits their integration to form a coherent scene" (ibid. 1987a: 278). Correspondence is a precondition for elaboration, a process in which one component specifies a yet schematic substructure of the second component with which it is to combine into the more complex pattern. That is one of the component units offers a schematic substructure (a so-called elaboration site/ e-site) which then is elaborated by the second component. The arguments of verbs, for example, can combine with their verbs precisely because each verb offers particular e-sites which are then elaborated by their arguments. Verbs which are traditionally classed as 'monotransitive' bring along two of these elaboration sites. The two prominent e-sites of a monotransitive verb are referred to as trajector (tr) and landmark (lm) and correspond to the roles of subject and object on the clausal level. To give a concrete example, kill has two inherent substructures that need to be elaborated for the phrase to completely depict an act of killing: the killer as the verb's tr and the victim as its lm 19 . The verb kill is said to be conceptually dependent on these two elements, since one cannot conceive of an event without its interacting participants. In the sentence (12) […] Connon killed his wife. (GUD: 1846) the symbolic unit [ CONNON ] elaborates the e-site of the tr of kill, whereas [ HIS WIFE ] spells out the e-site of the verb's lm. If this conception of 'syntactic relations' between the parts of a construction is adopted to COCs, the conceptual integration of the CO is only possible if the conceptual content of the verb provides a substructure which can be elaborated by the CO, i.e. an e-site corresponding to a lm. As has been stated before, many of the verbs that occur in combination with a CO are predominantly used intransitively. Intransitive verbs, however, are supposed to have one e-site only - that of the trajector. Hence, it needs to be explained which region of the verb's conceptual structure is elaborated by the CO if the fusion of components into a composite structure is based on the elaboration of latent structure of one component by the other. Lan- 19 This is a somewhat simplified account as valency must be considered as a matter of degree. One might postulate further e-sites for kill, including substructures that relate to the instrument of killing, the location, the time etc. This continuum, ranging from primary e-sites to less prominent substructures, is reflected in the continuum between arguments (complements) and adjuncts (modifiers) (see Langacker 1987a: 306ff.). <?page no="106"?> 94 gacker suggests that the CO profiles a substructure which corresponds to the schematic boundary of an abstract region which itself is defined through the interconnected component states of the process (cf. ibid. 1991a: 363). He gives the following account of how the integration of verb and CO into a more complex composite structure is permitted: In a cognate-object construction […], this abstract region is salient in both the verb and its object: it is profiled by the object, and the verb accords to it the status of a primary landmark (ibid.). This highly abstract relation is depicted in Fig.4.7. Fig. 4.7 Corresponding substructures between Cognate Object and verb (adapted from Langacker 1991a: 364) Langacker's model may very well explain aspectual features of telicity/ boundedness of the event, which, in contrast to variants with the 'intransitive verb + adverb' structure, is foregrounded by a COC (as suggested by Macfarland 1995, Felser & Wanner 2001): The profiling of an abstract region brings about the notion of boundedness, which, in the case of temporal relations, clearly implies temporal boundaries. Moreover, his description advocates a status of the CO which comes close to that of adjuncts if one considers the low degree of prominence of the verb's substructure which is claimed to be elaborated by the CO. The description offered by Langacker clearly underlines the borderline status of COs between arguments and adverbials, yet suggests solving this issue in favour of COs as arguments, though as peripheral representatives. As has already been indicated (cf. ftn. 19), the distinction between arguments and adjuncts must be seen as a continuum rather than a clear-cut distinction, which is due to the fact that valency, i.e. the potential of substructures of conceptual entities to function as prominent e-sites, is a matter of degree. The following section will briefly take up the treatment of the lm tr Verb Cognate Object Nominal <?page no="107"?> 95 argument - adverbial distinction as put forward in Langacker's CG, tested on COCs. 4.4.2 COC S : Arguments or adverbials? Recall Jones’ suggestion that COs should be treated "as modifiers of the VP, on a par with […] manner adverbs" (1988: 93; emphasis added, SH). Following from this proposal, the sentence variants (13) a. Brad smiled charmingly. and b. Brad smiled a charming smile. can be considered as identical expressions of what appear to be truthfunctionally equivalent situations. However, several fundamental assumptions of CL, which were discussed in greater detail in Chapter 3, rule out an interpretation of the sentence pair as synonymous: 1) Meaning is conceptualization 2) Grammatical constructions are meaningful 3) Change of form implies change of meaning. Thus, a speaker choosing the one or the other form of expression, e.g. favouring the COC over an adverb construction, does so in order to express a particular conceptualization of the same situation. Apparently, one of the reasons for the meaning differences between the two variants in (13) must be sought in the different conceptual relations that hold between the verb and the respective constituents following the verb. Charmingly is a manner adverb, and adverbs are understood in CG as profiling an atemporal relationship, i.e. interconnections that hold between two or more entities. In the specific example of charmingly, the profiled relationship is that between a process (P), which is the schematic trajector of the relation, and a defined region on a scale measuring friendliness/ attractiveness/ politeness (Fig. 4.8). Note that an adverb is conceptually Fig. 4.8 Conceptual content of charmingly P Friendliness <?page no="108"?> 96 dependent 20 , i.e. one cannot conceive of charmingly without evoking at the same time an event (even if highly schematic) to which the value expressed by the adverb is ascribed. This is a crucial observation with regard to the adjunct-argument distinction: The distinction between conceptually autonomous and dependent predications is crucial for characterizing a number of important grammatical notions. One of these is the contrast […] between what is variously called 'central' vs. 'peripheral' elements in a clause, 'complements' vs. 'modifiers' of a verb, or 'actants/ participants' vs. 'circumstantials'. (Langacker 1987a: 308; emphasis added, SH) In the above figure, the event (expressed by the verb) is indicated by the small square: P constitutes a salient substructure of charmingly which needs to be elaborated by a verb designating a process. This constellation of a more autonomous (verb) and more dependent (adverb) element on the one hand and the verb's functioning as a profile determinant are crucial features for defining the syntactic relation that holds between the two components. In this constellation, charmingly (and any other adverb) functions as a modifier, which is by definition that constituent that is conceptually dependent on the profile determinant, in this case the verb (Langacker 1987a: 309). In (13b), on the other hand, we have the nominal a charming smile, with the constituent as a whole profiling a thing instead of an atemporal relation. Of course, the relation that holds between the noun smile and the adjective charming is comparable to that of the verb smile and the adverb charmingly in that the adjective portrays an atemporal relation as well: It profiles an interconnection between an entity (smile) and a scalar value of friendliness. However, what is of interest here is the relationship between the full nominal phrase and the verb. If a CO is to be described as an argument, the presence of features of nominal complements (arguments) as defined by CG must be probed for. Langacker speaks of complements when a dependent structure functions as profile determinant whose inherent substructures are elaborated by more autonomous entities (cf. ibid.). It seems that this is the case for the relation between verb and CO: A process (verb) is conceptually dependent on its participants; it offers e-sites that need to be elaborated by other entities. As shown in the preceding section, the CO (an autonomous participant) is said to elaborate the lm of the verb, whereby the latter serves as the profile determinant. Therefore, we can argue that COs are complements instead of recognizing them as adverbials, contrary to Jones' (1988) practice of labelling them as modifiers of the verb. 20 'Conceptually dependent' is defined as follows: "One structure, D, is dependent on the other, A, to the extent that A constitutes an elaboration of a salient substructure within D" (Langacker 1987a: 300; emphasis added, SH). <?page no="109"?> 97 Hence, the difference between a verb-adverb structure and a verb-CO structure is one of A/ D asymmetry in the first place, and therefore one of conceptual dependence between the verb and other constituents of a profiled relation. These different constellations are summarized in Table 4.1. verb - adverb verb - CO verb adverb verb CO A/ D asymmetry A D D A Direction of elaboration elaborater elaboratee elaboratee elaborater Syntactic function profile determinant (head) modifier profile determinant (head) complement Table 4.1 Adverb-argument distinction from a CG perspective Unfortunately, the situation is not as straightforward as presented so far, because the definition of arguments/ complements in general hinges on the presence of a linguistic element that is conceptually dependent on another entity. As described above, in the case of verbal complements, the verb must provide e-sites, the need of elaboration of which makes it conceptually dependent. Since many of the verbs in COCs lack a prominent e-site for a landmark, i.e. they are conventionally associated with one participant only, one might argue that this is exactly what motivates the discussion of COs as adverbs. Yet, as Langacker (1987a: 303) points out and as is nicely illustrated by Croft (2001: 274f.), the autonomy-dependence distinction is gradient, as are the notions "salient substructure" (see ftn.20) or "prominent e-site". Therefore, the complement-modifier distinction may at best be treated as a continuum, ranging from clear cases of adjuncts (Brad smiled charmingly) over COs (Brad smiled a charming smile) to clear cases of arguments (Brad ignored her charming smile). If we return to the first sentence of the 'alternating' pair in (10), it seems rather obvious that charmingly provides a salient substructure that needs elaboration, namely that of a process that is assigned the quality designated by the adverb. Nevertheless, an event of smiling has certain qualities, too, and thus, it might be argued, it not only involves a " SMILER " (as a highly salient substructure) but also quality. Hence, the quality of the smiling can be said to constitute a substructure of the semantic structure of smile, while at the same time charmingly provides a highly salient substructure of a process, elaborated by smile. However, there is a difference in the degree of salience in that the schematic process in the structure of the adverb is much more salient than is the schematic quality in the verb. It is exactly the dif- <?page no="110"?> 98 fering degree of salience that makes charmingly conceptually more dependent on smile than vice versa and it thus should be treated as an adjunct/ adverbial. Accordingly, there should be an imbalance in the degree of salience of the substructures provided by smile (v) and a charming smile that shifts dependence towards the one or the other. This means that if a charming smile is an argument of the verb, the degree of salience of a schematic result/ event in the verb (i.e. the verb's lm-e-site) should be much greater than the degree of salience of the substructure within the noun referring to a schematic process (i.e. CO's e-site of a schematic process). As will have become obvious, the asymmetry in salience as observed for smile and charmingly is not as strong for smile and a charming smile, especially if the NP is interpreted as an Event Object, because in that case the processual component is not completely de-focused (see p. 80f.). For a result reading of the CO, one may claim that it is possible to conceptualize a charming smile as the result of an action without conceiving of the preceding action 21 . However, due to the fact that this type of CO expresses the results of an event, it might also be argued that the event is schematically present to some degree. The verb smile, on the other hand, is a verb of facial expression, and therefore the semantic structure of the verb smile includes a substructure relating to that facial expression, i.e. the result of friendly curving one's mouth upwards. The decisive question now is which substructure is more salient: The schematic process in the resultative nominal or the schematic result in the process? For an event interpretation of a charming smile, the degree of salience of a schematic process is no doubt higher than for the object of result variant, since it is hardly possible to conceptualize the event of a charming smile without conceiving of the simultaneous action. Langacker points at the possibility that there might be a 'state of balance' as regards degrees of dependence: Canonically the structures in a valence relation manifest substantial asymmetry, with one of them […] clearly dependent, and the other autonomous. As always, though, recognition of the prototype must not be allowed to obscure the existence of other possibilities. Nothing in the definition [see ftn. 20] precludes a relation of mutual dependence between the two structures, or guarantees that there will always be a significant relation of dependence in one direction or the other. (ibid.: 300; emphasis added, SH) Event-/ Event-Result-COCs obviously must be considered as deviations from the prototypical constellation in which one element clearly depends 21 Macfarland argues that for COs as result objects "the use of the verb entails the existence of the cognate object since negating the noun results in infelicity: ? ? John laughed, but in fact he didn't laugh a laugh. ? ? John was thinking, but in fact he thought no thoughts". (ibid. 1995: 101f., emphasis added, SH) <?page no="111"?> 99 on another. That is, according to Langacker's view, in COCs the relation of dependence is not clearly in one direction. Whereas for objects of results the pendulum indicating autonomy might swing into the direction of the CO (which speaks in favour an argument interpretation), the dependence relation between verb and Event-COs is close to mutual, which makes this type of CO a more peripheral member of the category "argument", marking the transition from argument to adjunct in the postulated continuum. A clear manifestation of the argument-adjunct (complement-modifier) continuum can be observed in the combination live + a life. Since the NP a life expresses a temporal dimension, i.e. the span of a lifetime, a clear distinction between CO and temporal adverbial is not unproblematic, as is shown in the following examples: (14) a. Now began a new phase of his life in which he tried to live a grand, filmstar-type life in a huge house on three acres of land. (CDG: 526) b. You could live your whole life in some tiny place and think it was the whole world. (CEU: 532) c. E.G. James was a farmer and self-taught fix-it person who lived all his life in Mungo County, Virginia. (EBT: 2675) d. Mr Kenneth McGarvey (I), aged 41, has lived in Norton all his life. (K4W: 4864) Jespersen, when commenting on instances such as he lived all his life / his whole life, notes that "we see as it were a cross between the cognate object and the substantival subjunct [i.e. adverbial, SH] of time (length)" (1954 [1927]: 236). It is indeed hard to decide for the cases presented in 14 b-c) whether we deal with a CO argument or an adverbial of time. (14a) undoubtedly qualifies as an instance of a COC, and considering the fact that usually the sequence verb + complement is not interrupted by an adverbial (i.e. the PP in Norton), the NP all his life in (14d) should be taken as an adverbial. (14b) and (c), however, allow for both interpretations, which once more illustrates the somewhat ambiguous status of some COs when it comes to the adjunct/ argument distinction. As was mentioned at the end of section 4.4.1 and particularized in this section, Langacker's CG approach to COs mirrors quite well their peripheral status as a verb's complement if we take COs to elaborate a rather latent region of the verb's conceptual 'equipment' as depicted in Fig. 4.7 (p. 94). However, several shortcomings of this description must be acknowledged: Langacker's characterization cannot account for the CO as a created object/ object of result because it renders its 'coming into existence' as independent of the AGENT involved in the process. As can be seen in Fig. 4.7, no direct relation between the AGENT of the process and the CO is depicted, <?page no="112"?> 100 the latter merely elaborates an abstract region comprising all the component states of the event. Represented in this way, the entity profiled by the CO rather seems to have come about as a "by-product" of the process designated by the verb. This comes as a drawback of Langacker's analysis of verb complementation which solely builds on the specific verb as the carrier of a particular valency. Moreover, Langacker's definition of COCs seems to be a narrow one, excluding such attested cases as build a building, weave a web, sow a seed - cases in which the CO undoubtedly elaborates a substructure of the verb which is rather salient (as discussed in section 4.3.2). An alternative mechanism of verb-object integration is built into Goldberg's framework of Construction Grammar, which explains the interaction between verbs and the arguments structures in which they occur. The model, to which we will turn now, offers a rather different account for the, in some cases unconventional, complementation pattern displayed by verbs occurring in a COC. 4.4.3 COCs as argument structure constructions - Testing Goldberg's model In my discussion of different views on the concept of construction in Chapter 3.7.4, I already pointed out that Goldberg's account of verbal complementation differs from Langacker's model to some extent: One of the most prominent characteristics of Goldberg's theory is that it reduces the number of lexical entries of verbs (caused by the different kinds of complementation they occur with), and instead considers constructions as those units which 'bring along' a meaningful distribution of arguments: In Construction Grammar, instead of predicting the surface form and interpretation solely on the basis of the verb's independent specifications, the lexical verb is understood to combine with an argument structure construction […]. Verbs constrain the type of argument structure constructions they can combine with by their frame-specific semantics and particular obligatory roles, but they typically can combine with constructions in several ways. (Goldberg 1997: 70) The semantic restrictions on verbs and their fusion with particular constructions, which are observed by Goldberg, are similar to Langacker's notion of correspondence of substructures of components in so far that they specify preferred argument arrangements for particular verbs based on their semantic specifications. As Goldberg argues, independent specifications offered by the main verb alone, however, are often not sufficient for the interpretation and form of a clause, but are part of the individual construction in which a verb is used. The more schematic argument structure constructions provide a "direct link between surface form and general aspects of the interpretation" (ibid.). In the default case, the number of par- <?page no="113"?> 101 ticipants a verb is usually associated with corresponds with the number of argument roles offered by the construction. Nevertheless, there can be mismatches between the specifications of a verb and the specifications of the construction, be that with respect to the profiling of roles or the number of roles. Goldberg makes explicit and demonstrates with a number of cases that a construction can enrich the participant constellation conventionally associated with a particular verb: "[I]t is not necessary that each argument role of the construction corresponds to a participant of the verb. […] the construction can add roles not contributed by the verb. (ibid.1995: 54)". It is on the basis of fusing the semantics of a particular construction with the verb that speakers can easily interpret sentences which include verbs 'equipped' with participants which are not determined by the verb's participant specifications, as in the following cases (underlined elements are roles added by the respective construction): (15) a. Anthony Everard tried to laugh away his daughter's fury. (G0S: 1756) (caused-motion constr.) b. A correspondent of that chain, that accompanies the British troops, assured that the allied soldiers were applauded to the entrance in the Iraqian city 22 . (caused-motion constr.) c. I've cried me a river, I've cried me a lake 23 . (ditransitive constr.) d. Daniel Craig dresses his way to fame. (way-constr.) (HindustanTimes.com; accessed: 3/ 4/ 2007) Likewise, the majority of verbs taking an Event/ Event-Result-CO can be described as cases where there is a mismatch between the number of participant roles associated with the verb and the number of argument roles of the construction. This implies that the argument structure is 'imposed' on the verbs by a meaningful argument structure construction: Thus, if COCs are treated as a special, non-prototypical type of transitive constructions (offering two argument slots), it must be assumed that the argument roles are inherent in the construction and not provided by the verb. Hence, we can conclude that the second argument slot for verbs such as cry or dream, which are conventionally associated with an intransitive construction and therefore with one participant only, is 'made available' by the monotransitive COC as a meaningful argument structure construction. The monotransitive construction is a highly polysemous construction in that the same formal arrangements can express a whole range of meanings. Prototypically, the construction is described as expressing a relation in 22 Source: www.atlasescolar.com.ar/ NewsArchives/ 50/ 10/ archivo-ze4123_esen.shtml, accessed 3/ 4/ 2007. 23 Source: Line of the song Thirty years of tears by John Hiatt. <?page no="114"?> 102 which an AGENT acts upon a PATIENT and thereby causes a change of state in the PATIENT , with the AGENT conventionally expressed as the subject and the PATIENT as the object. Adopting Goldberg's mode of notation, we can represent the prototype of monotransitive patterns as shown in Fig. 4.9.The semantic layer spells out the semantics directly associated with the construction, i.e. " X CAUSES Y to CHANGE " 24 , while the syntactic level presents the syntactic functions V , SUB and OBJ , to which the argument roles are linked. This description is most straightforwardly applicable to monotransitive constructions including an affected object which undergoes a change of state. In the case of effected (resultant) objects, i.e. objects that are created through the activity expressed by the verb, the meaning of the construction differs slightly: It verbalizes a scenario in which an action does not cause a change of state on the side of the PATIENT , but the coming into existence of some entity (which, nevertheless, might be interpreted as a change from non-existence to existence). The semantics of this constructional variant can be captured as ' CAUSE - EXISTENCE ' < agt pat >; it has the same syntactic representation as the monotransitive prototype. With COs frequently being described as objects of result, the monotransitive constructions with effected objects should be considered as the pattern which sanctions COCs (except for lm-COs of the "affected"-category, which fall under the monotransitive prototype, e.g. Many other individual gifts were also given. [HS1: 456]). However, even if it is the construction which provides an additional role, several other conditions have to be fulfilled in order to fuse the par- Fig. 4.9 The Monotransitive Construction (prototype) 24 While Goldberg (1998) represents the meaning of the monotransitive construction as " X ACTS ON Y ", many linguists emphasize a causative element in the prototypical meaning of the constructions (see Hopper & Thompson 1980, Rice 1987, Horita 1996). Therefore, I prefer the notation " X CAUSES Y TO CHANGE " as the representation of the semantics of the pattern. See Ch. 6 for an elaborate discussion of prototypical monotransitive forms. Sem CAUSE - CHANGE < agt pat > PRED < > Syn V < SUB OBJ > <?page no="115"?> 103 ticipant roles of the verb with the argument roles of the construction. Bearing in mind Goldberg's claim that there are semantic restrictions on the types of constructions a verb can occur with, we have to clarify the following points: How can live and dream, for example, verbs designating 'existence' and 'mental process', respectively, enter a construction which by definition expresses 'cause-existence'? How can it be explained that the CO, which has been characterized so far as an event or the final state thereof, can instantiate the role of an effected PATIENT ? Here, construal as a basic human cognitive ability plays a crucial role: In her Semantic Coherence Principle, Goldberg (1995: 50) notices that the participant roles of the verb and argument roles of the construction need to be "semantically compatible" in order to be integrated (i.e. fused), where semantic compatibility is a matter of construal: "Two roles r 1 and r 2 are semantically compatible if either r 1 can be construed as an instance of r 2 or r 2 can be construed as an instance of r 1 " (ibid.; emphasis added, SH). In the particular case of COCs this means that in order to meet the specifications of the monotransitive construction with an effected object, the reified event (expressed by the CO) needs to be construed as an entity which is effected by the action of the AGENT . Thus, we have to identify a construal process which plausibly explains how speakers come to perceive of an action or the result thereof as a concrete, effected entity. One such possible construal operation, conceptual metaphor a very powerful and ubiquitous cognitive tool shall be considered here: One basic type of conceptual metaphors are so-called ontological metaphors, 'means' of grasping intangible concepts such as emotions, experiences, ideas and events as bounded, concrete entities or substances. They represent mappings which have their source in our interaction with physical, clearly delineated objects and enable us to "refer to [our experiences], categorize them, group them, and quantify them and, by this means, reason about them" (Lakoff & Johnson 1980: 25). EVENTS / ACTIONS ARE OB- JECTS / CONTAINERS (alternatively, ACTIVITIES ARE CONTAINERS , as labelled, for example, by Kövecses & Szabo 1996) is one of the manifold ontological mappings which human beings constantly make use of to apprehend the complex nature of events and actions 25 . As such, they can be perceived as being created, manipulated, possessed and transferred 26 : 25 Note that for the metaphorical mapping EVENT / ACTIONS ARE OBJECTS , conceptual reification plays a crucial role. 26 In their more recent accounts on metaphor, Lakoff (1993) and Lakoff & Johnson (1999) discuss the Event-Structure metaphor, a systematic network of metaphors which structures various aspects of events such as states, changes, causes. Although they do no longer explicitly refer to a mapping EVENTS ( ACTIONS ) ARE OBJECTS ( CON- TAINERS ), they assume such a mapping when acknowledging the Object branch of the Event-Structure metaphor. Sentence (16a) serves as an illustration of a submapping of this branch. <?page no="116"?> 104 (16) a. I have a headache. (Possession) (Lakoff & Johnson 1999: 196) b. […] as soon as her back is turned, we give the dog a kick and it shoots off. (Transfer) (B24: 1975) Moreover, as Lakoff & Johnson (1980) observe, activities are viewed as "the containers for the actions and other activities that make them up. They are also viewed as containers for the energy and material required for them and for their by-products […]" (ibid.: 30; emphasis added, SH). The authors provide the example (17) I put a lot of energy into washing the windows (ibid.: 31) as a possible verbal reflection of a speaker's construing the activity of washing windows as a 'collecting tank' of energetic processes. Drawing on these insights of conceptual metaphor research, I propose the analysis of COs as objects that are effected through the action an AGENT is executing as a plausible explanation of the occurrence of entities denoting events or actions in the PATIENT slot of the monotransitive construction, if one thinks of them as containers for energetic transfers 27 . This, of course, has consequences concerning the types of verbs that can be used in a COC with an EVENT / RESULT -CO: The approach presented here suggests that all those verbs which imply some energetic exchange may occur in the construction, i.e. all those that can be construed as actions which require some amount of energetic input and create some 'output', be it sorts of sounds (laugh, cry, sob), some kind of verbal utterance (tale, song), a bodily movement (jump, dance, step) or the product of cognitive/ psychological processes (thought, dream). These qualify as the most likely candidates and should be considered as prototypical COC-verbs. The slots constructions offer are variable and are usually filled with lexemes from a particular category. Since categories have prototypical instances and peripheral members, it must be expected that verbs are used in the construction's slot which only marginally meet the specifications listed above and that one cannot fully predict each and every verb that will occur in a COC. As Bybee (2003: 158) comments: The possibility of adding new peripheral members to a category allows productivity and change. New items can be used in a construction if they are perceived as similar in some way to existing members" (emphasis added, SH). Chapter 5 will take up the issue of classes of verbs occurring in COCs again, discussing the results of a corpus analysis. 27 The notion of energetic transfer is one of the starting points of a CL description of transitivity. The relation between COCs as monotransitive constructions and the verbalization of energetic processes is discussed in Ch. 6. <?page no="117"?> 105 Furthermore, in order to account for the great (semantic) variety of verbs in COCs, one has to acknowledge the 'intervention' of so-called coercion effects, which shift the verb's meaning so that it is compatible with the meaning of the construction. Coercion can be described as the phenomenon of one linguistic unit "exerting an influence" on another unit if combined with it, thereby "causing to change its specifications" (Taylor 2002: 287). Goldberg (1995) already observes the need to recognize a particular process of coercion in order to account for cases in which "[…] a construction requires a particular interpretation which is not independently coded by particular lexical items" (ibid.: 159) 28 . Michealis, aiming at a description of shifts in aspectual characteristics of verbs in particular constructions, argues along similar lines: I assume a coercion mechanism whereby constructional requirements […] 'win out' over lexical features when the lexical item and the construction upon which it is superimposed have different values for a given attribute. This accommodation mechanism is described […] as the Override principle: […] If a lexical item is semantically incompatible with its syntactic context, the meaning of the lexical item conforms to the meaning of the structure in which it is embedded. (ibid. 2004: 51; emphasis added, SH) With respect to argument structure constructions, for example, coercion effects are claimed to be responsible for the motional meaning of the verbs urge and fit, evoked by the caused-motion construction, as in (18) a. Hugh urged Mrs Tobias into her taxi and walked off smartly in the opposite direction. (FB9: 576) b. How do you fit four elephants into a Mini? (A74: 684) Similarly, for COCs, the verbs dream and roar can be said to be coerced into having a creational meaning by the monotransitive construction in (19) a. [T]hey dream wildly beautiful, but sometimes impossible, dreams. (JY2: 3267) b. She roared the roar of a lioness celebrating her kill. (buffalospath.blogspot.com/ 2005_12_01_archive.html) It is apparently the concurrence of coercive effects a construction exerts on the meaning of lexemes occurring therein and a speaker's capacities for construal which enables the use of highly diverse, semantically unrelated verbs in the construction. Thus, even the verbs dream and roar, which hardly share a common semantic ground, can be used in the COC, 'eliciting' a creational sense for both verbs. 28 In her most recent work Constructions at work, Goldberg (2006: 22) comments on the notion "construal" as intended to also allowing for processes of accommodation or coercion. <?page no="118"?> 106 Adopting Goldberg's approach for the description of COCs holds several advantages: Firstly, it allows us to consider the referents of COs as true participants in an event, which are construed as coming into existence through an action performed by an AGENT . Secondly, we can easily account for the whole range of COCs found in the literature: build a building, paint a painting or weave a web are as good examples of the construction as smile a smile or walk a walk with respect to the semantic specifications of the construction given above. The only difference between these closely related examples is that the former group involves verbs whose number of inherent participants coincides with the number of argument slots provided by the construction, whereas the latter group inherits one role (i.e. the PATIENT argument) from the construction. Thirdly, examples such as drink a drink, feed food or fish fish, which involve affected objects and which lack the notion of 'creation', can also be accounted for as being members of the category 'COC': One of Goldberg's major contributions to the discussion of constructions is the acknowledgment of argument structure constructions as often highly polysemous patterns which are linked within a constructional network (ibid. 1995: 32-39). As mentioned above, the monotransitive construction can express a variety of meanings; similarly, if we look at the variety of instances of COCs cited in the literature, we might 'redefine' the pattern as a family of constructions which are related through instantiation or extension links. Here, too, the reader is referred to Chapter 5, which offers a more detailed account of a network of COCs. 4.4.4 A 'hybrid - model' for COCs Having discussed options for the integration of a CO with verbs which originally do not provide a second argument slot both from Langacker's CG perspective and within Goldberg's model of Construction Grammar, I suggest a different representation of the integration of verb and CO into a composite structure [V+CO], in particular for Event-COCs. The model is a hybrid form in that it incorporates ideas and notions of both Langacker's and Goldberg's approaches to verbal complementation 29 . While Langacker's CG account offers suitable procedures for the representation of the pattern's conceptual content, Goldberg's model is (for our purposes) more helpful for handling the issue of additional argument roles, which allows us to discuss COs as true participants in an event. Fig. 4.10 illustrates the merging of two elements of the COC, that of verb and object, by the elabo- 29 See Leino (2005) for a similar procedure in his description of the Finnish Permissive Construction in terms of construction grammar(s). <?page no="119"?> 107 Fig 4.10 Composite structure (cf. Langacker 1987a: 326) ration of shared substructures 30 . However, while Langacker suggests that the CO elaborates a rather latent substructure of the (in traditional terms) intransitive verb, in my model it elaborates a more prominent landmark esite (i.e. PATIENT slot). The latter, indicated in the illustration as dotted small boxes, is inherent in the argument structure construction and automatically "handed down" to the particular verb used in the construction. That is the construction enriches the conceptual content of the verb. The upper box shows the composite structure, emerging through the fusion of verb and CO; it very vividly depicts the unique (redundant) character of this type of construction which combines two different profiles of the same conceptual base. The illustration would be similar for the merging of EVENT - RESULT -COs with their respective verbs in that the lm is also added to the verb by the construction. Yet, the composite structure would display a lower degree of redundancy since by definition EVENT - RESULT - COs only profile the final component state of the verb in a reified form. For constructions in which "inherent lm-COs" or "product COs" are combined with their respective verbs, dotted small boxes representing the verb's lm would 30 For reasons of clarity, I do not depict the integration of the tr / AGENT . For the same reasons, the upper box in Fig. 4.10 shows only one component state of the composite structure. Verb (profile determinant) Cognate Object Composite structure V + CO tr lm <?page no="120"?> 108 give way to boxes with solid lines, indicating that the verb is already 'equipped' with such a substructure. In the above description of the emerging composite structure of COCs through the integration of verb and CO into a complex unit, the issue of redundancy, a crucial notion in Langacker's description of the construction (see section 4.1), has been re-introduced. The following sections will explore the nature and function of structures which are characterized by a 'repetition' of linguistic material, and evaluate the conceptual weightiness of COCs compared to seemingly simpler, non-redundant variants. Assuming iconic traces in the pattern, my considerations will, to a great extent, rest on basic assumptions and insights of studies on iconicity in language. 4.5 Repeated content: Redundancy vs. Iconicity 4.5.1 Iconicity in language Under the influence of the pioneering work of the American philosopher C. S. Peirce 31 , the preceding decades have seen a 'revival' of the idea that language cannot simply be reduced to a conventionalized, arbitrary system of signs of different complexity. Instead, linguistic signs and structures are described as being motivated to a considerable degree. Peirce explicitly points at language when he observes that "in the syntax of every language there are logical icons of the kind that are aided by conventional rules" (as cited in Givon 2001: 34). Running counter to the Saussurean principle of the arbitrariness of signs 32 , this view on language promotes a natural, 'nonartificial' relation between signs (words, structures) and the things they signify. The strict dichotomy between predictability and arbitrariness underlying the structuralist and post-structuralist principle of the arbitrary sign is no longer upheld, instead, linguistic forms and structures are evaluated along a scale of motivation. Considerable linguistic research has shown that iconicity as one form of motivation operates on all levels of language displayed in such phenomena as sound symbolism or ono- 31 Peirce set up a trichotomy of signs, based on the different relations holding between sign and signified. Next to icons (where the relationship is one of similarity) he defines indices (with the relationship being one of contiguity) and symbol (a conventionalized relationship) (see Chandler 2006, Nöth 1985: 39f.). 32 Most often, de Saussure's 'Principle I: The arbitrary nature of the sign' (1959: 67) is quoted to emphasize the absoluteness of his view. It must be admitted, though, that de Saussure distinguishes between degrees of arbitrariness, i.e. between absolute and relative arbitrariness, and explicitly states that "the sign may be motivated to a certain extent" (ibid. 131; emphasis added, SH). He identifies syntagmatic relations as the primary source of motivation in the language system (ibid. 132). <?page no="121"?> 109 matopoeia, but also in inflectional categories and word order (cf. Fischer & Nänny 1999: xix, Haiman 1994). For the description of language in relation to iconic principles, various types of iconicity must be acknowledged, as put forward by Peirce (1932) and applied to language most originally by Haiman (1980). The main distinction that is drawn is that between imagic iconicity and diagrammatic iconicity 33 . Imagic iconicity is characterized by a "'direct', one to one relation between the sign or signifier" (Fischer & Nänny 1999: xxii) and is found, for instance, in onomatopoeic expressions and sound imitations, such as in the form bowwow signifying the sound made by a dog. Diagrammatic iconicity, on the other hand, is of a more abstract nature and works on the relationship between several signs, e.g. the morphological and syntactical arrangement of linguistic units 34 : "There exists an iconic link between the (horizontal) relation(s) on the level of the signifier and the (horizontal) relations on the level of the signified" (ibid.). Thus, linear relations on the sentence or text level are interpreted as standing for temporal, spatial or causal relation in the described setting. The often-quoted statement by Julius Caesar, "Veni, vidi, vici", may serve as an illustration here, where the temporal order of the events in the real world is iconically mapped onto the order of the verbs as action-naming linguistic signs (cf. Haiman 1994: 1630). The most general phenomenon of language instantiating structural diagrammatic iconicity is known as isomorphism, the principle of one form for one meaning. As a corollary of this principle it follows that sameness of form indicates sameness of meaning and that difference in form signals difference in meaning. Claiming that it is a "natural condition" of a linguistic system to maintain such a balance of one form for one meaning, Bolinger (1977: x f.) states that […] any word which a language permits to survive must make its semantic contribution; and the same holds for any construction that is physically distinct from any other construction. (emphasis added, SH) 35 Similarly, in a plea to "take language seriously", Dik (1997a: 18) elaborates on the idea of an iconic relationship between linguistic form and meaning: Whenever there is some overt difference between two constructions X and Y, start out on the assumption that this difference has some kind of functionality 33 See Fischer & Nanny (1999: xxii) for an overview on different types of iconicity. 34 Fischer & Nänny (1999) further distinguish between structural and semantic subtypes of diagrammatic iconicity. Morphological and syntactic arrangements fall into the structural subgroup. 35 Considering common occurrences of homonymy and polysemy in languages, one must remark that Bolinger's view is rather idealistic. Givon (1985: 59) calls Bolinger a "naïve iconist" because of his assumption that substantive markedness is always paired with structural markedness. <?page no="122"?> 110 in the linguistic system. Rather than pressing X into the preconceived mould of Y, try to find out why X and Y are different, on the working assumption that such difference would not be in the language unless it had some task to perform. (emphasis added, SH) Both statements give rise to and lend support to the assumption that COCs and their simple verb variants convey different meanings due to their 'physical difference' (i.e. a difference in form), and it might already be hypothesized here that the 'more of form' in COCs signals a 'more of meaning'. This assumed correlation between a 'more of form' and a 'more of meaning' has come to be known as the 'quantity principle' in iconicity research (cf. in particular Givon 1985, 1990, 1991a). In the following I will demonstrate how several general iconic principles that have been identified and formulated in the course of research on iconic traits in language can provide us with a 'natural' explanation for the idiosyncratic structure and meaning of COCs. 4.5.2 The "more of form" Definitions of COCs describing the cognate object as a mere repetition of the verb and reducing the construction to a dummy form for the inclusion of additional modifying elements (which would have been impossible with the simple verb variant) may give rise to the impression that COs are elements void of any kind of meaning. This would appear to hold true in particular for instances of bare, unmodified CO phrases cases which not even fulfil the suggested 'dummy function'. Such an account is too simplistic, considering the attested occurrence of unmodified COs (notwithstanding the less frequent usage of such complements). In his discussion of iconicity in syntax, Givon (1985) formulates the socalled iconicity meta-principle, emphasizing the supposed one-to-one relation between linguistic forms and the extralinguistic units they signify: All other things being equal, a coded experience is easier to store, retrieve and communicate if the code is maximally isomorphic to the experience. (ibid.: 189) Elaborating on this relatively general meta-principle of iconicity, Givon postulates several sub-principles, some of which seem to lend evidence to the non-redundant nature of the CO phrase: a) [I]f a piece of information is communicatively either unimportant, irrelevant, or should not in the speaker's judgement be communicated, for whatever reasons, that piece is not mentioned. (ibid.: 204, italics in the original) With respect to COCs, subprinciple (a) labelled "relevance principle" by Givon (ibid.) can simply be reformulated as the claim that if the CO actually were an element empty of meaning or unnecessary, a speaker would not include it in the utterance and choose the simpler structure ' intr.verb + adverbial' instead. Two other subprinciples put forward by Givon, which <?page no="123"?> 111 refer to the general coding potential of the structural inventory of language, are quoted in (b) and (c): b) The more important an item is in the communication, the more distinct and independent coding expression it receives. (ibid.: 206; emphasis added, SH) c) The more stereotypical an object, instrument or manner adverb is as information, the less likely it is to be given independent coding expression, and the more likely it is to be incorporated into the verb. (ibid.: 207; emphasis added, SH) Both axioms can be considered as a 'legitimization' of the COC structure: It provides a slot for an item which is considered to be important and is thus given independent and distinct coding expression, viz. in the form of a COphrase. In particular subprinciple (c) expresses an implicit statement about the non-predictable or unexpected nature of independently coded linguistic material. To illustrate this claim, Givon compares the phrases work fast and run ('go fast'), where the latter expression shows semantic incorporation of a more stereotypical feature MANNER (i.e. speed), which is less predictable for the activity of working and which for this reason is expressed by means of an independent linguistic unit (cf. ibid). Likewise, Kemmer (2003), discussing the verbal elaboration of events with a focus on middle and reflexive constructions, introduces the notion of conceptual differentiation / separation of participant roles for the purpose of describing a constructional pattern in which the two participant roles (of a seemingly transitive construction) are unexpectedly filled by the same individual (e.g. He washed himself), as compared to the more expected middle construction where there is no such separation of participant roles (e.g. He washed). Kemmer (ibid.: 113ff.) regards the use of a separated, independent reflexive morpheme as the linguistic reflection of such conceptual separation and thus as an instance of iconic motivation: The formal separation of verb and reflexive morpheme is claimed to be iconically related to the semantic (conceptual) distance or separation. A similar motivation applies to the form of COCs: It seems that the Event-Result-COC is just another instance of a construction which enable speakers to elaborate an event to a higher degree by conceptually separating the result of an event (i.e. its final state) from the event itself, which, in turn, is expressed via the integration of a separate linguistic element (i.e. the CO). Moreover, Kemmer maintains that it is not only conceptual distance but also conceptual weight which is mirrored in the use of a separate reflexive morpheme in reflexive constructions: Just as degree of formal distance mirrors conceptual distance, degree of formal weight or substance also corresponds to the degree of conceptual substance associated with the respective semantic categories expressed by the two forms. <?page no="124"?> 112 The more conceptual material there is in a given expression, the more formal substance there will be to signal it. (ibid.: 113) In a somewhat similar vein, Tobin (1993) points out that lexemes which are in some way marked for the feature [+ RESULT ] carry a greater conceptual load 36 . Incorporating these observations into the present discussion, one might argue that COs, which encode a resultative component, reflect this more of conceptual content in their form. To sum up, it seems reasonable to consider the CO as an element contributing meaning instead of dismissing the phrase as expressing the obvious. Returning to Bolinger's claim of a strong correlation between form and meaning, we can conclude that a 'more of meaning' is iconically mapped onto a 'more of form'. The nature of such additional meaning needs to be specified, however. This question will be taken up in Chapters 6 and 7, where I present evidence for a clear functional contribution of the CO. 4.5.3 The iconicity of repetition Reviewing again the earliest description of COs in grammars of English, i.e. the one offered by Sweet, we find an explicit reference to the repetitive nature of the object phrase: "Sometimes an intransitive verb is followed by a noun in the common form which repeats the meaning of the verb" (ibid. 1891: 19; emphasis added, SH). Studies concerned with iconicity in grammar have discussed the repetition of linguistic elements and structures as clear instances of motivated linguistic form: A number of grammatical operations are iconically motivated. Perhaps the most widespread is grammatical reduplication, insofar as it expresses any of the broad categories of intensity, plurality and repetition. (Haiman 1980: 16; emphasis added, SH) Another and more clearly iconic manifestation of the quantity principle may be seen in the phenomenon of repetition or reduplication. (Rohdenburg 2003: 263; emphasis added, SH) [T]he repetition of a lexeme in the sound stream of natural languages […] is a rich source of iconic possibilities. (Conradie 2003: 203; emphasis added, SH) The diversity of forms and functions of repetition in language has been the concern of linguists working in the most different subdisciplines: Espe- 36 As Tobin (1993: 17) explicates: "The feature 'result' has been chosen as the marked feature because it involves a more complex cognitive process: an action, state or event must not be viewed in a partial or immediate manner, merely from its 'beginning' and/ or during its performance or execution, but must be viewed in a holistic way which necessarily includes its RESULT (goal, consequence, conclusion, destination, telic endpoint etc.)". (emphasis added, SH) <?page no="125"?> 113 cially in the branches of textlinguistics and conversational analysis, psycholinguistics and stylistics the phenomenon has been described from several different perspectives 37 (cf. Aitchison 1994, Johnstone 1994, Tannen 1989). Aitchison (1994: 15) offers a nice illustration and summary of the many different facets of "repetition": When parrots do it, it's parroting. When advertisers do it, it's reinforcement. When children do it, it's imitation. When brain damaged people do it, it's perseveration or echolalia. […]. When novelists do it, it's cohesion. […] When sounds do it, it's gemination. When morphemes do it, it's reduplication. When phrases do it, it's copying. When conversations do it, it's reiteration. Both Aitchison and Johnestone deploy a classification of forms of repetition which takes into consideration (among others) the form of the repeated material (exact vs. partial repetition), temporal relations (immediate vs. delayed), the size of the unit (phoneme, morpheme, word, phrase), the intentionality of repetition (intended vs. unintended), and the participants involved (self-repetition vs. other-repetition). Following these criteria and regarding the fact that COs are nominalised forms of their respective verbs, one could classify COs as intentional, self-repeated, partial repetitions on the morphemic level, which can be immediate or delayed (in cases where modifying elements precede the CO) 38 . English has been described as a language that rarely makes use of repetition, even less so in regard of grammaticalized repetitive patterns: [T]he intolerance of the language for repetition can be seen in pronominalization, indefinite agent deletion. Most of reduplication in English is lexical (choochoo, willy-nilly, shilly-shally). (Johnstone 1994: 17) None of the authors mentioned considers COCs as instances of codified forms of immediate repetition. The majority of studies on repetition in English focus on the "delayed" iteration of lexemes in discourse; yet, the few descriptions given of immediate lexical repetition (see, for example, Aitchison 1994, Knowles 1979) highlight the intensifying role of repeated adjectives or intensifiers 39 : 37 In textlinguistics, repetition is considered an important cohesive device; in childlanguage acquisition reiteration has been studied in particular as a strategy of reinforcement in guiding the child in his/ her handling of linguistic items and structures. 38 It should be noted that "delayed" has been primarily applied to repetitions across sentence boundaries, in the case of which the repeated element serves cohesive functions. 39 Aitchison (1994: 24) lists the intensification of intensifiers as one of the few cases of English in which repetition is obligatory. <?page no="126"?> 114 (20) * He ran extremely very fast. He ran very very fast. (Aitchison 1994: 24) Knowles (1979: 643f.) observes that there is a strong correlation between the effect of an iteration and the lexical category of the lexemes involved: Adjectives and adverbs are intensified through repetition, iterated nouns imply an increase in quantity, and the iteration of verbs results in aspectual changes, the nature of which he describes as either durative or repetitive: (21) He thought (and) thought (and) thought all day long. (ibid.: 644) In the case of COCs, it is evidently the verb which is 'repeated', nevertheless, a straightforward adoption of the observed regularities to COCs is complicated by the fact that the repeated element is not an 'exact copy' with respect to its lexical category. The notion of aspectual change, which Knowles identifies for the repetition of verbs, is prevalent in studies describing the effects of reduplication, i.e. the reiteration of morphemes. The outcomes of research on reduplication might give some useful hints concerning the nature of the CO, since the formal features of the reduplicational type of repetition come close to what one finds in the structure of COCs: partial repetition of a morpheme. Moravcsik provides a systematic account on types of reduplication found in the languages of the world: Reiteration may be either total or partial: it is total if it involves the repetition of the whole semantic-syntactic or of the whole phonetic-phonological string whose meaning is correspondingly changed; and it is partial if it involves the reiteration of only part of the semantic-syntactic or phonetic-phonological constituent whose meaning is accordingly modified. (1978: 304; emphasis added, SH) As regards temporal order she notes that for reduplicated structures, […] all the logically possible relations that any event can bear to any other are logically available, such as simultaneity, immediate precedence, nonimmediate precedence, overlap, inclusion, and interlocking. (ibid.: 305; emphasis added, SH) The highlighted features are applicable to COCs, demonstrating a certain similarity between "true" reduplications and COCs. This, however, is not to imply that I consider COCs to be instances of reduplication. Moravcsik (1978) draws the general conclusion that the two main meanings of reduplication are emphasis, in the case of which the increased amount of form iconically reflects an increase in intensity, and continuity. Regarding the latter she argues that a reduplicated element can suggest "increased temporal extent of an action" (ibid.: 312). Likewise, in a study on reduplication in Afrikaans, Conradie (2003: 203) observes a general tendency of repeated lexical material to signal "duration of action", "extension <?page no="127"?> 115 of an action or a state", and an "intensified state or emphasis". For a transfer of these findings to COCs, the following, more refined description of the relationship between repeated items and temporal duration should be of particular interest: [R]eduplication may be iconic of duration or continuation either because the two points of reference implied by the two constituents may signal the beginning and end of a time-span or because the repetition of a lexeme may indicate an ongoing situation or a sustained state. (ibid.: 215) Can we make similar observations for structures in which part of the verb is repeated in a nominalised form, i.e. in form of a CO? Effects on aspect due to the inclusion of a CO in contrast to an objectless predicate have already been commented on in previous chapters and will be discussed in more detail in Chapter 7. Although I do not go as far as to claim that through the addition of the CO phrase a notion of continuity is added to an expression, it cannot be denied that a certain form of temporal bounding is achieved: In the majority of cases 40 the combination of verb and CO does signal the extension of an action over a particular period of time with the temporal limits coinciding with the boundaries of the event expressed by means of the CO, the reified version of the event: (22) a. He sighed harshly and leaned back in his seat. b. He sighed a harsh sigh and leaned back in his seat. (JXS: 3513) Whereas in the former example no temporal limits are specified, the expression in b) describes a process which extends within the boundaries of "a harsh sigh". That means that the event of sighing takes precisely as long as it does to 'create' a harsh sigh. Thus, the CO does have an effect similar to what has been observed for forms containing reduplicated morphemes, and this effect is iconically represented in the structure of COC expressions: The use of a structure which involves a bounded event-entity mirrors the boundedness of the event. At the same time it shifts the focus on or zooms in a certain section of the action within these boundaries. Note, however, that this is not to propose that the action itself is presented as more extended than in the CO-less variant. The second function of repetition/ reduplication identified in most studies is that of intensification, which comes as a result of the re-newed reference to previously verbalized elements (cf. Aitchison 1994, Conradie 2003, Johnstone 1994, Shepherd 1994, inter alia). Aitchison (1994: 17) recognizes two facets of intensification through repetition, ascribing to it an increase in quantity or quality. For COCs the intensifying effect is rather of a qualitative nature, since (as has been argued above) the object phrase does not contribute to the expression of a temporally more extended event than 40 This statement holds for those COs which allow for an event interpretation. <?page no="128"?> 116 encoded by an objectless variant. However, with COs designating a created (even if abstract) entity their addition clearly leads to a shift in qualitative characteristics of the event: An action which is perceived as having some kind of outcome is of a higher quality than an action without a result. The coincidence of form of verb and effected object exhibits the most natural, iconic link between action and the result thereof and intensifies such a perception. In relation to a discussion of repetition a few remarks on the notion of redundancy 41 need to be made here. A "tolerance of redundancy" is claimed by Langacker (1991a: 364) to be an important prerequisite for the acceptance of COCs as somewhat deviating transitive structures (cf. 4.2.). Unfortunately, he includes a vague indication to his understanding of the term only, viz. as an overlap of the conceptual structure of two linguistic units. His handling of the term is just another instance of a rather common method of dealing with the notion of redundancy in linguistics: A proper definition of the concept is excluded from its discussion 42 . Most generally, redundancy can be thought of as the encoding of one piece of information more than once 43 a phenomenon which can be found at all levels of language. A proper example of redundancy in the morphological system of English is the marking of the third person present tense singular with the suffix -s attached to the verbal stem. The information of "third person" is already conveyed through the subject, which needs to be expressed since English is not a pro-drop language. This presence of the subject makes the morphological marking of the verb truly redundant, but it is obligatory in accordance with the rules of English morphology 44 . In cases where such multiple encoding of information is optional, it still cannot be dismissed as empty of meaning. Here, it either might serve the maintenance of communication, which otherwise might break down due to transmission inferences, ambiguity or deletions, or fulfil a more special function such as isolating features, intensifying certain aspects of a message, or the creation of a "poetic effect" (Gillette & Wit 1998). Repetitions of lexical items clearly are examples of linguistic redundancy, and the functions of iteration discussed above appear on the list of the functions that Gilette & Wit record for redundancy. The conclusion which can be drawn from the above discussion 41 Note that this term is missing on Aitchison's list of notions involving repetition. 42 Hurford (1994: 3481) speaks of redundancy as dealt with in linguistics as an "informal, pretheoretical notion". See Gilette & Wit (1998) for an overview of different approaches to redundancy. 43 Besides the sense 'unnecessary repetition', a second, related sense of redundancy is 'predictability' (cf. Hurford ibid.; Gilette & Wit 1998). 44 Other well-known and often-cited cases of redundancy are tautologies (i.e. redundancy on the lexical level) or the structural characteristics of English informational questions, where there is a combination of interrogative pronoun, word order and intonation to mark an utterance as a question in speech (cf. Gilette & Wit 1998). <?page no="129"?> 117 is that COs are redundant but meaningful elements which through their repetitive nature express meaning, namely intensification, temporal boundedness and conceptual load, in an iconic fashion. 4.6 Summary and conclusion Chapter 4 has taken up a discussion of several issues which are indispensable for a proper treatment of COCs. Applying the descriptive inventory of CL, I have offered a description of the conceptual content of COCs which vividly depicts the construction as a pattern which combines two linguistic units profiling different aspects of the same conceptual base. The result of the integration of these two units is a complex, more or less redundant expression. However, as was unmistakably contended, redundancy is not to be understood as 'empty of meaning' or 'unnecessary repetition': It could be demonstrated that the overlap of conceptual content of verb and object in COCs clearly has an intensifying effect for the verbal expression of an event. The special formal characteristics of the COC are motivated and display diagrammatic iconicity in that such an increase in intensity is reflected in a more of form. COs, which must be considered as a group of verbal complements of different but related types, have been portrayed as true participants in an event, which are - depending on the type of CO - either effected or affected by an AGENT 's activity. Especially in the case of those constructions including an Eventor Event-Result-CO, such an interpretation draws on 'special resources': It is evoked and facilitated through an interplay between the semantics of the monotransitive construction (eliciting coercive effects) and a speaker's ability to construe events as bounded entities which are created by an AGENT involved in a particular activity. Combining facets of Langacker's and Goldberg's models of verbal complementation, I have proposed a hybrid-version which unambiguously captures Event- / Event- Result-COs as central rather than as peripheral participants. In sum, with regard to the 'argument vs. adjunct' controversy, I clearly advocate a description of COs as arguments, although, as explicated in section 4.4.2, these two syntactic categories merge at their peripheries. Having suggested solutions to several problems which are inherently associated with a discussion of COCs on a more theoretical and intuitive basis, I shall now continue this investigation with considerations much more exploiting 'real usage' data an approach which, although postulated as lying at the core of the CL paradigm, is still far from being the default case in CL analyses of linguistic phenomena. <?page no="130"?> 118 5 Consulting the corpus: Towards a usagebased network of COCs 5.1 Introduction The preceding chapter was devoted to a cognitive-linguistic description of semantic and formal characteristics of COCs that had so far been given a detailed account only within the generative-linguistic framework and/ or had only been touched upon by linguists working with the CL paradigm. For the illustration of my claims and ideas, I quoted examples of COCs found in the BNC or on the Worldwide Web as evidence from naturally occurring texts, yet without going into detail about frequency of occurrence, distributional patterns etc. In this and the following chapters, the focus is shifted to a careful statistical analysis of usage data, i.e. my compilation of COCs extracted from the BNC, by which I hope to gain valuable insights into the actual use of the construction by native speakers of English. Such a methodological procedure enables the specification of characteristics of COCs which are not solely based on the linguist's intuition but reflect patterns which are conventionalised by and possibly entrenched in speakers. None of the studies on COCs discussed so far in this book drew on an extensive survey of real language data, the only exception being Macfarland's (1995) comprehensive discussion. However, she merely pointed at tendencies of usage and employed naturally occurring examples of COCs as a means of illustration for the claims postulated in her study, neglecting a statistical validation of formal and semantic characteristics in the use of the construction. The overall aim of the investigation conducted here is the development of a usage-based network of constructions which draws on both the outcomes of the more theoretical descriptions provided in Chapter 4 and the findings of my corpus research presented and interpreted in the following chapters. The constructional network will be worked out in a step-by-step fashion, from highly schematic constructional templates at the top, via midlevel schemas incorporating more subtle specifications of form and meaning, to highly specified, in some cases lexicalized, forms, found at the very bottom of such a network. The analysis involves a discussion of issues relating to constructional patterns, transitivity, aspect, and language register, the results of which will be incorporated into the particular levels of schematic representation. At the same time, the claims and constraints on form and meaning of the construction as they are proposed in the literature <?page no="131"?> 119 (concerning such aspects as the usage of passive COCs or the make-up of the CO nominal) will be checked against the corpus data compiled here. The present chapter is structured in the following way: After giving a short overview on the promising and fruitful interrelationship between Cognitive Linguistics and corpus-based descriptions of linguistic phenomena, I will, in the second part, describe methodological considerations and procedures relevant for my empirical investigation, and then go on to explore possible interrelations between different types of COCs and particular lexemes or group of lexemes, applying the method of collostructional investigation developed by Gries and Stefanowitsch (2003, 2004, inter alia). Special prominence will be given to highest-level representations of COC schemas and the outline of a possible network structure. 5.2 CL and corpus linguistics: The usage-based approach to language Corpus linguistics, that is the systematic exploration of naturally occurring language compiled in large (electronic) corpora, has for a long time been neglected or even rejected in certain branches of linguistic research 1 , especially under the influence of Noam Chomsky and his intuition-based generative theory of language. In this framework, the intuition native speakers in general and linguists in particular have about their language was not only seen as adequate for the postulation of general rules of that language, but even judged as "better evidence than samples of real-life usage would be" (Sampson & McCarthy 2004: 3, italics in original). Sentences which were made up by language researchers and tested against their own or native speakers' judgments on the acceptability of the constructed forms constituted the whole test material used within the generative linguistics paradigm. Although in some corners of generative linguistics, a certain hostility towards corpus-based investigations into language still prevails, it has been increasingly accepted by the majority of linguistic schools that intuition does not suffice as the sole basis for a comprehensive analysis and description of language, that means linguistic phenomena on all levels of the language system. This change in favour of the 'consultation' of collec- 1 Before the "Chomskyan Revolution", linguists and language pedagogues such as Fries & Traver (1940), Bongers (1947) and Fries (1952) had already based their descriptive grammars or foreign language teaching materials / syllabi on language corpora (see McEnery & Wilson 1996: 2f.). Also, field linguists and language anthropologists collecting utterances from exotic languages for a minute description thereof applied methods which are similar to those of corpus linguistics. For a concise overview on the development of corpus Linguistics, see McEnery and Wilson 1996: 2-25. <?page no="132"?> 120 tions of usage data has come as the result of an increasing body of evidence that native speakers are no reliable source when asked to make judgments about the acceptability of expressions (cf. ibid.; Boas 2003: 12, Boberg 2002: 12). As a consequence, the analysis of language corpora has undergone an immense increase of value in recent years. Its general recognition in the wide field of linguistics has been nicely summarized by Michael Hoey (quoted in Sampson & McCarthy 2004: 5): "Corpus linguistics is not a branch of linguistics but the route into linguistics". In Cognitive Linguistics, corpus analysis is on its way to becoming recognized as an indispensable method of linguistic research, owing its rising status to the "commitment to empiricism" by those working within the paradigm (Gries 2006: 1) 2 . As noted before, the preferred model of linguistic representation in CL is the usage-based model, that is, a speaker's knowledge of the linguistic system as reflected in his/ her actual use of linguistic structures is considered as the basis of all grammatical and lexical description and any kind of abstraction thereof. With language corpora being samples of actual language usage, there is a clear intersection between corpus linguistics and the usage-based approach to language description. The analysis of corpus data can be carried out from a qualitative and a quantitative perspective. While in qualitative research, the corpus is 'investigated' for the identification and description of linguistic categories, providing a rich and multi-faceted view on particular phenomena, the quantitative approach relies on frequency counts and statistical inquiries to depict a "reliable and generalisable" portrait of the structure under investigation and its distribution (cf. McEnery & Wilson 1996). The latter method has been found a useful tool in usage-based descriptions of language, as in such a model, the frequency of usage of a particular grammatical form is assumed to both reflect and affect the representation of grammatical units in the minds of speakers. That is, the frequency of usage events as recorded in a language corpus is taken to provide insights into the degree of entrenchment of a specific pattern in speakers and therefore permits judgements about the status of such a pattern as a construction. Schmid (2000: 39) describes this correlation between frequency of usage and pattern entrenchment in his 'From Corpus-to-Cognition Principle': "[f]requency in text instantiates entrenchment in the cognitive system". Croft & Cruse (2004) formulate similar generalisations when they observe that 2 It was already noted in Ch.3 that, although the commitment to empiricism is at the core of the CL programme and 'usage-based' is a commonly-used term in CL research, empirical methods in general and corpus analysis in particular still lack the prominence one might expect under the given presuppositions tied to the framework (cf., for example, Mukherjee 2004 and Gries 2006, who appeal to a greater integration of corpus analyses into usage-based accounts of language). <?page no="133"?> 121 [e]ntrenchment of a construction is proportional to the number of instances of the construction at any level of schematicity, and to the degree of formal and semantic coherence of the instances of the construction. (ibid.: 309; emphasis added, SH) Since frequency at any level of schematicity contributes to the entrenchment of linguistic patterns, both type and token frequency need to be considered as parameters to evaluate a pattern's potential as an entrenched construction. That is, when a linguistic construction is to be tested for its potential entrenchment, one not only has to take into account the overall number of instances (tokens) found in the chosen corpus, but also the number of different lexemes (types) occurring in the relevant slot of the construction. With regard to an analysis of COCs, the above observations imply that attention must be paid not only to the frequency of the construction with a particular verb or the overall frequency of instances in the corpus (i.e. token frequency), but also to the number of different verbs which occur in the construction (i.e. type frequency). Whereas the former data will enable the detection of more or less fixed lower-level schemas (i.e. idiom-like phrases and partially lexically filled construction which allow for some variation as regards the lexical items that occur in the slots offered by the construction), it is especially the latter type of frequency information that will permit conclusions about the entrenched status of more abstract, schematic constructs which language users can draw on in order to sanction infrequent and new instances of the construction. Taylor (2002: 227) offers a 'formalized' version of this interrelationship between type and token frequencies and degree of entrenchment: A schema gains strength (it becomes 'established' or 'entrenched') in proportion to the type frequency of the instances which elaborate it. High token frequency of an instance entrenches the instance and weakens (or at least, does not strengthen) the schema. High type frequency of the instances entrenches the schema and weakens (or at least does not strengthen) the instances. (emphasis added, SH) Furthermore, the type frequency, i.e. the variety of verbs occurring in the COC as found in the corpus will allow us to specify semantic characteristics of the construction: We can test for its 'semantic coherence' as spelled out by Croft & Cruse (see above), which in turn presents further information on whether one can really talk about the COC or rather about a family of related constructions. Having given a summary about the usefulness of integrating data retrieved by an analysis of language corpora into a usage-based model of language, and having described the types of data which necessarily should be considered for the postulation of general statements about the representation of linguistic knowledge in a speaker's mind, I shall briefly comment on some methodological aspects of my investigation of the BNC for COCs. <?page no="134"?> 122 5.3 General methodological considerations The World Edition of the British National Corpus (BNC 2) 3 served as the basis for my corpus investigation of COCs. The corpus, currently being one of the largest electronic corpora of British English available, comprises 100,000,000 words of contemporary written (90%) and spoken English (10%), covering the period from 1960-1993, in which the texts were produced. The corpus is judged to be both representative and balanced, that is, no particular medium, variety, register etc. is disproportionately overor underrepresented in the corpus. For the retrieval of data, I made use of the search software tool SARA (version 0.941), which was specifically developed as part of the BNC project. In order to guarantee an appropriate statistical analysis of COCs as represented in the BNC, an exhaustive search of the corpus, retrieving all instances of the construction, would have been necessary (as will be shown in section 5.5 & Chapter 9, especially for a collostructional analysis the precise number of instances of a construction in a corpus is needed). However, several factors made such an ideal exhaustive search impossible: Since the COC is not tied to one particular lexical item (as are, for instance, the to-dative construction, the way-construction, or the into-causative construction), the search for all instances of the construction could not be carried out in form of a word query, the results of which could then, in turn, have been narrowed down to all actual occurrences of COCs. Furthermore, since the texts forming the BNC are not parsed, it was not possible, for example, to enter a search syntax allowing for an exhaustive search yielding all monotransitive constructions, from which then all COCs could have been manually filtered out. Therefore, I had to find a search procedure with which I would come as close as possible to the described ideal of an exhaustive search and decided to proceed in the following way: First of all, I collected a list of all verbs that are mentioned in the respective literature as occurring in COCs (i.e. either in form of concrete example sentences or as suggestions as to which verbs might be good candidates for the construction 4 ). Moreover, some studies point out individual semantic classes of verbs which are most likely to be found in the construction: Rice (1988) particularly suggests categories such as verbs of bodily function, performance and elimination, while Levin (1993) especially emphasizes verbs of non-verbal expression, verbs of manner of speaking 5 , and waltz-verbs (verbs which are zero-derived from names of dances such as waltz, boogie, 3 This version, released in 2000, is an improved version of the original BNC, released in 1995. 4 For such suggestions, see especially Rice (1988) and Levin (1993). 5 Levin (1993: 96) takes up Zwicky's (1971) suggestion that verbs of manner of speaking occur with a CO, yet doubts their actual usage in a COC. <?page no="135"?> 123 polka, cancan). I tried to extend these classes, searching for more potential candidates of these classes in dictionaries, thesauri etc. In addition, I ran a search in the British component of the International Corpus of English (ICE-GB), an electronic corpus comprising 1,000,000 words, which is not only POS tagged, but also fully parsed. Here, an exhaustive search for all monotransitive constructions was carried out in order to find further instances of the construction or verbs that might possibly enter the COC. The online version of the OED served as another source for possible examples of the construction: This computerized form of the classic dictionary of English allows for a systematic search not only of the entry words but also of the etymological information, the quotation section, entry dates etc. given for each lemma. Using the "Simple Search" tool, I ran a search of the 'definition' and 'etymology' sections, looking for entries containing the phrase 'cognate object'. This search yielded 63 hits. Finally, I included verbs that I intuitively considered good candidates for the construction. Having exploited all of these procedures, I was able to draw up a list of almost 400 verbs which were to be tested for their use in the COC in the BNC. As for the concrete search in the BNC, the query builder tool provided by SARA was used. Here, I designed a query that searched the corpus for a particular verb from the compiled list in all its morphological forms (resulting in a lemmatized search) plus the cognate noun in its singular and plural form, both preceding 6 and following the verb within the range of a socalled sentence-like element. I preferred this range over a range with a fixed concrete number of words between verb and CO for the following reasons: Firstly, co-occurrences of verb and noun crossing sentences boundaries were excluded as results, and secondly, the range between verb and noun was 'granted' a possible maximum within a sentence, which ensured the inclusion of instances with heavily pre-modified NPs following the verb or examples in which other linguistic material interfered between verb and CO 7 . The application of the outlined procedure led to the following overall results: Of the 400 verbs tested, 109 verbs were found to actually occur in the COC, yielding an overall number of 3,139 instances of the construction. All instances were then manually tagged for selected features that I consider of interest not only for the general characterisation of the construction but also for a comparison between claims made in the literature about re- 6 The option "preceding the verb" was included to retrieve possible instances of passivized COCs and fronted or clefted COs as in It was a great song they sang. 7 See, for example, the following hits, in which we find a great amount of 'interfering' material between the verb and its cognate object: a) He was screaming the high-pitched, nerve-grating scream of agony that always turned Kathleen's blood to stone. (JYB: 282) b) Instead, he laughed one of those raucous Hancock's Half Hour snide laughs. (J0W: 2451) <?page no="136"?> 124 strictions on COCs (see Ch. 2) and their reflection in actual language data. The following characteristics, which will be described in more detail in this and subsequent chapters, were collected for further analysis: 1. Semantic class of verb 2. CO type 3. Type of determiner in CO-NP 4. Type of modification in CO-NP 5. Voice 6. Register (i.e. text type of source text) Before I now finally turn to the discussion of my findings with regard to the frequency and distribution of the listed features, one final word of caution with respect to my corpus analysis and the conclusions drawn from the results should be given here: The generalizations and conclusions I try to arrive at by a close and careful inspection of the collected data are abstractions of the material as recorded in the BNC. That is, even though I base my discussion on a large set of data, this does not mean that I am able to capture each and every facet of the construction. There might exist forms which are so rarely used that they 'escaped' their representation in the BNC. Similar remarks can be made for those verbs of my list of 400 lexemes that could not be detected in the COC in the course of my investigation of the BNC: The fact that instances involving these verbs are not recorded in the corpus does by no means rule out the possibility of their usage in the COC. Nevertheless, the fact that they do not occur must be taken as an important indicator of the usage of these forms. Finally, my findings will be restricted to the use of COCs in British English. Corpora representing other national varieties of English are available and I could have extended my analysis to these 'language collections'. However, the corpora usually vary in size and every corpus is composed in a different fashion as regards the proportion of texts of different language dialects, modes, register etc. In order to avoid a comparison of 'apples and oranges', I decided to restrict my study to one corpus only, i.e. the BNC, hoping to be able to depict a relatively comprehensive picture of the COC in British English usage - despite the shortcomings mentioned. 5.4 Semantic classes and COC types As noted above, the close analysis of the BNC yielded 3,139 instances of COCs, involving 109 different verbs. Table 5.1 provides an overview on the <?page no="137"?> 125 Verb CO Total COC Type of CO semantic class of verb (1) live life 699 EV/ R 1 existence (2) sing song 466 EV/ R 1 performance (3) tell tale 401 EV/ R 1 verbal communication (4) smile smile 202 EV/ R 1 non-verbal communication (5) sow seed 198 A putting (6) produce product 141 R 2 creation (7) give gift 128 A change of possession (8) build building 100 R 2 creation (9) die death 87 EV/ R 1 disappearance (10) think thought 78 EV/ R 1 mental activity (11) see sight 72 A perception (12) do deed 67 EV/ R 1 execution (13) name name 45 A verbal communication (14) dream dream 45 EV/ R 1 mental activity (15) weave web 26 R 2 creation (16) smell smell 24 A perception (17) feel feeling 23 A perception (18) drink drink 23 A ingesting (19) feed food 21 A ingesting (20) fight fight 19 EV/ R 1 social-interaction (21) grin grin 18 EV/ R 1 non-verbal communication (22) plant plant 17 A putting (23) sleep sleep 14 EV/ R 1 bodily process (24) dance dance 12 EV/ R 1 performance (25) laugh laugh 10 EV/ R 1 non-verbal communication … … … … … 3139 Table 5.1 Top 25 of verbs in COCs in the BNC frequency of the 25 most frequent verbs 8 occurring in COCs, including information about their semantic class (based on Levin's [1993] system, see below) and the type of the COC. As regards the types of COCs, I have come up with a threefold distinction, drawing on characteristics of the cognate object. This classification comes as a consequence of my observations that it is impossible to describe COCs as a single, homogeneous class rather they should be discussed as a family of constructions. With respect 8 See appendix for the complete table. All other tables presented in a condensed form only in this chapter are also listed in their full version in the appendix. <?page no="138"?> 126 to my classification, I follow the conclusions drawn in Chapter 4, where I spelt out different types of COs based on the semantic relationship between verb and object. I set up three classes of 'subconstructions' with the following labels: i) EVENT/ RESULT 1 (EV/ R 1 ). As typical examples we find live a life, smile a smile, die a death. As can be inferred from the literature discussed in Chapter 2, this type of COC seems to constitute the 'classical'/ prototypical type of the construction. In the majority of cases, the object can express either the event itself or the result of the action, depending on a speaker's profiling of the scenario (cf. section 4.3.1). In the case of an event interpretation, the object is construed as a bounded entity by summary scanning (weakened form) of the single component states of the verb plus a shift in profile from process to thing. Each single component state of the verbal process contributes to the existence of the object. In the case of a result interpretation, the final component state is profiled, a mental process known as 'end-point-focus'. This type of COC is assumed here to verbalize an 'intensified action' construal, as it relates the AGENT and the action in which he is involved in a twofold way (in the form of the verb and its nominal variant in the CO-NP). Many of the instances of this category which contain a CO that is premodified by an adjective allow for a 'simple verb + adverbial paraphrase (He smiled a cheerful smile. - He smiled cheerfully., see Ch. 9) and are frequently found in LVCs (He gave a cheerful smile., see Ch. 10). ii) RESULT 2 (R 2 ). Typical examples of this category are weave a web, build a building, produce a product. These COs correspond to the socalled objects of result or effected objects. They designate concrete entities which come into existence through the action expressed. Unlike the first class, they do not allow for an event interpretation. Also, instances of this category with premodified cognate NPs cannot be paraphrased by a 'simple verb + adverbial' construction. (He wove an intricate web. - *He wove intricately.) iii) AFFECTED (A). Typical examples of this class are sow a seed, drink a drink, smell a smell. It comprises those cognate objects which belong to the class of affected objects 9 . In contrast to the COs in the two other classes, these objects do not come into existence by the action/ event, but are affected by the action denoted by the verb. 9 The members of this class are not treated as COs in most studies. They are the kinds of COs that are, according to Huddleston & Pullum (2002: 305) "of no syntactic significance" (cf. Ch.2, p. 15). For reasons that have been discussed in Ch. 4, I include this type of object in my study, especially for the demonstration of category structures and the development of a network of COCs. <?page no="139"?> 127 Forms containing a premodified CO cannot be expressed by a 'verb + adverbial' form, yet, some of them occur in LVCs (have a drink/ a smell/ a feeling). Some of the nouns are borderline cases and certainly could have been categorized differently. Especially the COs song, dance and tale, which are labelled EV/ R 1 in my system, are not uncontroversial with respect to their classification. They deviate from most other nouns in this class in that they are ambiguous as to whether they come into existence only through the action designated by the verb or are already in existence: Songs, tales and dances, of course, usually have some 'predetermined' form (i.e. types), and thus can be said to exist independently of their performance. However, every time they are performed they are created anew through the action of singing, dancing and telling, respectively. That is, their performance creates unique tokens which do depend on the particular event. Moreover, in combination with their cognate verbs, all three nouns designate events rather than concrete objects, a feature which clearly distinguishes them from the items in the R 2 class. Tale and song can, of course, also refer to more concrete (i.e. prefabricated) entities the polysemy of these words can be explained along the lines of the so-called conduit metaphor. In a now classical paper, Reddy (1993 [1979]) investigates the metaphorical framework which shapes and affects the concepts which language users have of communication. At the core of this framework lies the assumption that human beings have a repertoire of thoughts, ideas and feelings ('repertoire members') which they can 'set free' and deliver to the external world (that is, transfer these thoughts to others) by packaging them into 'linguistic containers' (signals) such as words, phrases and sentences. Verbal realisations of the conduit metaphor (e.g. There is doubt in your words. It's hard to capture these feelings in words.) are permeating the English language, and Reddy demonstrates how difficult it is to find equivalent expressions which do not draw on this metaphor. This particular idea about communication is certainly responsible for the meaning of tale or song as fixed forms (concrete objects) since they can be written down, i.e. 'packaged' into containers. Furthermore, according to Reddy (ibid.: 179f.), the conduit metaphor has brought about the polysemy of all those lexemes which denote such containers. Words such as text, essay or poem not only refer to the letters, marks or sounds of these forms, but have developed a sense which relates to the "emotions and concepts assembled in [their] reading", the "conceptual and emotional material" (ibid.: 178f.). This second sense is primarily associated with the reception ('unwrapping') of such 'transferred messages'. However, Reddy does not consider the case of the performance of a poem or the deliverance of a tale, for example, that is, its active replication. It might be argued now that a third meaning for these signal-denoting <?page no="140"?> 128 lexemes exists in relation to such an active 're-packaging' of the substance. Due to different artistic abilities, mnemonic capacities and emotional states of the respective performer, the seemingly fixed form is altered during this re-packaging process. This active re-creation is mirrored in the event reading of the respective lexemes, which is 'triggered' in combination with the COC-verbs, verbs which denote performance. These considerations led me to categorize the nouns in question as EV/ R 1 -COs instead of assigning them to the R 2 category. Similarly controversial are the nouns feeling, smell, sight and taste (i.e. the COs of perception verbs), which I list as members of the category AFFECTED . All of these nouns designate stimuli which are perceived by an EXPERIENCER (i.e. 'affected' by his/ her perception). They might as well be described as entities which only arise through neural activity and thus labelled EV/ R 1 . Once more are we faced with fuzzy and overlapping categories. As regards the type and token frequencies of the single classes, the following figures could be ascertained: More than half of the 109 forms fall into the COC class comprising those instances which are most commonly described as COs in the literature, i.e. examples of the 'classical' type: 57 items (types) found in 2217 concrete instances of usage (tokens) are of the kind that designate abstract results of actions or events. 40 items (625 instances) are categorized as affected COs, and only 15 items (297 instances) are grouped as effected objects of the type R 2 (see Fig. 5.1). On the basis of these data it might be hypothesized that it is the EV/ R 1 type that forms the 0 10 20 30 40 50 60 EV/ R1 R2 Aff Types Fig.5.1 a The distribution of CO-types in the BNC <?page no="141"?> 129 0 500 1000 1500 2000 2500 EV/ R1 R2 Aff Tokens Fig.5.1 b The distribution of CO-tokens in the BNC core of a network of COCs, since both type and token frequency point at a prominent status of this subcategory. As for the ranking of the single verbs according to raw frequency, the findings of my corpus investigation are similar to the ranking set up by Macfarland (1995). In both my and her study the pairing live + life is found most frequently, and most of Macfarland's 'top twenty' verbs are listed among my 'top thirty' 10 . These include the combinations sing + song, smile + smile, tell + tale, die + death. The only 'outliers' in her compilation are the verbs shoot, walk and chuckle 11 , which appear considerably less frequently or not at all in the BNC; in my list, on the other hand, it is weave + web which is ranked much higher. The above data give a first indication of preferences of particular subtypes of COCs and of verbs that seem to be preferably used in these different types. These tendencies will be tested in a more elaborate analysis in section 5.5. The second feature I selected for discussion relates to the semantics of the COC-verbs. On the basis of general semantic characteristics the verbs occurring in the construction were grouped into categories, with the aim to detect semantic idiosyncrasies of COCs and to identify classes of verbs that are preferably used therein. My categorization of semantic classes of the verbs is based on Levin (1993), who set up 57 semantic classes of verbs, which are in turn made up of more than 190 semantic subcategories. For the classification of 'my' verbs I decided to make use of Levin's more gen- 10 As noted in section 4.3.4, the criteria I used for the selection of COCs differ from those applied by Macfarland. Therefore, several of 'my' thirty most frequent verbs (sow, produce, build) are not included in Macfarland's discussion. 11 It must be kept in mind that Macfarland's (1995) collection partially builds on examples quoted in reference grammars, which, in turn, are often constructed by the respective author. <?page no="142"?> 130 eral categories (e.g. verbs of gestures) instead of applying her very subtly distinguished subcategories (e.g. wink verbs) in order to recognize some semantic clustering. It must be noted that for some verbs (11) it was difficult to find a more general group they could be put in. These cases were left uncategorized. The picture that emerges after labelling the verbs hints at a rather heterogeneous character of the COC as regards the semantics of the verbs used therein, although some tendencies can be observed. All in all, the 109 verbs belong to 31 semantic classes of the most different nature, ranging from verbs of creation (build, weave, produce) over verbs of perception (feel, smell, see) or of mental activity (think, dream) to verbs of existence (live) and of disappearance (die). No general underlying notion can be abstracted for all these groups. However, some of the classes include more items than others a fact which allows for a comment on some tendencies. Creation verbs and verbs expressing some notion of verbal utterance/ communication are the two classes which clearly stand out from the other classes with respect to the number of items in each category. Since COs most often designate created objects, it should come as no surprise that it is the class of creation verbs that occurs more often than others in a COC: 14 verbs fall into this class, all of them taking COs of the R 2 type. The group of verbs referring to different types of verbal communi cation 12 (e.g. tell, talk, yell) is ranked second and comprises 11 items, here, however, all but one of the verbs appear with COs of the EV/ R 1 class. Surprisingly, one semantic category of verbs which is not mentioned at all in any of the studies consulted could be singled out in the analysis: the class of verbs of manner of motion (e.g. walk, jump, fly), amounting to 6 different verbs found in a COC. Two of the classes suggested as typically appearing in the construction by Rice and Levin (see p. 122) are represented in my collection, too, but not to such a great extent as might have been expected from the authors' predictions: The classes of verbs of bodily processes (e.g. sleep, cough, yawn) and of non-verbal communication (e.g. smile, grin, frown) both comprise 5 different verbs, almost exclusively taking EV/ R 1 -COs. The same number of items is found with the groups of verbs of putting (e.g. load, spray, drop), verbs of perception (see above) and verbs of attaching (e.g. attach, button, nail), all of them occurring in the type of COCs taking affected objects. All in all, it can be observed that there is a strong correlation between the type of COC and the semantic verb class, as most of the semantic categories occur in one particular type of COC: Whereas the classes of verbs of communication, verbs of bodily processes, verbs of gesture and non-verbal communication and verbs of manner of motion exclusively occur in the construction of type EV/ R 1 , perception verbs and put- 12 This group includes 5 verbs of manner of speaking, a class that was suggested by Levin but at the same time considered as unlikely to occur in actual usage (see ftn. 5). <?page no="143"?> 131 semantic verb class frequency semantic verb class frequency creation 14 obtain 2 throw 2 bodily impact 1 verbal communication (*types of verb. commun.) (*manner of speaking) 11 (5) (5) characterize 1 manner of motion 6 combine 1 bodily process 5 cut 1 non-verbal communic. 5 disappearance 1 perception 5 dress 1 attaching 5 existence 1 putting 5 group existence 1 change of possession 4 killing 1 ingesting 3 provide 1 social interaction 3 psych verbs 1 change of state 2 separating 1 gesture 2 weather 1 measure 2 animal sound 1 mental activity 2 (uncategorized 11) Table 5.2 Frequency of semantic verb classes 13 ting verbs, for example, are only found in the A-type. The construction type R 2 entirely consists of verbs of creation. Such a correlation between semantic verb class and COC class, comes, of course, as the logical result of my description and distinction of COC types into categories which are semantically defined. There are various other semantic verb categories that are represented by one or two tokens only, resulting in the heterogeneous compilation of verb classes mentioned above. Thus, summarizing my observations from the analysis which was directed at the question of whether COCs are preferably used with particular semantic classes of verbs, the following can be stated: Different COC subtypes obviously prefer some semantic classes of verbs, but cannot be restricted to these. Especially for the EV/ R 1 class it seems that the only restriction for verbs to occur in this type of COC is that the event or action expressed by the verb can be construed as an abstract object that is created by an AGENT involved in that action. Because of this restriction, an explanation can be given for why virtually no verbs which, for example, designate permanent states 14 or body posture are found in this 13 The class "verbal communication" comprises the two outstanding subclasses "communication" and "manner of speaking" which are here listed separately. 14 The status of the verb live is intermediate between a stative and a dynamic interpretation, and common reference grammars disagree on this point. While Quirk et al. (1985) treat it as a stative verb; Downing & Locke (1992) assign to it a dynamic reading if an endpoint is established. It is the latter interpretation which permits the use <?page no="144"?> 132 constructional subtype: Since these verbs refer to relations which are perceived as stable through time, single component states are unlikely to be construed and thus cannot undergo summary scanning or an end-point focus transformation. Moreover, the relations designated by these verbs are non-energetic, a fact which shows in the unlikeliness of the following examples (which were judged highly unacceptable by native speakers of English): (1) a. *He sat a patient sit in the waiting room. b. *They stayed a 5-day stay at the hotel. c. *The couple owns a huge own. This is in accordance with the predictions made in Chapter 4 (p. 104f.) concerning the energetic characteristics of the verbs most likely to occur in a COC. There it was claimed that it is verbs designating relations which can be construed as actions requiring some amount of energetic input and create some 'output' which must be expected in the constructions. The corpus data clearly support this claim. Issues on the construal of energetic transmission will be taken up again in Chapter 6. 5. 5 A collexeme analysis of COCs - Measuring the association between verb and construction The discussion in the above section is based on raw frequency data, which can be taken as a first indicator of usage patterns if we consider type and token frequencies as a reliable source for such judgments. Therefore, one might be tempted to draw conclusions on the basis of these data concerning collocational preferences between the construction and instantiating verbs. However, as Stefanowitsch and Gries (Stefanowitsch & Gries 2003; Gries & Stefanowitsch 2004a) demonstrated, raw frequency data of verbs in constructions can be misleading to some extent, as the interpretations of such figures do not take into account the fact that verbs with a high frequency in the corpus in general are obviously more likely to occur in a particular construction than less frequent verbs. Moreover, these data might give rise to the assumption that verbs which occur less often in the construction are attracted to a lesser degree than verbs which appear with a higher frequency, neglecting the fact that, for example, a verb which has an overall frequency of 298 instances 15 in the corpus, occurring 122 times in a constructions is seemingly attracted to a much higher degree by this construction than a verb which is used 550 times in the construction, but has of the verb in a COC, which - through its aspectual characteristics to be laid out in Chapter 7, contributes a temporal boundary. 15 These figures are hypothetical and merely used as means of illustration. <?page no="145"?> 133 an overall frequency of 12,800 in the corpus. Driven by these considerations, Stefanowitsch & Gries developed and introduced several different, yet related methods of studying the relationship between constructions and the words used therein, known by the cover term "collostructional analysis". Besides collexeme analysis (ibid. 2003), which is utilized in this section, they also established distinctive collexeme analysis (Gries & Stefanowitsch 2004a) and co-varying collexeme analysis (ibid. 2004b) as tools for measuring such associations (see Ch. 9 for an application of the distinctive collexeme analysis). In order to describe the associations between verbs and COCs, I conducted the so-called collexeme analysis, which measures the collostructional strength between a construction and its collexemes (i.e. those lexemes which are attracted to a particular slot in the construction). The advantage of this procedure is that it takes into account the overall frequency of all verbs in COCs and in all other sorts of constructions in the corpus, the overall amount of constructions in the corpus and the ratio between these figures (cf. Stefanowitsch & Gries 2003: 216; Gries, Hampe & Schönefeld 2005: 646). The method, which is based on the Fisher-Yates Exact-Test 16 , provides results which indicate whether a particular lexeme occurs in a construction more or less often than expected by chance and thus can be used as a measure of the strength of attraction or repulsion between word and construction. Moreover, as could be experimentally demonstrated by Gries, Hampe & Schönefeld (ibid.), rankings obtained by a collostructional analysis represent actual language usage more adequately than rankings elicited through raw frequency counts. For my calculations, I 'fed' the different types of frequency needed for such an analysis into the program Coll. analysis 3. A program for R for Windows 2.x (Gries 2004). The below 2x2 table shows, for example, which frequencies of occurrence are necessary to measure the collocational strength between the COC and the verb live: live live Row Totals COC 699 2,440 3,139 - COC 26,802 10,176,359 10,203,161 Column Totals 27,501 10,178,799 10,206,300 Table 5.3 Frequency data necessary for the analysis of collostructional strength 16 See Stefanowitsch & Gries (2003: 217f.) for a discussion of the advantage of the Fisher-Yates Exact Test as compared to other statistical methods for measuring collocation strength, such as z-score, t-score or ²-test. See Oakes (1998) for an overview of advantages and disadvantages of available tests. <?page no="146"?> 134 The table must be read as follows: Of the 27,501 times the verb live is recorded in the BNC, it occurs in the COC 699 times, while in 26,802 cases the verb is used in other types of constructions. The COC, on the other hand, has a frequency of 3,139; that is, all other verb constructions (i.e. 10,203,161) of an overall amount of 10,206,300 verb constructions in the BNC (as measured by Stefanowitsch & Gries 2003: 219) belong to other categories of constructions. 10,176,359 of these constructions neither involve a COC nor the verb live. The figures in bold print served as input for the software tool Coll.analysis 3. Table 5.4 shows the outcomes of my calculations. The data in the column 'collostructional strength' are given in the form of the negative logarithm to the base of ten of the p-value, for which, if they are 'translated' into p-values for standard levels of significance, the following correspondences hold (cf. Gries 2004): coll. strength > 3 = p <0.001 coll. strength > 2 = p < 0.01 coll. strength > 1.30103 = p < 0.05 The results must be interpreted as follows: The higher the index is, the stronger is the attraction or repulsion between the verbs and the COC; any value which is bigger than 3 indicates an attraction or repulsion between Collexeme OF 17 coll. strength Collexeme OF coll. strength 1 live (EV/ R 1 ) 699 Inf 17 feed (A) 21 16.0352252 2 sing (EV/ R 1 ) 466 Inf 18 dance ( EV/ R 1 ) 12 11.0907643 3 tell (EV/ R 1 ) 401 Inf 19 fight (EV/ R 1 ) 19 10.1506220 4 sow (A) 198 Inf 20 farm (A) 6 9.1245645 5 smile (EV/ R 1 ) 202 297.9094632 21 sleep (EV/ R 1 ) 14 8.7065932 6 produce (R 2 ) 141 118.8687254 22 sigh (EV/ R 1 ) 8 6.1915389 7 build (R 2 ) 100 83.1064042 23 think (EV/ R 1 ) 78 5.9205229 8 dream(EV/ R 1 ) 45 74.1631838 24 light (A) 8 5.6867855 9 die (EV/ R 1 ) 87 67.0985713 25 pray (EV/ R 1 ) 7 5.0026088 10 name (A) 45 49.3171758 26 tie (A) 7 4.2810390 11 weave (R 2 ) 26 42.5363268 27 edit (A) 5 4.0938225 12 give (A) 128 32.8022754 28 paint (R 2 ) 7 3.8935361 13 smell (A) 24 30.6824942 29 laugh (EV/ R 1 ) 10 3.4179565 14 grin (EV/ R 1 ) 18 20.5684481 30 yawn (EV/ R 1 ) 2 2.4074298 15 drink (A) 23 19.2846741 … … … 16 plant (A) 17 18.3931186 Totals 3139 Table 5.4 Top 30 of significantly attracted collexemes in the COC 17 OF = observed frequency <?page no="147"?> 135 verb and construction which is highly significant. The direction of association, i.e. whether a verb is attracted or repelled, can be deduced from a comparison of expected and observed frequency: Verbs that occur more often in the construction than expected by chance tend to be attracted, whereas verbs that occur less often in the construction than expected by chance tend to be repelled. Bearing in mind that a collostructional strength of 3 equals a p-value of 0.001, pointing at a highly significant association, we can conclude that of the 109 verbs 29 verbs are attracted to a great extent by the COC, as shown in Table 5.4: The data indicate that the collostructional strength between the construction and the verbs live, sing, tell and sow is infinite (all of them are also in the 'top five' of frequency of occurrence), so that these verbs can be considered as the most 'reliable' representatives of the construction. They form the group of those candidates which are most likely to appear in the construction, i.e. they possibly represent entrenched instances of the COC. However, it must cautiously be noted here that so far no measure for a pattern's degree of entrenchment is available and that neither raw frequency nor such measures as collostructional strength correlate in a 1: 1 fashion to the degree of a construction's entrenchment in speakers. The values (based on p-values) used here only serve as an indicator of the relative importance of a collostruction and as an orientation for the postulation of possible candidates for entrenchment (cf. Gries & Stefanowitsch 2003, ftn. 6). If the data yielded by the calculation of the collostructional strength between verb and construction are compared with the raw frequency list, it seems that the results do not differ to a great extent. The analysis brings to light, however, that although the verbs sow and smile are used similarly frequently in the construction (198 vs. 202), the former shows a much higher index of attraction to the COC. Other verbs, too, undergo a 'revaluation' or 'devaluation': Whereas weave and grin turn out to be 'stronger' candidates of the COC than could be read off from their raw frequency of occurrence, especially the verbs think, feel and see seem to 'lose' of their importance. The latter observation doubtlessly highlights the advantage of the collexeme analysis of taking into account the frequency of occurrence of the search items in other constructions besides the investigated pattern. Think, feel and see are recorded in the BNC as highly frequent verbs, being more strongly associated to verbal constructions other than the COC. Returning to the 'top 4' candidates, we note that except for sow, all of these infinitely attracted verbs belong to the EV/ R 1 category that was already alluded to as the 'prototypical' or 'classical' form of COCs. Considering the other verbs that are strongly associated with the construction, we find that here, too, the majority of them are members of the EV/ R 1 category. Altogether, 15 of the 29 highly significant verbs are members of the <?page no="148"?> 136 EV/ R 1 category, with verbs of non-verbal expression (4) being most frequently represented; 10 belong to the A type, with no semantic verb class prevailing; 4 are instances of the R 2 type, all of them being verbs of creation. While the procedure applied here is not suited as a statistical measure of prototypicality 18 , the findings are, however, not in conflict with the statements and assumptions made so far on the more prominent status of COCs involving COs of the type EV/ R 1. Moreover, the results allow for a semantic description of COCs in which the meaning 'intensified action', which was described to be characteristic of EV/ R 1 -COCs (see Ch. 4), is granted a central role. The formulation of a more concrete meaning is impossible, since even in the group of the significantly attracted verbs no semantic class prevails to a considerable extent. This confirms Stefanowitsch & Gries' (2003: 229) assumption that the semantics of the strongest collexemes of a construction are not in any case a sufficient indicator of the construction's meaning. Of the 109 verbs that were tested for their collostructional strength with COCs, 30 verbs are measured as being repelled by the construction, 4 of them to a significant extent (see Table 5.5). Surprisingly, the relatively frequent pairing do + deed (raw frequency: 67) is the second most strongly repelled verb, an observation which would not have been possible on the Collexeme Obs. Freq. Coll. Strength 1 say (EV/ R 1 ) 1 40.8088888 2 do (EV/ R 1 ) 67 18.0000159 3 find (A) 2 9.4672782 4 work (EV/ R 1 ) 1 6.0269432 5 run (EV/ R 1 ) 2 2.7963248 6 develop (R 2 ) 1 1.9016093 7 offer (A) 2 1.7625744 8 decide (A) 2 1.4338407 9 cut (EV/ R 1 ) 1 1.3674997 10 lose (A) 3 1.1659835 11 describe (R 2 ) 3 1.0300209 12 love (A) 1 0.9467163 13 talk (EV/ R 1 ) 5 0.7556484 14 serve (EV/ R 1 ) 2 0.7284872 15 walk (EV/ R 1 ) 3 0.6960773 Table 5.5 Strongly repelled collexemes of the COC 18 See Gries (2003b) for an attempt to determine constructional prototypes by means of highly complex statistical procedures. <?page no="149"?> 137 basis of raw frequency data, which shows do on rank 12 of the most frequent verbs found in a COC. This repulsion between verb and construction is due to the high overall frequency of DO in the BNC and its infrequent usage in the COC if compared to other constructions in which the verb is used. The majority of the rejected verbs were found only once or twice in a COC so that these negative associations between word and construction are intuitively plausible. As with the class of significantly attracted verbs, no semantic verb category is dominant in the group of repelled verbs. 0 5 10 15 20 25 30 35 40 45 EV/ R1 R2 A Attr. Rep. In the course of the discussion of the data that were obtained from a collostructional analysis the advantages of this method have clearly shown: Firstly, it could be demonstrated that particular verbs are (most) strongly attracted or rejected by the COC - the analysis yielded results which could not necessarily be predicted on the sole basis of raw frequency data and enabled us to detect verb + CO pairings which may qualify as entrenched structures. Secondly, grouping the verbs which are significantly attracted by the COC according to the type of CO they occur with lends further support to the hypothesis that the EV/ R 1 -subcategory must be considered the core form of COCs. In accordance with the type and token frequencies presented in section 5.2 and with the observation that the majority of verbs that are most strongly attracted by the COC belong to this group, prototype tendencies must be recognized for this type. 5.6 A usage-based network of argument structure constructions So far, the discussion has aimed at portraying the COC as a family of subconstructions rather than describing it as one single construction with a highly abstract meaning. The results gained from the more general frequency analysis and the collostructional analysis, which yielded important insights into the distribution and semantic specifications of these construc- Fig. 5.2 Distribution of attracted and repelled verbs according to COC type <?page no="150"?> 138 tional subtypes, will serve here as the basis for the development of a network of COCs which in turn forms part of a larger network of argument structure constructions, including, among others, the double object construction, the resultative construction and the intransitive motion construction (cf. Goldberg 1995). Representing grammatical patterns in a network of related and overlapping forms with various levels of specificity constitutes a promising alternative to representing grammar as a list of more or less arbitrary 'rules', as done in the generative paradigm. The networkform of presentation captures the insight that [c]onstructions that share a particular feature (or a set of features) form clusters of mutually related generalizations about linguistic competence going from the most abstract and unconstrained to the most restricted (Fried & Östman 2004: 72). The network (see Fig.5.3), which 'zooms in' COCs, shows the different patterns as (peripheral) subtypes of the monotransitive construction and depicts the linkage of the subcategories of COCs with other monotransitive subtypes by means of instantiation links 19 . As spelt out in Chapter 3, the various nodes in such a schematic network are linked through categorizing relationships - cognitive routines which are defined as an "established comparison event assessing one node in relation to another" (Langacker 1987a: 379). For a construction A to be an instantiation of another construction B means that A includes all the features of B plus specifications which are idiosyncratic to A. Through these instantiating links, the COC types, for example, inherit a major part of their meanings from the respective monotransitive subtypes: As depicted in Fig. 5.3., the A-type is a particular type of the prototypical form of monotransitive constructions, having a fully elaborated AGENT - ACTION - PATIENT schema. What makes it a separate subcategory is the special formal/ morphological and semantic relationship between verb and object. Similarly, the R 2 -type has a link to the nonprototypical monotransitive construction with effected objects, being a special form of this pattern. The EV/ R 1 -type, however, which has been identified as the core form of the construction, has a somewhat different anchorage in the network: Besides being linked to the effected object construction as one of its instantiating forms, it is directly tied to the R 2 -type by a metaphorical extension link. The conceptual metaphor on which the connection between EV/ R 1 and R 2 draws is EVENTS / ACTIONS ARE OBJECTS , an ontological metaphor human beings constantly make use of to grasp the complex nature of events and actions (see Lakoff & Johnson 1980: 30ff.). 19 See Goldberg 1995, Lakoff 1987, Langacker 1991a, for different types of networks and links therein. In my network, instantiation links connect higher order schemas with more specific forms, extension links connect schemas which are related through semantic extension (e.g. via metaphor), following Langacker's approach to constructional networks. <?page no="151"?> 139 Whereas the created objects of the R 2 class are discrete, spatially bounded entities, in the EV/ R 1 class, the concept of boundedness is 'transferred' via metaphor to events and actions, construing them as delimited objects which can be manipulated, transmitted, created even. Through this special kind of link the EV/ R 1 type develops its unique meaning, construing a scenario in which an Agent is intensely involved in an action with the reified form of the action (either event or result) as its outcome. Yet another link, which is introduced here as 'similarity link' 20 , connects special kinds of COCs of the type EV/ R 1 to the intransitive construction, which is to account for those instances of this subcategory that have an 'intransitive verb + adverbial' (VAC) equivalent. Since such a constructional correspondence presupposes an adjectively premodified CO-NP, the similarity link is established between the VAC and a less schematic template of EV/ R 1 -COCs at the mid-level of the network. This arrangement, specifying the constructional make-up of the latter in more detail, already anticipates a more elaborate presentation of constructional schemas which is to be discussed in later chapters. It is included at this place to betoken the richness of information and links captured by such a type of network 21 . As Capelle (2006, esp. pp. 22f.) has discussed and exemplified by means of the particle placement alternation, links between alternating, semantically similar constructions can also be captured in form of categorizing relationships as described by Langacker (1987a: 379ff.), since speakers are capable of perceiving the similarities of such alternations. He views such a relation as an essential part of speakers' linguistic knowledge when he claims that "the similarity between two or more allostructions [i.e. formal variations of one constructional schema, SH] […] can itself be 'objectified' and included in the whole of what speakers (have to) know about their language" (ibid.: 22). In our case, the nature of the similarity link between the EV/ R 1 - COC and the VAC is somewhat controversial with regard to its directional specifications. Langacker specifies a bidirectional relationship, which holds for cases in which speakers perceive the similarity between two entities "without attributing primacy to either one" (1987a: 380). As neither of the construction (i.e. the COC or the VAC) is derived from the other or is consid- 20 I avoid the label 'synonymy link' used by Goldberg (1995: 90f.) to describe, for example, the relation between a ditransitive form and its to-dative paraphrase (e.g. Jane gave Marc a peach/ Jane gave a peach to Marc) for reasons extensively discussed in the present study: There is obviously a semantic and pragmatic difference between COCs and their adverbial counterparts. Whereas in the ditransitive -to-dative alternation no semantic roles are changed, added or 'lost', the COC clearly has an additional PA- TIENT role if compared to its adverbial paraphrase. 21 Mid-level schemas are, of course, available for all other COC-subtypes as well. For reasons of space and in order to include the similarity link, only one (still quite schematic) mid-level schema is depicted for EV/ R 1 -COCs, representing information on the elaboration of the CO-NP (see Ch. 8). <?page no="152"?> 140 ered 'more basic' than the other (a basic principle of CL), the postulation of such a bidirectional link seems most plausible. Yet, Langacker also posits a unidirectional relationship for cases in which speaker intuition or the prototypicality of one of the two entities grant primacy to one form. As will be shown in Chapter 11, the use of COCs (in particular of the EV/ R 1 type) is more or less restricted to certain types of registers, which yields a somewhat marked status of the constructions as compared to the VAC form. Thus, the 'intransitive verb + adverbial' construction might count as more natural, the COC as more peripheral, which would then justify a similarity relation with a unidirectional link from the VAC to the COC. However, drawing on Langacker's (ibid.) comment that nothing can guarantee a nonarbitrary choice of the categorizing direction, I represent the similarity link between VAC and COC as bidirectional, though 'weighted' towards one particular direction (indicated by the black, filled arrow), although, possibly, the question of the directionality of a similarity link can only be answered on a case-to-case basis. For verbs which are strongly associated with the COC, such as live, sing or smile, the weighting might be less strong (or even cease to apply) than for verbs only moderately attracted by the construction. A more detailed account on the nature of the similarity relationship between the two constructions will be given in Chapter 9. Note, furthermore, that with respect to alternating forms and the similarity links between these forms, Light Verb Constructions (to be discussed at length in Chapter 10) would need to be given some space in the network, too. As was noticed in the research overview in Chapter 2, scenarios expressed via COCs can be conveyed with subtle changes in meaning by LVCs. However, for the sake of clarity, LVCs are not included in the representation in Fig. 5.3. The columns on the lowest levels so far depicted in the network comprise constructions filled with those verb lexemes which have been identified by means of the collexeme analysis as most strongly attracted by the respective constructional subtypes. Note, however, that the network needs to be extended to even lower levels, capturing more detailed lexical specifications (i.e. preferred types of determiners or modifiers), which cannot be spelt out at this point of discussion. These more fine-grained characteristics of COCs as represented in the BNC will be taken up in later chapters of the study. Although the links between schemas and instantiating subschemas are represented in Fig. 5.3 by downward arrows, this does by no means imply that the model is couched into a top-down approach to linguistic structure; the arrows merely indicate the categorizing relationship of elaboration. In <?page no="153"?> 141 Intransitive Construction [ NP S V ] Monotransitive Construction [NP S V NP DO ] Affected Object Construction [NP S V NP DO/ A ] COC- EV/ R 1 [NP S V CO-NP DO/ E ] Effected Object Construction [NP s V NP DO/ E ] NP S + LIVE + LIFE NP S + SING + SONG NP S + TELL + TALE NP S + SMILE + SMILE NP S + DIE + DEATH NP S + PRODUCE + PRODUCT NP S + BUILD + BUILDING NP S + WEAVE + WEB NP S + PAINT + PAINTING NP S + RECORD + RECORD NP S + SOW + SEED NP S + GIVE + GIFT NP S + SMELL + SMELL NP S + SEE + SIGHT Simple Intransit. [NP S V ] Intrans. + Adverb [NP S V AdvP] Argument Structure Constructions Instantiation link Metaph. extension link EVENTS ARE OBJECTS Similarity link Eff. non-cog. O [NP S V NP DO/ E ] COC-R2 [NP S V CO-NP DO/ E ] COC-A [NP S V CO-NP Do/ A ] Aff. non-cog. O [NP S V NP DO/ A ] Modified EV/ R 1 [NP S V Det Adj CO DO/ E ] Fig. 5.3 A network of argument structure constructions <?page no="154"?> 142 fact, usage-based models of language as promoted by CL are framed into a bottom-up-orientation, "the schemas spring[ing] from the soil of actual usage" (Langacker 2000: 3). That is, low-level schemas (actual usage) provide the basis for the abstraction of higher-level pattern once speakers have recognized sufficient commonalities of similar usage event. As can be seen in the network, there is no highest-level schema "COC" which comprises all COC-subcategories. This is due to the fact that the representation in Fig.5.3 draws on a classification of subcategories of the monotransitive construction which is based on semantic characteristics (in particular semantic features of the direct object), but neglects formal similarities to some extent. An integration of formal commonalities of the constructional subtypes into the same representation would have resulted in a taxonomy interfering with the semantic category. Therefore, an alternative depiction is given in Fig. 5.4, which sketches a network representation giving priority to formal features of the different construction types and subtypes. This depiction includes a highest-level schema "COC", unifying formal and semantic characteristics shared by all COC subtypes, such as 'morphological cognateness of the object noun' (formal) or 'externally directed energetic transfer' (semantic, cf. Ch. 6). The networks developed here stand in full accordance with the CL approach to conceptual and linguistic categories organized as radial or proto- Fig. 5.4 Alternative representation of a constructional network of COCs Argument Structure Constructions Intransitive Construction [NP S V ] Monotransitive Construction [NP S V NP O ] Cognate Object C. [NP S V NP CO ] Non-Cogn. Obj. C. [NP S V NP Non-CO ] Effected COC [NP S V NP CO-E ] Affected COC [NP S V NP CO-A ] Effected N.-COC [NP s V NP Non-CO-E ] Affected N.-COC [NP s V NP Non-CO-A ] … COC-EV/ R 1 [NP S V NP CO-EV/ R ] COC-R 1 [NP S V NP CO-R ] Instantiation link Metaph. Extension link EVENTS ARE OBJECTS <?page no="155"?> 143 type categories: The close interrelation both between subcategories of one category and between subcategories of different categories, even allowing for overlaps at the periphery, mirrors the fact that category boundaries may not be clearly delineated, and that the transition from one (sub-) category to another may proceed gradually. 5.7 Summary and conclusion This chapter has provided a first outline of a possible network of COCs with specifications on subtypes of the constructions, basically motivated by the verbs which are most strongly associated with these patterns. Applying the method of collostructional analysis, not only those verbs which are strongly attracted by the construction could be detected, but also those which are repelled by it, that is, those verbs which, although they occur in the pattern, are more typically found in other constructions. These results clearly emphasize the advantage of collostructional analysis over raw frequency analyses, since the findings could not have been obtained on the basis of the latter. The examination of the semantic categories of the verbs associated with the different types of COCs could cast only little light on possible semantic features of the construction, since no clear tendencies concerning preferred semantic classes became apparent. Therefore, an extensive discussion of semantic / functional aspects has not been taken up in the above argumentation. The following chapters will consider questions which are indispensable for an elaborate description of the 'functional load' the construction carries, e.g. functions associated with the highest-level schemas in the network. As has been noted several times before, the notions of 'transitivity', 'energetic interaction' and 'verbal aspect' play a crucial role for the portrayal of COCs. Chapter 6 and 7 will offer accounts of COCs which discuss the construction against the background of these concepts. <?page no="156"?> 144 6 The Transitivity of COCs 6.1 Introduction Most, if not all, descriptions of COCs build on transitivity as a crucial definitional feature of the types of verbs that may enter such a construction. This chapter will provide a description of the construction's meaning incorporating a cognitive account of transitivity, which is closely tied to the structuring of events on the basis of the construal of energetic characteristics of the respective scenario. Different types and degrees of transitivity, usually reflected in the distinction between intransitive-unaccusative, intransitive-unergative and transitive verbs, will be scrutinized for the full characterization of COCs and the kind of construal the constructions may express for or impose on a specific event. Further, the question whether passive COCs are acceptable will be pursued in greater detail, by testing the data obtained from my corpus study against claims on the restricted use of such expressions, which are stated in the majority of studies on COCs. This topic is included in the argumentation, as passive constructions are often described as presenting a mirror image of the participant constellation of transitive events and therefore are of high relevance for a discussion of event construal as well. The description offered here applies to higher-level schemas of the network, with the distinction between the different subtypes of COCs, i.e. the forms EV/ R 1 , R 2 and A, being taken into consideration in the following discussion. As will become apparent, however, the classes of verbs associated with the different subtypes do not fully coincide with the unaccusative-unergative-transitive distinction. Although there is a strong correlation between the groups R 2 and A on the one hand, and verbs that are conventionally used in a transitive construction on the other, the group EV/ R 1 is of a "mixed" character with respect to transitivity features of the verbs occurring in this pattern. 6.2 Transitivity in Cognitive Grammar In contrast to the approaches surveyed in Chapter 2, in which the features 'transitive' or 'intransitive' were considered as inherent to the verb, defining its complementation characteristics (i.e. valency characteristics or subcategorization frames), Cognitive Grammar accounts of transitivity phenomena incorporate cognitive models and construal. Hence, it is not only the choice <?page no="157"?> 145 of an individual verb and its particular lexical specifications that sanction its use in a particular argument arrangement (i.e. construction) but as has already been alluded to in the preceding chapters a speaker's conception of an event in its specific setting. Rice (1987: 434) nicely summarizes the CL perspective on transitivity when she observes that "[t]he way we construe a scene or event, the way we conceive of participants interacting within an event, the way we access and experience cognitively is what determines its transitivity profile" 1 . Events, as understood in this study, are best described as "units of conceptual structure that allow human beings to quickly and conveniently structure complex, temporally extended experience for purposes of thought and communication" (Kemmer 2003: 94). In relation to verbal communication it has been observed that it is the linguistic unit of a clause which most closely corresponds to speakers' conceptualizations of simple events (cf. ibid.), with different linguistic categories and patterns being available to speakers for conceptually organizing events. Here, the transitive construction as a grammatical category operating on clause level constitutes one particular pattern which structures an event in such a way that the interaction of two participants is focused on. Following from this, it does make a difference whether a speaker utters Rick fought furiously or Rick fought a furious fight since the use of an intransitive construction, as in the former example, structures the event in a different way than does the transitive construction in the latter case. This view clearly stands in contrast to those descriptions that treat COCs as stylistic variants of expressions including manner adverbials, where the above sentence pair is said to have an equivalent meaning (cf. Jones 1988, Huddleston & Pullum 2002). One descriptive construct for the representation of how humans possibly conceptualize events (which, in turn, is thought to motivate clause structure) is Langacker's 'canonical event model', diagrammed in Fig. 6.1. The model combines several submodels which represent forms of interaction (the billiard-ball model) and perspectivization and grounding (the stage model), and so-called role archetypes, which shall be discussed in more detail in the following. 1 As regards the definition of transitivity as a grammatical phenomenon, Rice (1987: 422) puts forward the following view: "I agree that transitivity is something above and beyond the lexical or logical definition: a verb taking a direct object or one sustaining two arguments. But I'll argue that transitivity is not inherent to or derivable from properties of the morphology of the clause[…] The relevant parameters of transitivity are not always objectively given in the content of the sentence per se, but may be imposed from without, i.e. subjectively, by the speaker" (emphasis added, SH). <?page no="158"?> 146 Fig. 6.1 The canonical event model (Langacker 1991a: 285) Role archetypes are cognitive entities relating to participants in events: The archetypes defined by Langacker, such as AGENT and PATIENT (as the most prominent ones), but also INSTRUMENT , EXPERIENCER or MOVER may evoke notions of the analysis of thematic relations (i.e. semantic roles) of arguments in clauses, yet, as he argues, these roles must be understood as "pre-linguistic constructs grounded in everyday experience" rather than as linguistic concepts (ibid. 1991a: 285). They emerge from our interaction with the world in which we initiate actions ( AGENT ), move from one place to another ( MOVER ), manipulate objects ( PATIENT ) or make use of tools to achieve certain aims ( INSTRUMENT ). While these role archetypes characterize participants of events, the metaphorical billiard-ball model 2 captures the most basic constellations of interaction, involving these role archetypes. The model is organized around elements that correspond to such concepts as space, time, material substance and energy in the extralinguistic world and serves to reflect how human beings perceive and grasp interactions of any kind - namely in terms of "discrete physical objects that move about and interact energetically when they come into contact." (Langacker 1991a: 545). These 'basic ingredients' are employed for the definition of grammatical constructs such as nouns and verbs: 2 Note that the use of metaphors / metaphorical models for explanatory purposes is not an idiosyncrasy of cognitive linguistic research, but is common, "omnipresent" even (Ungerer & Schmid, 1996: 147) in science in general. Well-known examples are the VIRUS metaphor in computer science, the description of the spatial distribution of bound electrons as "electron cloud", the ascription of 'orbitals' to atoms as if electrons were planets spinning around a nuclear sun, the model of the brain as a 'computer' and thought as 'processing' etc. (cf. Boyd 1993, Kuhn 1993). AG PAT setting V <?page no="159"?> 147 Physical objects and energetic interactions provide the respective prototypes for the noun and verb categories, which likewise represent a polar opposition among the basic grammatical classes. (ibid.: 283) Taken together, the notions of role archetypes and the prominence given to them in energetic interactions as displayed in the billiard-ball model can explain basic characteristics of clause structures in English. To give an example: If one considers interactions as energetic chains, it seems plausible why the AGENT , i.e. the entity which initiates the energetic transfer, is much more likely to occur in the subject position of a clause than a PATIENT or an EXPERIENCER , which take only secondary or tertiary positions in such a chain. Moreover, as will be shown below, the energetic interactions captured in the model are crucial conceptual analogues of the concept of transitivity. A second metaphorical model Langacker introduces for his conceptual approach to clause structure is the stage model. It captures the fact that speakers, when verbalizing their perceptions, can usually only focus their attention on a limited section of their surroundings, especially on the objects occurring in this region. This focused section is compared to a stage, 'occupied' by the main characters (i.e. the participants focused on, most naturally AGENT and PATIENT ) which are involved in an interaction, taking place in a particular setting hence the name 'stage-model' (ibid. 1991a: 284 ). In diagram 6.1, the speaker is indicated by an encircled 'V', i.e. s/ he has the role of an offstage viewer, watching the 'play'. Each verbal expression made by the speaker can then be seen as the description of one discrete event of the 'play' performed on the 'stage': Langacker nut-shells the conceptual impact of the model as follows: [T]he stage-model idealizes a fundamental aspect of our moment-to-moment experience: the observation of external events, each comprising the interactions of participants within a setting. (ibid.) As has been pointed out above, specific linguistic patterns or categories serve to highlight particular structures of an event, and Langacker recognizes transitive finite clauses as those encodings that come closest to a full representation of the canonical event model, i.e. they verbally reflect the prototypical action: The notion of transitivity involves a number of parameters […] all interpretable as reflecting some aspect of the canonical event conception […]. There is then a natural correlation between this idealized cognitive model and the structure of a transitive finite clause. When a canonical event is coded linguistically in the maximally unmarked way, the clausal head is a verb that designates the agent-patient interaction, the agent and patient being coded by the subject and object nominals, respectively. (Langacker 1999a: 25, emphasis added, SH) <?page no="160"?> 148 While transitive clauses verbalize AGENT - PATIENT interactions, intransitive constructions prototypically code such conceptual archetypes as an object's motion along a path (Mary ran to the train station) or that of a static situation in which an entity is ascribed some kind of quality or inherent property (Michael is a good student). These encodings deviate to some degree from the canonical event model which typically involves an energetic transfer between entities: Whereas the former example encodes an energetic action but 'does without' the construal of a transfer of energy from one entity to another, the second example depicts a relation in which an energetic impulse is completely absent. In the quote given above, Langacker hints at a "number of parameters" of transitivity which can be seen as indicators of an event being a more or less prototypical instance of a canonical event. A whole compilation of such parameters is discussed in a study by Hopper & Thompson (1980), in which the authors attempt to isolate component parts of transitivity, intuitively understood by them as an activity carried over from an agent to a patient (ibid.: 251). Since their understanding of transitivity is similar to that of CG, drawing on the concept of a transaction of energetic momentum between two participants, a short discussion of the study shall be included here to illuminate what is not explicitly discussed by Langacker. 6.3 Hopper & Thompson (1980): Features of transitivity In a cross-linguistic study, Hopper & Thompson (1980) extracted 10 components of transitivity, which are rated 'high' or 'low' with respect to a transitivity scale; these factors are summarized in Table 6.1. The employment of such factors clearly suggests that transitivity is not to be seen as an absolute, all-or-nothing phenomenon, but rather as a gradation of more or less. That is, transitivity comes in different degrees, determined by the presence or absence of the specific features listed and by different combinations thereof. Accordingly, an expression reflecting the highest degree of transitivity (i.e. maximally representing the canonical event model) is a clause incorporating all the features listed under 'high', as in e.g. : (1) The hunter shot the boar. According to Hopper & Thompson's study, it is not only the presence of a direct object alone that yields a high degree of transitivity of a verbalized event. Thus, although both sentences in (2a-b) contain a direct object, the second example displays a higher degree of transitivity since it comprises <?page no="161"?> 149 Table 6.1 Features of transitivity (Hopper & Thompson 1980: 252) more transitivity features ranked 'high' 4 than the first example (cf. ibid.: 253): (2) a. Jean loves music. b. Bridget hit Anne. On the other hand, since transitivity must be regarded as a continuum, the inclusion of an object is according to the authors' observations but one feature that contributes to the transitivity of an expression, and the absence of such does not necessarily render a verbal expression less transitive than a verbal expression containing an object (cf. ibid.: 254). Compare the following examples: (3) a. Amy laughed out loud. b. Jean loves music. Here, (3a), involving only one participant displays a higher degree of transitivity than (3b), where two participants are involved, since the first example has more high-transitivity features 5 . Obviously, Thompson & Hop- 3 Hopper & Thompson (1980: 253) explain this feature as follows: "The component of individuation […] refers both to the distinctness of the patient from A [...] and to its distinctness from its background". Individuation is high when the referents of the objects have the following features: (1) proper, (2) human, (3) animate, concrete, (4) singular, (5) count (6) referential, definite. "An action can be more effectively transferred to a patient which is individuated than to one which is not; thus a definite O is often viewed as more completely affected than an indefinite one". 4 The decisive parameters for the examples are the following: Kinesis Aspect Punctuality Volitionality Affect. of O Individ. of O 2a) low low low low low low b) high high high high high high 5 For the two examples, the decisive features are: Participants Kinesis Aspect Punctuality Volitionality 3a) low high high high high b) high low low low low High Low Participants 2 or more participants, Agent (A) and Object (O) 1 participant Kinesis action non-action Aspect telic atelic Punctuality punctual non-punctual Volitionality volitional non-volitional Affirmation affirmative non-affirmative Mode realis irrealis Agency A high in potency A low in potency Affectedness of O O totally affected O not affected Individuation of O 3 O highly individuated O non-individuated <?page no="162"?> 150 per do not necessarily consider transitivity as a constructional phenomenon, i.e. as the quality of a particular syntactic arrangement describing the interaction between at least two participants, but view it as semantic phenomenon of the clause which can be reduced or increased depending on the contribution of the features given above. It must be recognized, however, that several of the features listed by Hopper & Thompson crucially depend on the presence or absence of an object. Thus, if a clause lacks a second participant expressed as the direct object, it not only rates low on the feature 'Participants' but also on those of 'Object Affectedness' and 'Object Individuation' 6 . Moreover, as the authors observe themselves, "[T]otal affectedness of O follows from the semantic perfectivity of the verb" (1980: 287): Hence, one must also assume an interdependency between the factors 'Telicity' and 'Affectedness of O'. All in all, four of the suggested factors of transitivity depend directly or indirectly on the presence of a direct object. Therefore, one can expect that the use of a verb in a construction including any kind of direct objects increases the transitivity of an expression to a great extent in comparison to an objectless construction. 6.4 The transitivity prototype It is quite obvious that many of the factors isolated by Hopper & Thompson also matter in the billiard-ball model, suggested by Langacker as fundamentally reflecting conceptual organization. The model undoubtedly hinges on the presence of two or more entities interacting with each other, a fact which clearly explains the importance of the presence of a clausal object for the degree of transitivity of an utterance. Furthermore, the motion of the entities in space, as depicted in the billiard ball model, correlates with the factor 'Kinesis' isolated by Hopper & Thompson, while the change of the spatial location of the entities (or a change in the direction of their movement) as one result of their collision mirrors in the factor 'Affectedness'. Finally, the short temporal extension of such a collision has its equivalent in the feature 'Punctuality'. It is only logical then to consider the two approaches described here as similar, with Langacker's model presenting the conceptual correlates of Hopper & Thompson's factors of transitivity, which are mainly based on features marked linguistically, either morphologically or syntactically 7 . Having shown that many of Hopper & 6 For a full discussion of the interdependencies of the features suggested by Hopper &Thompson (1980), see Olsen & Macfarland (1996). 7 Note that Rice (1987: 422f) sees the mere focus on morphologically or lexically expressed indicators of transitivity as a major shortcoming of Hopper and Thompson's approach to the phenomenon: "Hopper and Thompson's 10 components fall short of providing a coherent prototype of a transitive event because their factors are by and <?page no="163"?> 151 Thompson's transitivity features have conceptual analogues, I will, in the course of this chapter, make use of their collection of features as a 'checklist' when evaluating the degree of transitivity of an event. In the framework applied in this study, transitivity will primarily be treated as a conceptual notion, i.e. as the construal of energetic interaction between most typically two or more entities in a perceived event. Accordingly, transitivity is not so much understood as a grammatical feature present or absent in a verb (verb phrase) or clause, instead, the grammar of a language must be seen as providing means to express transitivity to varying degrees. The subject-object (trajector-landmark) asymmetry on the clausal level as an encoding of a particular participant constellation, for example, is one primary reflection of how syntax represents transitivity. With the facts and factors of transitivity explicated above and taking into account that number and constellation of participants, energetic interaction and control of action are core features for a definition of transitivity, we are now able to describe a prototypical transitive scenario more specifically, based on three outstanding features (cf. Rice 1987, Horita 1996, Taylor 1995): (1) Two discrete participants, AGENT and PATIENT (expressed as subject and object respectively) (2) An interaction of AG and PAT , in which the former is the initiator of an energetic transfer and controller of the action (3) A change of state on the part of PAT as a result of the interaction (i.e. an affected object). The below pictogram (Fig. 6.2), comprising these three features, adequately illustrates the conceptual content of a prototypical transitive clause. The examples in (4) provide full verbalizations of such a prototypical transitive scenario, displaying all of the features listed: An AGENT interacts with a PATIENT , energy is transmitted from a source ( AG ) to a sink ( PAT ), and due to that interaction the PATIENT undergoes some change: (4) a. Cheeky thieves broke the window of Anthony Gordon outfitters in Westborough […]. (K55: 1035) b. They stabbed and slashed him 45 times with hunting knives. (ECT: 2105) large couched in linguistic terms. They seek transitivity in the overt morphology of the clause, in transitivity markers, perfective markers, definite articles etc. These factors do not always correlate nor do their values (presence or absence) reliably covary within a clause" (emphasis added, SH). <?page no="164"?> 152 Fig. 6.2 Prototypical transitive scenario (Horita 1996: 227) If we now consider COCs, having the same syntactic form as prototypical monotransitive clauses, we find that only few instances represent scenarios which come close to the transitive prototype, the majority of forms, however, verbalize events which obviously do not qualify as prototypical transitive scenarios: (5) a. We drank the drink and both of us went to bed and sleep. (K66: 217) b. He put five really thick swords down his mouth, then he nailed two six-inch nails up his nose. (CK5: 527) c. He smiled an enigmatic smile. (H97: 1459) d. One or two English kings were murdered and one died in battle in this period; no French or German king died a violent death. (BMV: 51) e. To have run the most glorious run of your life […], to do it in an Olympic final and then to be falsely accused of drug-taking by an officialdom that was frantically sniffing out Ben Johnsons like a Salem witchfinder! (FBL: 1431) While (5a) and (b) surely encode events which are high on the transitivity scale (which holds good for the majority of constructions involving an affected CO-NP), the remaining instances in (5c-e) exemplify nonprototypical transitive expressions. Their deviation from the transitive CAUSE BECOME AG PT AG : Agent PT: Patient Participant Ultimate State Transmission of Energy Change of State Cognitive scope <?page no="165"?> 153 prototype can mainly be put down to the rather abstract, non-concrete 8 nature of the entities expressed in the object phrase. For these events, it is hard to picture an energetic interaction between two participants -they seemingly lack the notion of an intended manipulation of the PATIENT by the AGENT . Yet, in order to understand the full impact of the construal of an event verbalized by a COC, one must first examine the more 'basic' scenarios expressed by these verbs without their respective COs (i.e. in an intransitive use) and then compare them to their COC counterparts. A finer analysis of the settings captured by the constructions the CO verbs are more conventionally associated with might yield answers to the question whether features of transitivity form the motivation for COCs, i.e. if the usage of a COC mirrors a change in construal and results in the foregrounding of energetic interaction. 6.5 Other event scenarios Typical examples of COCs quoted in the respective literature usually involve verbs that are labelled 'intransitive' (or 'used intransitively') in traditional terminology, with die, dance, fight, laugh, live, run and smile listed most frequently as representative verbs. The following citations illustrate the use of these verbs in an intransitive construction, i.e. without direct objects and thus lacking one of the central transitivity features. This places them more towards the intransitive end of the transitivity scale, i.e. they display a low degree transitivity: (6) a. It was just possible […] that Maidstone had died naturally. (ASN: 2416) b. She ran faster and faster, afraid to look back, but with hope in her heart. (EFJ: 2510) c. She looked back at him and smiled gently […]. (AB9: 336) d. They lived comfortably in a house built in King Edward's reign. (ADP: 1756) e. Eva was jubilant, thinking, see how you limited yourself when you did not think positively 9 . (H7E: 2398) 8 That is, they lack clear, concrete boundaries, which corresponds to a low value in Hopper & Thompson's factor "Individuation". 9 The verb think is more typically associated with that-complementation or PPcomplementation (about, of) with the complements paraphrasing/ summarizing the content of the thought in the PP. Note that in the following, I describe the verb in the sense of 'to form in the mind ', neglecting other meanings such as 'believe' or 'have an opinion'. <?page no="166"?> 154 f. Julia was dreaming again when Comfort got back to her bedside and her skin felt hot and dry. (FSC: 910) In all of the examples only one participant is expressed, which verbally indicates the absence of any conceptualization of an energetic transfer from a source to a (separate) sink, and logically, a change of state in a second entity. With respect to the three features of a prototypical transitive scenario listed above, none of the examples would qualify for the label 'transitive', at least not in its prototypical sense. If the sentences are checked against the criteria suggested by Hopper & Thompson (1980), as summarized in Table 6.2, one easily recognizes that the expressions combine several of the features the authors consider as indicative of a low degree of transitivity. Due to the one-participant constellation, i.e. the absence of an object, some of the criteria (O affectedness, O individuation etc.) are not even applicable and are thus ascribed the value '0'. According to these features, sentences (6b) and (6c) can be considered as having a higher transi- Ex. Verb Participants Kinesis Aspect Punctual. Volitionality. Agency / Role 10 A. of O I. 11 of O 6a die low low high high low low / THEME 0 0 6b run low high low low high high / MOVER 0 0 6c smile low med. low med. med. high / AGENT 0 0 6d live low low low low med. med./ AGENT - EXPER . 0 0 6e think low low low low med. med. / AGENT - EXPER . 0 0 6f dream low low low low low low / EXPER . 0 0 Table 6.2 Features of transitivity for example sentences 10 I add here a further criterion for the purpose of further discussion in sections 5.5. The roles are based on Langacker's (1991a: 285-287) classification, who defines them as follows: agent = participant who wilfully carries out a (typically physical)action mover = participant that changes position, patient = participant that undergoes an internal change of state, experiencer = locus of some mental process, theme = participant in thematic relationship, i.e. autonomous core of an event conception in an absolute construal (see p. 156 for a clarification of the term 'absolute construal'). To denote the more or less ambivalent participant role of the tr of the verbs live and think, I introduce the 'hybrid'-role 'agent-experiencer', accounting for the fact that these events can be both actively initiated/ controlled and experienced. 11 A. of O = affectedness of object, I. of O = individuation of object <?page no="167"?> 155 tivity potential than the rest of the examples. Sentence (6b) clearly encodes an activity which involves energetic input. Examples (6d) and (6e) describe activities whose trajectors display features of experiencers, undergoing kinds of intellectual, perceptive or emotive experience and being able to control these only to a certain extent- thus the features 'volitionality' and 'agency' are rated as 'medium' for (6d-e), and 'low' for (6f): A certain way of life can be volitionally chosen but is always exposed to some uncontrollable external factors; a cognitive process such as thinking can be consciously initiated and controlled, but the mind may be distracted by external entities or subconscious processes, which may result in the loss of complete such as a dream, which comes about as a synthesis of signals from brain activity during sleep, lies almost completely outside the control of an EXPERIENCER (unless the event describes lucid dreaming or daydreaming, forms of dreaming which can be controlled to some degree 12 ). With the exception of die, all verbs have in common that they belong to the unergative type of intransitive verbs, i.e. the action expressed by the verb is perceived as actively initiated and partly controlled by its trajector, the trajector is responsible for the coming about of an event. Since there is no outer sink towards which the energy (which must be exercised to carry out the action) flows, the subjects are supposed to "supply energy to themselves in order to bring about the activity" (Horita 1996: 231). Horita depicts this form of event construal as shown in Fig.6.3, arguing that in such a scenario the AGENT is both source and sink of the energetic transmission. Fig. 6.3 Unergative scenario (e.g. x run) (cf. Horita 1996: 231, adapted) In the case of run, the agentive MOVER undergoes the process of motion and, as a consequence thereof, a change of position. Yet the participant not only functions as the experiencer but at the same time also as the provider of the amount of energy that is necessary for muscular contraction (which, in turn, enables motion) (cf. Langacker 1991a: 389). The construal in Fig.6.3 seems a likely one for all the verbs in sentences (6b-e), i.e. the subjects cause themselves to participate in some kind of activity, having more or less control over it. (6f) is slightly different in that there is no such notion of 'self- 12 Note that utterances such as Will you stop dreaming? ! reflect the belief that we can control our (day)dreams to some degree. X X ' <?page no="168"?> 156 causation', still, the subject must be seen as the source of energy for the mental activity of dreaming. In contrast to unergative verbs, unaccusative verbs such as die or vanish entirely lack the notion of their subject's control over the action, i.e. the trajector does not actively initiate or is not responsible for the action the verb expresses; rather, it has properties which it shares with the PATIENT object of a transitive verb in that it undergoes some process. There is the general claim that only unergative verbs can enter a COC (see Ch. 2), yet since some authors (e.g. Jones 1988, Dowty 1991) list die as a representative of the intransitive unaccusative type, I will include this verb class in my discussion 13 . My decision to do so is also based on the fact that in the semantics of die, any notion of a causative impulse or control emerging from the subject is absent, which makes it different from the verbs of the unergative group. Die, vanish, appear, emerge and fall are common examples of unaccusative verbs in English. Horita (1996) observes that it is this class of verbs which mainly allows for a verbalization of the so-called absolute construal. According to Langacker's description (1991a: 291f.), in this specific type of construal the event component is profiled as independent of any other participants. One indication of such an absolute construal is the complete absence of energetic transmission (in the case of die, one may speak of a loss of energy [Iwasaki, 2007: 22]) and the notion of causation. Thus, as regards events designated by unaccusative verbs, speakers construe their coming about without conceptualizing an energetic source; the processes are perceived as independent of an initiating participant 14 (ibid.: 232). Fig. 13 There is considerable inconsistency throughout the literature whether to classify die as an unergative or unaccusative intransitive verb. Macfarland (1995: 184-216) presents evidence for die as an intransitive unergative verb, Lemmens (1998: 43, 46) even lists it as transitive non-effective (a cover term for one-participant constructions of the type ACTOR - PROCESS . Besides die, Lemmens only explicitly names run as another verb belonging to this category. As Levin & Rappoport Hovav (1995) observe, different languages offer different categories to encode the equivalent of die. However, they remain inconclusive about the status of the verb die. Dowty (1991: 607) labels predicates which are inherently telic and combine with a subject that is non-agentive as "definitely unaccusative". Possible candidates mentioned by him are the verbs bleed, blush or vomit. Some researchers (see esp. Macfarland 1995) have argued that it is the verb's occurrence in a COC that can be taken as strong evidence for its status as an unergative verb, since unaccusative verbs are defined as not being able to occur with any surface object, including a CO (cf. Levin & Rappaport Hovav 1995: 40). Interestingly enough, Horita avoids a classification of die by not discussing it as a representative of any of the types of construal she suggests. 14 Many verbs allow for more than one construal with respect to energetic flow and control by an agent. Compare: <?page no="169"?> 157 Fig. 6.4 Unaccusative scenario (e.g. X die) (cf. Horita 1996: 232) 6.4 portrays an event typically encoded by an unaccusative verb. As the illustration shows, an entity undergoes some change of state, yet there neither is an outer source that causes that change through a transfer of energy nor is it the entity itself that functions as such a source (as is the case with unergative verbs); therefore an arrow symbolizing energetic transmission is absent. To sum up the discussion so far, there are three basic types of construal with respect to the depiction of energetic transaction: (a) the (prototypical) transitive construal, which profiles the flow of energy from a separate source to a separate sink, (b) an intransitive unergative construal, in which source and sink of the energetic flow coincide, and (c) an intransitive unaccusative scenario, which completely lacks energetic transmission. These basic types are also suggested by Horita (1996); however, I would like to mention a fourth type of construal to account for expressions that have unexpressed objects 15 , containing verbs which also appear in COCs. Consider the following examples: (7) a. So she learns to give all her stories happy endings and take her troubles to bed with her after she has cooked, cleaned, washed and ironed […]. (C8Y: 1002) b. But we were five against double as many, and armed villains very willing to kill. (G0M: 21) i a) The wax melted. b) The sculptor melted the wax. (causative level included) ii a) The mirror broke. b) The artist broke the mirror. (causative level included) Verbs which occur in both constructions in nominative languages are commonly referred to as ergative (unaccusative verbs) (cf. Bussmann 1998). Note, however, that in the present study the notion "unaccusative" is restricted to the class of verbs which (conventionally) are used only in the pattern exemplified in the (a)-versions of the above examples (i.e. die, fall, emerge etc., see p.159). 15 Different labels for this type of object are found throughout the literature: null complement (Fillmore 1986), omitted object (Rice 1988), deprofiled object (Goldberg 2006) or unexpressed object (Huddleston & Pullum 2002). Fillmore suggests a finer distinction when he differentiates between indefinite null complements for cases in which the referent's identity is unknown or irrelevant (He ate all evening), and definite null complement for cases in which the identity of the referent must be recoverable from the context for the full comprehension of an expression (The Argentines lost). In a more recent account, the terminology has been changed to definite/ indefinite null instantiation (Fillmore & Kay 1993, so quoted in Croft 2001: 276). X <?page no="170"?> 158 c. Rachaela played music, Ruth painted. (GUM: 2933) d. The young woman had been out drinking with friends. (K1U: 3869) All of the sentences in (7) include verbs for which a transitive use is semantically more basic in that "the intransitive is interpreted as having an unexpressed object" (Huddelston & Pullum 2002: 300). Cook, clean, wash and iron entail some kind of object at which the activity is directed, and so do kill, drink and paint, the latter two being COC-verbs. Usually, the unexpressed object can be recovered from the context and is restricted by the semantics of the verb, i.e. the lexical content of the objects is automatically triggered by the meaning of the verb (cf. Rice 1988: 204). For sentence 6c), for example, the most likely candidate to fill the slot of the omitted object is painting or picture, whereas the entity whose image is reproduced by the activity of painting would have needed to be expressed explicitly, e.g. Rachaela played music, Ruth painted flowers. For the type of construal verbalized in the sentences (7a-d) I propose the following depiction: Fig. 6.5 Conceptual scenario for verbs with omitted objects Rice (1988: 205) observes that omitted objects […] are still objects, which is to say that they are still present at some level of organization, perhaps not at the lexical or syllabic level, but certainly at the conceptual one. Most importantly, the object does not really go away when it is omitted (emphasis added, SH). This kind of conceptual organization is captured in Fig. 6.5, where broken lines indicate participants/ relationships that are still present at the conceptual level but not verbalized. In contrast to an unergative intransitive construal, for which a self-reflexive flow of energy was suggested (i.e. AGENT = energy source and sink), in a construal verbalized by an omitted object construction the flow of energy 'departs' from the agent, however, the concrete direction and sink of the energetic interaction remain schematic. Note furthermore that some of the verbs, when used in an omitted object construction, are ambiguous or take on a special sense 16 : Drink, as in 16 Verbs taking on specific meanings in the omitted object constructions are, for example, expect (as in She is expecting ) - to be understood as "being pregnant", or wash, where the intransitive use applies to clothes, sheets or towels, while in the transitive AG PT <?page no="171"?> 159 (8a-b) must be interpreted in the sense of "being addicted to alcohol", "drinking alcohol regularly and excessively"; in this case, the omitted object is understood as an alcoholic drink. Example (8c), however, shows that such an interpretation is not necessarily the only one: Here the unexpressed object, most probably "water", can be recovered from the surrounding cotext. Clearly, coand context bias the one or the other reading. (8) a. In the novel itself, where we might expect Marmeladov to speak of solace, respite, forgetting, companionship, he grasps the paradox that he drinks because he is in search of suffering. (A18: 130) b. I rather wondered whether she drinks, as when I went in in the morning there was a large bottle of beer on the table. (AP7: 67) c. For instance, Oggie puts one or both feet into his water bowl when he drinks. His puppies do the same thing. (C8U: 747) Having given an overview about possible transitive and intransitive scenarios and different ways of construing the interaction of participants within an event (based on forms of the energetic transfer or the absence thereof), I will now go on to consider to what extent COCs reflect a conceptual reorganization of event structure. 6.6 COCs - Means of elaborating event structure 6.6.1 Unaccusative verbs Intuitively, the most remarkable effect should be observed for the unaccusative verb die, which is recorded 87 times in a COC in the BNC and ranked 9 th of those verbs which are most strongly attracted by the construction. It verbalizes an event which not only lacks a second participant but is also characterized by the complete absence of any energetic flow and any initiating / causative momentum. Compare the following sentence pair: (9) a. Is it surprising that he should die naturally from a heart attack? 17 b. Is it surprising that he should die a natural death from a heart attack? (GUF: 1974) use the object can designate various entities which can be washed (i.e. dishes, cars, windows, floor etc.) (cf. Huddleston & Pullum 2002: 304). 17 Here and in other examples provided in the following for the purpose of contrasting COCs with their CO-less variants, the latter are 'constructed' by me on the basis of the respective COC variants found in the BNC. <?page no="172"?> 160 Here, the use of a construction including a CO phrase radically changes the scenario of an absolute construal involving a thematic subject only. The most obvious change concerns the participant constellation: Instead of a setting with one THEME -participant experiencing death, the COC presents a constellation with two participants the subject-participant and a reified form of the event of dying as the clausal object, which stand in an asymmetrical relationship. Drawing on the studies that argue for COs as objects of result, we may consider a natural death in (9) as a result brought about by the trajector through some action. Thus it seems that we are now dealing with a construction which resembles to some degree the transitivity prototype in which an AGENT affects or in the case of objects of results effects a PATIENT . However, can we assume that the former THEME (subject) is now depicted as having agentive features in the event, i.e. are we dealing with a construal of energetic interaction between an AGENT and a reified event? It appears rather unlikely that the COC construes the THEME as an AGENT initiating his own death. But what then might motivate the occurrence of the verb die in a COC, i.e. 'equipped' with a subject and an object of result? A look at the history of the phrase die a …death may offer an explanation for the motivation of such a construction. Visser (1963: 415) provides the following observations relating to the origin of die + CO: That in the original *diegan (sweltan) deaþe the complement was realized as a kind of adverbial adjunct of manner (cause? circumstance? ) may be inferred from the fact that in Middle and early Modern English the noun death is often preceded by a preposition. The phrases that occurred with such a type of complementation in these earlier periods of English include the following: die with… death, die by… death, die of … death, and die on … death. The latter phrase, Visser speculates, might be considered as the origin of die a death: "In 'die a death' […] a was probably the preposition on, o, which came to be treated as the indefinite article" (ibid.). Hence, it seems that the complement a…death actually did start life as an adverbial 18 . The examples Visser quotes from the literature all illustrate instances with prepositional phrases which specify the reason for or manner of dying 19 . 18 Recall that Jones (1988) suggests that COs are equivalent to adverbials. His suggestion, however, is not restricted to COs containing the noun death, but serves as a general account of COs (see Ch. 2.3.2). 19 a) 'With deth thou shalt die ; b)[…] dye of bitter deeth ; c) […] what deth this floe shal die on; d) […] died al by violent and miserable deaths. The personification of death as the 'Grim Reaper' might have motivated these forms, especially those in which the PP denotes the cause of passing away. This personification is discussed as a result of the conceptual blending of several metaphors (e.g. EVENTS ARE ACTIONS , PEOPLE ARE PLANTS WITH RESPECT TO THE LIFE CYCLE etc.) in great detail by Turner (1996: 76ff.). <?page no="173"?> 161 However, all this is knowledge a contemporary speaker of English does not have, and the great variety of modifications of the CO a/ the death show a creative use of the construction and deviate from the original motivation, with most of the phrases allowing for an 'event' or 'result' interpretation: (10) a. […] and dies a hero's death in the process. (A7L: 549) b. I only died a small death at his words and the way he said them. (FEE: 1992) c. Then she died a thousand deaths and was born again instantly as at last his lips came down to cover hers. (JXS: 3575) d. Chatterton died the romantic and traditional early death of the divided, the invaded man, while Eliot did not. (A05: 574) e. But the one true R'n'R death was died by the one true original R'n'R star, Elvis, who died on the lavatory of what boiled down to a lethal dose of constipation. 20 Especially example (10e) underlines the successful reanalysis of the former PP into an object-NP, where the CO appears as the subject of a passivized sentence, a syntactic slot which can only be 'occupied' by true objects (as already pointed out in earlier sections). The examples also provide some counterevidence for Iwasaka's (2007: 22) suggestion that die exclusively expresses the manner in which someone dies, a claim which is at least weakened by (10c). The solution I suggest for the construal expressed by a COC including a verb that is originally associated with an absolute construal is the following: Instead of rendering the clausal subject as an initiating and controlling participant, the COC depicts a scenario in which an intensive relation is expressed between an event and a more or less passive participant, who is deeply involved therein. If the CO is interpreted as an event (which seems more likely in the case of die), it adopts the role of a THEME which is experienced by the trajector. That is, the THEME -subject of the unenergetic scenario is re-construed as a somewhat more involved EXPERIENCER in the COC variant. The relation between the participants is of a non-energetic nature, the focus is put on the relation or involvement itself. Note that the absence of any notion of control or initiation on the part of the trajector enables the use of inanimate subjects; instances of die-COCs as given in (11a -c) are commonly found: (11) a. Thousands of research reports for example, annually die a sad, but predictable death, because the authors mistakenly believe their 20 (http: / / findarticles.com/ p/ articles/ mi_qn4158/ is_20021017/ ai_n12649630; accessed 02/ 06/ 07) <?page no="174"?> 162 efforts should end once the report has been delivered to management. (AYJ: 1528) b. The new Sun was clearly going to provide a Conservative voice for the tabloid market and would eat further into its declining sales, so the Sketch died a quiet and unsurprising death. (CRY: 371) c. Nice guys, then, seem bound to decrease in numbers: niceness dies a Darwinian death. (ARR: 281). However, an interpretation of the trajector as a controlling AGENT who 'creates' for himself a particular kind of death (result reading) is not completely ruled out and seems plausible for those special cases in which the end of one's existence is to be understood as staged or planned: (12) a. During Tracy's death scene in Test Pilot (1938), Tracy lay in Gable's arms dying the slowest, most lingering death in history, as Gable described it. (CDG: 1542 ) b. Early on, Ti Lung dies a staged death as part of his theatre per formance. Later these scenes are inserted with scenes of his real death […]. (www.aintitcool.com/ ? q=node/ 15358, accessed 06/ 26/ 07) In such a usage, die no longer has the unaccusative, 'passive' meaning of fading away, but denotes active performance, which, however, is not incompatible with the construction: Other verbs of performance such as dance or sing are well-described COC verbs and are among the verbs most frequently found in and most strongly attracted by the construction. Notice that die, if classified as an unaccusative verb, is the only one of this verb class that frequently appears in a COC and the only one of which I could detect examples in the BNC. Kuno & Takami (2004: 116; see Ch. 2.4.3) and Iwasaki (2007) report on the grammaticality of the below sentences, containing other unaccusative verbs such as blush, grow, bounce, drop or fall: (13) a. The apples fell just a short fall to the lower deck, and so were not too badly bruised. b. The tree grew a century's growth within only ten years. c. The stock market slid a surprising 2% slide today. However, these sentences are contrived by the authors and not extracted from a corpus of naturally occurring language. The reason for the special status of die is likely to lie in the diachronic emergence of the verb's CO as discussed above, i.e. the reinterpretation of the preposition on as the determiner a, and thus from an adverbial phrase to an argument phrase. The virtual absence of other unaccusative verbs <?page no="175"?> 163 from COCs can be readily explained by the transitivity-evoking effect of the construction: Since unaccusative verbs are associated with an absolute construal of scenarios which inherently lack any form of energetic flow, it is difficult to construe the events expressed by these verbs as involving energetic exchanges. 6.6.2 Unergative verbs Compared to the class of unaccusative verbs, the class of unergative verbs taking a CO comprises a greater number of instances and is relatively heterogeneous as regards semantic aspects. Commonly quoted representatives of unergative verbs with CO complements which are also attested in the BNC are laugh (sound emission), grin (facial expression), think (mental state), run (agentive motion) or live (verb of existence), the latter being the 'top COC verb' both with respect to raw frequency and collostructional strength. All of the unergative verbs in COCs found in the corpus have a cognate noun which can evoke an event reading and a result interpretation, i.e. COCs involving these verbs usually instantiate the type EV/ R 1. To repeat what has been explicated before, unergative verbs differ from unaccusative verbs in that the trajector, expressed as the clausal subject, is also the initiator of the action or event designated by the verb. In many cases, the verbs express actions evoking a scenario in which the agent simultaneously functions as source and sink of the energetic transmission. The integration of unergative verbs into a COC reflects a more elaborated participant constellation of the scenario as we find it in the objectless variant: (14) a. He smiled harshly and stepped away from the door. b. He smiled a harsh smile and stepped away from the door. (JXS: 1702) The AGENT from (14a) remains an active force in the COC; however, a second participant, the reified event, is introduced (as in the case of the unaccusative die, cf. above). With this new participant, the flow of energy takes on a new direction and is no longer construed as being self-reflexive. Since the second participant is the reified form of the process expressed by the verb, I suggest that the COC construes the trajector as an AGENT who invests the amount of energy he has to expend into the process. Thereby the construction depicts him as more consciously 'committed' to the process, having more control over the action and its outcome than in the CO-less construction. Thus, for the above example the COC variant yields a reading in which the AGENT , by the action of smiling, simultaneously 'creates' an abstract entity which is the event itself or the result thereof. Such an analysis can be applied in a similar fashion to the alternation think / think a thought. Think has been rated lower in transitivity than smile <?page no="176"?> 164 in Table 6.2, which is especially due to the verb's low value of the feature 'Kinesis'. However, mental activity is activity, even though the energetic momentum is not as obvious (since not visible) as with physical activities (e.g. run). The construction with the CO profiles such a momentum by focusing on the created entity / outcome of the process, functioning as the energetic sink: (15) a. Take a deep breath and think positively. b. Take a deep breath and think positive thoughts. (B10: 1651) In the COC variant in (15b), the subject is construed as an AGENT , creating concrete objects. Since any creation of 'something' presupposes a creating force investing energy into the formation of that 'something', the event described in (15b) can be discussed in terms of energetic transfer, and as such as expressing a transitive scenario. Since the energetic flow is depicted as running from the subject/ trajector to the object/ landmark, the trajector can be considered as an active AGENT , initiating and at least partially controlling the mental process of thinking and its result, namely 'positive thoughts'. In chapter 4, I already commented in more detail on two particular conceptual operations of speakers which enable a construal as described for smile a smile or think a thought. One of these, conceptual metaphor, was shown to be at work for the not uncommon perception of activities as energetic sinks, which is part of and entailed by the conceptual structure of the mapping EVENTS / ACTIVITIES ARE CONTAINERS (Lakoff & Johnson 1980: 25ff.; cf. Ch. 4.4.3 of this study). Utterances such as (16) a. They are all into fixing cars and other guy things. (www.whiteninjacomics.com/ forum/ viewtopic.php; accessed 09/ 12/ 06) b. You can get a lot out of reading the words of Sun's executive bloggers. (blogs.sun.com/ woodjr/ date/ 200702; accessed 03/ 03/ 07) are verbal clues to this particular mode of conceptualizing events or activities. The concept of CONTAINER , having boundaries and thus 'insides and outsides', permits the notion of putting something into the 'bounded region', which in turn motivates the conceptualization of investing (i.e. putting) energy into an activity. The other mental operation which possibly contributes to construing an intransitive scenario as a transitive event is that of conceptual separation (Kemmer 2003; cf. ch.4.5.2), by which one concept (participant) is conceived of as two separate entities, as is verbally mirrored in the English reflexive construction. Event/ Event-Result-COCs might be reflections of conceptual separation, in which components of the process designated by the verb are construed as separate entities. The rela- <?page no="177"?> 165 tion between these now separately conceived phenomena, that is 'creation' (verb) and 'created' (final state/ event), is expressed via the monotransitive construction, as in (17) He laughed a harsh laugh and shook her impatiently. (creation) (created) (JXS: 4082) If the criteria suggested by Hopper & Thompson (1980) for measuring the degree of transitivity of expressions are tested on COCs with unergative verbs, we see that the elaboration of events by means of COCs goes hand in hand with an increase of transitivity. With the construal of a second participant, which is verbalized as the direct object, the number of participants increases, and factors such as 'Aspect' and 'Affectedness of O' take on a higher value. As has been noted by several researchers (see Ch. 2), COs can be described as measuring out an event: Since the activity is completed with the full creation of the entity expressed by the CO, a sentence with such a type of object is considered as temporally bounded (i.e. telic), which is different with the CO-less version, expressing atelic situations. With regard to 'Affectedness of O', it might be argued that the creation of an entity is an extreme form of a change of a state - from 'zero' to 'something'. Hence, the integration of a CO into the verbal structure has an effect on this factor as well, increasing its value. As has been argued in the above sections, COCs can be described as means for the elaboration of events: Firstly, they permit the construal of directed energetic processes (which is a crucial component of event structure) and, secondly, they express a constellation of participants of an event which is more intricate and complex than that construed by the unaccusative and unergative patterns most of the CO verbs are more commonly associated with. 6.6.3 Transitive verbs So far, the analysis has concentrated on the effect of integrating verbs into the COC that are conventionally used without an object. Let us now turn to the third group of verbs, viz. those verbs that come close to or actually represent the transitivity prototype like build or drink, for which the CO is only one of the different complement-NPs the verbs can be used with: (18) a. Jerry told a tale/ a story/ a lie/ the truth/ a joke. b. The team produced a product / a study/ a mess/ a document/ a show. c. They built a brick building / a tower/ a bridge / a city / a new model/ a raft. <?page no="178"?> 166 d. The soldiers felt a (terrible) feeling / despair / the rain / the ground beneath them. The majority of these verbs have a cognate noun which - if used in a COC - denotes a result of the type R 2 or an affected entity of the type A, except for tell, dance and sing whose nominals tale, dance and song denote created entities which are event-result-like (R 1 ). If verbs that are used transitively, i.e. with an object other than the CO, occur with a CO, the effect of their usage in a COC is different from the cases discussed in the preceding sections. The integration of these verbs into the construction does not change anything about the original participant constellation or energetic transfer, as the CO simply designates one particular type of possible effected or affected entities. What is special, however, is the relatively generic level of the object if compared to the other possible objects. One could argue that it is with these instances that one would expect further modification of the CO phrase to avoid a certain degree of underspecification. Some of the verbs which take other objects besides the CO (e.g. dance, sing) show a particular behaviour as regards the restrictions seemingly associated with these objects: The entities referred to by the nouns have to stand in a hyponymic relationship with the referent of the CO, i.e. they denote subtypes of the CO referent (e.g. dance waltz, Samba, Salsa, Mambo etc.) For an analysis of the effect of the use of these verbs in a COC, compare the following examples of possible expressions: (19) a. The school choir sang Ø / a song / a gospel/ an anthem/ a canon. b. The pair danced Ø / a dance / a tango / a waltz/ a polka. c. We fed them Ø / food / carrots / pieces of pear / cream of wheat. A song, a dance and food occur as the COs of their respective verbs, all other possible complements are hyponyms of the COs. As indicated in the examples in (19), these types of verbs are also commonly used without an object, an observation which suggests a continuum within COCs: Verbs of the type used in (19) mark the transition from the more intransitive class to a more transitive class (tell, smell), the latter even involving prototypical transitive verbs (e.g. build). One could even go as far as to say that The school choir sang, The pair danced and We fed them are instances of constructions with deprofiled/ omitted objects, which then may be treated on a par with Ruth painted or Stephen drank. I suggest that in cases of verbs that are found both in COCs and omitted object constructions, it is the CO that is omitted in the latter type. This seems plausible because the CO designates a rather unspecific and therefore more 'schematic' entity (compare song vs. aria). The difference between the two variants, i.e. COC and omitted-object form, is obvious and once again can be explained in terms of interaction or creation. <?page no="179"?> 167 Whereas the omitted object construction focuses on a subject's activity alone, the explicit mentioning of the (cognate) object, which is now brought centre stage, allows for the representation of an event which not only depicts the subject as active, but as actively involved in the creation of an entity as the outcome of the activity. A concrete interaction with a second entity is profiled, a notion which is only schematically present at the conceptual level in the omitted object construction. That is, where the omitted object construction only adumbrates an energetic flow away from the AGENT (i.e. not self-reflexive), the COC explicitly profiles the actual sink of that flow. 6.7 The transitivity continuum The preceding discussion of the transitivity of COCs has yielded important insights for an exhaustive description of the construction: First of all, further evidence could be obtained in favour of the treatment of COs as the landmarks of their verbs, that is, as a true participant in an event. As one characteristic feature of a landmark Langacker (1991a: 355) defines: "[…] this participant lies downstream from the trajectory […] with respect to the transmission of energy". That is, the status of COs as landmarks is corroborated by the fact that they are conceptualized as an energetic sink. The CO's construal as such a sink motivates the use of a transitive construction. Furthermore, it could be shown that the depiction of an energetic interaction and the construal of the trajector as an initiating and to some extent controlling entity can be considered as one of the 'driving forces' for the usage of COCs. This effect is most remarkable for verbs that are more conventionally associated with an intransitive/ unergative setting. Although the integration of unaccusative verbs into a COC does not result in the depiction of a flow of energy from one participant to another, the use of the construction reflects a different participant constellation and their perceived roles in the event: The trajector is construed as an EXPERIENCER (as compared to THEME in the CO-less variant) who undergoes an event that is denoted by the CO (i.e. the second participant). Given the variety of types of verbs that are found in COCs (as occurring in the BNC), the polysemic nature of the construction, already pointed out in Chapter 4, must be acknowledged once more: We find one form (verb + cognate object) with several, yet related meanings and effects. The fundamental features shared by all subtypes of the construction are their profiling a relationship between two participants involved in an energetic transaction, and the combination of two different profiles of the same conceptual base (cf. ch. 4.2) Further effects of the use of verbs in a COC depend on the more basic, less elaborate scenario depicted by the construc- <?page no="180"?> 168 tion a particular verb is most conventionally associated with. These effects are summarized in Table 6.3. Scenario of CO-less construction Effect of COC Intransitive unaccusative Ex: He died naturally. -restructuring of participant constellation, introduction of landmark (event, result of event) -establishment of a relationship between trajector and event in the sense of experience -trajector construed as an experiencer of an event (=CO) He died a natural death. Intransitive unergative Ex. He thought positively. -introduction of landmark (event, result of event) -construal of energetic transmission from AGENT ( DOER ) to the outcome of the action, functioning as energetic sink -energetic flow no longer self-reflexive but directed towards sink He thought positive thoughts. Omitted/ inherent object Ex. He painted. -schematically present object is elaborated -energetic flow takes on a concrete direction towards a clearly delineated sink -focus shifted from AGENT - ACTIVITY relation to AGENT - ACTIVITY - RESULT relation He painted a painting. Transitive Ex. He felt strong pain. -explicit specification/ designation of an inherent object of particular event instead of an 'external', less related entity (e.g. feeling vs. pain ) He felt a tingling feeling. Table 6.3 The effect of COCs as event elaborating constructions A concluding point which should be taken up with respect to the transitivity of COCs concerns their contribution to the phenomenon of transitivity as a continuum. Not only do the different subclasses of COCs vary in their degree of transitivity but also does the class as a whole mark the transition between the various types of intransitive and transitive constructions. The whole continuum is illustrated in Fig.6.6, which is, of course, an idealized model of transitivity, as a clear delineation between the different types of construction is impossible and not even intended: Since transitivity is to be understood as a continuum, the different categories naturally overlap and merge by definition. The existence of such overlaps only lends <?page no="181"?> 169 support to the rather broad definition of COCs advanced here and hence the great variety of different types of verbs and objects that were included in the study as candidates for the construction. Earlier discussions of the construction built on attempts of depicting COCs as strictly delineated categories with the consequence of sometimes inconsistent accounts of the phenomenon. In presenting COCs as a family of different but related, even overlapping constructions, these inconsistencies can be avoided. Fig. 6.6 The Transitivity Continuum 21 Two other types of constructions which have not been discussed previously in this chapter are included in the chart: the middle voice construction and the reflexive construction, illustrated in examples (20a) and (b) respectively: (20) a. […] and then the door opened a tiny crack. (APR: 1399) b. At least he can shave himself now, I gather. (APM: 877) Just like COCs, reflexive constructions have been labelled "unprototypical transitive constructions", placed somewhere in the middle between the transitive and the intransitive pole of the continuum (cf. Kemmer 1992: 149). Their status as peripheral members of the category of transitive constructions is due to the specific nature of the participant which is assigned the patient slot. In contrast to prototypical transitive constructions, on the one hand, the reflexive construction encodes a scenario in which energy source and energy sink coincide, that is AGENT and PATIENT are embodied by the same entity. On the other hand, in contrast to sentences like He ran 21 Note that for reasons of graphical representation, the Reflexive Construction is in a somewhat 'deranged' position. This representation is to indicate that the degree of transitivity of the Reflexive construction lies somewhere between that of Unerg.+ COs and Trans. + COs. Unaccusative Constr . Unergative Constr . Middle Voice Constr . Prototypical Trans. Constr . TRANSITIVITY CONTINUUM Omitted Object Constr . Refl. Constr. COCs Unacc. Unerg. Trans. + CO + CO + CO <?page no="182"?> 170 through the woods or He laughed harshly, which have also been described above as involving some degree of self-reflexive energetic interaction, in the reflexive construction the sink of such a flow of energy is overtly expressed in form of a reflexive pronoun. Since this type of PATIENT is of a more concrete nature, reflexive constructions are judged to be of a slightly higher degree of transitivity than COCs involving conventionally unergative intransitive verbs 22 . Middle voice constructions are similar to unaccusative constructions in that the syntactic subject corresponds to the PATIENT of the activity described by the predicate. What makes them distinct, however, is the fact that middle voice constructions, containing verbs which are commonly described as 'ergative' (see p.156, ftn.14), have a corresponding transitive sentence, in which the 'middle-voice subject' is verbalized as the PATIENT object. This does not apply to unaccusative constructions: (21) a. The door opened a tiny crack. (middle v.) b. He opened the door a tiny crack. (22) a. Chris died. (unacc.) b. *Richie died Chris. Givon (2000: 117f.) discusses middle voice constructions as detransitivizing devices because the focus is shifted away from the AGENT , who, nevertheless, is schematically present on the conceptual level. Thus, middle voice constructions are set further away from the intransitive pole on the transitivity scale than are unaccusative and unergative constructions, and are found in close neighbourhood to omitted object constructions, with which they share the feature of containing a schematic participant. As regards COCs and their value on the transitivity scale, it might be debated whether transitive verbs occurring with a CO are less transitive than prototypical transitive constructions, as indicated in the above figure. What speaks in favour of such a 'ranking' is the transitivity factor labelled 'Individuation of O' by Hopper & Thompson (1980), which, if applied to COCs, can explain their particular position on the transitivity scale. 'Individuation', which hinges on notions such as 'Animacy' or 'Concreteness', 22 Kemmer (1992: 149) assigns reflexive constructions an intermediate status within the transitivity continuum. Interestingly, different language types have different means for the expression of reflexivity, which might "shift" the constructions either more towards the transitive pole or more towards the intransitive pole. Some languages make use of personal pronouns which occur in the patient slot of the construction. Here, the tendency goes more towards the transitive prototype. Other languages use reflexive particles such as clitics or affixes, that is, the patient slot remains empty. These languages encode the reflexive more towards intransitive prototype, as no second participant is overtly expressed. <?page no="183"?> 171 describes the salience of an entity in a scenario with regard to the degree of how distinct it is from the AGENT or other components of the setting (cf. p.149, ftn.3). Obviously, the more distinct the referent of the object is, the easier it is to conceptualize its involvement in an energetic transfer. Olsen & Resnik (1997) observe that an essential element that contributes to an object's individuation is its distinctness from the verb, "measur[ing] the extent to which the direct object replicates information provided by the verb" (ibid.: 4). The more distinct the direct object is from the verb, the higher is its degree of individuation and the more it contributes to the transitivity of an event expressed in a clause. Following from this claim, (23b) displays a slightly higher degree of transitivity than (23a): (23) a. I'm just going to drink my drink. (KDW: 4250) b. I'm just going to drink my milk. This also explains why on the above scale all of the different subtypes of COC construction are found to the left of the transitivity prototype, since by definition COs replicate information encoded by the verb. With these summarizing remarks on the transitivity continuum and the placement of COCs in such a model I conclude my discussion of COCs as transitive constructions which serve as a means for the elaboration of event structure and turn to the issue of passivized COCs. In studies on the construction the acceptability of such passives, presenting a mirror image of the participant constellation of active COCs, is often doubted or reported to be highly restricted 6.8 To passivize or not to passivize? One of the characteristics of COCs which has attracted great attention is their occurrence (or rather: non-occurrence) in a passivized form. Several scholars (Jones 1988, Moltman 1990, Pinker 1989) have put forward the very general claim that instances of COCs fail to occur in the passive voice: "[…] cognate or 'fake' objects are unpassivizable, as in *A hearty laugh was laughed; A horrible death was died" (Pinker 1989: 92; emphasis added, SH). With passivization being the traditional test for objecthood, the alleged resistance of COCs to occur in the passive voice has been considered as evidence for the non-object status of the CO-NP. That is, a clash between the COC and the passive construction on syntactic grounds, caused by the assumed weak or non-object status of the CO-NP, is considered the major reason for the rare or non-occurrence of passive COCs. However, Pinker, trying to account for the broad range of verbs in English that are used in a passive construction, holds that "passivization seems to apply when the object either is the patient or is capable of being construed as one" (ibid.: 136; emphasis added, SH; cf. also Bolinger 1977). With <?page no="184"?> 172 our overall claim that the CO-NP in the COC is construed as an entity that comes into being through the AGENT 's activity, the occurrence of expressions in which the CO is construed as a passive PATIENT subject should not be too surprising. Indeed, as was already shown by Rice (1987), Macfarland (1995) and Kuno & Takami (2004), passive forms of COCs can be found in naturally occurring language, albeit - according to these authors rather infrequently. Just like any other construction, the passive construction comes along with a prototypical form and more or less marginal instantiations. The prototypical passive construction verbalizes a situation in which a subject with the semantic role of PATIENT is directly affected physically by an external AGENT , who is expressed either in a by-phrase 23 or is completely deleted (cf. Takahashi 2000: 248). This prototype appears as the reverse image of the transitive prototype, instances of which are said to passivize most easily. Peripheral cases of the passive incorporate non-patient subjects (e.g. THEMES ) which are affected in a non-physical way. Compare the following passive sentences, with the example in (24a) representing the prototype and the one in (24b) being a non-prototypical instance of a passive, involving a THEME subject and an external EXPERIENCER as participants in an event which can hardly be described as bringing about a physical effect. Example (24c) shows a passivized COC: (24) a. Mrs Swami, who is in her 40s and lives in Kingsbury, north-west London, claims she was beaten almost daily by the two sisters of the Emir of Kuwait. (A96: 972) b. The proceedings were watched by Michael Crook, principal Adult Education Organiser for Norfolk, along with other AE officers […]. (KAB: 165) c. The opening prayer was prayed by a Muslim woman who lived in Jerusalem around the year AD 800. (CDC: 603) A similarly non-prototypical participant constellation as described for (24b) holds for the passive COC example in (24c), where the notion of a physical effect brought about by the event (i.e. praying) is clearly absent. Thus, rather than assuming syntactic characteristics of the COCs to block their combination with a passive construction, we should consider a conceptual clash between the prototypical passive construction and COCs, which 23 Although the expression of the AGENT via a by-periphrasis is infrequent in speech (see ftn. 24, p.173), there might be reasons to overtly indicate the AGENT in this non-subject position: the PATIENT might outrank the AGENT in definiteness (cf. Langacker 1991a: 337), or the AGENT might represent discourse-new information and thus has to follow the element(s) expressing discourse-old information (cf. Huddleston & Pullum 2002: 1443f.). <?page no="185"?> 173 hampers passivization. With the majority of instances of COCs being nonprototypical, "construed" transitive forms, one simply must expect their passive forms to be of a non-prototypical kind, used, if at all, less frequently than prototypical instantiations. This should hold true in particular for the subcategory of EV/ R 1, which has been treated in the present study as the form that is found at the very periphery of transitive constructions. Taylor (1995, esp. ch.11), treating syntactic constructions as prototype categories, observes the following: "What we can identify on semantic grounds as more central members of the transitive construction exhibit a number of syntactic and distributional characteristics not shared by marginal members" (ibid.: 209; emphasis added, SH). The ability to passivize is such a syntactic characteristic which is most distinct with central instances of the transitive construction and decreases towards the periphery. The analysis of my corpus of COCs with the focus on voice features confirms that for all of the COC subtypes passive forms are recorded in the BNC, as shown in Table 6.4. Compared to COCs in the active voice, passive COCs are used much less frequently, yet, this is a general feature of the passive construction. Frequency counts in English texts have demonstrated that passives occur with a much lower frequency than actives (Greenberg 1966: 45f., Givon 1991b: 352; Biber et al. 1999: 446), with the studies reporting on frequencies of passive constructions that range between 2-25% of all finite verbal constructions in the texts 24 . Thus, with an average frequency of 13.8%, the use of passive COCs does not deviate from that of other passive forms. However, the distribution is significantly dependent on the type of active passive Sum EV/ R 1 2058 (92.3%) 159 (7.7%) 2217 R 2 189 (63.3%) 108 (36.75%) 297 A 468 (75.1%) 157 (24.9%) 625 Sum 2715 (86.2%) 424 (13.8%) 3139 Table 6.4 Distribution of the Passive Voice in COCs 24 The frequency distribution of the passive heavily depends on the text type. Biber et al. (1999: 447) observe that the passive is used most frequently in academic writing (25%) and least frequently in conversation (2%). Givon (1991b) reports on a similar distribution of the passive construction. <?page no="186"?> 174 COC, as a ² -test 25 of the data substantiates ( ²=281.12, df = 2, p< 0.001).The assumption that the passive occurs least frequently with EV/ R 1 is confirmed by these findings. With this type being the central form of COCs, I will discuss its use in a passive construction first and in more detail. It must be mentioned here that the verb types of the EV/ R 1 group which are used in a passive COC is restricted to a small number. Of the 54 verbs in this class, only 12 are found in a passive form, most frequently tell (43), live (33), sing (33) and do (25), making up 85.3% of all passive examples. Except for live, these are verbs which rank high on the transitivity scale (cf. Ch. 4), as they are complemented by non-cognate objects as well. Of the less and least transitive EV/ R 1 -CO verbs, only fly (2), dream (1) and think (1) are attested in a passive instantiation in the corpus. For other frequent CO verbs which have been measured to be strongly attracted by the COC, such as smile, laugh, sleep and die 26 , no passive forms are recorded in the BNC. The below sentences illustrate passivized forms of the EV/ R 1 type: (25) a. The days of the Prize Ring were such a brutal business, fought under rules which enabled men to inflict terrible injuries on each other, and fights were fought to a finish when a man was downed and had been battered into insensibility. (BM9: 146) b. He dubbed them cheats, and told them they were swindling their audience out of what he called the last great procession, when, after the wedding songs had been sung and the dances danced, the characters took their melancholy way off into darkness. (CRE: 12) c. It was calculated that the estimated time en route would have been approximately three hours and thirty-one minutes with a fuel reserve of thirty two minutes if the flight was flown at an altitude of 31,000 feet, and a fuel reserve of nineteen minutes if the flight was flown at 28,000 feet. (EWS: 318) 25 Put very simply, a 2 test is a statistical hypothesis test which is used to prove or disprove the so-called null hypothesis. This null hypothesis supposes that any difference in the distribution of a tested variable between two or more groups is purely due to chance, i.e. it assumes an even distribution of the data. Thus, the 2 - test allows an estimation whether the distribution and/ or frequency of the data obtained for the variable in question differ significantly from each other (cf. Oakes 1998: 9; 24f.). 26 As Kuno & Takami (2004: 133) and Iwasaka (2007: 5) argue, COCs containing unaccusative verbs do not passivize, thus die is excluded from entering a passive construction. See, however, example (10e), repeated here for reading convenience: But the one true R'n'R death was died by the one true original R'n'R star, Elvis, who died on the lavatory of what boiled down to a lethal dose of constipation. <?page no="187"?> 175 d. So, although interpreting her character nowadays, she is quite immoral and dishonest, to the Christian audience in those days, she would have been a heroine. Much the same would have been thought of Jessica as she would have been seen as someone who walked away from an evil religion and stepped into the right path.However, these thoughts would only have been thought by a Christian audience of that time. (HUB: 980) How can these uses of these passive COCs with CO-NPs which are very peripheral candidates for the AFFECTED / PATIENT role (occurring in the subject slot) be accounted for? Most studies on the passive agree on the following pragmatic and discourse functions of this voice construction (e.g. Jespersen 1954[1927], Shibatani 1985, Croft 1994, Givon 2001): Firstly, the agent of the corresponding active construction becomes defocused, which, secondly, permits the topicalization of a non-agentive element, occurring in subject position. Thirdly, the causal chain of the event expressed by the verb is presented in reverse order, which has the effect of "shifting the perspective […] from beginning to the end" (Shibatani 1985: 841). That is, in terms of transitivity as expressing a flow of energy, this flow is construed from the reverse angle and the energetic sink rather than the source is focused on. Considering the sentences in (25a-c), we see that they constitute agentless passives, focusing on the result of an activity (the energetic sink) and the activity (in exactly that order) alone, leaving the initiator unspecified 27 . With the verbs in their more common intransitive form, the defocusing, even deletion, of the agent would not be possible, as only syntactically transitive verbs allow for passivization. While German, for example, has the impersonal passive, a "dummy-it"-construction which yields agentless expressions such as (26) Es wurde getanzt, gelacht und gesungen. (*It was danced, laughed and sung.) , English lacks such an 'auxiliary' construction. Only by means of the COC, a passive structure can be verbalized, admitting, in turn, the demotion and deletion of the agent and the foregrounding of result and action 28 . 27 It has been observed in various text counts that 70-90% of passive sentences contain no mention of the agent (Shibatani 1985: 831; Biber et al. 1999: 937f.). In this respect, passive COCs do not behave differently, with an average of 89.5% of our instances being agentless constructions. 28 As discussed in Ch. 2.4.4, Kuno & Takamuro (2004: 129) argue along similar lines when they postulate a strong correlation between unmodified CO-NPs and passive COCs. The authors claim that unmodified CO-NPs are easily acceptable in passives, enabling passivization. I do not see any reason to restrict such a function to unmodified CO-NPs, as modified CO-NPs expressing a particular (i.e. modified) result can be focused on as well: <?page no="188"?> 176 Example (25d) shows a use of the passive (with the agent included) which can be said to be primarily motivated by 'textual factors' such as information structure and cohesion. The NP these thoughts takes up information which has been introduced in the preceding sentence and now serves as the topic which is commented on. Since there is always a preference for a topic to be expressed in the subject position, the passive is a most suitable means to verbalize the effected or affected entity in this prominent, topicalizing syntactic slot. Clearly, as the examples presented above show, particular pragmatic and discourse factors (i.e. topicalization of result of an action, defocusing of agent, information structure, cohesion) sanction the use of passive EV/ R 1 -COCs. An interesting phenomenon occurs with passivized COCs containing the verb live, which, as has been noted above, is found in the passive relatively frequently (this is, of course, due to the high frequency of live in COCs in general). In a great number of examples, the subject CO-NP is accompanied by a possessive construction as its determiner: (27) a. Victoria's whole life had been lived within the aura of the Royal Family, and she reacted by instinct to nurture that relationship. (APU: 321) b. I passed to her the wisdom I picked up eavesdropping on our two eldest sisters, whose lives were then lived in romantic turmoil. (K5M: 6260) c. Anyway, with Selina here, my life is being lived in white underpants. (H0M: 2946) As will be argued and illustrated in Chapter 8.2 (see p. 206f.), such a type of modifier refers to a participant, most commonly the agentive participant, of the event which is expressed by the CO in a nominalised form. In the above sentences, these participants are not expressed as agents in the form of a byphrase, but are 'incorporated' into the passive subject. This has the effect of not only passivizing the action but the actual agent as well, which gives rise to the impression that the formerly agentive participant expressed now in the possessive determiner phrase has given away control of the action or is influenced by aspects external to the subject which dictate a particular way of life. Notice that all of the examples include a circumstantial element (within the aura of the Royal family, in romantic turmoil, with Selina here) which expresses the source of such an external influence. The suggested interpretation is supported by the absence of a by-phrase which explicitly denotes Readers who enjoyed your issue on Green consumerism (NI 203) might be interested to stay with a working-class family in Kerala, India, for a month to discover how a happy life can be lived at very modest levels of consumption. (HH3: 9263) <?page no="189"?> 177 an agent who is not co-referential with the participant expressed in the possessive construction (e.g. His life was lived by his father). In this case, the 'copied action' reading might have been another possible reading, verbalizing the scenario of an AGENT who is involved in copying someone else's action with its (hoped-for) result. However, with such an external agent missing, the passive COCs in (26) do not only demote or deprofile the agent, but due to a particular arrangement of the CO-NP, the understood agent assumes a patient-like role. Having commented on selected factors and effects of using the coretype of COCs in a passive 'make-up', I will now briefly consider the data relating to passive forms within the other two COC-classes. Table 6.4 shows a significantly high use of the passive voice with R 2 -COCs. Since, by definition, these types of COs refer to concrete entities which can be construed as affected objects more easily than COs of the EV/ R 1 , the higher frequency of the passive seems logical. Surprisingly, this frequency is even higher than that of passive forms of the A-type (which stands closest to the transitive prototype and thus should be expected to occur in the passive form most frequently of all COC types). A closer inspection of the instances of R 2 passives reveals that it is the combination build + building which is responsible for the great number of passive constructions in this class. This pairing makes up more than 61 % of the passivized instances of R 2 and is the only one in my corpus of COCs which is used more often in a passive construction (64.0 %) than in the active voice. This high frequency is due to the great amount of architectural descriptions of buildings in my corpus, in which the noun functions as the topic-subject, followed by information on the time of its erection, builder, style etc., as in The oldest building, St Margaret's Chapel, was built in the eleventh century. (ECS: 1557). All in all, the R 2 group sees a greater amount of verb types used in the passives than the EV/ R 1 class (see Table 6.5): 7 of the 15 verbs are recorded in such a voice construction, with products produced and webs woven following buildings built in frequency. Similar remarks can be made on the A-type of COCs: Here, 18 of the 40 verbs appear in a passivized form, most frequently we encounter seeds sown (83), gifts given (28) and plants planted (10). As has been noted before, the use of the passive construction with this latter COC subtype does not require an elaborate explanation as it constitutes a particular form of the transitive prototype which passivizes most easily. Concluding the discussion of the usage of passive forms, we must note that the claims on the unacceptability of a combination of COCs with the passive construction cannot be maintained. Passive constructions involving COs do occur, even if there seems to be a clash between the COC (especially those of the EV/ R 1 type) and the prototypical passive construction on <?page no="190"?> 178 EV/ R 1 R2 A tale told life lived song sung deed done kick kicked flight flown execution executed dance danced attempt attempted dream dreamt prayer prayed thought thought 43 (10.7) 29 33 (4.7) 33 (7.1) 25 (5.4) 2 (66.7) 2 (33.3) 2 (66.7) 2 (16.7) 1 (100.0) 1 (2.2) 1 (14.3) 1 (1.3) building built product produced web woven painting painted drawing drawn writing written plot plotted 64 (64.0) 24 (17.0) 6 (23.1) 5 (71.4) 1 (25.0) 2 (100.0) 1 (100.0) seed sown gift given plant planted publication published farm farmed sight seen light lit name named tie tied charge charged food fed loss lost edition edited investments invested description described findings found recruits recruited 83 (41.9) 28 (21.9) 10 (58.8) 5 (71.5) 6 (100.0) 4 (5.5) 4 (50.0) 3 (6.7) 3 (42.9) 3 (50.0) 2 (9.5) 2 (100.0) 2 (40.0) 2 (40.0) 1 (33.3) 1 (50.0) 1 (100.0) Table 6.5 Frequency of verbs in passive construction semantic grounds. As has been shown, the same functional and pragmatic factors which are listed for the general usage of the passive forms sanction the use of passive COCs. It must be admitted, though, that for the central type of COCs, only a fifth of all verbs found in this class account for the recorded passives. Examples such as laughs are laughed, screams are screamed, smiles are smiled, which are instanced in the literature (see Kuno & Takami 2004, Macfarland 1995, Langacker 1991a) have no entry in the BNC and seem to be marginal cases and/ or constructed for use of illustration. The occurrence of passive forms of COCs speaks in favour of my analysis of the CO-NP as part of an energetic chain which links AGENT , AC- TION and RESULT , and thus as a participant in the event. The passive construction depicts this chain from a different perspective and foregrounds its final stage, focusing on the reified action as an entity created by an action. The number of passive examples found in the BNC invalidates those ap- 29 Number in brackets = percentage of instances of particular verb in passive construction <?page no="191"?> 179 proaches which describe COs as 'fake' objects, pseudo-arguments or adjuncts on the basis of their assumed non-passivizability. 6.9 Summary and conclusion The present chapter has been concerned with a detailed explication of one primary expressive function of COCs on the basis of corpus data: It was argued that one major contribution the constructions makes to the verbalization of events is that of elaborating or restructuring the participant constellation, thereby bringing about a change of energetic characteristics of the event. This mainly applies to COCs involving unaccusative and ergative verbs, which are exclusively found in the EV/ R 1 type of the construction, i.e. the one proposed as the most central type of the COC in this study. A few verbs associated with this type (e.g. sing and tell), however, have been shown to have a similar effect as those verbs which were categorized as instantiating constructions of the R 2 and A type, which is either that of elaborating an otherwise schematic (omitted) object or that of specifying a quite general class of entities, in the case of which no change in the participant constellation is observable. A second point of analysis tried to answer the question on the acceptability of passive COCs. On the basis of BNC data and their statistical evaluation common claims on their low acceptability or restricted usage could be rebutted and/ or differentiated: While the overall percentage of passive COCs does not differ from that of the general occurrence of passive forms in language usage, the type of COC seems to matter, with the central form EV/ R 1 being passivized least frequently. The next chapter will consider the issue of aspectual characteristics of COCs, a topic which is closely related to that of transitivity (as verbal aspect and the degree of transitivity of an event interact [cf. Hopper & Thompson's criteria]) and which complements the description of the function of COCs as event-elaborating patterns. <?page no="192"?> 180 7 ‘Packaged’ events: Aspectual characteristics of COCs 7.1 Introduction As was explicated in my overview on different approaches to COCs in Chapter 2, several authors discuss aspectual differences as one motivating factor for the usage of COCs. Especially in the studies by Massam (1988), Macfarland (1995) and Felser & Wanner (2001) verbal aspect is identified as a primary notion to consider in the comparison of COCs with their simple verb variants. The accounts of these authors centre on the concept of 'telicity', an aspectual category which refers to the temporal boundaries or limits of an event as construed by a verbal expression. In all three studies, a change from an atelic (simple verb construction) to a telic reading (COC) of the event is observed, implying that the use of the COC evokes the profiling of a temporally limited event, possibly including its terminal point. However, the analyses presented in these studies only focus on the contributions of the CO to the aspect or Aktionsart of the predicate, neglecting a discussion of aspectual features of the verb phrase which are possibly added by further aspect-denoting lexemes or constructions, aspectual features of the CO itself, and the consequences of an interaction of all these characteristics for an interpretation of the whole COC. Complementing the picture of COCs as aspect-indicating constructions outlined in Chapter 4, where I commented on iconically reflected aspectual idiosyncrasies of COCs, and spelling out further functional characteristics of the pattern, the present chapter will consider different aspectual categories of verbs and focus on likely aspectual changes evoked by their integration into the COC, which, above all, seem to relate to temporal bounding. Moreover, I will test in how far the assumed bounding effects of the construction interact with other aspectual categories built into the grammar of English (e.g. the progressive aspect): It will be shown that the distinction between result reading and event reading of COCs can be attributed to particular combinations of aspectual characteristics of the verb phrase and those of the CO. Before I take up these matters, however, some general observations on the notion of aspect shall be expounded. <?page no="193"?> 181 7.2 Aspect, aktionsart, aspectuality It is hardly possible to give a concise overview of the abundant literature on aspect and aspectuality, and the many different definitions and treatments of the phenomena offered in the studies 1 . Thus, I will concentrate on those notions that play a decisive role in the treatment of COCs as possible aspect-changing constructions. The most important distinction that is maintained by the majority of studies (although given up by some linguistic schools) is that between aspect (also termed "aspect proper", "tempo-aspectual class", "perspective point" [Sasse 2001: 6f., Tobin 1993: 3f.]) and aktionsart (often labelled 'action', 'aspectual character', 'situational type', 'event sort' [Sasse 2001: 206f.]). In order to avoid any terminological confusion or to indicate preferences for one or the other model, I will adopt Sasse's (2001) more 'neutral' terminology in this overview and speak of ASPECT 1 in relation to aspect proper, and of ASPECT 2 when referring to what is commonly known as aktionsart. Both concepts are subsumed under the more general notion 'aspectuality'. ASPECT 1 is traditionally described as the 'perfective-imperfective' dichotomy, which is generally realized through the morphological and syntactic system of languages and frequently indicated by the addition of particular affixes or the use of different verb forms, often in combination with different auxiliary verbs. This aspectual level is described as expressing either an external viewpoint (perfective) on a situation, i.e. including the startingand the endpoint of a process, or an internal viewpoint (imperfective), presenting the event from within its temporal boundaries, i.e. neglecting starting and endpoint. In English, the progressive aspect, which is expressed analytically by a combination of a tenseand voice-inflected form of be and a lexical verb with the inflectional affix -ing (i.e. the verb's present participle form), encodes an action which is presented in its continuation, 'cutting out' temporal boundaries. A SPECT 2 generally relates to more subtle temporal characteristics of situations and events, which are most often lexically intrinsic to the verb and are generally described in terms of dynamicity, stativity, durativity, punctuality and telicity (Sasse 2001: 6) 2 . Many of these terms stand in close relationship to the concepts introduced in Vendler's (1967) system of socalled time schemata, in which he deploys the features [+/ -punctual], [+/ durative], [+/ -telic], [+/ -dynamic] to distinguish between the aspectual 1 See Sasse (2001) for a general overview of major 'schools' of the study of aspect. 2 Some more recent approaches (the so-called uni-dimensional approaches) doubt the necessity of distinguishing between these two aspectual levels and have established an all-comprising aspect category which is an amalgamation of aspect proper and Aktionsart (cf. Sasse 2001: 6). <?page no="194"?> 182 classes of states, activities, accomplishments and achievements 3 . Although originally these categories had been set up by Vendler to describe the temporal characteristics of verbs, later studies suggested that this system rather be extended to the temporal structure of whole events, which, in turn, can be encoded by linguistic expressions of different complexity. Thus, a verb originally categorized as 'activity', i.e. lacking an inherent terminal point and hence the feature [+telic], may be used in the depiction of an event which is an 'accomplishment' ([+telic]), in which case telicity might be 'induced' by the introduction of further arguments, temporal adverbials etc. To give an illustration, eat alone describes an activity, i.e. there is no inherent terminal point beyond which the activity cannot continue, yet, eat a slice of pizza verbalizes an accomplishment, defined by the presence of a terminal point - the act of eating is finished when the slice of pizza has been fully consumed. More recent approaches to the level of ASPECT 2 have put special focus on the contribution of sentence arguments to aspectual values of event descriptions and promoted models on aspect-driven argument selection. Such a view is most strongly expressed in Tenny's Aspectual Interface Hypothesis (1987, 1994), in which she claims that "[t]he universal principles of mapping between thematic structure and syntactic argument structure are governed by aspectual properties" (1994: 2). Following other linguists who set up a particular class of arguments as determinants of the telicity of a predicate 4 , Tenny (ibid.) introduces the concept of an argument which "measures out" an event. She identifies the CO as an element that belongs to this distinctive class of arguments. According to her claims, in (1) They sang a cheerful song. the temporal extensions of the unit a cheerful song can be considered as delimiting the event of singing. However, it has been argued that the mere inclusion of such a measuring argument does not automatically result in a change from an atelic to a telic presentation of an event. The make-up of the argument NP cannot be neglected, since, for example, the absence of a quantifying element within that phrase yields an argument whose quantitative characteristics are left unspecified. The integration of such an unquantified NP into an event-describing construction still creates an atelic representation of the event (cf. Levin & Rappaport Hovav 2005: 91). Compare the following examples, in which the second involves an unquantified NP and the third a quantified NP: 3 See Ch. 2.5, pp. 43f. for a more detailed description of these aspectual classes. 4 Dowty (1991), for instance, suggests an "Incremental Theme" argument, which is defined as expressing the physical extent of an object and which serves as a scale for the temporal progress of an event. <?page no="195"?> 183 (2) a. Alec ate. b. Alec ate carrot soup. c. Alec ate a bowl of carrot soup. It is questionable whether the inclusion of an argument suffices to completely change the atelic reading of sentence in (2a) to a telic one in (b). Only the integration of a quantified element in (c) is supposed to clearly lead to an aspectual shift. These and other observations, somewhat weakening Tenny's strong claims on the correlation between aspect and argument structure, lead Levin (2000) and Levin & Rappaport Hovav (2005) to the conclusion that, although some degree of interdependency is undeniable, telicity and aspectual notions in general cannot be considered the "sole determinants of argument expressions" (Levin 2000: 420). As concerns the rough ASPECT 1 : ASPECT 2 distinction introduced above, it goes without saying that both levels of aspect are closely intertwined in the overall interpretation of a predication's aspectuality, especially since both rely on the concepts of boundaries and boundedness. Sasse (2001) offers the following summary of the 'state of the art', recognizing, however, the usefulness of keeping the two levels apart: Nevertheless, a good case can be made for a distinction between two types of boundedness which can be associated with the dimension of ASPECT 2 and AS- PECT 1 : ASPECT 2 involves inherent boundedness while ASPECT 1 involves grammatically established boundedness. (ibid.; emphasis added, SH) Table 7.1 provides a simplified outline of the correlation between the two dimensions of aspectuality. ASPECT 1 imperfective perfective ASPECT 2 stative / atelic punctual / telic cover term non-bounded bounded Table 7.1 Relationship between aspect 1 and aspect 2 (adapted from Sasse 2001: 8) In my analysis of aspectual features of COs and COCs I will make use of the bounded - non-bounded terminology rather than the telic-atelic distinction, since the latter is traditionally understood as referring to terminating boundaries only, i.e. a process is described as telic only if it involves a terminal point (cf. Talmy 2000: 54). I, on the other hand, understand 'bounded' to refer to boundedness 'at both ends', which, however, does not necessarily mean the inclusion of the terminal point within these boundaries. That is, I consider an event to be bounded if its component states are conceptualized as being distributed over a limited span of time (cf. Langacker 1987a: 80ff.). With such a conception, the unbounded-bounded dis- <?page no="196"?> 184 tinction can be applied to a description of aspectual characteristics of both the CO and the verb phrase the CO is part of. 7.3 Boundedness: Concepts and implications In Langacker's CG approach to aspect (1982b, 1987a: 255ff.) the notion of boundedness plays an essential role for his description of different verb categories, labelled "perfective" and "imperfective" by him. Some terminological confusion might arise here, since he makes use of these terms in describing aspectual classes (i.e. ASPECT 2 ), referring to what in some cases seems to be the traditional 'stative-dynamic' and 'telic-atelic' distinction. While the perfective/ imperfective contrast is usually found with descriptions of morphologically or syntactically manifested aspectual oppositions (see above), Langacker's contrast is strictly semantic in nature (i.e. a clear division between the levels ASPECT 1 and ASPECT 2 is not undertaken). The notion of 'change in time' serves as the primary criterion for the distinction between "perfective" and "imperfective": Processes that are characterized by a change through time belong to the perfective type (e.g. break, kick, build), whereas imperfective processes describe temporal relations which are construed as stable, i.e. unchanging through time (e.g. know, resemble, see). Temporal bounding is set up by Langacker as a second criterion for his aspectual distinction, when he defines that […] a perfective process is specifically bounded within the scope of predication, whereas the imperfective process is not; the relation profiled by the latter may completely occupy the scope of predication and even overflow its boundaries. (ibid. 1987a: 261) Like other scholars (cf. e.g. Brinton 1995, Paprotté 1988, Talmy 1988), Langacker delineates the opposition between perfective and imperfective processes in analogy to the distinction between count and mass nouns. This conception is undoubtedly driven by the all-pervasive conceptual metaphor TIME IS SPACE , subsuming a large number of more specific submappings, such as SITUATIONS ARE MATERIAL ENTITIES / EVENTS ARE OBJETCS , TIME PASSING IS MOTION , ACTIVITIES ARE JOURNEYS etc. (cf. Haspelmath 1997, Janda 2002, 2004; Lakoff 1987, Lakoff & Johnson 1980, 1999). Imperfective processes such as sleep, know and think "occupy time the way substances occupy space" (Janda 2002: 6); viz. they can be likened to mass nouns as regards a) their lack of inherent boundaries, b) their homogeneity and infinite subdivisibility, and c) the fact that they can only be measured out in portions (cf. Brinton 1995: 28). Compare these characteristics in the below table: <?page no="197"?> 185 mass noun ( water) imperfect. process (sleep ) inherent boundaries none none homogeneity each part of water is itself water each part of sleeping is itself sleeping countable/ measurable in portions only a glass of water sleep for an hour/ some 5 Table 7.2 Analogy between mass nouns and imperfective processes Similarly, perfective processes such as dress, conquer or marry can be compared to count nouns in that they a) have inherent boundaries, b) cannot be subdivided, and c) are countable due to their consisting of distinctive phases: count noun ( table) perfect. process (dress) inherent boundaries yes yes heterogeneity/ subdivisabiltiy not any part of a table is a table not any state of dressing is an equivalent of the whole action countability/ measureability in its entirety a table dress in 2 minutes 6 Table 7.3 Analogy between count nouns and perfective processes To sum up, as regards the construal of imperfective processes as bounded and imperfectives as unbounded forms we can identify particular submappings of the TIME IS SPACE metaphor as providing the respective motivation: The metaphor pair PERFECTIVE IS A CONCRETE SOLID OBJECT vs. IMPERFECTIVE IS A FLUID SUBSTANCE is advanced by Janda (2007: 9) as marking the most obvious distinction between perfective and imperfective aspect. Just as mass nouns can be reconstrued as count nouns and vice versa, which is signalled by the addition of certain linguistic elements (She ate a peach [count noun] vs. There was some peach on her dress [mass noun]), boundaries can be imposed on inherently imperfective processes, and perfective processes can be "debounded". Paprotté (1988) and Jackendoff (1991) nicely describe these aspectual shifts through the metaphorical labels of "packaging" (i.e. bounding) and "grinding" (i.e. debounding). Bounding and debounding are illustrated in the below examples: 5 Talmy (1988: 179) terms such measuring phrases "portion excerpting", as they demarcate a portion of the unbounded event. 6 In contrast to "portion excerpting" expressions, phrases such as "in NP extentof -time " express boundedness (cf. Talmy 1988: 179). <?page no="198"?> 186 (3) a. I walked. I walked a mile. (bounding) b. She sighed. She kept sighing. (debounding) (cf. Brinton 1995: 32f.) Returning to COCs, a discussion of COs and the whole constructions in relation to boundedness is of interest on two counts: Firstly, integrating a CO into the verb phrase/ clause has been pointed out to have a bounding effect on the whole predication. Hence, COCs belong to the category of "packagers". Secondly, the CO is a bounded entity itself. As spelt out in preceding chapters, COs (and nominalizations in general) are the result of cognitive reification, by means of which an action or event becomes conceptualized as a bounded object. Such an analysis renders nominalization itself as one form of boundary-imposing ("packaging") processes. As regards the latter, Brinton (1995) takes up the question in how far the various nominalised suffixes found in the English language 7 differ in their bounding effect on the nominals they derive from the base verbs, and in how far the inherent temporal characteristics of these bases are preserved in the nominalised forms. Her approach implies that the notion of aspectuality is transferred from the category of verbs to the category of nouns, which is a logical assumption with respect to the list of parallels between the two classes as discussed above. Brinton concludes that [i]n English, both Latinate derivational suffixes and the native zero affix serve to perfectivize or package the aktionsart of the verb, much as the simple tense does in the verbal domain, with events [i.e. accomplishments] presented as bounded wholes, activities treated either indeterminately or determinately bound, and states remaining inherently unbound […]. The gerund seems to imperfectivize or grind the aktionsart of the verb […], resulting in unbounded situations. (ibid.: 39; emphasis added, SH) More precisely, the zero-suffix, the derivational suffix which is most frequent with COs, is described as a "unit-extracting" device for iterative activities such as laugh, breathe or cough. For verbs denoting continuous activities, e.g. run, walk, dream, Brinton observes an extraction of an arbitrarily bounded stretch of activity (ibid.: 35f.). In the following, I will test Brinton's observations against a description of the inherent structure of selected CO nouns and examine in how far there is an interaction of temporal and aspectual properties of the verb and the CO phrase if combined within a COC. 7 These include the zero-affix, the Latinate affixes -age, -al, -ance / -ence, -ation, -(e)ry, -ment, and -ure, and the native gerund ending -ing (cf. Brinton 1995: 29). <?page no="199"?> 187 7.4 Iterative processes Iterative processes can be described as events that are carried out in incremental units, i.e. ideally they consist of identical phases (component states) which can be repeated constantly and infinitely. Through their repetition they form a continuous activity which lacks inherent boundaries (see Fig. 7.6.a). Although Langacker (1991a: 208) notes that principally every verb in English allows for a replicate construal, and that many verbs are ambiguous between a punctual interpretation or an iterative reading (flash, blink, sneeze, kick), for some verbs, such as snore, giggle, breathe or laugh, the 'repetitive' interpretation comes along more naturally since human beings experience these activities most often as replicate processes. According to Brinton (1995), if these verbs are reified by means of a zero-suffix, one single phase is extracted out of the whole process. The resulting 'construct' bears resemblance to a prototypical count noun. The below diagrams contrast the respective construals of the verb breathe (7.6a) and its zero-derived noun breath (7.6b): Fig 7.6 Effect of the zero-suffix as nominalizer on iterative imperfectives Fig. 7.6 depicts the effect of the zero-suffix as a nominalising device in combination with imperfective iterative verbs, which is - as spelt out above - "unit excerpting": While the verb potentially construes a sequence of nearly identical component states, i.e. in the case of to breathe a repeated cycle of inhaling and exhaling air, the noun breath conceptually evokes only one of these components. Notice that the component profiled by the noun is picked out randomly. Such a random selection is possible since each of the units is roughly identical to the others, i.e. there is no change through time. My collection of COCs extracted from the BNC includes the following verbs denoting processes that are of an iterative nature or at least naturally Time Time a) X BREATHEV b) BREATHN + Ø <?page no="200"?> 188 occur as a sequence of iterated phases: breathe, laugh, cough, and possibly shoot, purr, nod, drop, jump. All of these, except for drop, instantiate COCs of the type EV/ R 1 . What are the consequences when an iterative verb and the respective CO are integrated to form a complex predication? The following observations can be made: Firstly, the inclusion of the CO as a perfectivizing element (i.e. the use of a COC) provides the event expressed by an imperfective verb with temporal demarcations. With the CO referring to one unit of the event denoted by the verb, the COC not only imposes temporal boundaries on the event, but also spells out the "quantity" inside such boundaries: The action extends over exactly one such incremental unit. Secondly, the integration of potentially iterative verbs into a COC results in the depiction of a punctual event and thereby dissolves the ambiguity of these verbs. Consider the following instances of COCs including iterative verbs: (4) a. As she walked the woman breathed a great breath of warm night air. (HG9: 2413) b. He […] coughed a harsh, chest-tearing cough before asking further. (HWE: 877) c. I mean that you did not jerk the trigger once, you didn't hang up on any first shot during rapid fire, you didn't shoot a shot that you didn't want.... (www.bullseyepistol.com/ zoyola.htm; 27/ 03/ 06) In (4a), for example, the action of inhaling air is limited to one 'portion' if compared to a possible CO-less variant She breathed warm night air. Hence, we can conclude that combined with iterative verbs, the CO not only provides the event with temporal limits, but possibly reduces the extension within these limits to a minimum. The 'sinlge-act' 8 reading is the only plausible interpretation for the below example, in which the CO cough is used in its plural form, emphasizing iteration: (5) She coughed […] not the hacking, tearing deep coughs of a bad attack but persistent little coughs, as though trying and failing to clear her throat. (ADS: 163) Clearly, the CO-NPs the hacking, tearing deep coughs and persistent little coughs are construed as a repetitive series of single acts - their interpretation as reifications comprising several component states seems an unlikely option. Note that in (5) the reading in favour of one single unit of the cyclic process is further supported by the modifying adjective phrases hacking 8 Cf. Janda (2007: 4f.), who distinguishes between and 'Complex-Act-Perfectives' and 'Single-Act-Perfectives' of Russian verbs. <?page no="201"?> 189 and persistent little, which both carry the idea of 'abruptness' or 'limited temporal extension'. The conceptual motivation for the construal of a single-unit-scenario can be found again in the TIME IS SPACE metaphor network. In a recent account on aspectual forms of Russian verbs, Janda (2007) identifies a cluster of submappings of this conceptual metaphor which seem to underlie and link different perfective and imperfective forms of a verb stem. The motivation for the so-called 'Single-Act-Perfective', described by Janda as a perfective verb form that "extract[s] a single cycle from a repetitive activity" (ibid.: 4) is to be seen in the metaphor A SINGULARIZABLE ACTIVITY IS A GRANULAR SUBSTANCE , where the identical component states are conceptualized as 'granules' of the substance and can be discretely singled out. In contrast to Slavic languages, where these aspectual distinctions are built into the morphological system, English has to express similar aspectual nuances by lexical means or constructional means beyond the morphemelevel. As Janda (ibid.: 7) concisely summarizes: "The shapes of events in English are not prior given, but often interpretable only upon reflection, as a result of factors beyond the verb itself". Constructional forms like the COC can be considered such factors that go beyond aspectual characteristics inherent to the verb in isolation. It has been pointed out above that many of the assumed iterative verbs are of an ambiguous nature as regards their aspectual values, i.e. besides the repetitive reading a punctual-construal-interpretation is possible as well. Quirk et al. (1985: 208), Langacker (1991a: 208) and Huddleston & Pullum (2002: 166) observe that these verbs evoke an unequivocally iterative construal in the progressive aspect construction, as the construal of an internal perspective on point-like events or their depiction as durative is highly implausible. Therefore, it is logical to assume that the use of COCs with iterative verbs is avoided in a progressive form if the 'single-unit' construal is the intended version, since the progressive, imposing a durative construal on an event, would then suspend this reading. The instances in my corpus of COCs confirm this assumptions; none of the examples of iterative verbs in a COC comes in a progressive form. Yet, since the number of instances of constructions involving iterative verbs is rather small (38 corpus hits), a generalizing conclusion at this point would be premature. 7.5 Continuous processes The situation depicted above for iterative verbs changes when it comes to verbs designating continuous, unbounded activities (i.e. activities that do not exist of iterated phases). Here, Brinton describes the effect of the nominalising zero-suffix as "portion-excerpting". That is, through the imposition <?page no="202"?> 190 Fig. 7.7 Effect of the zero-suffix as nominalizer on activity-denoting verbs 9 of this suffix a whole stretch of the activity is profiled within boundaries. Fig. 7.7 depicts this situation for fight: As is displayed in the figure, several component states are included within the excerpted portion of the process. Since these component states differ from each other, the entity denoted by a zero-derived noun is similar to a count noun with respect to its internal heterogeneity. The count-noun-like nature is also reflected in the type of determiners the derived form can be combined with: One can fight one fight, several fights or a couple of fights, just as one can read one book, several books or a couple of books. Due to the inclusion of several component states within the boundaries some processual character is still preserved in the nominalised form of the verb fight. It is precisely this feature which makes an eventreading of the CO possible, since some development within the bounded stretch is allowed for. However, the interpretation in favour of the eventive type (in contrast to the result type) depends on other features of the verb phrase the CO combines with, as will be shown below. With regard to the aspectual characteristics of a verb designating a continuous activity in combination with the respective CO, the following points can be put forward: Through the integration of the CO, i.e. a bounded entity, the event receives temporal demarcations as well, it is perfectivized. Keeping in mind Brinton's comment that the zero-suffix marks "some arbitrarily bounded stretch" (1995: 35), two different types of boundedness are possible: Firstly, we can think of a scenario in which this bounded stretch is construed as fully overlapping with the component states of the event designated by the verb, i.e. there is a coincidence of the 9 The different vertical positions of the component states symbolize a change through time. time time X FIGHT FIGHT + Ø <?page no="203"?> 191 respective boundaries 10 . In this construal, the action completely unfolds within the boundaries imposed by the CO. Hence, the final component state of the action lies within these boundaries. As this terminal component of an activity corresponds to the ultimate outcome of this activity, a result interpretation is available. Such an interpretation is stimulated if the verb phrase in which the CO is embedded carries the perfective marker PERF, which is itself a packaging device in that it singles out the final component state of an event. Langacker describes the influence of PERF as follows: The effect of PERF with perfectives is to collapse the temporal profile into the end-point of the process - in other words it derives a state defined in terms of the completion of a process, designating the point at which a trajectory is fully instantiated. (ibid. 1982b: 279; emphasis added, SH) Hence, in COCs in which a CO is combined with a verb phrase carrying a PERF marker, the result interpretation of the CO is coerced by aspectual characteristics of the verb phrase. Compare the following sentences: (6) a. Despite the heat they fought a phenomenal fight. b. Despite the heat they have fought a phenomenal fight. c. Despite the heat they are fighting a phenomenal fight. The versions in (6a) and (b) both illustrate the type of scenario in which the boundaries of the event designated by the verb and those of its reified form, the CO, potentially coincide. That is, the temporal extension of the action of fighting is included in the CO. Nevertheless, in the particular case of (6b), a result reading of the CO is more prominent than for (6a) due to the influence of PERF, which zooms in the final component state of the event. Version (6c), on the other hand, depicts a completely different situation, instantiating the second type of the two possible 'boundedness scenarios', to which the notion of a 'coincidence of boundaries', as described for the first two sentences, does not apply. In this example, we have a combination of the progressive aspect PROG 11 , usually imposing an unbounded, internal perspective on an event, and a CO, which, as a perfectivizing element, has been described as evoking a bounded construal of an event. Combined with a debounding device such as PROG, the boundaries the CO imposes are partly suspended. The temporal setting of the scenario now comes close to what Brinton describes as "bound, but of 'indeterminate limits' " (ibid.: 38), with the boundaries being of a more schematic or potential nature. The progressive aspect can be said to, firstly, weaken the effect of the CO, and secondly, to coerce the event-CO reading of the nomi- 10 See Talmy (2000: 54f.) for a slightly different use of the notion of "boundary coincidence". 11 In the following, PROG is to be understood as referring to the whole progressive construction, i.e. be +V-ing. <?page no="204"?> 192 nal form: It 'zooms' into the event (i.e. imposes a "restricted immediate scope" [Langacker 1991a: 210]) and thereby 'cuts out' the boundaries and thus the final state, which is not profiled within such an immediate scope. This type of construal interferes with (or at least backgrounds to a considerable degree) a result-interpretation of the CO. The interaction of the "grinding" progressive aspect and the "packaging" CO phrase is depicted in Fig. 7.8: Fig. 7.8 The interaction of grinding -ing and packaging CO As illustrated, the CO imposes boundaries on the imperfective event, which are 'toned down' by PROG 12 . Within these potential boundaries, the process is depicted as evolving through time. Such a 'setting' is similar to that verbalized in the sentence She was drinking a glass of wine, as discussed by Krifka (1992: 39f.) and Filip (1995: 672). Here, the NP is described as providing the event with potential boundaries (where the terminal point of the action coincides with the consumption of the last drop of wine contained in the glass). Yet, the process of drinking is presented as ongoing within these boundaries. Note that the CO-less variant of (6c), They are fighting, lacks such potential temporal limitations. One of Macfarland's (1995) essential observations is that the main overall effect of the COC is a change in telicity of the event designated by the verb (cf. ibid.: 145ff.), yet her description completely lacks a discussion, even acknowledgement, of the fact that there is an interaction of the CO as an 'aspectualizer' with other aspect-denoting linguistic elements. It is exactly the scenario described in the preceding paragraphs which has made me treat the notion of 'telic' with some caution in a description of the effect 12 The schematicity of these boundaries is indicated through a dotted line of the boundary-imposing ellipsis. Grinding be + -ing Packaging CO X is fighting a fight <?page no="205"?> 193 of COCs. The use of inherently unbounded verbs in the COCs has a bounding effect, but not necessarily a 'telicizing' effect, if 'telic' is understood as including the ultimate state of an event. Such a construal is just one special option, in the case of which the temporal extensions of the event are construed to fully coincide with the stretch profiled by the CO. An important observation, which lends support to my view, is made by Filip (1995), who discusses the influence of aspectual markers on the interpretation of nominal object arguments 13 : Quantized or telic events involve directionality, an inherent orientation toward a definite state. If a perfective operator imposes bounding over an atelic event, it does not add telic structure. The notion of simple bounding with no directionality in the verbal domain parallels bounding of mass nouns, for example, in the nominal domain: Just as a bathtub can be full of water, a jar full of beans, so can a sustained stretch of activity, such as blushing, be thought of as being simply associated with a certain bounded interval of time. (ibid.: 674; emphasis added, SH) Considering the multiplicity of types and tokens in my corpus of COCs which occur in the progressive form, Macfarland's claim must be evaluated as too strong. The following sentences exemplify the use of COCs with progressive aspect: (7) a. What expression do they have when they're thinking black thoughts about someone they're trying to ingratiate themselves with, or when they're being praised for qualities they're aware they don't possess. (G0F: 1466) b. No doubt, too, he was dreaming pleasant dreams of his dinner companion of last night! (JYF: 1955) c. No, they're still fighting the good fight. (C87: 1416) d. Ward was smiling a bright, happy, almost drunken smile, playing the innocent Scot and putting on his broadest accent. (GV6: 16) Taken together, a total of 9.2 % (289) of the COC expressions in my collection are embedded in a PROG construction. As in the case of passive forms, the use of progressive COCs considerably depends on the type of construction ( 2 = 17.81; df = 2; p< 0.001) : While the A-types occur significantly less often than expected in a PROG form (37/ 5.91%), EV/ R 1 do so significantly more often (228/ 10.71%) (see Table 7.4). The not uncommon usage of a 13 In her discussion, Filip describes the influence of Czech perfective markers on the reading of nominals which are underspecified for definiteness as the language system lacks this type of determiner. Czech vino, for example, can both refer to "wine" (mass noun) and "the wine" (count noun). The addition of a perfective marker (denoting boundaries) to the verb taking the NP as its argument, leads to the interpretation of the NP as denoting a bounded entity (Filip 1995: 658f.). <?page no="206"?> 194 combination of COC and progressive construction stands in accordance with the observation that common progressive aspect verbs typically take as their human subject an AGENT , actively controlling the action (cf. Biber et al., 1999: 470 ff.), a function that corresponds to that observed here for COCs, i.e. their profiling the event's trajector as a controlling AGENT , which is particularly pronounced with examples of the type EV/ R 1 . That is, a combination of COCs and PROG enforces the notion of agentivity. Moreover, as Biber et al. observe, "verbs like think and wonder, which commonly occur with progressive aspect can be interpreted as invoking an active agent controlling the thought process" (1999: 473, emphasis added, SH.). Thus, for think, which was said to take an AGENT - EXPERIENCER in the COless (default) variant, the combination of COC and PROG unequivocally depicts the trajector of the process as an AGENT . Furthermore, since the COC adds agentive traits to the trajector, verbs which take a human EX- PERIENCER as subject and thus are found to rarely occur in the progressive (CO-less form) (cf. ibid.), would seem to be more felicitous in a progressive EV/ R 1 R 2 A sing smile tell think live dream die sleep grin fight talk dance run do purr shoot laugh 84 (18.0) 14 28 (13.8) 27 (6.7) 27 (34.6) 25 (3.5) 10 (22.2) 7 (8.04) 5 (35.7) 4 (22.2) 3 (16.6) 2 (40.0) 1 (8.3) 1 (50.0) 1 (1.5) 1 (100.0) 1 (25.0) 1 (10.0) 228 (10.7%) produce build weave 15 (10.6) 5 (5.0) 2 (7.7) 22 (7.4%) sow smell name give drink publish plant offer feed operate repeat invest charge feel nail rain 11 (5.6) 5 (20.8) 3 (6.7) 3 (2.3) 2 (8.7) 1 (14.3) 1 (5.9) 1 (50.0) 1 (4.8) 1 (50.0) 1 (100.0) 1 (50.0) 1 (16.7) 1 (4.3) 1 (50.0) 1 (33.3) 37 (5.19%) Table 7.4 Frequency of verbs in progressive construction 14 Number in brackets = percentage of instances of particular verb in progressive construction <?page no="207"?> 195 COC variant. With regard to the examples in my corpus, this applies to the verbs dream and sleep, which occur in such a constructional combination in 22.2 % and 35.5% of their uses, respectively. Regarding the observation that PROG weakens the bounding effect of the CO, the notion of 'force dynamics', in particular that of 'agentivity', then qualifies as one plausible motivational factor other than aspectuality for the use of a COC in such instances: In the examples in (7), the profiling of the subject as an active creator, i.e. the depiction of a directed energetic flow from AGENT (source) to PATIENT - EVENT (sink), appears to be more prominent than the imposition of temporal demarcations. Furthermore, one cannot neglect modification or stylistic reasons as factors driving the preference of COCs over other available constructions: Both (7a) and (d) do not even have a corresponding simple verb + adverbial variant which would express the same conceptual content as 'elegantly' or 'compactly' as the COC. 7.6 Peripheral cases So far, the focus of the discussion has been put on COs which are zeroderived from their respective verbs. My corpus of COCs also attests CO nominals which are derived by other suffixes, e.g. -ment (arrangment) or ion (decision), most often described as denoting an 'action', 'process' or 'state'. Brinton (1995: 30) claims that the meaning of the derived nominal depends on the original verb, there usually being a "one-to-onecorrespondence in Aktionsart between verbs and the corresponding nouns". The over-all effect of an imposition of these derivational suffixes is, similar to the zero-suffix, perfectivizing (i.e. bounding). However, as the author notices, the majority of nouns formed through this type of derivation are built from telic verbs, viz., in her understanding, verbs that have an inherent terminal point. This observation has two implications: Firstly, the stretch of the event which is profiled can be taken to include the final state, which entails that the nominalizations of telic, i.e. perfective verbs, present the event completely. This in turn facilitates a result interpretation when these nouns occur as a CO of their respective verbs. In fact, the majority of the COs derived by regular suffixation do not even allow for an event interpretation (see below). Secondly, as regards the base verbs it seems unlikely that their use in the COC is solely for the purpose of imposing boundaries on the event, as their semantics already provides them with terminal points. In such cases the CO phrase can be described as explicitly spelling out these inherent temporal demarcations. In my collection of COCs, the majority of those verbs whose corresponding CO nouns are derived by suffixes other than the zero-morpheme fall in the category of inherently bounded verbs: arrange, attach or create <?page no="208"?> 196 belong to the class of accomplishments and so do their nominal counterparts, while decide can be classified as an achievement, i.e. a punctual event. Moreover, all of these verbs belong to the group of peripheral CO verbs associated with the constructional types R 2 and A, i.e. those verbs that conventionally occur in a transitive pattern. For this more peripheral class, the CO has been defined in this study as either an object of result (i.e. effected object) and as such describing the (potential) final state of the process, or an affected object (cf. Ch.4, pp. 79ff.). See the following examples: (8) a. What does my right hon. Friend think that we could do to stop the drug being imported into this country, or arranging some reciprocal arrangement whereby we would stop importing Dutch cheese or some such thing? (HVV: 20660) b. An alternative method is to attach an electric drill water pump attachment to the tap. (HH6: 1402) c. Campaign Level lies at the core of the product, and is where the major decisions and tactics are decided. (HAC: 1902) As was already pointed out in Chapter 6, in such arrangements the CO is simply one of the many possible arguments filling the PATIENT role. Therefore, the use of the respective verbs in a COC does not result in a shift from unbounded to bounded or a change in other aspectual features if these forms are compared with their non-cognate object counterparts. Here, the use of a verb complemented by a cognate object could be driven by expressive or stylistic reason, where the partial repetition of form has an intensifying effect (as in 8c), or might simply be due to an incidental, stylistically less elaborate 'coincidence of form' between verb and a rather general object (as in 8a-b). The discussion of perfective, i.e. inherently bounded verbs occurring in a COC naturally leads to an analysis of the verb die and its CO death, although the morphological process deriving the noun from the verb is different from the cases discussed in the preceding paragraphs 15 . In major reference grammars and studies on Aktionsart or verbal situation types, the verb die is described as denoting a conclusive punctual event (Quirk et al. 1985: 201) or an achievement, i.e. a telic and punctual event (Dowty 1979, Pustejovsky 1991), which, however, allows some extension in time (Huddleston & Pullum 2002: 166). Recalling Macfarland's suggestion that in some cases the inclusion of the CO adds the temporal feature 'durative' (see Ch.2, p. 43f.), we might build on her observation in testing for whether the 15 The etymology of death can be summarized as follows: death > O.E. deað, from PreTeut. *dauthuz, from verbal stem *dau- "die" + *-thuz suffix indicating "act, process, condition." (cf. "death", OED, online version; "death"; Online Etymological Dictionary, www.etymonline.com). <?page no="209"?> 197 CO death can be interpreted as adding aspectual nuances. Although Macfarland does not explicitly say so, it could be argued that the use of die with a CO adds a notion of greater temporal extension. It does not suspend the temporal boundaries semantically inherent to the verb, but rather widens them. The corpus provides some interesting examples which seem to support this suggestion: (9) a. An extreme form of the first is the fur trapper in the Russian and Canadian Arctic who lays a line of traps across country which he visits once a fortnight collecting the victims who have died a slow and agonising death. (B03: 103) b. The lines were coming out as written, but the play was dying a slow death. (H92: 934) c. We are dying a slow death, and the players are frustrated and con fused. (K4T: 5858) Similarly to the examples of punctual verbs in (5), however, it must be pointed out that the adjective slow, modifying the COs in (9), considerably contributes to the idea of particular temporal extensions. Moreover, as exemplified in (9b) and (c), the NP slow death occurs often in a die-COC in combination with the debounding PROG-construction 16 , in the case of which the notion of durativity is added in the first place by PROG and not by the CO. Thus, instead of seeking the motivation for the use of punctual die in a COC in the aspectual characteristics of the construction, I refer to the change in the participant constellation associated with the COC: The primary reason for using die in the construction is an emphasis of the trajector as an EXPERIENCER of an event- THEME , aspectual notions seem to play a secondary role only. The cognitive-linguistic description of aspectual characteristics of COs and COCs presented in the above discussion also provides an explanation why COCs are found only rarely in combination with -ing-derived nominals. The reason must be sought in the aspectual nature of these derived forms: Brinton's description of -ing as a nominalizing suffix explicitly renders it as a "grinding", i.e. a debounding device, and as such it functions like the progressive-marking verbal suffix -ing, zooming into the event and cutting out the terminal components (see Fig.7.9). The nouns resulting from ing-nominalizations behave like mass nouns, which shows, for example, in the type of determiners they occur with: (10) a. the/ some/ much/ *a/ blood; *two/ *many bloods b. the/ some/ much / *a running; *two/ *many runnings 16 In my corpus of COCs, 4 of 6 (66%) of the instances of die + slow death occur in combination with the progressive construction. <?page no="210"?> 198 Fig. 7.9 Conceptual content of -ing-derived nominals (here: running) 17 If one of the functions of COCs is to mark a shift in aspect from 'unbounded' to 'bounded', the choice of -ing-derived nominals would be quite infelicitous as they denote unbounded, imperfective events. The few examples attested in the corpus building, painting, drawing) must be seen as exceptions: They are lexicalized forms which denote concrete results of the action instead of reified events. 7. 7 Conclusion Summing up this discussion of aspectual characteristics of COs and COCs, we can record that the inclusion of a CO, itself a 'packaged' event, provides boundaries for the event expressed by the verb it is combined with if the verb belongs to the class of imperfectives, i.e. verbs designating inherently unbounded events. As could be shown for this class, the inherent aspectual structure of the base verb has a decisive effect on the nature of the CO's inherent structure. COs formed from iterative verbs most likely denote one single unit of the process, while COs derived from continuous, noniterative verbs construe a bounded stretch of the event, i.e. they comprise several component states. In combinations of COs with their verbs, different interpretations of the CO are possible, depending on whether there is a boundary coincidence or a lack of such a complete overlap: If the boundaries of verb and CO are construed as coinciding, the CO is ambiguous for a result or event reading. As has been argued, the combination with a PERFmarked verbal head triggers a result reading. If the COC comes in a progressive form, an event-reading is more plausible since PROG dissolves the temporal demarcations the CO imposes and merely leaves the event with 17 Note that the dotted ellipse indicates a highly schematic boundary, which represents the mass-noun like character of the derived form. (the) RUNNING <?page no="211"?> 199 potential, schematic boundaries. With this latter scenario, the usefulness of distinguishing between bounded construals and situation types which are traditionally described as telic (i.e. including the terminal point of an event in the profiled stretch of activity) could be demonstrated and consolidated. Moreover, the analysis has clearly proven that distinguishing COs into objects of results and event objects can only be done by also taking into account aspectual characteristics of the verb phrase it occurs with. In isolation, the zero-derived CO is unspecified for these different readings and thus of an ambiguous nature. For a third class of verbs - the class of perfective verbs, inherently endowed with terminal points - no changes could be noticed in the aspectual characteristics if the verb occurs in a COC. Here, the CO merely 'fleshes out' these temporal limits. With these concluding remarks on the effects the construction exerts on aspectual features of an event, the investigation of COCs will now leave the higher levels of constructional schemas suggested for our network of constructions and move to mid-level schemas, studying in more detail conventions on the make-up of the CO nominal from a usage-based perspective. <?page no="212"?> 200 8 Mid-level schemas: The make-up of the COnominal 8.1 Introduction The network of COCs developed and fleshed out in the preceding chapters so far integrates relatively abstract constructional templates only, comprising information on available COC subtypes, verbs (and consequently CO nouns) which are strongly attracted by these different construction forms, and semantic/ functional factors (such as event elaboration and temporal bounding) which are associated with these higher-level schemas. However, further specifications as to types of determiners or modifying elements which typically elaborate the bare CO noun and which are part and parcel of a proper description of the CO-NP, and hence of the whole construction, have not been made yet. These types of constructional characteristics are represented in so-called mid-level schemas of the network, i.e. schemas which, for example, contain more specific information on types of specification (e.g. definite vs. indefinite) or modification (e.g. adjectival premodifier vs. prepositional postmodifier) found with the CO-NP, but still allow for some degree of lexical flexibility within the construction. Possible forms of mid-level schemas will be scrutinized in the subsequent sections, which will investigate in great detail the make-up of CO-NPs. In the course of this survey, the results of a close inspection of some more fine-grained characteristics of COCs and their distribution over the different subtypes of the construction will be presented and discussed. Once more, special attention is paid to the EV/ R 1 type, the assumed core type of COCs. 8.2 Determiners in the CO-phrase As mentioned in Chapter 2, there has been some controversy on the use of definite determiners in combination with CO-NPs. Whereas Massam (1988) doubts their use with a class she labels true COs 1 , pointing out that these arguments denote events which come into existence through some action and thus cannot be taken to pre-exist, Macfarland (1995) considers COs to be definite in any case: She understands the notion of definiteness in terms of presupposition, that is, any entity that can be presupposed is definite. From her point of view since verbs of creation (the prototypical class of 1 This class roughly corresponds to my EV/ R 1 -COC type. <?page no="213"?> 201 verbs occurring in COC) presuppose a created object these objects are definite by their very nature. In short, Massam considers an indefinite determiner within the CO-NP the default case, Macfarland, however, does not impose any such constraints. Before I turn to findings from my investigation of the corpus I will briefly summarize how determiners, definiteness and consequently, the use of definite determiners, are explained within the Cognitive Linguistics paradigm. The general function of determiners consists in the profiling of an instance of a type: Whereas a bare or modified noun (e.g. [full] life) merely designates an abstract type, the addition of a definite determiner (e.g. the [full] life), for example, is said to instantiate and ground 2 that abstract type, singling out a uniquely identifiable entity or set of entities and defining a particular 'location' of the entity in reference to the ground (cf. Langacker 1991a: 55f., 97ff., 549). The two more general groups of such nominal grounding predications, i.e. definite forms (including the definite article the, demonstrative pronouns, possessive pronouns) and indefinite forms (a, some, any, quantifiers) are distinguished on the basis of whether they instantiate an entity that can be uniquely recognized by both speaker and hearer (definite), or an entity which is arbitrarily picked out of a typeclass as a representative and which the hearer cannot uniquely identify in the ground (indefinite) 3 . Thus, by choosing the definite determiner the, the speaker basically assumes the hearer to be able to mentally contact the same entity the speaker has singled out (Taylor [2002: 354] speaks of "joint identification"). The use of indefinite determiners, on the other hand, excludes such a "coordination of reference" (Langacker 1991a: 98), the hearer cannot access an individual, unique entity in his mental discourse space. In sum, while both types of determiners establish reference, the contrast between definite and indefinite reference shows in the selection of the respective determiner. In the following, I will elaborate on the data obtained from my analysis concerning the frequency and distribution of different types of determiners with COCs and offer possible explanations of my observations, building on the above considerations on the conceptual content of the various grounding predications. 2 'Ground' is understood as the context of a speech event (including participants, situational context, knowledge shared by the participants etc.); and 'grounding' as the process of 'pinning down' an entity in relation to this ground (cf. Langacker 1991a: 96f., Taylor 2002: 346). 3 This is, of course, a very general description of the different types of determiners. See Langacker 1991, chapter 3, for a fine-grained analysis of grounding predications. The CL description of the definite article stands in close relationship to the many studies into the meaning of the which either analyze the determiner in terms of unique identifiability or in terms of familiarity. See Epstein 2001 for an overview on these studies, and Hawkins 1991 for a typology of sources of definiteness. <?page no="214"?> 202 My analysis of the corpus instances yielded the results which are summarized in Table 8.1. In each of the subgroups of COs the indefinite determiner a (Ø in the case of plurals) occurs most frequently, followed by the definite determiner the. Whereas the frequent usage of the first type of determiner should not pose a problem, as it is assumed to be the 'default case' for COs, the data observed for the definite type are worth being discussed in more detail, keeping in mind Massam's consideration that COs as event objects do not pre-exist or cannot be presupposed and can thus not be referred to by definite noun phrases (which induce the presupposition that their referents exist). There is obviously a considerable difference between the two subgroups including effected objects (i.e. EV/ R 1 & R 2 ) on the one hand and the subgroup including affected objects (A) on the other, with the latter appearing almost twice as often with definite the. The ²test shows that the distribution of the types of determiners is highly significant, determined by the type of COC ( ² = 321; df = 20, p < 0.001). That is, the null hypothesis, which - in this case assumes an equal distribution of determiners in the Type 1 2 3 4 5 6 7 8 9 10 - EV/ R 1 59 2.6 4 436 19.6 1095 49.3 120 5.4 419 18.9 7 0.3 34 1.5 9 0.4 20 0.9 5 0.2 14 0.7 R 2 3 1.0 57 19.3 183 61.7 6 2.1 14 4.8 0 30 10.0 2 0.7 0 0 2 0.7 A 8 1.1 254 40.6 260 41.5 13 2.0 39 6.2 0 19 3.0 1 0.16 19 2.8 7 0.9 5 0.8 Total 70 2.2 747 23.9 1538 48.8 139 4.4 472 15.0 7 0.2 42 1.4 12 0.4 39 1.2 12 0.4 21 0.7 1 = no determiner (in the case of singular forms) indefinite 2 = definite 'the' definite 3 = indefinite (singular: 'a', plural: Ø ) indefinite 4 = demonstrative pronouns definite 5 = possessive pronouns definite 6 = universal pronouns (each, every) indefinite 7 = quantifiers indefinite 8 = relative/ interrogative 'wh' indefinite 9 = assertive/ non-assertive (any / some) indefinite 10 = negative 'no' indefinite - = no categorization possible Table 8.1 Distribution of determiners in COCs 4 Percentage of respective CO type <?page no="215"?> 203 different sub-categories, must be rejected. It is especially the values of variation (between expected and observed frequency) of definite the in type A (74.5) and quantifiers in R 2 (63.2) (e.g. Essex Water Company have applied for permission to build five new buildings [CF5: 21]) which are 'responsible' for the high ² value. The different nature of effected and affected objects might serve as an explanation: Affected objects are already in existence and a joint identification of a unique referent by speaker and hearer can be arrived at more easily than with created objects as they do no exist yet. Thus, it is more 'natural' to produce utterances of COCs involving effected objects with an indefinite determiner (1a-b), whereas a definite determiner with affected objects (1c-d) is as likely as an indefinite one. (1) a. He practises what he has always preached respect and tolerance for people of whatever religion, who try to live a good life by their own creed. (CAR: 1326) b. As the roars of approval reverberated around the ground he grinned an uncomfortable grin, almost one of embarrassment. (HJ3: 2676) c. The plants can be planted in flower-pots and embedded in the gravel. (CBL: 797) d. Now the seeds had been sown. (H7E: 1496) It seems that the evidence from the corpus supports Massam's assumption at least with regard to the indefinite article a 5 . However, the corpus also includes definite 'counterexamples'. The use of definite the with a CO is unproblematic in those cases in which the surrounding co-and context provides sufficient information to establish unique reference. Hawkins (1991: 408) describes this source of definiteness as a situation in which an entity is a member of the "previous discourse set". For the majority of cases of do the deed (31 instances), or the relatively common fixed phrase live/ survive to tell the tale (23 instances), for example, a plausible explanation can be given on these grounds. As for the former type, we are most often introduced into settings which evoke scenarios of crimes, so that it can be easily inferred that the deed refers to a criminal act, as illustrated in (2a). In 5 If we add the percentages of definite and indefinite forms, we arrive at the following figures: EV/ R 1 : Definite: 43.9 % Indefinite: 56.1 % R 2 : Definite: 26.2 % Indefinite: 73.4 % A: Definite: 48.8 % Indefinite: 51.2 % That is, only in the R 2 type of COCs, the use of an indefinite determiner clearly prevails a result one should have expected first and foremost for the EV/ R 1 type. The ²test confirms a highly significant distribution of the usage of definite vs. indefinite determiners depending on the type of COC ( ² = 44.76, df = 2, p < 0.001). The values of variation between observed and expected frequencies are highest for the usage of both determiner types with COCs of the A-category. <?page no="216"?> 204 other cases, the deed simply is a more general, superordinate term, substituting for a more specific NP in preceding discourse, as in (2b): (2) a. Shifting loyalties once more, he determines to kill Custer, but cannot do the deed when he has the chance, and becomes a vagrant. (C9U: 1175) b. Apparently Hank's divorce had come through sooner than expected, the deed was done one Saturday afternoon in the local Registrar's Office […]. (B3F: 1004) The phrase live/ survive to tell the tale is found in contexts which report on wars, accidents and other catastrophes or dangerous situations. The CO the tale then designates reports on these terrible events: (3) a. In order to survive the jungle and live to tell the tale it is important not only to have good companions, but also to have the best available equipment. (AYP: 169) b. Inevitably many books and much schoolwork was destroyed by the fire but there is at least some good news: the Holy Trinity pet gerbils were rescue and survived to tell the tale. (K21: 3152) The above examples are unexceptional and could be easily complemented with instances of other COC verbs in combination with a definite CO-NP. Yet, the corpus also provides examples which are more problematic cases for which the obvious context does not offer an adequate amount of information for mentally accessing a unique referent of the CO: (4) a. He had fought the good fight, he had finished his course, he had kept the faith. (ALK: 767) b. When people in the 60s decided to get away from it all and live the good life in the Cotswolds they were treading a well-worn path. (K1J: 1421) c. If you intend to fight a missile duel, the chances are that most of your highly mobile force will attract a disproportionate amount of your enemy's firepower and will die the death. (CM1: 1610) d. […] Balor simply turned his head and grinned the gap-toothed grin and occasionally nodded amiably. (G1L: 1357) How can the uses of the definite article in the above examples be accounted for? As for (4a), the use of a definite grounding predication can be explained by the somewhat proverbial character of the phrase. Fight the good <?page no="217"?> 205 fight is a biblical quote 6 and as such constitutes a fixed phrase which does not leave a choice as regards lexical items. Moreover, it is an elliptical form, as in the original quote the good fight is postmodified by the PP of faith. It is precisely this phrase which enables the use of the definite article, as it pinpoints the reference uniquely, i.e. the necessary information is provided in the NP itself (Hawkins 1991: 410). This so-called "cataphoric reference" 7 justifies many of the instances of COs determined by the definite article in our data. Other more or less idiom-like phrases such as sow the seeds of + abstract NP (66 instances), tell the tale of + NP (15) or sleep the sleep of + deadjectival NP (7), which contribute to the relatively high frequency of the in a CO-NP, can be explained along these lines. In example (4b), the CO-NP the good life can be seen as representing a commonly and socially agreed-on best form of different forms of the type GOOD LIFE , which thereby appears as a unique, mentally accessible entity and thus allows for the use of the definite article 8 . The seemingly problematic collocation die the death 9 in (4c) can be justified in a similar fashion. The OED circumscribes the phrase as 'to be put to death', with Dr. Samuel Johnson (quoted ibid.) explaining that "'die the death' seems to be a solemn phrase for death inflicted by law". That is, the CO-NP designates a particular type of death, which was legislatively and judicatively approved of and recognized by a cultural group, a development which is mirrored in the relatively fixed phrase with an unambiguous, conventionalized meaning. Considering example (4d), it is the introduction of the proper name of the character, Balor, which provides sufficient context to understand the gap-toothed grin as the type of grin that is specific of and unique to the person. That is, Barlor functions as a type of reference point (see below) that provides the hearer with clues for an unmistakable identification of the kind of smile produced. 6 Quoted from Timothy 6.12.: "Fight the good fight of faith, lay hold on eternal life, whereunto thou art also called, and hast professed a good profession before many witnesses". The abbreviated version fight the good fight is found in a hymn by J.S.B Monsell (cf. Oxford Concise Dictionary of Quotations.) 7 Quirk et al. (1985: 268) explain this type of use of the by cataphoric reference, in the case of which the postmodification of the NP "restricts the reference of the noun, so that its referent is, for the purpose of discourse, uniquely identified". 8 It may be added here that the good life can be said to represent the stereotype of the class GOOD LIFE . A stereotype comes along with a particular set of characteristics which makes it an outstanding representative of the whole class. This uniqueness then enables the use of the. See Lakoff [1988: 79f.] for a discussion of prototype effects on stereotypes. 9 Die the death started life as a variant of the biblical quote die by the death, which in turn is the literal translation of the original Hebraic Bible line. (Marsen, Richard. "Cain’s Face, and Other Problems: The Legacy of the Earliest English Bible Translations". http: / / www.tyndale.org/ Reformation / 1/ rmarsden.html. Accessed 30/ 10/ 2006). <?page no="218"?> 206 The above considerations allow for an account of the use of the definite article with a CO, which also shows that Massam's objection (i.e. that created objects cannot be presupposed and thus cannot be used with a definite article) does not rule out such a combination: Even if the CO refers to a newly created entity, this entity might be of a type which is generally known by or familiar to speaker and hearer. That is, the CO can be construed as the replication of an acknowledged type, consciously and purposefully "copied" by the AGENT . Such a reading, which is exemplified in (4b c), is fully compatible with the semantics of the construction as portrayed in this study. Having spelt out possible settings which sanction the use of definite the with COs, I now turn to another class of determiners, prenominal possessives, the use of which seems to be especially common to the class of EV/ R 1 -COCs (18.9%): (5) a. ST came to me and rubbed against my side, purring his rattling, wheezing purr. (G02: 1724) b. He was squatting at our feet grinning his wide sheepish grin. (FEM: 1957) c. He sighed her sighs and, in his gentlemanly manner, raised his eyes against the mocking Fates. (FP1: 2098) d. Our collective unconscious does not dream Martin Luther King's dream, or pray Pope John Paul's prayer, it endures the nightmare of the lost souls, the raving of the unclean spirit. (B1J: 2004) e. To rediscover the life of Jesus in our own times is to come before him, poor and naked, and to live his life, not ours. (ARG 1069) Like the definite article, prenominal possessive constructions might be considered as members of the class of grounding predications which can establish specific, definite reference 10 , i.e. the referent is construed as being uniquely identifiable by both speaker and hearer. Several linguists working in the CL paradigm have analyzed the meaning of prenominal possessive constructions in terms of Langacker's (1991a) reference-point model (ibid. 1993, Janssen 2003, Taylor 1996b). This model, succinctly described by Langacker (1991a: 552) as "a fundamental model based on our capacity to invoke one conceived entity (a reference point) for purposes of establishing 10 See Langacker (1991a: 168ff.) and Taylor (1996b: 136) for a discussion of possessives as determiners. It must be noted, however, that possessives are close to or even overlap with the category of modifying predications (note that in traditional approaches, articles are considered as a subgroup of adjectival modifiers).Whereas the conceptual content of the is rather schematic, the possessive construction carries more specific content sufficient to modify an element (see section 8.3). <?page no="219"?> 207 mental contact with another (the target)" accounts for possessives, such as in Dido's ear, as follows: In order to enable the hearer to mentally access the intended referent (ear), another, more salient entity, the possessor Dido, is invoked, in whose neighbourhood the referent of the possessee is located and thus can be traced via the reference-point. That is, unlike definite morphemes such as the or that, which convey direct reference, possessives do so only indirectly. Returning to the examples in (5), we see that the nature of the relationship between 'possessor' and 'possessee' is of a highly specific kind. As Langacker (ibid.: 171) notices, in cases of the use of possessives with nominalizations, as in Lincoln's assassination or her arrival, the possessor (i.e. the reference-point) relates to a participant in the nominalised action or event. Choosing the participant as a reference point comes naturally, as it is next to impossible to conceive of an action/ event autonomously, without conceptualizing at the same time the entities involved therein. One might go one step further and describe the possessive determiner in such a case as a grounding relation 11 , profiling the relation between an event and a participant - a description which has been postulated by Langacker for nominal periphrasis with of (1999a: 83ff.), as in the chirping of the birds. Here, of is said to profile an intrinsic relationship between an event (chirping) and the AGENT (birds). This description is applicable to COs with a possessive determiner, in the case of which the 'possessor' corresponds to the AGENT of the event / action named by the CO. Fig. 8.1 (c) depicts such a possible relation as exemplified in her smile: The big, bold-printed circle around the process marks its reification ( SMILE ), the smaller circle represents the AGENT ( SHE ), with the possessive form her (represented in the figure by a double line) establishing this relationship 12 . In the majority of examples from the corpus, the 'possessor' is coreferential with the AGENT of the verb phrase, as in (5 a-b). Where this is the case, we have a conceptual overlap of entities not only between the process expressed by the verb and its reified form in the CO-NP, but also between the agent of the process, verbalized as the verb's subject, and as the 'possessor' in the CO-NP. This results in the depiction of a relation between AGENT , action, and outcome which is construed as most 'natural': The result of the action is in full accordance with the agent and his/ her specific way of performing the particular action expressed by the verb. With this de- 11 It could be argued that the reference point-construction profiles a relation as well, if we consider the mental path that leads from the reference point to the target as relational. 12 This description is analogous to that of possessive constructions involving the genitive -'s (e.g. Linda's smile). Here, the morpheme -s can be described as profiling the relation between process and participant. <?page no="220"?> 208 Fig. 8.1 Participant-Event relationship in nominalized possessive constructions, e.g. her smile (adapted from Langacker 1999a: 86) scription it can be readily accounted for why a CO-NP which includes a possessive determiner that is co-referential with the AGENT occurs so frequently with live+ life (201 instances). The construction depicts a situation in which the AGENT is involved in and controls a kind of life which conforms to his/ her most intrinsic demands and expectations of having and passing his/ her existence. (6) a. Then you can decide to live your life fully getting as much out as you put in. (ED4: 883) b. You've got to live your life too, you know, said Keith with a grin. (A0F: 3591) However, as exemplified in (5c-e), co-referentiality between 'possessor' and AGENT is by no means necessary. These sentences present construals in which an AGENT performs an action which 'creates a copy' of someone else's action and the result thereof. Similar to the findings on the definite determiner the, we are faced with a scenario in which an action is carried out that aims at the reproduction (or the prevention of such, as in 5d)) of a particular result. Again, the semantics of the COC, depicting an intensive link between Agent-Action-Result, are perfectly compatible with such a construal. Besides, these findings invalidate a claim put forward by Horita (1996: 234), expressing the necessity of co-referentiality between subject and a possessive within the CO-NP, and the claim by Halliday (see Ch. 2, ftn. 38), who completely rules out the use of COs with possessives. In this section, I have considered the distribution of types of determiners accompanying COs, focusing in particular on the question whether COs can occur with definite determiners. As I have shown, the object phrase A EVENT (a) EVENT (b) Reification (c) Intrinsic relationship EVENT - AGENT <?page no="221"?> 209 allows for the same range of determiners 'normal' direct objects can occur with, including definite forms, although the distribution of the different types varies with the class of COC. Focusing on EV/ R 1 -COs as the core type of the construction, I have discussed their use with the definite article the and prenominal possessive as representatives of the class of definite determiners. I have pointed out arrangements which render their usage natural, thereby identifying their special semantic contribution to the meaning of the whole construction. The following section will pursue the question of modification, which many scholars have proposed as a necessary element of CO-NPs. 8.3 Modification patterns As has been noted before, many researchers concerned with COCs have put forward the claim that the CO-NP should include a modifying element, since otherwise the expression would be highly redundant. The frequent occurrence of utterances such as (7) a. But it's the sorta thing I'd expect to hear from older people who've been fighting a fight and aren't interested in anything I'm doing. (CD6: 1109) b. I am hoping this will die a death. (CH8: 1888) was questioned and their use described as marginal or unacceptable (Jespersen 1954 [1927]); Massam 1990; Kuno & Takami 2004). In what follows, I consider and evaluate the results I have obtained from my investigation of the BNC, discussing the different types of modification within CO-NPs and their distribution according to the different COC-types. Table 8.2 gives a general survey of my findings. Surprisingly, for all three subtypes a usage without a modified NP was found in more than 30 percent of the respective instances. Partly, this high percentage can be explained with the relatively high frequency of unmodified COs which are specified by a possessive pronoun (cf. ft. 10). As has been discussed in the preceding section, these determiner constructions in a way enrich the conceptual content of their head (unlike the determiners a and the) in that they establish a relation between the (in the majority of cases) acting participant, a particular manner/ form of performing his/ her/ their action and the outcome of such a particular performance. In these cases, the unmodified phrases cannot be described as mere repetitions of the meaning of the verb. Moreover, the combinations tell a tale and sing a song, which make up a great amount of those unmodified forms, <?page no="222"?> 210 1 2 3 4 5 6 EV/ R 1 774 (35.5) 945 (43.2) 99 (4.5) 195 (8.9) 55 (2.5) 120 (5.5) R 2 88 (30.3) 116 (40.0) 47 (16.2) 15 (5.2) 9 (3.1) 14 (6.2) A 212 (32.8) 146 (22.6) 51 (7.9) 117 (18.1) 60 (9.3) 60 (8.9) Total 1074 (34.4) 1207 (38.6) 197 (6.3) 327 (10.5) 124 (3.9) 194 (6.2) Types of modification 1 = no modification dream a dream 2 = Adjectival Phrase think deep thoughts 3 = compound tell fairy tales 4 = Prep. Phrase breathe the breath of life 5 = Relative Clause live a life that pleases him 6 = Combinations of 2-5, eg: Adj. + Comp. sing a very famous Air Force song Adj. +PP die the early death of romance Adj. + Comp. + PP build an elegant fourteenth century building of black and white marble Comp + PP sing skipping songs about soggy semolina Adj. + RC smile that predatory smile she remembered so well Table 8.2 The distribution of different types of modification with COs involve verbs which allow for a small range of other complements (e.g. tell a story, sing an anthem). In tell a tale or sing a song, the CO names only one of the possible results (tell: a story, a lie, a joke; sing: a ballad, an aria, a solo) and thereby specifies the action instead of giving obvious information. This observation also applies to most, if not all, instantiations of the R 2 and A types, since all of the verbs in these classes can be complemented with several NPs other than COs. Additionally, one has to take into consideration whether the unmodified CO-phrase comes along with a definite or an indefinite determiner. As spelt out above, the definite determiner construes an entity as being uniquely identifiable and singles out one specific object. This uniqueness brings in itself some kind of 'indirect characterization', in that it evokes characteristics and specifications that are unique to the designated entity. In such examples, too, the information conveyed by the CO phrase is anything but redundant. Table 8.3 gives a survey of the distribution of the most frequent types of unmodified COs in correlation to the type of determiner they occur with. <?page no="223"?> 211 COC type unmodified + definite unmodified + indefinite unmodified + possessive EV/ R 1 tell the tale sing the song live the life die the death dream the dream 81 36 14 4 3 sing (a) song(s) tell (a) tale(s) die (a) death(s) dream( a) dreams think a thought 82 61 10 5 2 life x's life sing x's song tell x's tale dream x's dream think x's thought 168 34 21 10 10 R 2 produce the product build the building weave the web 13 11 2 build (a) building(s) produce (a)product(s) draw (a) drawing(s) 14 8 3 build x's 2 building A sow the seeds see the sight drink the drink 37 6 3 name (a) name(s) give (a) gift(s) sow (a) seed(s) 28 23 21 drink x's drink sow x's seed smell x's smell 6 5 3 Table 8.3 Most frequent unmodified forms In a number of instances, it seems that 'structural parallelism' might serve as the primary explanation for the use of an unmodified CO: (8) a. I'll be sitting at the back and I don't want anybody showing up the firm, asking questions, telling tales. (AEB: 2106) b. They waved their banners, sang songs, made speeches and shouted slogans in unison, mostly in Vietnamese. (AE8: 316) c. Your sons and your daughters shall prophesy, Your old men shall dream dreams, And your young men shall see visions. (G3A: 350) d. He dubbed them cheats, and told them they were swindling their audiences out what he called the last great procession, when, after the wedding songs had been sung and the dances danced, the characters took their melancholy way off into darkness […]. (CRE: 12) In all of the above examples the COC is surrounded by other verb phrases that take a direct, mostly unmodified object. Here, the construction could best be described as a means to preserve the 'rhythmic, structural flow' of the sentence and can be considered a stylistic means. According to the descriptions found in the majority of studies on COCs, most 'troublesome' and in need of explanation would then be those forms in which we find an unmodified COC with an indefinite determiner. Since this type of determiner singles out an arbitrary instance of a category without establishing unique reference, the notion of a more particular characterization is completely absent, and we are apparently faced with what Sweet (1891: 91, see Ch. 2, p. 9) rendered as a mere repetition of the meaning of the verb: <?page no="224"?> 212 (9) a. We do not think a thought in the same sense in which we think Herodotus is unreliable. (A7C: 321) b. […] she had been sleeping badly, dreaming dreams, longing all day for Tristram to come and take her in his arms and love her, when instead she had had to be content with no more […]. (EWH: 1746) c. If you tell him to jump a jump twice, he does it […]. (AR5: 1536) However, with the approach to COCs as proposed in this study, it is exactly these cases which can be accounted for fairly easily if the construction is considered as meaningful in itself. The sentences in (9) can then be described as instantiating the COC in its 'purest' form, depicting an energetic chain between AGENT , ACTION and RESULT , thereby construing the AGENT as a 'creator' and presenting an intensified version of an event of an otherwise non-interactional nature. For example, in (9a) the process of thinking is given a certain 'direction', and not construed as a mere mental state. In fact, think as a verb of cognition is categorized in several reference grammars (Quirk et al. 1985, Downing & Locke 1992) as a stative, non-dynamic verb. As used in (9a), it clearly assumes a more dynamic meaning. Similarly, in (9b), instead of presenting the sleeping person as simply being exposed to a mental experience, the sentence depicts her as being involved in a goaldirected process, resulting in the creation of trains of fancies or vision. Moreover, carrying the aspectual features I have explicated earlier, the COC imposes a temporal frame on the setting which results in a bounded representation of the event, with the event unfolding within those boundaries. Thus, even without modification, the CO clearly contributes meaning as an essential part of the construction. In a number of cases, an unmodified CO-NP is recorded with the verb die, which even native speakers that were presented with such instances judged as "odd" or "never heard of ": (10) a. Last night we looked back at the history of the MG, which appeared to have died a death when the factory at Abingdon closed in 1981. (K28: 629) b. THE WAR movie died a death during the Gulf War. (CAD: 2023) According to Visser (1963; see Ch. 6), the combination of die with a CO originated as a form expressing the cause or manner of deceasing, which makes it highly plausible to assume a modified CO-NP (with the modifying element expressing cause or manner) as the only acceptable form with die. What then, however, explains the use of such expressions as given in (10)? The phrase die a death seems to have become established in English (in particular in British English) as an idiom with the sense of 'to fail and end' (cf. Cambridge International Dictionary of Idioms). The objectless verb form die can as well be used to express a sense of failure; the COC, however, which <?page no="225"?> 213 in the case of die has been described as expressing an intensified/ ultimate effect of the event on the involved entity, is a very suitable construction to emphasize the notion of complete, ultimate failure. Considering the overall distribution of the different types of modifiers shown in Table 8.3, there are only three values which clearly stand out, i.e. cases where there is a high discrepancy between observed frequency and the expected even distribution: First of all, it is striking that COs of the Atype occur twice as often with a PP-postmodification as the EV/ R 1 - type and even three times as often as the R 2 type. This 'deviation' can be traced back to the frequent usage of the fixed phrase sow a/ the seed(s) of, recorded 70 times. Secondly, of all three CO-types, the R 2 -COs occurs most frequently in compound forms, which is due to the fact that the most often used nouns of this category, building and product, refer to rather general conceptual categories, subtypes of which are expressed in the form of compounds (e.g. estate-, school-, stone buildings; haircare-, software-, farm products). Thirdly, affected COs are used only half as often with adjectival modification as each of the other two CO-types. For this last observation I cannot offer a plausible explanation. As reflected in Table 8.3 (column 6), CO-NPs frequently come along with highly complex forms of modification, in which several types of modification are combined: (11) a. Cunningham smiled a broad slow smile of undisguised provocation. (H8T: 1139) b. Chatterton died the romantic and traditional early death of the divided, the invaded man, while Eliot did not. (A05: 574) c. Jenny grinned that wide, slightly toothy grin which she tried so hard to avoid and which filled her whole face with an animation and glow that turned Antony's heart upside down. (GUD: 2015) As was pointed out before, such richness and complexity of modification would be impossible to express by means of a simple verb + adverbial variant or could be verbalized in rather clumsy way only. The forms with such heavily loaded NPs point at the COC as a means for an 'elegant' arrangement of information, which partly supports those accounts on COCs which describe the construction as a 'dummy' construction for the expression of information which lack an adverbial counterpart (Jespersen 1954[1927]; Huddleston & Pullum 2002). 8.4 Summary and conclusion The observations made in the discussion of the types of determiners and types of modification that are most frequent with the three different types <?page no="226"?> 214 of COCs and the 'circumstances' of their usage are part of the information found in the constructions of intermediate levels of the suggested constructional network of COCs. Specification of such kind (i.e. particular determiners, modification patterns) is integrated, however, into still rather flexible schemas of representation. Chapter 11 will investigate the fully lexicalized phrasal patterns representing the lexical level of the constructional network, some of which have already been alluded to in the preceding analyses (e.g. do the deed, tell a tale, sow the seeds of). Before a discussion of such low-level schemas is taken up, however, Chapters 9 and 10 will be concerned with contrasting the COC with apparently similar constructions, which, after having discussed adjectival modification of the CO-NP in the prior section, I consider natural and necessary at this point of the argument: This is a) a comparison of COCs involving an adjectively modified CO-NP with verbal constructions modified by adverbs which are semantically and morphologically related to the respective adjectives, as in the sentence pair He laughed a harsh laugh. / He laughed harshly, and b) a comparison between COCs and so-called light verb construction, as in the pair He laughed a harsh laugh / He gave a harsh laugh. The main reason for this excursus into constructional alternations lies in the common practice of grammarians to consider these variants as synonymous, a claim which is to be tested and - if necessary - falsified in the following. Moreover, the discussion will focus on links within the network which were suggested to hold between mid-level schemas of special types of COCs and constructions lying outside the actual COC domain (see Ch. 5.6) so that the results gained from this analysis might add further details to the structural design of the network at this level. <?page no="227"?> 215 9 COCs vs. VACs: Putting alternations to the test 9.1 Preliminary considerations In Chapter 5.5 one form of collostructional analysis (i.e. collexeme analysis) was utilized in order to determine significant relations between verbs and constructional subtypes of the COC. The results obtained added to the depiction of a constructional network of more and less schematic levels of this construction and the description of links holding between these schemas. In this chapter I will focus on a non-COC, which, however, is often described as an alternative or equivalent construction in the relevant literature, and which I have assumed to be part of the network of COCs as a possible paraphrase structure (see p. 139): This is the simple verb + adverb construction (VAC), and I will discuss in what way it may be seen as similar to mid-level schemas of COC-subtypes. The discussion addresses the relationship that holds between modified EV/ R 1 -COCs (e.g. He laughed a harsh laugh) and the 'simple verb + adverbial' construction (VAC) (e.g. He laughed harshly). As argued before (see p. 139f.)., links between semantically and/ or functionally similar constructions are a fundamental part of a speaker's linguistic knowledge and therefore constitute an essential element in a usage-based constructional network. The VAC has already been alluded to in foregoing chapters, in which it was claimed that the use of either of the available forms is motivated by a particular construal a speaker has of or wants to impose on a situation. Because of their evoking different construals, the two constructions cannot be treated as synonymous, but instead should be discussed as paraphrases or semantically similar constructions which differ with respect to form and, of course, more subtle shades of meaning (possibly, in their pragmatic functions). Lambrecht (1994: 6) takes up the term 'allosentence', introduced by Daneš, referring to "semantically equivalent but formally and pragmatically divergent sentence pairs such as active vs. passive, canonical vs. topicalized, canonical vs. clefted or dislocated" etc. Most recently, Capelle (2006) has proposed the term 'allostruction', a notion which is to cover formal variation at all levels of grammatical description (i.e. morphemes, phrases, sentences and, as argued by the author, possibly even phonemes). Capelle explicitly suggests the label 'allostruction' with respect to the placement variability of verbal particles (put on the clothes / put the clothes on), arguing that the two forms are allostructions of a more schematic verb- <?page no="228"?> 216 particle construction which is unspecified for word order. However, whereas the alternating forms discussed by the author do serve different functions, they do not verbalize strikingly different construals of the same basic scenario, unlike the COC - VAC alternation. This is one of the reasons why I will not discuss COCs and VACs as allostructions of a more abstract constructional schema. Furthermore, as already mentioned, Capelle defines allostructions as formal variants 1 of a more abstract, formally underspecified schema. An application of this conception to the COC-VAC alternation would imply that every VAC could be realized by a COC and vice versa, which - as has been shown and will be shown in the course of the book is not plausible at all due to the formal and semantic restrictions that hold for COCs, in particular for EV/ R 1 -COCs. When I speak of alternations here, I use the term in a different way than it is done in studies on equivalent syntactic structures within the generative/ transformational grammar framework. In such approaches, one of the two (or more) available patterns is seen as more basic; by means of a transformational rule which is applied to this basic form, the other (less basic) pattern is derived. In this framework, passive sentences, for example, are derived from active ones, continuous verb-particle combinations (take off the label) serve as the underlying structure for discontinuous patterns (take the label off), and the VAC is described to serve as 'input' for the COC (Jones 1988, see Ch. 2, p. 22f.). In my approach to such an alternation, I consider the two patterns COC and VAC as 'independent' constructions with their own semantic, syntactic and pragmatic properties, which are, however, linked by a similarity relation (as described in section 5.6). As was argued in Chapter 5.6, these constructional relations and links between schemas are an essential part of a speaker's linguistic knowledge, with speakers being able to consciously reflect on the similarity between constructions. Fig. 9.1 depicts that particular extract of the constructional network proposed here which is of interest in the present discussion (cf. Fig 5.3, p. 141); the schemas and links which are to be focused on in the analysis are highlighted in bold print. As Stefanowitsch (2006: 258) emphasized, the linguistic knowledge speakers have acquired on available, possibly similar constructions also includes information "as to which words best distinguish between the constructions in question" (emphasis in original). The evidence he can provide 1 The 'primacy' seemingly given to form in Capelle's concept of allostructions allows him to discuss allomorphs and allophones under the same heading, since in cases of allomorphy and allophony no semantic variation is signalled. Or, put differently, if formal variation is inherently associated with semantic / functional variation, which is a primary tenet of CL, allomorphs and allophones could not be considered as allostrcutions. <?page no="229"?> 217 for this claim is to defend the psychological interpretation of results yielded by collostructional analysis. Following from this, the lexemes which are distinctive for the two 'alternative' constructions will be a major subject in their discussion, which is based on the comparison between particular COCs and VACs. Lexemes which are highly characteristic of either of two available patterns can be identified by the use of another method of collostructional analysis, yielding data which admit predictions about the usage of the one or the other alternative construction: Gries & Stefanowitsch (2004a) propose the so-called distinctive collexeme analysis for the description of semantic differences between alternating constructions. The authors de- Intransitive construction Monotransitive construction COC-EV/ R 1 [NP S V CO-NP DO/ E ] A RGUMENT S TRUCTURE C ONSTRUCTIONS … … COC-EV/ R 1 / modified CO [NP S V Det Adj Cog. N DO/ E ] NP s LAUGH Det Adj Cog. N DO/ E Intr. + Adverb [NP S V AdvP] NP S LAUGH AdvP … elaboration link bidirectional similarity link NP S LAUGH HARSHLY NP S LAUGH A HARSH LAUGH Fig. 9.1 Similarity links within the constructional network <?page no="230"?> 218 scribe it as a method which "can reveal subtle differences between seemingly synonymous constructions" (ibid.: 97) in that it singles out lexemes which are strongly preferred by either of the alternating forms. By examining the semantics of the lexemes that are preferred by the one or the other construction, inferences can be drawn about aspects of meaning of the whole construction. Applying this technique, Gries & Stefanofanowitsch could gain valuable insights into differences between, for example, the ditransitive and to-dative constructions, or the active and the passive constructions. Similar to the collexeme analysis, the index of distinctiveness of a collexeme is calculated on the basis of the Fisher-Yates exact test, which in turn builds on four types of 'input frequency', which are the lemma frequency of the collexeme in construction A, the lemma frequency of the collexeme in construction B, and the frequencies of construction A and construction B with words other than the collexeme in question. (ibid.: 101) The values resulting from the computation, which are used to measure the distinctiveness of a lexeme, are again given in form of -log(p-value Fisher-Exact 10). My application of the described method to the COC and its supposedly alternating form aims at filtering out lexemes that distinguish between the use of selected verbs in the COC and the VAC. In order to detect lexemes which might be characteristic of, i.e. strongly preferred by the COC or the VAC, as in alternations of the type (1) a. He smiled a lopsided smile. b. He smiled lopsidedly., I decided to focus on the adjective/ adverb pair as possible distinctive collexemes. That is, interest is given to the question whether the most distinctive adjectives (or adverbs, respectively) of either form might give clues about slight semantic nuances of the alternating constructions. Moreover, this procedure might reveal candidates for entrenched patterns, i.e. enable us to identify particular adjectives/ adverbs that 'drive' the choice of one of the available patterns. My decision to consider the adjective/ adverb pair (e.g. lopsided/ lopsidedly) as distinctive collexemes makes it necessary to treat the forms of such a pair as lemmas of the same lexeme (e.g. LOPSIDED ). The alternating constructions can thus be represented in the following way: (2) a. COC: V + X adj -CO b. VAC: V + X adverb For the collection of the required data, the BNC instances of COCs were checked for verbs which frequently occur with an adjective-modified CO and which, in turn, allow for a paraphrase by the alternating construction <?page no="231"?> 219 V + X adv (thus, they must belong to the EV/ R 1 -type in any case). The verbs live, smile and die were found to be promising candidates since my corpus research yielded an appropriate number of COCs containing these verbs in combination with an adjectively modified CO. In order to obtain frequencies for the respective VAC variants, a collocational search for these verbs was run with SARA, by means of which all adverbs which are used with the verbs could be identified. I designed a search syntax which allowed me to extract all clauses from the BNC in which the respective verbs are either followed or preceded by the adverbial forms of the collocation list. The extracted sentences were manually checked and edited to exclude instances with 'coincidental' collocations of verb and adverb. The obtained frequencies were submitted to the Fisher-Yates exact test, run on the program Coll. analysis 3. A program for R for Windows 2.x (Gries 2004). As an example, the frequencies which are necessary for the alternating pair smile a slow smile / smile slowly are shown in Table 9.1. The frequencies in bold print served as input for the software tool Coll. analysis 3. SLOW - SLOW Row totals smile X adj smile 16 159 234 smile X adv 22 2,042 2,064 Column totals 38 2,201 2,239 Table 9.1 Frequency information necessary for distinctive collexeme analysis The figures in the table must be understood as follows: Of the 234 COCs involving the verb smile, 16 instances occur with the premodified CO-NP slow smile, compared to 22 instances of smile slowly of a total of 2,064 VACs containing smile. These raw frequency figures seem to be indicative of the preferred usage of smile a slow smile over smile slowly, an assumption which will be tested and spelt out below. Note that for the calculation of distinctive associations between lexemes and alternating constructional varieties, only those lexemes were considered that can theoretically occur in both constructions. Thus, for example, the relatively frequent collocations of the smile-COC, smile a small smile, smile a little smile or smile one's best smile, could not be integrated into a discussion of alternating pairs since they lack equivalent forms in the VAC (*smile smally, smile little 2 , ? smile best). Similarly, in cases of alternations of the verb live, for the COC live a married life the alternative form live marriedly is unacceptable or highly marked 3 , and live wild cannot count as a genuine paraphrase of live a wild life. 2 Smile a little smile is by no means equivalent to smile little. 3 Although examples of live marriedly can be found in the WWW, native speakers of English (p.c.) consider this phrase unacceptable. <?page no="232"?> 220 The results I gained for the respective verbs are discussed in the following. Note that in the discussion of the data I base my descriptions and evaluation on a significance level of p < 0.01 (i.e. distinctiveness index > 2) instead of p < 0.001 due to the small amount of lexemes available and suitable for discussion. Lexemes, whose value is only slightly below the distinctiveness index of 2 (i.e. 1.9) will be included in the discussion and interpretation of distinctive collexemes. 9.2 Live happily or live a happy life? Live, the verb most frequently found in a COC in the BNC (699 entries), is complemented by an adjective-modified CO in 284 instances, involving 123 different adjectives. Of the more than 23,500 clauses containing live in the 'intransitive verb + adverbial' construction as contained in the BNC, about 1,200 see the verb modified by an adverb (i.e. these uses represent instantiations of the VCA). In the majority of cases, live is modified by an adverbial PP, most often expressing spatial ( 12,500 examples), circumstantial (5,000 examples) and temporal ( 2,000 examples) information. Apparently, one of the major functions of this constructional type is the expression of 'biographical' information 4 . Besides these adverbial constructions, in a few cases live is also complemented by NPs other than COs, yielding phrases such as live a nightmare, live a dream etc. Only 40 of the modifying lexemes are found in both constructions (e.g. HAPPY , DANGEROUS , COMFORTABLE ), the majority of which, however, occurs more frequently in the VAC (e.g. INDEPENDENT 14 : 37; QUIET 14 : 30; SEPA- RATE 8 : 15; SIMPLE 5 : 9), if judged on the basis of raw frequency. This is, of course, due to a much higher overall occurrence of the VAC in the BNC, a factor which is accounted for by the distinctive collexemes analysis: Since the analysis incorporates the absolute frequencies of the two constructions and the frequencies of the two constructions with words other than the collexemes investigated, it can detect asymmetries in the relative frequencies of co-occuring lexemes. Therefore, the question whether INDEPENDENT or QUIET are in fact lexemes preferred by the VAC can only be answered after the application of the distinctive collexemes analysis. The below sentences illustrate the usage of live with selected modifiers in the constructions under investigation: (3) a. While the vast majority of older people are able to live independently, significant minorities experience considerable difficulties. (AP5: 436) 4 Note that in no instance a COC was used to verbalize such biographical facts, i.e. ? He lived his life there in 1967. <?page no="233"?> 221 b. However, there was also an increased tendency for the elderly to live independent lives. (FP4: 574) (4) a. But they live simply and have even stopped collecting their nineteenth-century vases. (CD6: 1314) b. The smaller resident landowners lived a simple life and were, like the Basque nobility, interested in the reform movement of the late eighteenth century and supplied many of its leaders. (FB7: 73) (5) a. He worked as an engineer with a construction company and lived quietly with girlfriend Lorna Burns in Eltham, south London. (CEN: 6462) b. Back in England I found a house and began to live a quiet life. (FRX: 565) The collexemes which were measured to be most distinctive of each construction are listed in Table 9.2. As can be read off from the figures given, only few of these items are distinctive for the respective alternating patterns to a significant extent (i.e. their distinctiveness index is > 2). Whereas normal, full and private are the most distinctive collexemes for the live + CO variant, alone, longer, happily, locally and long are associated most strongly with the VAC. Judging by these results, it can be assumed that these patterns are candidates for entrenched schemas of their preferred constructions 5 . As described above, a second step for the recognition of possible meaning differences between the constructions at issue involves the grouping of the distinctive collexemes of each pattern on the basis of their semantics. An analysis of the semantics of the distinctive collexemes of each constructional variant as listed in the table brings to light a highly diverse semantic character of the items so that their grouping seems next to impossible. However, the semantics of these lexemes are fully compatible with the description of the COC suggested in this study so far. Keeping in mind that the COC has been described as a construction that presents an AGENT as more intensely involved in an action and more responsible or in control of the outcome of the action, one would expect adjectives which describequalities whose degree can be influenced by the agent. The adjectives normal, full, healthy, solitary and decent meet this criterion to some extent. This idea might also serve as an explanation why long and longer are pref erably used in the VAC, as the length of someone's life can be controlled to 5 Note again that the degree of entrenchment of a pattern cannot be measured by statistical tests and that the values used here only serve as an orientation for the postulation of possibly entrenched schemas. <?page no="234"?> 222 COC VAC collexeme COC VAC IoD 6 collexeme COC VAC IoD NORMAL 34 1 20.41 ALONE 7 6 350 38.49 FULL 14 5 5.31 LONGER 3 141 13.52 PRIVATE 5 0 3.14 HAPPY 6 117 8.22 DIFFERENT 6 2 2.51 LOCAL 0 40 4.75 EVIL 4 0 2.51 LONG 8 90 4.41 ORDINARY 4 0 2.51 COMFORTABLE 3 35 1.96 HEALTHY 5 1 2.45 DANGEROUS 1 21 1.69 SOLITARY 5 1 2.45 CHEAP 0 11 1.29 DECENT 4 1 1.90 ROUGH 0 8 0.93 EXCITING 3 0 1.88 HARMONIOUS 0 5 0.58 MISERABLE 3 0 1.88 MAGICAL 0 5 0.58 RELIGIOUS 3 0 1.88 COMMUNAL 0 4 0.46 AUTHENTIC 2 0 1.25 SUCCESSFUL 0 4 0.46 CONTEN- TED 2 0 1.25 VICARIOUS 0 4 0.46 NOMADIC 2 0 1.25 ETERNAL 0 3 0.35 Table 9.2 Most distinctive collexemes of the live+ X adj life / live X adv alternation a very limited extent only. Yet, happy and comfortable, which also designate characteristics whose extent might be consciously and intentionally influenced, show up as more distinctive in the VAC. Of course, the frequent and strong collocation live happily can be explained on the basis of the proverbial fairy tale ending They lived happily ever after a reason which does not account for live comfortably. In summary, it seems that in the case of live, only few items are clearly distinctive for the construction in which they are preferably used, that is, either the COC or the VAC. The semantics of the lexemes favoured by the COC conform to the more general meaning components of the construction, while some of those lexemes that preferably occur in the VCA are simply less compatible with the semantics of the COC. However, the obtained data do not suffice to establish more subtle semantic specifications of the constructional variants. Therefore, I argue that the choice of one of the alternating constructions with particular adjectives/ adverbs is either a matter of habit (which is most likely the case for those lexemes that have been measured distinctive to a highly significant extent) or is determined 6 For reasons of space, the figures in the column "IdO." ('index of distinctiveness') in Tables 9.2.- 9.4 are rounded. See Appendix for exact figures. 7 The lexeme ALONE is considered to represent the lemmas alone and lonely in this analysis. <?page no="235"?> 223 by the specific construal a speaker chooses for an event. The latter holds true in particular for verbalizations which involve collexemes frequently occurring in both constructions and whose index of distinctiveness indicates only a slight preference for one of the constructions, so-called 'freely alternating' collexemes ( see Gries & Stefanowitsch 2004a). Although in this case study, the distinctive collexeme analysis could yield only vague results for the attempt to more subtly specify the semantics of the two similar constructions, the advantages of statistical methods such as collostructional analysis have become apparent once more: Returning to the lexemes INDEPENDENT , QUIET , SEPARATE and SIMPLE (discussed at the beginning of this section) which are used in both constructions but - based on raw, absolute frequency - occur much more frequently in the VCA 8 , we note that they could not be confirmed as preferred lexemes of the construction by the distinctive collexeme analysis. All of them were calculated to be more strongly attracted by the COC, even though to an insignificant extent only. Therefore, they can be classified as freely alternating lexemes, a status which could not necessarily have been read off from raw frequency data. 9.3 Die naturally or die a natural death? The BNC lists 74 examples of die with an adjective-modified CO (39 types). The verb occurs 12,000 times with adverbial modification, with 820 of them being of the VAC type. As in the case of live, in the majority of cases the adverbials modifying the verb die consist of a PP expressing temporal ( 3,700 examples), spatial ( 1,700), causal ( 1,000) and other circumstantial ( 600) information. Only 15 of the lexemes appearing in the slot that we are interested in are used in each of the related constructions (e.g. NATURAL , NOBLE , VIOLENT , TERRIBLE ), as selectively exemplified in the following sentences: (6) a. It was just possible, Sandison thought, that Maidstone had died naturally if choking to death in your sleep after a huge binge could be called natural […]. (ASN: 2416) 8 A comparison of the relative frequencies (arrived at by dividing the number of occurrences of the lexeme in the construction by the total number of instances of the construction) of the respective collexemes does indicate a preference of INDEPENDENT , QUIET , SEPARATE and SIMPLE by the COC. However, relative frequency does not fully detect asymmetries in the distribution of lexemes in alternating constructions as it does not allow for direct conclusions about the statistical significance of the calculated values. <?page no="236"?> 224 b. It occurred to him that Newley might have died a perfectly natural death a heart attack, perhaps, brought on by the blackmail. (GUU: 1155) (7) a. Also, several friends have died violently in drug battles. (CH1: 9284) b. What amuses me is that many of those who claimed I would hang, died violent deaths themselves in some pot-holed alleyway, bleak battlefield or gory execution yard. (HH5: 157) (8) a. [S]ince they had nothing left but their lives and their bodies, they chose to die nobly rather than to betray and abandon their king'.(HPT: 1455) b. Alas, he was lying ill and semi-paralysed and not expected to recover, though that great man, whose reputation today has regrettably fallen, survived to die a more noble death in the Occupied Paris of 1941. (H9X: 788) Although it seems that the lexemes NATURAL (16: 9) and VIOLENT (5: 6) freely alternate between the COC and the VAC, if judged on the basis of their raw frequency, the distinctive collexeme analysis shows them to be (highly) distinctive for the COC (see Table 9.3). However, considering the high value of the index of distinctiveness, only die a natural death might be considered a candidate for a possibly entrenched pattern. In Chapter 6 I pointed out that the COC die a death underwent a different historical development than the typical COC forms of the EV/ R 1 type, and therefore I postulated a slightly different meaning for a COC involving die: Here, the AGENT is not depicted as being responsible for the outcome of the action, but is construed as an EXPERIENCER being more intensely involved in / affected by the event (which is also part of the meaning of 'normal' COCs). Do the adjectives that are preferably used in the COC variant of die elaborate this rather general meaning? If we look at these collexemes, listed in Table 9.3, they mainly fall into two categories: While NATU- RAL , UNNATURAL and VIOLENT describe causal circumstances of dying in this context, UGLY , HORRIBLE and TRADITIONAL name a particular manner of passing away, as do the less significantly distinctive or even insignificantly preferred items ROMANTIC , CRUEL , IGNOBLE and NASTY 9 . The majority ofthese manner adjectives have a rather emotional connotation, and thereby contribute to the intensifying function of the COC. Moreover, the two classes of adjectives correspond to the semantics of the original form of the 9 Note that with a significance level of p < 0.05, ROMANTIC would still be considered as significantly distinctive (index of distinctiveness > 1.30103 = p < 0.05). <?page no="237"?> 225 COC VAC collexeme COC VAC IoD collexeme COC VAC IoD NATURAL 16 9 11.08 SUDDEN 1 146 3.79 HORRIBLE 4 1 3.46 QUICK 0 38 1.67 VIOLENT 5 6 2.71 TRAGIC 0 24 1.04 TRADITIONAL 2 0 2.05 ALONE 0 23 1.00 UGLY 2 0 2.05 UNEXPECTED 0 17 0.73 UNNATURAL 2 0 2.05 IMMEDIATE 0 14 0.60 ROMANTIC 2 1 1.60 PREMATURE 1 21 0.43 CASUAL 1 0 1.02 NEEDLESS 0 10 0.43 CRUEL 1 0 1.02 EARLY 2 31 0.41 IGNOBLE 1 0 1.02 EASY 0 6 0.25 NASTY 1 0 1.02 MYSTERIOUS 0 6 0.25 POLITICAL 1 0 1.02 PEACEFUL 0 6 0.25 PREDICTABLE 1 0 1.02 COMPLETE 0 5 0.21 STRANGE 1 0 1.02 FAST 0 4 0.17 UNHEROIC 1 0 1.02 UNNECCESSARY 0 4 0.17 UNJUST 1 0 1.02 RAPID 0 3 0.12 Table 9.3 Most distinctive collexemes of the die + X adj death / die X adv alternation phrase die a …death, which evolved from 'verb + PP-adverbial' construction in which the PP was described as expressing the cause or manner of death (e.g. with deth thou shalt die [quoted in Visser (1963: 415)], see Ch.6, p. 160).The ostensible preference of the COC variant for manner adjectives is also in accordance with the claim by Iwasaki (2007) that die a … death can only indicate the manner of death, but not the resultant state of the event (see, however, the discussion in Ch.6, p. 161). Yet, concerning the rather small number of lexical items which also occur only once in the data and which, moreover, are statistically insignificantly preferred only by the construction, it is only tentative to draw the general conclusion that the COC attracts the described types of adjectives. The observed regularities rather point at a tendency of preferred usage; the question whether the COC variant of die profiles a stronger involvement of the affected subject in the process can be answered neither satisfyingly nor sufficiently on the basis of the available set of data. For the VAC, only the adverbial form of SUDDEN is singled out as a highly significant distinctive collexeme, the adjective form being used only once in a CO-NP. This result stands in conformity with aspectual characteristics of the COC. As was explicated in Chapter 7, the COC lends a frame to the setting in which the described event unfolds within a certain temporal boundary, whose extensions - in turn are determined by the nature of the <?page no="238"?> 226 CO-NP. The semantics of SUDDEN go contrary to the setting up of such a frame that supports the unfurling of an event in time and thus makes the adjective rather unsuitable for usage in the CO-NP. Viewed from such a perspective, the strong preference of SUDDEN in the VCA is highly plausible. The same explanation can account for the frequent occurrence of die quickly (which does not have a COC variant in the BNC at all), although the distinctiveness index ranks it insignificant. If we go down the column of collexemes being distinctive for the VCA, we see that semantically similar adverbs such as immediately, fast and rapidly complete the list of such temporal adverbs (even though they were not measured significantly distinctive, the BNC represents them as exclusively used in the VCA). Note that none of the adverbs denotes causal circumstances, and only peacefully names the manner of dying. Prematurely, unexpectedly, needlessly and unnecessarily might be grouped as adverbs which express a subjective evaluation from the speaker's point of view rather than the 'victim's' way of dying. This group, however, cannot exclusively be ascribed to the VCA, as the semantically similar forms predictable and unjust occur in the COC. Once again, the data can be merely interpreted as tendencies instead of considering them as entrenched patterns. Summarizing the above observations on collexemes which are distinctive for COC and VAC variants involving the verb die, we can record the following tendencies: While on the one hand, the COC seems more strongly associated with lexemes expressing cause and manner of dying (conforming to the historical development of the phrase), the VAC apparently attracts evaluative and temporal adverbs, with the latter seemingly incompatible with aspectual characteristics of the COC. The distribution of the long list of lexemes which occur only once in either of the alternating constructions, however, must be left uncommented; they simply are not typical for either of the variants. 9.4 Smile slowly or smile a slow smile? The final case study will be concerned with a discussion of those lexemes which might reveal differences between the use of the verb smile in a COC and in a VAC. The CO-NP smile occurs with adjectival modification 234 times 10 in the corpus, 122 different adjectives (types) being used. 2064 instances of the verb smile in a VCA could be detected, involving 352 different adverbs (types). Of this great variety of adjectival and adverbial modifiers, 75 lexemes are found in both constructions (e.g. THIN , WRY , WAN , 10 This figure exceeds the overall frequency of smile-COCs since I counted each adjective. Thus, smile a long, thin smile, for example, was counted as two instances. Adverbial forms were dealt with in the same way. <?page no="239"?> 227 GRIM ). Some of these appear frequently in both the COC and the VAC and are not distinctive of either construction (e.g. ENIGMATIC 3: 10, BITTER 4: 11, CROOKED 4: 11, CHARMING 8: 3) their use in the one or the other construction depends on the speaker's specific construal of the event. An illustration of the semantically similar forms involving selected, seemingly freely alternating collexemes is given in (9) - (11): (9) a. 'Swipe him one,' a fellow marketeer suggested, but she smiled enigmatically, remembering Gabriel. (CA3: 349) b. He didn't say anything, he just smiled his quiet, enigmatic smile and returned to his paper. (G36: 1237) (10) a. Tolonen met Karr's eyes and smiled bitterly. (G04: 1126) b. He smiled a small bitter smile, and pushed the white folder aside. (C8S: 1354) (11) a. 'However-' he smiled charmingly - 'it is a risk I am prepared to run.' (H8N: 3132) b. 'We are surprised, Rain,' said Maurin. He smiled his most charming smile. (GV2: 3425) Other lexemes, however, are significantly distinctive for one of the alternating forms. SLOW and PREDATORY are distinctive of the COC to a highly significant extent, while RUEFUL , POLITE , SUDDEN and FAINT are of the same status for the VCA. What is a plausible explanation for these strong associations between construction and lexeme? Grouping the items used with the COC variant is an almost infeasible 'venture', as they do not follow any semantic pattern. What is more, many of the items preferably used in the COC have semantically related forms in the VAC (e.g. FRIENDLY , LOVELY > WARM , SWEET , POLITE ; INNER > INWARD ). A different picture evolves, however, when we consider the lexemes which were calculated as (more or less) distinctive for the VAC. FAINT , WEAK , SLIGHT , THIN and BRIEF 11 all ex press 'low intensity', and they seem rather infelicitous in the COC, which was said to emphasize the agent's involvement. Moreover, both SUDDEN and BRIEF describe a quality 'of little temporal extension', a meaning which appears to be rather incompatible with the aspectual characteristics of the COC as spelt out above. For the remaining items I owe an explanation other than chance. 11 Note, however, that although the analysis measures them as distinctive for the VAC, for most of them the degree of distinctiveness falls under the significance level used as a basis of discussion here (i.e. index of distinctiveness > 2 ; p < 0.01). <?page no="240"?> 228 COC VAC collexeme COC VAC IdO collexeme COC VAC IdO SLOW 16 22 6.63 RUEFUL 0 53 2.50 PREDATORY 4 0 3.97 POLITE 0 48 2.26 BRIGHT 3 0 2.98 SUDDEN 0 47 2.21 HARSH 3 0 2.98 FAINT 1 63 2.10 TWISTED 3 0 2.98 SWEET 1 58 1.89 LOPSIDED 5 5 2.76 BROAD 2 73 1.87 AMUSED 3 1 2.41 WEAK 0 38 1.78 FRIENDLY 2 0 1.98 SLIGHT 0 37 1.74 INNER 2 0 1.98 WARM 1 44 1.32 LOVELY 2 0 1.98 NERVOUS 0 24 1.12 MISHAPPEN 2 0 1.98 SAD 1 38 1.09 MOCK - STERN 2 0 1.98 BRIEF 0 23 1.07 DISARMING 2 1 1.53 ENCOURAGING 0 22 1.03 F LEETING 3 5 1.40 THIN 5 80 0.92 S ECRET 3 5 1.40 INWARD 0 19 0.88 Table 9.4 Most distinctive collexemes of the smile + X adj smile / smile + X adv alternation As in the cases of live and die, we cannot clearly identify a semantic tie holding between those adjectives that are listed as strongly preferred in the COC and thus cannot extract nuances of meaning which elaborate the more general semantics of the COC. Nevertheless, a logical explanation of why some items show a strong association with the VAC and are not or less preferably used in the COC could be arrived at: The semantics of these lexemes are more or less incompatible with the general meaning of the COC in the above case 'unfolding in time' and 'intensity' and thus tend to occur in the alternating construction which lacks these specifications. 9.5 Summary and discussion The application of the distinctive collexeme analysis to COC/ VAC alternations involving the verbs live, die and smile aimed at an identification of modifying lexemes which, firstly, are strongly preferred by either of the constructions and which, secondly, due to their semantic coherence, can give clues about finer nuances of meaning of the favoured construction. In several cases I could pinpoint lexical items which are significantly distinctive for one of the constructions and which - because of their strong association with the respective form - can be considered candidates for possibly entrenched forms. Moreover, in the majority of cases it could be shown <?page no="241"?> 229 that the semantics of the distinctive collexemes conform to the more general aspects of meaning of the respective construction, or it could at least be plausibly explained why particular lexemes prefer one construction and repel the other. For both die and smile, for example, I have demonstrated that the aspectual characteristics of the COC make the use of some temporal adjectives in the construction infelicitous, which results in a stronger association of these lexemes with the alternative form of a VAC. The attempt, however, to group all of the distinctive lexemes into clear semantic categories in order to deduce shades and flavours of the meaning of the alternating patterns remained less successful. As noted, for all three verbs there exist long lists of collexemes whose distinctiveness index falls below the significance level (i.e. their distribution might be due to chance), or which more or less freely alternate between the two constructions. Their use in either of the alternating forms depends on the deliberate choice of the speaker how to construe the perceived event by means of the available linguistic structures. That is, it is clearly the construal the construction imposes on a setting which motivates usage and not so much a 'habit of usage' of automatized forms. In addition, the numerous lexemes found in both forms justify the similarity link between COC and VCA I have chosen to include in my network of constructions in Fig. 8.3. (see also Fig. 9.1). The items which were identified as distinctive, on the other hand, support my choice of establishing a 'similarity' link rather than a 'synonymy' link. It should also be pointed out that although the study of distinctive collexemes addressed - among other things the question whether there exist possible entrenched patterns concerning modifying elements, the high number of adjectives and adverbs which occur only once in the respective construction (see Appendix) suggests a high degree of flexibility of the constructional patterns. It thereby justifies the postulation of flexible midlevel schemas, i.e. patterns offering slots which can be filled according to the speaker's expressive needs. Finally, it must be noted critically that a 'cost-benefit'-analysis of the methodology applied in this chapter would probably yield an unfavourable balance: Compared to the great efforts put in the extraction of data necessary for the application of the distinctive collexeme analysis, the results obtained for a more exact differentiation of the two seemingly similar constructions are rather meagre and unsatisfying. Although for my comparison I have chosen those instances of COC-VAC alternations which are recorded most frequently in the BNC (i.e. those containing the verbs live, die and smile), the available data might still be insufficient for an adequate comparative analysis. The comparatively low overall frequency of COCs in the BNC, even more so of COCs with adjectively modified NPs which have a VAC variant, is a primary factor which hampers a more successful ex- <?page no="242"?> 230 amination. The analysis is nevertheless included here as part of an extensive investigation of COCs since it solidly demonstrates advantages of statistical procedures which make possible observations which could not have been made by a simple, traditional analysis of collocational patterns based on raw frequency. Besides the VAC, COCs are frequently discussed in relation to another type of construction which is also claimed to equivalently paraphrase the verbal content expressed by COCs: so-called 'light verb constructions', such as have a romantic dance, give a sweet smile or take a long walk. Chapter 10 is to take up an investigation and comparison of these constructional patterns, remaining in the 'realm' of mid-level schemas and similarity links assumed to hold between such constructional templates. <?page no="243"?> 231 10 Don’t have a run, run a run! - LVCs and COCs compared 10.1 Introduction Whereas the preceding chapter focused on a contrast of COCs with their VAC 'counterparts', the following sections aim at a comparison of the construction (primarily of the type EV/ R 1 ) with the so-called light verb construction, because the two patterns have repeatedly been described as similar or related in some fashion (Fillmore 1968, Jespersen 1954 [1927], Macfarland 1995, Quirk et al. 1985). That is I will examine alternating forms of the following type, which, as a most prominent feature, have in common that the nominal following the verb designates a conceptually reified version of an event (i.e. a dream, a cry): (1) a. Marcy had a wonderful dream. b. Marcy dreamt a wonderful dream. (2) a. Kelly and I had a cry. b. Kelly and I gave a cry. c. Kelly and I cried a cry. The discussion of the two constructions is presented in this part of my investigation of COCs, since the 'alternation' mainly concerns expressions with a modified cognate noun. That is, a discussion of a possible relation between the two patterns is for the most part relevant for mid-level schemas of COCs. Although, as can been taken from the examples in (2), modification of the NP is by no means necessary for either of the construction to be acceptable, the occurrence of a modified NP in a COC seems to be the default case, as was observed in Chapter 8. While the discussion of constructional characteristics in this chapter is still driven by usage-based considerations, the ideas about and descriptions of the constructions, however, do not involve elaborate statistical (e.g. collostructional) analysis, as the elicitation of the necessary data would have been too extensive for the research purposes pursued here: An exhaustive analysis of LVCs and their comparison with COCs goes beyond the aim of the present investigation and could be the subject of a separate, comprehensive study. Therefore, the description provided in this chapter focuses on selected aspects of the form and meaning of the two constructions which I consider most relevant and interesting for their comparison. <?page no="244"?> 232 Noun LVC Frequency COC Frequency have 42 bite take 57 bite 2 breath take 701 breathe 4 cough give 19 cough 2 give 70 cry have 21 cry 2 do 32 dance have 21 dance 12 dream have 284 dream 48 have 684 drink take 102 drink 18 feeling have 1017 feel 26 fight have 93 fight 16 grin give 111 grin 18 do 12 jump make 36 jump 5 have 142 laugh give 141 laugh 10 do 3 make 7 prayer give 10 pray 7 do 15 have 25 run make 120 run 2 scream give 25 scream 2 sigh give 172 sigh 8 have 62 sleep take 7 sleep 14 smell have 2 smell 24 smile give 598 smile 203 talk have 282 talk 5 have 30 give (it) 1 taste take 1 taste 2 have 311 thought give (it) 583 think 81 do 21 have 60 take 117 walk make 9 walk 3 6064 490 Table 10.1 Frequencies of selected LVCs and COCs (as occurring in the BNC) <?page no="245"?> 233 An investigation of the BNC confirmed that a great amount of nominals found in COCs are also used in LVCs, many of them even occur with several light verbs. Table 10.1 gives an overview of selected nominals which are both found in COCs and LVCs, and their frequency of occurrence in the corpus: With the exception of drink and taste, all of the COC forms listed in the table belong to the EV/ R 1 class, where the CO can designate both the result of an event or the event itself. As is obvious from the data, LVCs are used much more frequently in naturally occurring language than COCs, a fact which may be indicative of a more restricted (marked) or elaborate nature of the latter variant: While the selected nouns occur 6064 times in a LVC, they are used in a COC only 490 times. In my comparison of the two constructions, I shall mainly concentrate on their semantic potentials, thereby neglecting the question of the syntactic status of the nominal element following the verb in the LVC (which, as will be explained below, used to be one major issue in studies of the construction). Before taking up the discussion of semantic differences between the constructional patterns, I will provide a short portrayal of the LVC, discussing general issues that have been of principal interest in studies of the construction, and summarize observations on the general relationship between COCs and LVCs. 10.2 A brief portrayal of LVCs 10.2.1 General observations Only recently has linguistic research granted light verbs some contribution of meaning to the verb phrase they head (see below). In earlier studies and general descriptions found in reference grammars of English, the label "light verb" (originally used by Jespersen 1954 [1927]) is done justice with scholars presenting the members of this verb class (i.e. do, give, have, take, make) as semantically empty, delexicalized or desemanticized (see Allan 1998; Grade ak-Erdelji 2004). This tendency is illustrated, for example, by the definition of LVCs found in Huddleston & Pullum (2002): "The main semantic content is located not in the light verb, but in the noun functioning as the head of the direct object" (ibid.: 290). As Cattell (1984: 2) notes, these verbs seem to mean "little more than that a verbal action occurred". The main contribution of LVCs is seen in their yielding a "significant increase in syntactic versatility" (Huddleston & Pullum 2002: 291), providing speakers with a linguistic structure in which the action expressed by the deverbal nominal phrase can be modified more elaborately. Quirk et al. (1985: 751) ascribe to the construction some 'greater impact' than to the simple verb alternant, suggesting that <?page no="246"?> 234 [t]he construction type provides greater weight than the corresponding SV type, especially if there are no optional adverbials, and is often preferred to the SV construction in informal language (emphasis added, SH). The latter claim on the actual usage of LVCs is corroborated by Dixon's (1991) findings, who examined about 700 of the most common English verbs and found that about 25% can occur in at least one of the constructions have, take or give a + deverbal noun. These forms are described by him as "carry[ing] an overtone of friendliness and intimacy" (ibid.: 337) and as occurring far more frequently in colloquial styles of English. This is a characteristic which clearly distinguishes LVCs from COCs, the latter having been claimed to convey a "rather orotund style" (Quirk et al. 1985: 750). Brugman (2001) adds functional qualities to the list: She observes that through the nominalized element in LVCs not only quantification and modification of the event are facilitated, but also discourse coherence and information processing, for LVCs "serve to disperse verbal ideas over several lexical units" (Hopper 1991, quoted in Brugman 2001: 556). She confirms Dixon and Quirk et al. in their remarks on the colloquial use of LVCs, backing up her claim by an analysis of unplanned and planned discourse where she found a higher proportion of LVCs in oral/ unplanned speech (ibid.) 10.2.2 Formal characteristics of LVCs LVCs are commonly divided into different classes of constructions which have distinctive syntactic and semantic properties (Butt 2002, Kearns 2002 [1988], Wierzbicka 1982). Thus, similar to the case of COCs, there is not one light verb construction that can be uniformly dealt with, but rather a family of types of constructions, each having unique properties. In most of those studies which advocate different types of LVCs, the distinction is based on the syntactic category of the complement following the verb. Kearns, for example, puts forward two subclasses, labelled "true light verbs" (TLV), as in give the floor a sweep, and "vague action verbs" (VAV), as in make an offer. The main difference between these two types of patterns is that whereas VAVs take derived nouns as their heads, the complement of the TLV is headed by a 'stem noun' which is identical in form to the corresponding verb and has an action reading (2002: 1, 3f.). In contrast to the label 'stem noun' one also finds 'verb stem' for the characterization of LV complements, as used by Wierzbicka (1982) in her comprehensive study of haveconstructions. She distinguishes between the 'have a V' frame, headed by a verb stem, and those constructions in which a light verb is combined with a deverbal noun. Have a thought is treated as an instance of the latter category, and so is have a cough, even though cough coincides in form with the infinitive of its base verb (ibid.: 755 f.). Have a walk, have a swim or have a laugh, in contrast, are described as instances of have-constructions headed <?page no="247"?> 235 by a verb stem. These considerations parallel to some extent those analyses of COCs which attempt at an object-of-result / object-of-event / affected object distinction of the respective COs (as pursued in the present study), the distinctions being based on characteristics of the nominal element 1 . Grade ak-Erdelji (2004), however, questions the distinction between have + deverbal noun (e.g. have a cough) and have + bare verbal stems ( e.g. have a swim), seeing as one commonality between the two types their inclusion of an element which profiles a single episode of a process. Brugman (2001), who is aware of the difficulties of finding clear criteria for a classification of LVCs, opts for 'constructional polysemy' 2 as a solution to the question whether one can describe LVCs as one single constructional pattern, and argues: […] I will assume that light verb phrases constitute a natural class, […], that class I will refer to as 'the light verb construction' (LVC) or 'the light verb constructions'(LVCs). Whether we take the collection of light verb phrases as comprising a single construction or more than one depends on how detailed we wish to make the syntactic specifications; in the terminology of Fillmore […], a construction may be a collection of constructions […].( ibid.: 555) Drawing on Brugman's description, one can recognize constructional polysemy as one crucial characteristic shared by COCs and LVCs: Both can be captured in a network of constructions, comprising lower level schemas that reflect the distinction between result objects and eventive objects (i.e. objects expressing action). The question in how far the suggested different types of LVCs might be presented as metaphorically or metonymically related, however, goes beyond the scope of this summarizing section. 10.2.3 Syntactic and semantic issues in the study of LVCs Similar to their treatment of COCs, linguists working within the generative framework are primarily interested in the syntactic relation between the NP and the verb, focusing on the question whether the nominal complements receive case marking or a theta role from their verbs. Kearns (1988/ 2002: 12f.) claims that according to their Syntactic Argument Structure, the verbs in a LVC can project phrases and assign case, yet, due to an inactive Lexical Conceptual Structure (LCS) 3 (which leaves the verb semantically empty), they are devoid of their theta-marking capacities. Hence, the 1 Note however, that, while the distinction of the different CO-types is based on semantic criteria, that between 'verb stem' and 'deverbal noun' in the case of LVCs draws on formal characteristics. 2 The source of the polysemy of the constructions resides in the polysemy of both the respective light verbs and the various readings of nominal forms following the verbs. 3 In contrast to their 'light counterparts', the LCS is active in 'heavy verbs', which in turn is claimed to guarantee the assignment of thematic roles to the respective arguments. For a brief description of 'LCS', see Ch. 2.3.3., p. 25f.). <?page no="248"?> 236 NP (i.e. the stem noun) in such a construction is not assigned a thematic role. However, having an action noun reading, the stem nouns "save the structure" (ibid.) of the VPs they are part of by providing them with an LCS and thematic array. Accordingly, in John gives the floor a sweep, give assigns objective case both to the floor and a sweep. The thematic roles of John and the floor are provided by sweep, i.e. John is assigned the agent-role, whereas the floor is assigned the patient-role. Kearn's description draws on an analysis put forward by Grimshaw & Mester (1988), according to whom light verbs are subject to a process called "Argument Transfer". By this process, theta roles and case assignment characteristics are transferred from the verbal noun, which has an intact argument structure, to complete the defective structure of the light verb (ibid.: 229-230). Apart from these issues of thematic role and case assignment, other idiosyncrasies of LVCs, concerning their inability to passivize, or restrictions on the form of the NP (no definite NPs, no pronominal forms) are generally dealt with in analyses with a generative-grammar background (see Butt 2003, Kearns 2002 [1988], Macfarland 1995). Fundamentally different approaches to LVCs are offered, for example, by Newman (1996), Round (1998) and Brugman (2001, see above) 4 , their studies being couched within the cognitive linguistic paradigm and thus quite naturally focusing on the semantic-conceptual subtleties of the construction. Most of the authors approaching LVCs from a CL perspective emphasize that all of the 'full' verbs from which the light forms are derived encode meanings which relate to fundamental or primary events of human experience. These basic facets of meaning constitute the motivation for and figure to a certain extent in the meaning of the respective LVCs. In his cross-linguistic study on the verb give, Newman proposes a description of the 'light form' of give which is based on Langacker's analysis of periphrastic do (1991: 205f.; see below). The meaning that Newman ascribes to the light verb is merely of schematic verbal nature, which then is integrated "with the more richly elaborated deverbal predicate to form a semantic whole" (1996: 202). Nevertheless, he does not deny the verb any kind of semantic contribution and presents evidence for meaningful elements, such as intentionality, energetic flow, and punctual or perfective aspect, in the verb's usage as a light form. The slight form of energetic exchange that Newman describes for LVCs might lead to the assumption that they, just like COCs, can be treated as peripheral members of transitive constructions. In fact, Taylor (1995: 212) actually undertakes exactly such a description, noting that the inability of LVCs to passivize is simply an indication of their marginal status within the 'transitive category'. Round (1998) 4 Accounts by Wierzbicka (1980) and Dixon (1991), who primarily focus on the semantics of LVCs, will be included in my more detailed discussion of the semantics of LVCs in the following sections. <?page no="249"?> 237 analyses periphrastic constructions involving the light forms take and make as meaningful constructions (in the understanding of Goldberg 1995), which preserve aspects of the sense of their respective full verbs. These aspects then serve to impose a particular profile on the scenes the 'base' verbs from which the deverbal nouns combining with the light verb are derived designate. Round (ibid.: 340f.) suggests, for example, that the pattern take a + N attracts deverbal nouns of INGESTION , and here in particular nouns which can be construed as having a particular quantity. This property of the construction can be traced back to features of full take, which usually takes a direct object denoting a concrete entity and refers to actions which involve the separation of an entity or groups into smaller pieces or sections. Hence, one finds the perfectly acceptable form to take a bite, where bite measures out a particular quantity of a substance which is separated from a 'base substance/ mass'. *To take a munch, on the other hand, is judged to be unacceptable, as munching involves "acting on the food […] as a whole" (ibid.). A similar description, i.e. an approach to LVCs which relies on the semantic properties of the full forms of the particular light verbs, is proposed by Brugman (2001), who examines constructions incorporating the verbs give, take and have. In sum, studies conducted more recently, and mostly with a cognitive linguistic background, have shifted attention to LVCs as form-meaning pairings which receive part of their meanings from the respective light verbs they include, which in turn maintain semantic components of their 'heavy' counterparts. Meaningful components have been identified for most light verbs, some of which will be discussed in more detail in section 10.4. 10.3 Parallels between COCs and LVCs? Having briefly summarized major points of interests pursued in studies on LVCs, I shall now continue to contrast the constructional forms with COCs. In my summary of Macfarland's (1995) analysis in Chapter 2, several characteristics that distinguish the two constructional types from one another have already been presented and discussed. Macfarland, principally applying syntactic features as main test criteria (i.e. syntactic manipulation such as fronting and passivization) for a discrimination between the constructions, concluded that whereas COs can be considered true arguments of their verbs, the NP in LVCs has the status of an adverbial event argument. Thus, on the basis of their syntactic characteristics, the two constructions are treated as inherently different. <?page no="250"?> 238 Nonetheless, other scholars, paying attention to semantic and functional characteristics, describe the two constructions as similar. Jespersen (1954 [1927]: 117) notes that [s]uch constructions [LVCs, SH] also offer means of adding some descriptive trait in the form of an adjunct […].They thus in some way form a parallel to those with a 'cognate object'. (emphasis added, SH) 5 It cannot be denied that a good amount of examples of both LVCs and COCs found in the BNC contain information that could hardly be expressed via the simple verb construction or, if so, only in a highly cumbersome and longwinded style: (3) a. Instead, he laughed one of those raucous Hancock's Half Hour snide laughs. (J0W: 2451) b. This was not obvious to most people in the later nineteenth century, and the general effect of the Second Law was to encourage the gloomy belief that in time the universe would die the heat death of general tepidity […]. (GU7: 930) c. Lady Roscarrock gave a sigh of pleasurable anticipation. (EWH: 1524) However, as will become evident in the course of this chapter, the function of LVCs and COCs - viewed in contrast to their respective simple verb variants cannot be reduced to their offering a means for the inclusion of descriptive lexical material. The particular choice of one of the available constructions also depends on how speakers want to represent the action expressed by the verb or deverbal nominal respectively. Quirk et al. (1985: 751), too, notice a similarity between the two construction types, based on the semantic content of verb and object: The more frequent eventive object 6 [in LVCs, SH] can sometimes be related to a cognate object in that it substitutes for the major lexical meaning of the verb whereas the cognate object repeats the lexical meaning. One might assume then that the two constructions involve two opposite 'mechanisms': In LVCs, the deverbal object inherits the conceptual content of the 'original' verb and the 'substituted' light verb simply serves to preserve the processual character of the described event. In COCs, on the other hand, the object landmark repeats major parts of the verb's conceptual 5 However, unlike COs, which Jespersen lists as objects of results, the nominals in a LVC are not classified by him as such but as referring to "the action or an isolated instance of the action" (1954[1927]: 118; emphasis added, SH). In other words, Jespersen suggests an event interpretation of LVC nominals. 6 Quirk et al. (1985: 751) term the deverbal noun following the light verb "eventive object". <?page no="251"?> 239 content and is construed as the outcome of the process designated by the verb. Presumably, the COC much more profiles a relation between a process and its result, while the LVC completely shifts focus on one episode of a process (as an indication of perfectivity; cf. Chapter 7) - a notion of 'creation' being absent. These assumptions will be considered again in the course of this chapter. One further similarity between LVCs and COCs lies in their potential to change the basic argument structure of the more 'basic' simple verb construction. As noted before (see p. 33), COCs can elaborate the structure by providing a slot, for example, for the inclusion of another participant. Instances of constructions involving the verbs sing and smile shall serve as an illustration: Here, a recipient/ beneficiary profiting from the action can be verbalized with the help of the COC; the mention of such an additional participant is not possible with the simple verb alternant, unless it is expressed periphrastically by a prepositional phrase: (4) a. I'll sing you a song. but: *I'll sing you (I'll sing for you) b. She smiled him a smile. but: *She smiled him (She smiled at him.) As the examples given suggest, it is possible to verbalize the additional participant by means of a prepositional phrase. Such a periphrastic expression, however, where a preposition 'interferes' between the verb (action) and the object (recipient), lacks the 'immediacy of benefit' implied by the double object construction, where there is no such interruption. As Thompson & Coide (1987) show, the greater syntactic distance between verb and beneficiary corresponds to a lesser degree of personal involvement, or, more generally, to a greater degree of conceptual distance. For some verbs, such as teach and show (and arguably, sing and smile) "conceptual distance involves whether A's [the agent's, SH] actions actually impinge on the R [the recipient, SH]" (ibid.: 402). The authors argue that the selection of either of the alternative expressions follows iconic principles and might count as evidence for Haiman's claim that "the linguistic distance between two expressions corresponds to the conceptual distance between them" (1983: 782). Thus, speakers opting for the ditransitive COC instead of the simple verb + PP variant verbally express their perception of a certain impact the action has on a participant who is construed as a recipient and thus as possibly benefiting from the action. Comparable observations with respect to changes in the participant constellation can be made for LVCs, which, however, display the opposite effect. In some cases, the verbalization of an event via this construction enables the omission of an argument which is otherwise necessary for an acceptable simple verb construction. As demonstrated in (5), the LVC can be used to present an 'intransitive' version of the event: <?page no="252"?> 240 (5) a. I tried to make a correction. vs. *I tried to correct. b. I consulted all judges and took a decision. vs. *I decided. As regards the assumed close relation between the COC and LVC, Fillmore (1968: 85) even puts forward an approach in which both constructions possess a (underlying) CO and claims the LVC to be derived from the COC. In his discussion of factitive case (i.e. the case assigned to effected objects) he considers contexts in which this case category remains empty (with intransitive verbs, for example). Both COCs and LVCs can then be seen as dummy constructions for such instances, with COCs offering a dummy noun and LVCs a dummy verb, both enabling the expression of factitive case. Fillmore sets up rules whose step-by-step application results in the respective sentence types: First, the noun object is copied from the verb, giving rise to the CO form: (6) dream + Ø dream + dream (V) (NP) (V) (NP) In the resulting construction, the NP is assigned factitive case. In a second step, the original verb is replaced by a dummy verb (pro-V), yielding a LVC with a nominal still carrying factitive case: (7) dream + dream have + dream (V) (NP) (V) (NP) In Fillmore's description of the link between the two constructions, it is not so much a relationship of semantic or functional similarity that is claimed to hold between them but rather one of derivation in the sense that COCs form the base for LVCs. It goes without saying that such an analysis is rejected in a CL approach, for within this paradigm each construction is considered as having 'a life of its own', i.e. no construction is assumed as serving as the basis of derivation for another construction. Having discussed assumptions on general commonalities between LVCs and COCs, I will now turn to an elaborate description of semantic nuances of selected LVCs, working out the differences between the several patterns for those cases where there is a choice available to speakers. 10.4 The semantics of selected LVCs in comparison to COCs 10.4.1 Have-LVCs: Have a dance vs. dance a dance The most comprehensive work on the meaning of light have was carried out by Wierzbicka (1982), who focused on constructions in which the ver- <?page no="253"?> 241 bal component of the complement following have is still preserved. Excluded from her discussion are instances such as to have a cough in the sense of 'being ill', She has a cute smile, describing a trait of the subject, or She had a visit from her parents, in which the subject is not the agent of the action described by the verbal noun (cf. ibid.: 755). Moreover, the forms to have a quarrel/ an argument etc. are discussed as a separate construction type in which have combines with a noun denoting "continuous purposeful reciprocal interaction" (ibid.: 756) 7 . According to Wierzbicka and as also noted by Jespersen (1954 [1927]) and Dixon (1991) the main difference between the type of have-LVCs investigated by her and its simple-verb counterpart is of aspectual nature. The periphrastic construction is said to impose a reading which presents the action expressed in the construction as limited in time, however, not as momentary. Accordingly, it would be unnatural to interpret She had a swim as an event that stretched over hours, whereas She swam does not preclude such an interpretation. Similarly, following from Wierzbicka's claims, the occurrence of an utterance such as She had a long swim should be rather unlikely 8 . The temporal limitation of the event is reinforced by the use of the indefinite article a in front of the verb stem, giving the impression that only a portion of the activity (which can be pluralized or repeated) is carried out (cf. Wierzbicka 1982: 759). Such a general remark cannot be transferred to the description of COCs because the verbs typically found in such expressions are of a different nature with respect to temporal extension. That is, in contrast to the have-frame, the COC itself does not impose particular limits on the stretch between beginning and ending of the action, except for the fact that some bounded portion of the event is presented. Another meaningful component typical of have a-constructions is a notion of the agent's initiation of and control over the described action (Brugman 2001: 573, Dixon 1991: 349). This aspect of control is one element the light verb conceivably preserves from its heavy counterpart, heavy have being most closely associated with "alienable, physical or physical/ psychological possession, where the subject referent may have physical control" (Brugman 2001: 571; emphasis added, SH). Moreover, as Wierzbicka (1982: 757f.) notes, the LVC forces an interpretation of the activity as being aimless/ goal-less or self/ experiencer-oriented: In other words, one can have a walk or have a play (self-oriented), but not have a walk to the station (goal) or have a work (goal). 7 Although Wierzbicka does not explicitly discuss these examples, the following COC candidates would fall into this category: have a dance / a fight / a talk. Note, however, that for talk and dance, all of the semantic components described in the course of this section apply as well. 8 A quick web search with Google does, however, not confirm this claim: The query "had a long swim" yielded more than 15,600 hits (accessed: 09/ 03/ 2007). <?page no="254"?> 242 Brugman approaches these issues in terms of force-dynamic aspects, noting a coincidence of energy source and sink in the physical domain, but she also includes in her considerations the psychological domain (as the source of joy or relief), claiming that [i]n the psychological domain, the force dynamics are self-oriented: that is, the subject referent is an energy source, but also an energy sink; the energy is expended with the expectation of some outcome for the self, rather than for another. (Brugman 2001: 572; emphasis added, SH) 'Joy' or 'relief' brought about by the action is taken up by both Wierzbicka and Dixon as one further meaningful component of the have aconstruction. Consequently, one frequently finds verb stems expressing recreational activities (have a swim/ a lie-down / a stroll), activities that promise enjoyment or pleasure (have a taste/ a drink/ a smoke) or bodily actions resulting in physical outlets (have a yawn, a vomit). Expressions such as have a study / a spit 9 , on the other hand, seem odd, as they neither express actions of joy nor of relief (cf. Wierzbicka 1982: 763, 769). Having given a survey of the semantic nuances conveyed by the LV pattern, I will now present a systematic description of the different semantic effects of COCs and have-LVCs, and comment on major differences between such alternating forms as exemplified in (8): (8) a. The next dance was a foxtrot, but they continued to dance slowly. (BP7: 1331) b. Her body was warm through the layers of clothing they wore, when they dance slow dances. (H7F: 1117) c. We also had a slow dance to "Eternal Flame" and a kissing booth 10 . (www.artintheage.com/ post; accessed 05/ 07/ 2007) 9 Once again, counterexamples can be easily found on the WWW - in these cases, the notion of 'short duration' seems to override that of 'joy/ relief', as in the following example: a) When I finished researching about the Nissan cold air intake and Nissan fuel filters, I had a quick study on the CR-V’s engine and its specifications. (www.aeonity.com/ engine-101/ archive/ date/ ; accessed 09/ 01/ 2007) b) Unfortunately for Richard when Alex had a quick spit, the wind caught it and it hit him head on (so to speak). (www.grahamhowe.com/ endtoend.swf; accessed 09/ 01/ 2007) 10 For have a dance, expressing a reciprocal interaction, Brugman (2001: 573) observes that "while a single person can do a dance, only two or more people can have a dance. Wierzbicka (1982: 785) notes that, if two people do a dance, they can be dancing at a distance from one another; however, two people having a dance together must be dancing fairly close to each other. This feature is inherited from a subtype of the havea-V construction, the so-called 'joint-mutual action' type as displayed in have a hug/ a kiss/ a cuddle (cf. ibid.). <?page no="255"?> 243 In contrast to the simple verb variant in (8a), both the COC and the LVC represent a temporally bounded version of the activity. In the above examples, the activity ends when the dance is completed. Also, both (8b) and (c) can have a reading in which the type of dance (i.e. a slow dance) 'predetermines' the manner of action, which is not the case for the first variant. However, although the latter two utterances both encode actions with initiating Agents, the COC variant focuses on the involvement of the Agent(s) in the action and the outcome of the activity in which they are participating, whereas the LVC much more emphasizes the (expected) pleasant effect of the activity on the Agent(s). Wierzbicka (1982: 761) clearly identifies such a meaning potential in the have a construction: "[I]t shifts the focus from the action to the potential effect of that action on the agent; moreover […] it implies a beneficial effect." It can be argued that in both constructions the emphasis or de-emphasis of the action follows principles of iconicity: In LVCs, the original verb is replaced by a light form, resulting in a 'fading' of the particular action; in COCs, on the other hand, a repetition of the action by the nominal following the verb results in an emphasis of the action by profiling a relation along the energetic chain Agent-Action-Result. It needs to be noted here that, although both constructions are explained in the literature as a kind of "auxiliary" construction in order to express manner, i.e. to enable some modification of the action, this claim seems to be more valid for COCs: Of the 3351 hits of light haveconstructions (as listed in Table 10.1), almost half of the examples (44.6 %) occur with a bare nominal, i.e. including nouns without any kind of modification. For these unmodified cases, the meaning components described above possibly serve as a major motivation for speakers making use of the periphrastic construction instead of the simple verb pendant. The attested frequent use of a bare nominal within the have-LVC lends support to Wierzbicka's assumption of an iconic correlation between shortness of form (resulting, among others, from the absence of a modifying element) and shortness of duration, the author speculating that "the bareness, and therefore the shortness, of the verb stem has an iconic effect, emphasizing the limitation of the time of the action" (ibid: 1982: 760). 10.4.2 Give-LVCs: Give a smile vs. smile a smile One of the prominent features of LVCs involving give is their ability to express a ditransitive relation, a feature which the constructions 'inhere' from the full verb give. This gives rise to a 'ditransitive' subtype of give- LVCs, in which the deverbal NP is preceded by another NP which refers to an entity being the aim (recipient) of the action or benefiting from it. A second subtype appears in the form of a 'monotransitive' construction, which consists of the light verb and the verbal NP (cf. Dixon 1991: 384). The <?page no="256"?> 244 following tokens exemplify the two different versions (example [9c] contains both forms): (9) a. The solicitor gave an embarrassed cough. (FS1: 2320) b. Charlie gave his son a wry smile. (AN7: 206) c. […] Ashley gave the matter some thought and then gave a triumphant shout. (CAE: 1517) The two constructions preferably occur with different verbs: While the monotransitive pattern most frequently attracts verbs expressing various forms of 'releasing sounds' (cough, shout, cry, sigh), the ditransitive type is more heterogeneous as regards the types of verbs that are used therein. Concerning the meaning of give-LVCs, several semantic components are identified in the respective literature. Just like the have-construction, combinations with light give represent self-initiated, temporally short and limited (i.e. telic) actions without an external goal (Dixon 1991: 348f., Newman 1996: 203). If the verbal noun in the construction is of an incremental nature, the give-construction isolates one unit of the activity; no such limitations hold for have: Give a laugh most often relates to a single 'ha', whereas have a laugh could describe someone laughing for a minute or two at something he found excruciatingly funny. (Dixon 1991: 348) For the ditransitive subtype, the light verb can foreground the aspect of volition. Both Dixon (ibid.) and Newman (ibid.: 203) describe that, whereas the simple verb construction can represent both an intentional and an unintentional act, the LVC variant must be taken to refer to an intentional act only. Compare the following sentence pair: (10) a. Jenny pushed the door. b. Jenny gave the door a push. While (10a) does not rule out an interpretation of an 'accidental action', (10b) clearly describes the act of pushing the door as intended by the Agent. This subtype of light give-constructions preserves traits of the full verb give, which describes a force-dynamic scenario of an energetic flow from a source (Agent) to a sink (Patient/ Recipient). This scenario is transferred to light give in that the second participant (expressed in the NP immediately following give) is to be seen as an affected entity which is the "target" of the action. Notice, however, that the concept of "volition" is not necessarily implied by the monotransitive subtype (cf. Dixon 1991: 351). Jespersen (1954 [1927]: 118) even observes an element of unintentionality in this construction, claiming that "the phrase often denotes an involuntary reaction". As a con- <?page no="257"?> 245 sequence, the monotransitive form also allows for inanimate agents (The engine gave a final splutter, The bell gave a little ping), in contrast to the ditransitive subtype, By what has been said about the light give-construction so far, the different interpretations evoked by the different patterns as exemplified in (11) (involving verbal and nominal forms of smile) can be given a straightforward explanation. (11) a. Nick smiled. b. Nick had a smile (at the cartoon). c. Nick gave a smile. d. Nick gave her a smile. e. Nick smiled a smile. The main difference between example (11a) and the other four examples is primarily one of aspect: Only the simple verb variant elicits an atelic reading, all the other variants evoke a telic reading, i.e. the activity is presented as temporally bounded. As regards the difference between the have-version in (11b) and the give-example in (d), it is worth noting that the respective constructions induce slightly different meanings of the NP, which are 'transferred' from the verb smile (Dixon 1991: 349f.). As a verb, smile can mean i) "expressing pleasure or amusement" (OED); or ii) "looking on, upon, at, or to a person with a smile or pleasant expression" (OED). Only the first meaning, conveying amusement, is captured in the haveconstruction, whereas the latter surfaces in (11d), expressing an act of nonverbal communication. The variant in (11c) can as well refer to an act of communication, although in this case the recipient is not verbalized or made explicit. The main differences between the two variants involving light give relate to the action chain or force dynamics, and the degree of volition on the side of the agent. While construction (11d) allows for the inclusion of a recepient/ beneficiary epressed in the form of an indirect object and evokes a notion of transfer of an abstract entity (i.e. a smile) to this recipient, the construction in (11c) leaves the beneficiary schematic and hence lacks the notion of transfer. Moreover, the monotransitive giveconstruction verbalizes an action of smiling which might be done spontaneuosly or involuntarily, a shade of meaning which is absent from (11d), which describes a fully volitional and self-initiated activity. Finally, the CO construction in (11e) represents the 'strongest' version as regards the involvement of the agent in the action and its outcome 11 . In accordance 11 It needs to be kept in mind that the agent is schematically included in the CO (see p. 78). <?page no="258"?> 246 with what has been said before, it is this construction which maximally highlights the action since, first, there is a repetition of the verbal element in form of the CO (neglecting for a moment the issue whether it relates to the event or the result of it), and second, no other aspect of the activity is foregrounded (i.e. neither is volition backgrounded nor an effect on a beneficiary emphasized). What (11e) has in common with (11d) is a forcedynamic scenario in which we find an explicit profiling of the endpoints of the energetic chain. These, however, are instantiated by different entities with different semantic/ conceptual roles: In (11d), a participant with the role of a beneficiary marks the end of the profiled transfer, whereas (11e) includes the result smile, i.e. the entity created by the action, as the endpoint of such a chain. It has been noted before that a corpus analysis of the light-haveconstruction in combination with selected nominals brought to light the high frequency of examples with unmodified nominals. These findings weaken the claim that this construction type is an auxiliary construction permitting the inclusion of modifiers describing the manner of the activity. However, the results of an analysis of the light-give-construction present a different picture: Of 1720 examples, only 229 (13.3 %) include unmodified NPs. Thus, it seems to be a pecularity of the have-construction to attract unmodified nominals. 10.4.3 Take-LVCs: Take a breath vs. breathe a breath For the usage of light take, several authors note that there is an overlap between this construction and the have-construction as regards the nominals they usually combine with. The only difference observed between the two constructions is one of dialectal preference, with the takeversion favoured in AmE, and the have-version in BrE and AusE (Dixon 1991: 351, Huddleston & Pullum 2002: 295) 12 . Yet, as will be shown below, there are also semantic differences between the two constructional variants which cannot be neglected in a proper description of these two patterns. Similar to light have and give, light take preserves meaningful elements from its lexical counterpart. Full take, being the reverse of full give, expresses an activity in which the Agent is at the same time the Recipient of some entity / action. Brugman discusses the meaning of the takeconstruction in the light of its relation to the force dynamic setting of the lexical verb. She arrives at the conclusion that "all light verb phrases with TAKE seem to deploy aspects of the force dynamic schema of TAKE 1" (ibid.: 12 As can be seen in Table 10.1, this preference of usage is displayed by most of the LVC constructions in which both have and take are possible. The only exceptions are have/ take a walk, where take is clearly preferred in usage, and have/ take a bite, where we find an almost even distribution. <?page no="259"?> 247 565), TAKE 1 being specified as evoking a scenario in which an agent simultaneously functions as energetic source and sink and the deictic centre of the action 13 . Thus, one might suspect that one of the semantic shades carried by the light take-construction is that of a benefactive effect the agent experiences from his/ her action. This suggestion is confirmed by Dixon, who observes that the take -frame "refers to some volitional activity, done for its own sake" (1991: 351). This is a crucial feature it shares with the have-construction (as opposed to the give-variant, which carries a notion of transfer). The main difference between the light haveand take forms emerges from the focus (or absence thereof) on the physical effort the agent undertakes. Such a difference becomes already obvious when one compares the lexical verbs have and take. Have, being a stative verb, does not express physical effort at all; take, on the other hand, clearly involves physical activity in its central literal sense ("to touch", "to seize, grasp, capture", [OED]). These features of the lexical verbs are maintained by their light counterparts and result in the semantic properties of the respective LVCs: While both LVC-types are claimed to emphasize volitionality on the part of the AGENT subject, it is the take-version which much more clearly depicts the agent as an active initiator. In this way, the take-frame overtly profiles the energetic chain source-action-sink, which is not so much in the foreground for the have-construction, which lacks goal-directedness and thus a clear sink-orientation (see above). Since in the case of take source and sink coincide, the take-LVC can be described as fully focusing on the subject. Because of the semantic impact the heavy forms of have and take have on their LVC forms, one can have a look without paying full attention (according to Dixon [1991: 352], the have-frame suggests 'half-heartedness' ), yet, take a look implies some more dedication to or involvement in the action. This semantic difference between the variants is caused by the higher degree of 'physical effort' as part of the meaning of take, which is absent from have. Also, as Quirk et al. (1985: 752) point out, it is more appropriate to describe a baby having a bath instead of a baby taking a bath, where the latter would imply that the baby took over the initiative for the activity. As suggested by Wierzbicka, there is also a difference in the usage between taking a walk and having a walk, which the author traces back to the fact that the former refers to a unitary action, having a definite starting and end-point, and the latter to "an arbitrary chunk of an activity" (1982: 795). Therefore it is more felicitious to speak about taking a walk around the pond 13 In contrast to T AKE 1 ( as in He took the book from her), T AKE 2 (as in Mary took her parents to the restaurant) is characterized by a force dynamic schema in which the agent does not function as energy source and sink simultaneously, nor is s/ he the deictic centre. <?page no="260"?> 248 (definite and bounded unit of activity) or having a walk in the park (aimless, no clear startingand end point) than ? having a walk around the pond and ? take a walk in the park (cf. ibid.) 14 . When one now compares light take-constructions with COCs, breathe is the only nominal for which the take-construction is the only available alternative for the CO version (neglecting the simple verb construction). The above remarks on the semantic characteristics conveyed by the takeconstruction may serve as an explanation of the meaning difference between (12) a. I breathe a deep breath and try to relax. (www.bondage-story.co.uk/ blog/ archives/ 2006/ 02/ outside_looking.html; accessed 30/ 06/ 2006) and b. I take a deep breath and try to relax. Whereas the version in (12a) is neutral with respect to the "direction" of the breathing, i.e. it can refer to the inhaling or exhaling of air or the complete cycle of breathing in and out, the sentence in (b) focuses on the initial part of breathing, i.e. it supports the "inward-reading", with the agent functioning as the source and sink of the activity of inhaling air. As with other CO examples discussed before, the version in (12a) represents an intensified version of the activity with the agent depicted as a volitionally acting entity, involved in an action with an outcome that he/ she is responsible for. More generally, where there is a choice between the take-LVCs and COCs (e.g. take a bite vs. bite a bite, take a sleep vs. sleep a sleep, take a walk vs. walk a walk) the former emphasizes the agent as an experiencer of a selfinitiated activity, whereas the latter highlights the relation 'agent-activityresult'. Exceptions are constructions containing the nominals drink and taste, where the CO variants do not evoke a result-reading, but rather one of respectively affected or perceived entities. 14 Round (1998: 341), however, notes that the respective adverbials can be used interchangeably in the two constructions without creating less felicitous expressions, and thereby contradicts Wierzbicka's claim. A look into the BNC yields divided results: Of the 10 expressions involving either of the LVC types and the string "walk/ s …in…", only 3 use have, but 7 include take as their light verb, which lends support to Round's claim. In the case of "walk/ s …around…", all of the 8 constructions recorded in the BNC include take as the respective light verb, which seems to lend evidence to Wierzbicka's generalization. <?page no="261"?> 249 10.4.4 Do-LVCs: Do a dance vs. dance a dance Since there is no analysis of light do available that is comparable to what one can find in the literature on the light verbs discussed so far, I will attempt a brief analysis of this construction, following Newman's (1996) suggestion to apply Langacker's description of periphrastic do. Langacker (1987a, 2003) characterizes auxiliary do as a higly schematic verb; in fact, he notes in a footnote that it is maximally schematic for process predications, putting it on top of a hierarchy of schematic processes 15 . Due to its maximal schematicity, do can be elaborated by any non-auxiliary verb (cf. ibid. 1987: 354). The auxiliary verb provides a schematic process, aspects of which are then specified by a less schematic complement: (13) Does Janet like music/ play soccer/ eat pasta/ go to the concert? The exact procedure of elaboration of the schematic notion inherent in DO is described by Langacker as follows: […] the standard for [ DO ] is a schematic process […]. When the standard is elaborated by another verb in a valence relation, therefore, the overlap between [DO] and its complement is complete: all the specifications of [DO] (both standard and target) are inherent in the less schematic complement and are "absorbed" in accordance with the schematic-transparency principle. (ibid.: 354f.) The lexical verb do, on the other hand, is portrayed as always conveying some notion of activity and some kind of control or volitionality on the part of the subject (ibid. 1991: 238) and carries semantic components that can be paraphrased as "carry out" or "execute". Moreover, it emphasizes a "causative facet of some action" (ibid. 2003: 269). All these features limit its combinability with other verbs: (14) *Janet likes soccer, and Mike does so, too. (cf. ibid.) In (14), the anaphoric replacement of like, lacking the feature of 'activity', by an action-designating do renders the sentence as less acceptable. It is conceivable to discuss the light-do construction as a 'hybrid' of the two forms, in that do refers to a highly schematic process, elaborated by a verbal noun which expresses an activity that is initiated and controlled by the subject, preserving the components "carry out"/ "execute" of heavy do to some degree. Hence, the feature of maximal schematicity and thus its flexibility of combining with any nonauxiliary verb is lost in the light-do con- 15 The auxiliary do […] can be regarded as a version of the active do in which all sense of volitional control is bleached away (as well as that of physical activity)[…]. In fact, it is the most abstract of the process predicates, designating a process but making no specifications as to its nature, not even whether it is perfective or imperfective. Thus it sits at the top of the schematic hierarchy for processes, with the schemata for perfective and imperfective processes as immediate instantiations" (Langacker 1982a: 65). <?page no="262"?> 250 struction, as the construction does not readily integrate with any deverbal noun: (15) *He did a quick drink/ a long sleep/ an intense smell. Transferring the above observations to the light-do-construction, as displayed in (16), we see that instead of a full verb elaborating schematic do in the periphrastic auxiliary construction, we find a deverbal nominal which accomplishes elaboration in the light-do construction. (16) a. They do a run of one and a half to two hours, then return for breakfast. (HPP: 398) b. They wore stupid clothes and did stupid dances. (CK4: 2720) Compared to the other LV constructions discussed before, it seems to suggest itself that light do presents the most neutral version with respect to foregrounding or backgrounding aspects of the process expressed by nominal. If we contrast the constructional variants, as illustrated in (17), the following picture emerges: The only feature that is profiled by the light-do construction is that of the subject's initiation of and control over the activity. The main contrast between this form and the simple verb construction is, once again, that of aspect, where the LVC presents the temporally bounded counterpart of the unbounded simple form. (17) a. They danced slowly. b. They did a slow dance c. They danced a slow dance. It seems, however, that of all light-verb constructions the light-do variant is most similar to the COC in that it lends focus to an agentive, actioncontrolling subject in relation to the result achieved through the action. The COC presents the more intense form of the two variants due to the repetition of aspects of the event (captured in both verb and deverbal noun). Concerning the possible function of the light-do construction as providing a placeholder for modifying expressions, it must be noted that 49% of the nominals following light do occurred unmodified in my compilation of do-aconstructions. However, since the number of analysed hits was rather small (83 instances), generalizations about this construction and a possible preference for unmodified nominals would be premature and only insufficiently supported by empirical data. 10.5 Summary and conclusion In the preceding chapter, I have gone further into the question of possible similarities between Light Verb Constructions and Cognate Object Con- <?page no="263"?> 251 structions, which are claimed to exist by several authors. Instead of focusing on syntactic issues such as the status of the elements following the verb (i.e. verbal stem vs. deverbal noun, adjunct vs. argument), or questions related to the function of both constructions as dummy forms allowing the inclusion of modifying material, I have provided an overview and comparison of the semantic nuances of selected light verb constructions and contrasted them with their simple verb patterns and CO alternatives. As could be shown, both LVCs and COCs differ from simple verb constructions with regard to aspectual qualities, in that the former represent a temporally bounded version of an event. Also, both construction types permit (in some cases) the introduction or exclusion of further arguments, i.e. their use might bring about a change in the argument structure (participant constellation) as expressed the simple verb pattern. As for the semantic potential of the several light verb forms and COCs, clear differences in meaning could be worked out. In all LVCs, the particular type of action is not expressed in the verb, but in the noun-like element following it. The transfer of such meaningful content from the verb to its complement goes along with a weakening of the verbal element and iconically represents a de-emphasis of the particular action. At the same time, the choice of a particular light verb enables the speaker to add facets of meaning as regards the specific nature of the action. As has been demonstrated, each of the light verbs give, have, take and do introduces a meaningful component, 'inherited' from their lexical counterparts, into the expression. From such a perspective, COCs stand in polar opposition to LVCs, as the relation between content of verb and object is almost that of a 1: 1 projection. As was discussed at length in Chapter 4, such an almost complete 'reduplication' intensifies the action not only by verbalizing the result of the action, but in particular by the repetition of components of the conceptual content of the verb by the CO noun. Although these subtle semantic differences might guide and motivate speakers in their choice of verbalizing a particular construal of a scenario, it should, however, not be forgotten that COCs are used much less frequently and are subject to stylistic conventions. To anticipate some results of Chapter 11, the construction occurs more frequently in the written domain of fiction or in 'prefabricated' public speeches; in casual speech domains instantiations of COCs occur most often in the form of idiomatized expressions, where their use parallels to a certain degree an automatized action. With respect to the extension and elaboration of our network of constructions developed in a step-by-step fashion, the analysis and discussion raised in this chapter supports the establishment of another similarity-link between COCs and semantically comparable patterns. The 'growth' of the proposed network is captured in Fig.10.1. It only presents a section of pos- <?page no="264"?> 252 Cognate Object Constructions [NP S V CO-NP ] Light Verb Constructions [NP S LV DV-NP] COC-EV/ R 1 [NP S V CO-NP DO/ E ] Argument Structure Constructions COC-EV/ R 1 / modified CO [NP S V Det Adj Cog. N DO/ E ] GIVE -LVC [NP S GIVE DV-NP] DO -LVC [NP S DO DV-NP] GIVE -LVC/ mod. DV-NP [NP S HAVE Det Adj DV-NP] … NP S SMILE a short smile NP S GIVE a short smile … elaboration link bidirectional similarity link LV = light verb DV-NP = deverbal nominal Fig. 10.1 Similarity links between COCs and LVCs sible interrelations between the two patterns, focusing on COCs of the type EV/ R 1 , which are, as shown above, most relevant to the alternations described in the present chapter 16 . The links are arranged at different levels of schematicity for they hold between less specified patterns as well as for 16 As commented on in section 10.1, drink and taste - being CO nouns of the type A - are exceptions. Thus, more specified links between these instances of A-COCs and LVCs would have to be guaranteed some space in the network. For reasons of clarity, they are excluded in Fig. 10.1. <?page no="265"?> 253 more specific ones, in particular those patterns containing a modified NP. Since this modified form appears to be the default case for COCs, the relationship between the two constructions is highlighted at this level in bold print. For reasons of space and clarity, the links are presented for one type of LVCs only (in this case give-LVCs), but would need to be set up for all of the other types described above as well, linking them to EV/ R 1 -COCs. As in the case of COC-VAC alternations, which were scrutinized in Chapter 9, the similarity link between COCs and LVCs is a bidirectional, yet weighted one, indicated by the black filled arrow towards LVC schemas. The decision to give some primacy to LVCs rests on the observation by several scholars that LVCs are most typically found in colloquial speech and can thus be assumed to be of a less marked status than COCs, which are often associated with more elaborate registers of language use (see section 11.4). However, as stated before, the actual direction of the similarity relation might only be decided on in a case-to-case fashion, depending on the strength of associations of particular verbs to each of the construction, or the existence / availability of automatized, lexicalized patterns for which there is no real choice as to what kind of constructions is used. These most fully elaborated schemas, which are to be found on the lowest levels of a constructional network, shall be the subject matter of our next stage of this comprehensive investigation of COCs: The following chapter will examine more closely patterns of COCs which are specified in great detail and for which the degree of flexibility in form and meaning is limited to a great extent. <?page no="266"?> 254 11 Lower-level schemas: Collocations, idioms, register 11.1 Preliminary considerations It lies in the very nature of a usage-based network of constructions that it comprises all kinds of schemas along a continuum ranging from highly abstract constructional templates to concrete verbal expressions: A general feature of linguistic organization is the existence of complex categories, in which multiple variants are linked in networks. The individual nodes in such a network can be structures of any size or any kind. As a special case, each node consists of an entire symbolic assembly. The network then defines a category whose members are related constructions. Complex categories of this sort are important in describing both lexicon and grammar, which can be seen as forming a continuum. (Langacker 2008: 237, emphasis added, SH) The preceding chapters have covered a great deal of those linguistic facets and elements which Langacker considers as fundamental parts and pieces of a network: By means of collostructional analysis, measuring the strength of attraction between particular verbs and a construction, a possible representation of COCs within a constructional network at a relatively schematic level could be worked out, including subtypes of the COC and verbs most strongly attracted by these (Ch.5) Moreover, the discussion on patterns of specification and modification of cognate NPs has provided insights into types of determiners and modifiers that are preferably used with the different types of COs on a more specific, yet still schematic, level (Ch.8). The two preceding chapters investigated possible similarity links between particular mid-level schemas of EV/ R 1 -COCs and the semantically similar VACs and LVCs, thereby extending the network to related constructions. For the completion of the still somewhat fragmentary constructional network, the focus shall now be put on constructional schemas which allow for no or little flexibility in regard to the lexical items which occur in the slots offered by the construction. These are forms which are assumed to be stored as prefabricated wholes in the speaker's lexicon (i.e. either partially or fully lexically filled constructions) and which are granted special importance in CL as a usage-based approach since they form the basis for the abstraction of higher-level forms as a result of their "reinforced commonality" (Langacker 2008: 239). Besides an exploration of the particular make-up of these lowand lowest-level schemas of COCs, the discussion raised in this chapter will also <?page no="267"?> 255 involve an analysis of the distribution of such patterns over different types of language registers. The major questions to be pursued then are whether particular types of schemas can be associated with particular registers, and whether different types of schemas are employed to a greater or lesser extent for the creation of novel utterances in different speech situations. For the detection of possible lowand lowest-level schemas in my corpus of COCs I used the Cluster tool provided by the corpus analysis software Wordsmith and identified recurring, fully specified units of CO-NPs 1 : Among the candidates for such patterns are the seeds of ( SOW ), a web of ( WEAVE ), a thousand deaths ( DIE ) or a song ( SING ) etc. (see below) which appear to have the status of fixed elements within the COCs of their respective verbs. Here, a strong association between these specific NPs and their verbs can be assumed. For an evaluation of the strength of attraction between the verbs in the COC and such particularly patterned CO-NPs a collostructional analysis was conducted, comparing the attraction between these NPs and the respective COC-verbs and all other verbal constructions in the BNC. Table 11.1 gives an example of the data used for the calculation of the collostructional strength between the (still schematic) CO-phrase the seeds of and the verb sow, showing that of the 291 times this phrase occurs in the corpus, 66 times it is contained in a verbal construction headed by sow. the seeds of the seeds of row totals sow 66 481 547 sow 225 10,205,528 10,205,753 column totals 291 10,206,009 10,206,300 Table 11.1 Frequency data necessary for the calculation of collostructional strength The respective data, as represented in the table for sow + the seed of, were collected for all the CO-NP patterns that could be identified with the Cluster tool. These patterns mainly come in two types, which will be discussed separately in what follows: Firstly, there exist low-level schemas which are variable to some degree, and, secondly, we find fully fixed lexical arrangements 2 . Both forms are assumed to have unit status. 1 For a particular search word, the Cluster tool identifies strings of words which repeatedly occur together in the same syntactic arrangement. For tell, for example, the strings tell the tale (50) , to tell the (49), tale to tell (28), tales to tell (27) are, among many others, identified as such patterns. 2 These two types roughly correspond to Fillmore et al.'s (1988: 505) distinction between "substantive" and "formal" idioms. Whereas the former are idioms for which <?page no="268"?> 256 11.2 Variable low-level schemas By variable low-level schemas I understand COC templates of particular verbs for which the type of modification or determination of their cognate NP, or even both, is fixed. Based on their representation in the BNC, such patterns could be observed for the verbs sow, weave, sleep and live, which frequently collocate with an NP that is postmodified by an of-PP and specified by a particular determiner, in the cases discussed here either the or a. Interestingly, as regards the patterns involving PP-postmodification, the NPs within the PP show semantic similarities, designating, for example, abstract concepts or human qualities. In addition, for the verb live, the combination with a noun preceded by a possessive pronoun was found highly frequently. The structures in (1) describe and exemplify these constructional schemas, the respective frequency figures and values of collocational strength are provided in Table 11.2: (1) a. sow + the seeds of + abstract, mostly negatively connotated NP Examples: sow the seeds of disease/ hatred/ doubt The truth is that the seeds of disaster were sown early on: THE Duchess spent far too much of her marriage wanting to have fun. (CH1: 3214) b. live + a life of + abstract NP denoting state, condition, quality Examples: live a life of luxury/ hardship/ poverty These chose to live a life of poverty and charity supported by their own manual labours and loosely affiliated to local religious houses, from whom they might receive spiritual guidance, and to whom they gave the proceeds of their work. (HY6: 369) c. weave + a web of + abstract, mostly negatively connotated NP Examples: weave a web of intrigues/ lies / delusion So he wove a web of lies. But what's important right now is Thomas,' she said, growing impatient of these explanations about her foster brother. (JY9: 2563) d. sleep + the sleep of + deadjectival NP designating human characteristics Examples: sleep the sleep of the just/ the soulless/ the young/ the virtuous the "lexical make-up is (more or less) fully specified", the latter involve forms in which parts of the phrase can be realized by lexical items which are syntactically and semantically appropriate for the slot. Fillmore's distinction reflects the syntax-lexicon continuum. <?page no="269"?> 257 You'd think she'd deserve to sleep the sleep of the virtuous, but forget it. (HGN: 2037) e. live + the life of + animate NP denoting types of human beings Examples: live the life of a monk / a slave / a soldier […], and the fact that I've chosen to live the life of a monk for the past six months is because until the past few days I have felt no desire to give in to the temptations paraded before me. (HA5: 2108) f. live + PossPr. life/ lives Examples: live my/ your/ their live I need to get myself back on my feet and forget all this and get it all sorted out and so I can live my life again. (FXT: 515) verb CO -NP Percent. of total 3 NP in corpus obs. freq. of NP in COC exp. freq of NP in COC coll. strength sow the seeds of 31% 291 66 0.02 217.249 live a life of 6.5% 377 45 1.02 57.354 weave a web of 33% 74 9 0.005 25.852 sleep the sleep of 50% 23 7 0.01 17.642 live PossPr life / lives 17% 32313 177 87.07 16.831 live the life of 21% 1601 21 4.31 8.194 Table 11.2 Collostructional strength between verbs and fixed CO-NPs In order to enable the reader to evaluate these results, some comparable data should be offered here: The phrase the seeds of (1a), for example, also occurs in the BNC as the complement of the verbs contain, carry, plant and cultivate. The calculation of the collostructional strength between verb and NP yields the following results, showing that the verb sow attracts the phrase most strongly: (2) sow + the seeds of (66 hits): 217.24970 contain + the seeds of (26 hits): 34.93006 carry + the seeds of (11 hits): 8.67435 plant + the seeds of (5 hits): 7.76063 cultivate + the seeds of (1 hit): 1.64627 3 Percentage figures refer to frequency in relation to total amount of COCs of the respective verb. <?page no="270"?> 258 Note that some of the instantiating forms of the schema live + PossPr. life / lives (1f) are themselves candidates for fixed, completely filled constructions (see Table 11.3), which, however, does not invalidate the status of the higher-level schema as an entrenched unit in itself. Rather, this observation is in accordance with CL's assumption of a highly redundant representation of linguistic structure with high and low level schemas existing 'side by side'. live + obs.freq coll.strength their lives 64 42.146546 our lives 29 19.991594 her life 23 6.895116 your life 17 5.941293 my life 23 5.606500 Table 11.3 Coll. strengths of different instances of the schema 'live + Poss Pr.life/ lives' As has been noted already and as can be taken from the description of the patterns in (1), most of the combinations in which a PP-modified cognate NP was identified as one component, the PP contains NPs which come from the same semantic sub-set. This phenomenon is known as semantic preference (Sinclair 2004: 142), i.e. the tendency of certain lexical/ semantic units to co-occur with items that share a semantic feature. As can be taken from the data, the phrase live a life of preferably collocates with NPs expressing states and conditions (luxury, poverty hardship, humanity), whereas the phrase live the life of is strongly associated with NPs denoting categories of people (a monk, a beggar, a gentleman, a slave). Such a semantic preference is most obvious and most consistent with the phrase sow the seeds of: Surprisingly, it mostly attracts NPs which describe negative emotions (doubt, hatred, frustration), states and conditions (disease, distrust, discontent) and events (war, revolt, attack, rebellion), although the literal act of sowing a seed is generally connected with prospects of something to thrive and prosper and eventually a positive and fruitful harvest. Of the 66 instances of sow the seeds of + NP, only 12 involve nouns with a clearly positive connotation (success, career, cooperation), while 50 examples occur with negatively connotated nouns. A similar tendency can be noted for the phrase weave a web of + NP, which occurs with nouns carrying negative associations (lies, deception, illusion, intrigue, intrigue) in 7 of the 9 recorded instances of this schema. The amount of data, however, is too small to give a solid description of <?page no="271"?> 259 preferred collocates of the NP web of in a COC 4 . As regards the schemas live a life of + NP and live the life of + NP, the observations on the semantic characteristics of the slot-filling NP have important implications for the differentiation of possibly available schemas: The fact that these two forms attract nouns belonging to completely different semantic groups is taken to justify their treatment as two separate patterns in addition to a common schema live DET life of + NP. While the form with indefinite a is characterised by an NP denoting states or conditions, the schema with definite the seems to be closely associated nouns which name particular persons or groups of persons, whose particular way of life is taken as a model or reference point for comparison. To repeat what has been said already, these variable low-level schemas are characterized by a relatively fixed "base component" which offers one flexible slot only for the insertion of an NP into a modifying PP, for example, or a possessive pronoun, as in the case of live + PossPr. Fully fixed phrases, which do not provide for any choice, are to be considered next. 11.3 Fixed units The second type of patterning involves those constructions which are completely lexically filled, re-occurring phrases. With those combinations, the form of the CO-NP is fixed with respect to modification, type of determiner, number specification of the noun etc. The combinations listed in Table 11.4 could be retrieved with Wordsmith's Cluster tool as the most promising candidates for such fixed expressions (including also the instantiating forms of live + PossPr life) with their collostructional strength confirming the assumptions on their status as potentially entrenched forms. The expressions provided in (3) exemplify 'real usage' of the respective patterns. (3) a. 'I wanna sing now,' said Liam. Groans all round 'I wanna sing a song for my poor wife who's … who's gone off her head.' (ATE: 2765) b. But his philosophy that when your 'time's up your time's up' saw him through and he's back to tell the tale, though sadly he chose not to include the pictures he took at that time. (APL: 307) 4 A complementary search of the Corpus of Contemporary American English (COCA) (http: / / www.americancorpus.org/ ) confirms this tendency of weave + a web of to be complemented by negatively connotated NPs: Of the 12 entries of this pattern, 7 contain nouns such as deception, chaos, perjury or destruction. <?page no="272"?> 260 c. And, of course, he had been right. If she had done as he had suggested, become a doctor maybe, or a teacher, he would be alive today and she would be living a normal life. (BP7: 1018) d. The liberals hoped that the Inquisition, like the monasteries, would die a natural death as the result of French legislation. (FB7: 1260) e. Imagine him appealing to America to come in quickly, fight the good fight with us, shoulder to shoulder. (HWA: 3263) f. If the burden was placed on the parents to prove they did not do the deed, their failure to convince the jury of this would result in their conviction. (K96: 866) g. 'Look, I'm sorry…' she managed to say huskily, forcing a tight smile while she died a thousand deaths inside. (JY3: 3812) h. In 1967 ('the best year ever'), a Five Year Plan was conceived but seemed to die a death. (AMY: 1135) The indices of collocational strength in Table 11.4 undoubtedly reveal that the frequency of co-occurrence of verb and respective noun phrase is significant and not due to chance. Some of these expressions have acquired an idiomatic meaning ([b] 'to reveal a secret', [h] 'to fail and cease') or count as proverbs (f), while the others appear to be formed freely on the basis of more abstract schemas. However, the close association between verb and particularly arranged NP speaks in favour of their analysis as 'prefabricated chunks'. Some, but by no means all, of these identified patterns are verb CO phrase percent. of total COC NP in BNC obs.freq. NP in COC exp. freq. NP in COC coll. strength a) sing a song 10.9 587 51 0.33 92.08813 b) tell the tale 16.9 519 68 3.63 61.69577 c) live a normal life 4.3 82 30 0.22 54.80462 d) die a natural death 18.4 17 16 0.03 42.14789 e) fight the good fight 52.6 12 10 0.01 28.91498 f) do the deed 46.0 136 31 7.13 11.45158 g) die a thousand deaths 4.6 6 4 0.01 9.66912 h) die a death 9.2 309 8 0.60 6.63903 Table 11.4 Collostructional strength verb + CO-NP in fixed phrases <?page no="273"?> 261 already listed in common dictionaries of idiomatic phrases or dictionaries of collocations, which is a sign of the lexicalisation and idiomatization of these phrases (e.g. tell the tale, die a death, do the deed). No entries, though, could be found for live a normal life, die a thousand deaths, fight the good fight and the more flexible patterns weave a web of and sleep the sleep of. As is obvious from the examples of patterns listed in the Tables 11.2 and 11.4, most of the detected forms belong to the EV/ R 1 class of COCs. This is most probably due to the fact that verbs of the R 2 and A subcategories permit a wider range of NP complements and thus constitute generally more flexible patterns. Note that the phrases talk the talk ('to speak convincingly or effectively, or in a manner consistent with the image one projects' OED) and walk the walk ('to behave in a manner consistent with the image one projects', OED), which qualify as instances of COCs, have found their way into (phraseological) dictionaries of English. The two phrases commonly collocate or are used in a contrastive way, as in (4) Dail deputies can talk the talk, but McDowell walks the walk. (www.forumoneurope.ie/ index.asp? docID=1129&locID=113, accessed 20/ 05/ 07), or are even blended / condensed into walk the talk ('to act on your speech', cf. Siefring 2004), as exemplified in (5) Pakistan will have to walk the talk, to back up its promise to cooperate on terrorism. (www.khaleejtimes.com/ .../ subcontinent/ 2006/ October/ ; accessed 20/ 05/ 07). However, because of their low frequency in the BNC, these verb + CO combinations could not be identified as fixed, lexicalized patterns on the basis of pure corpus data. It is precisely the idiomatic character of these phrases which lends evidence to their assumed entrenchment as units. As Langacker (2008: 238) states, "[w]e have more direct evidence of unit status when an expression consistently displays some idiosyncrasy that does not follow from any regular pattern". The special semantics some lexemes have acquired in the context of fixed COC patterns, e.g. tale meaning 'secret' or die expressing 'fail' and, of course, the non-predictable overall meaning of some expressions (e.g. walk the walk 'to act on your speech), clearly are constructional idiosyncrasies which point at the unit status of the respective phrases. The incorporation of the described lowand lowest-level schemas into the network of COCs results in a highly specified arrangement of templates as illustrated in Fig. 11.1, which - for reasons of space and clarity can only <?page no="274"?> 262 C OGNATE O BJECT C ONSTRUCTIONS EVENT / RESULT 1 [NP S V CO-NP DO/ EV_R1 ] RESULT 2 [NP S V CO-NP DO/ R2 ] AFFECTED [NP S V CO-NP DO/ A ] LIVE + LIFE SMILE + SMILE DIE + DEATH BUILD + BUILDING WEAVE + WEB GIVE + GIFT SMELL + SMELL SOW + SEED live + POSS PR LIFE live my life weave + a web of + ABSTRACT NP weave a web of intrigues … sow + the seeds of + ABSTRACT NP sow the seeds of hatred … … die a death die a thousand deaths die a natural death PRODUCE + PRODUCT … weave a web of lies weave a web of deceit live their lives … Fig. 11.1 A network of Cognate Object Constructions <?page no="275"?> 263 present a small selection of the identified patterns. This network comprises the most abstract schemas of the family of COCs, the categories EV/ R 1 , R 2 and A, on the highest level of abstraction. On the next lower level, theparticular verbs are listed that occur in the respective COC subtypes, or-dered according to their collocational strength. All further lower levels then represent constructional schemas with additional specification concerning determination, modification, number etc. That is the links that hold between the different level are instantiation links, the next lower-level patterns all being instantiations of the next higher-level schema. Constructional information is inherited through these links. The lowest-level schemas are lexically filled constructions, allowing for no or only little variation (variation is possible as regards, for example, grammatical categories of the verb such as person, tense and voice). Note that there are two different types of constructions at the lowest level: Boxes marked with solid lines indicate idiomatic phrases (or potential candidates thereof) and strong collocations, whereas the boxes marked with dotted lines represent more flexible phrases which are possibly arrived at by a slightly more "compositional" route. The choice speakers have here is to select from the field of (often negatively loaded) abstract NPs. 11.4 Register In their characterization of COCs, Quirk et al. (1985) comment on the 'orotund' style of the construction, while Jones (1988) describes them as stylistically elaborate. Judging by such comments, one must expect a more frequent occurrence of COCs in more formal text categories (e.g. fiction, academic writing, and public speeches) and fewer usages in conversation and less formal written categories. Therefore, the focus of analysis shall be laid on the distribution of the construction over different text types and registers in this section, also considering whether schemas of differing degrees of abstraction are exploited differently by speakers in the respective types of register. As these research questions address high-level schemas as well as lowest-level patterns, their discussion is included in this part of the study. In order to compare data across subcorpora or different text categories, raw frequencies must be normed first, as the size (i.e. sum of words of the text in each category) of the different register categories vary in the BNC to a great extent, making the raw frequency data incomparable. For normalization, Biber et al. (1998: 236) suggest to divide the raw frequency count by the number of words of a text (or category) and multiply it by whatever basis is chosen for norming. In the present analysis, the basis was set to 1 million words, a basis which comes closest to the number of words of the smallest text component (i.e. the category 'spoken/ educational').Thereby, <?page no="276"?> 264 an "artificial inflation" of usage events, warned against by Biber et al. (ibid.), is avoided. Choosing a norming basis of 1,000 or 10,000 words would have yielded very small figures due to the rare occurrence of COCs 5 . For the comparison of the usage of COCs in different register domains, the categories distinguished in the BNC were adopted and slightly adapted in that the three different science categories (natural science, applied science, social science), were grouped together as one group labelled 'scientific writing'. The text types resulting from this grouping and employed for the comparison of data are the following: • the written categories 'fiction' (wf), 'arts' (wa), 'commerce' (wc), 'leisure' (wl), 'belief and thought' (wp), 'science' (wsc) and 'world affairs' (wwa) • the spoken categories 'education' (se), 'business' (sb), 'institutional' (si) and leisure (sl). Tables 11.6 a b give an overview of the findings on the distribution of frequencies of COCs over those register categories. : wf wa wc wl wp wsc wwa written EV/ R 1 x/ 1 mill 818 41.2 265 36.4 38 5.2 240 23.9 156 47.2 293 59.6 225 11.9 2035 23 R 2 x/ 1 mill 18 0.9 30 4.1 67 9.3 31 3.1 8 2.6 80 3.2 28 5.7 262 6 A x/ 1 mill 147 7.4 33 4.5 19 2.7 138 13.5 48 15.7 110 4.1 89 5.34 584 13.2 x/ 1 mill 983 49.9 328 45.0 124 17.2 409 40.5 212 65.5 483 66.9 342 23.0 2881 42.2 Table 11.6a COCs in written text categories 5 The exact numbers are: Written component Words Spoken component Words Imaginative 19,664309 Educational 1,265318 Arts 7,253846 Business 2,321844 Belief and thought 3,053672 Institutional 1,345694 Commerce 7,118321 Leisure 1,459419 Leisure 9,990080 Natural science 3,752659 Applied science 7,369290 Social science 13,290441 World Affairs 16,507399 Table 11.5 Size of text categories in the BNC (Source: Burnard 1995: 11; 24) <?page no="277"?> 265 se sb si sl spoken EV/ R 1 x/ 1 mill 16 12.6 4 3.1 28 20.0 113 76.0 161 25 R2 x/ 1 mill 7 5.5 2 1.5 10 6.7 10 6.2 29 4.2 A x/ 1 mill 8 6.3 9 6.8 21 15.6 30 19.2 68 10.3 x/ 1 mill 31 24.4 15 11.4 59 42.3 153 101.4 258 39.5 Table 11.6b COCs in spoken text categories The application of the ²-test shows that for both spoken and written registers, the distribution of the usage of COCs significantly depends on the respective text category (written texts: ² = 508.5; df =12; p< 0.001; spoken texts: ² = 29.9; df = 6; p<0.001). However, when the overall categories "spoken" vs. "written" are compared, the significance of the distribution diminishes ( ² = 5.4, df = 2, p< 0.01). That is the use of COCs depends to a greater extent on particular register categories than on the general distinction between language modes. This also shows in the fact that the figures of overall usage do not diverge too greatly: 42.2 COCs/ 1 mill. words in written language vs. 39.5 COCs/ 1 mill. words in spoken language. Considering the obtained data more closely, we find that the highest usage of all COC-types is that of the EV/ R 1 type in the 'leisure' category of spoken language (76 COCs/ 1 mill.). This finding runs counter to the above assumptions in which it was hypothesized that the core type of the construction appears most frequently in the more formal categories. The categories judged as most formal - 'scientific writing' and 'belief and thought' - only rank second (wsc: 59.6 COCs/ 1 mill. words) and third (wp: 47.2 CCOs/ 1mill. words) in the frequency of usage of EV/ R 1 -COCs. Instances of the EV/ R 1 occur least frequently in texts of the written category 'commerce' and the spoken category 'business'. Although these two registers belong to the more formal domain, one can assume that, due to the very nature of the topics dealt with in these text types, they are characterized by a more or less sober and unembellished (i.e. 'un-orotund') language. For a plausible explanation of the unexpected result, i.e. the most frequent usage of COCs in the least formal text category, the data needs to be scrutinized more closely. As it turns out, the 113 EV/ R 1 in the 'leisure' category of spoken language involve only a small number of verbs: 65 (57%) forms are instantiated by the verb sing, followed by live (17/ 15%) and tell (16/ 14%), meaning that these three verbs, which were, moreover, shown to be part of fixed phrases, occur in 86% of all examples in this category. The other COC-verbs found in this domain are die, dance and dream, used three <?page no="278"?> 266 times each. In analogy to measures of morphological productivity (productivity understood here as the availability of a schema to sanction novel forms), which draw on the relationship between number of tokens, number of types and number of so-called hapaxes 6 , the above figures can be interpreted in the following way: The relatively high number of tokens (113) as compared to the low number of types (6) is a strong indicator that in conversation, the abstract schema which sanctions EV/ R 1 -COCs is used rather unproductively. This observation holds likewise for all other spoken categories, which leads to the assumption that in conversation, speakers primarily rely on those forms which were earlier identified as possibly entrenched patterns, and only very rarely (nearly never) exploit the abstract schema for the creation of new expressions. An analysis of the written category 'fiction', on the other hand, gives a somewhat different picture of the usage of COCs of the EV/ R 1 type. Here, the 818 examples are instantiated by 29 different verb types. Although the four most frequent V + CO pairings found in this category add up to a high percentage (76.8%) of all the forms found there (live: 191/ 23.3%; smile: 191/ 23.3%; tell: 125/ 15.2%; sing: 123/ 15.0% ), the remaining instances see a much greater variety of verbs as it is the case for the category of conversation. If the type: token ratio of EV/ R 1 -COCs of the categories 'fiction' (29 : 818) and 'leisure'(spoken) (6 : 113) are compared, it appears that in the spoken category the construction is used more productively with regard to this aspect of our definition of productivity. However, if the number of socalled hapaxes is applied as a second criterion, this assumption must be modified: Several of the instances in the category 'fiction' are equivalents of such hapaxes, i.e. constructions with verbs that occur only once in the corpus (e.g. purr, step, yell, dress), whereas no such 'unique instances' are found in the 'leisure' category. This finding undoubtedly points at a more productive use of the prototypical COC in fiction writing. That is, the higher-level schema COC-EV/ R 1 is used for the production of new and infrequent instances of this type, and is utilized for verbal play, as in the following example, where the two idiomatic phrases talk the talk and walk the walk are combined and extended by a third (non-idiomatic) COC (i.e. dress the dress): 6 Taylor (2002: 295; see also Plag 2003) defines morphological productivity along these lines as follows: "The more productive a morpheme, the more types we will encounter relative to the number of tokens […] and the higher the number of hapaxes [i.e. instances that occur in a given corpus only once, SH] as a proportion of types. A nonproductive morpheme will be characterized by a relatively high number of tokens of each type […] and by a virtual absence of hapaxes". Since in our approach, both morphological and syntactic patterns are captured as abstract schemas, I see no reason to not apply this notion of productivity to constructions above the morpheme-level. Furthermore, although the term 'hapax' or 'hapax legomena' is used to describe the frequency of words in a corpus, we take its literal translation 'something said only once' as a legitimization to apply this concept to syntactic constructions. <?page no="279"?> 267 (6) Before he served two years in prison, Bobby Lavender had, as they say in LA, dressed the dress, walked the walk and talked the talk of the black urban gangster. (FBL: 2530) Similar results can be observed for the usage of R 2 and A types in the fiction category: No other register category in this study contains a comparatively high number of types and also of forms that appear only once (in sum, for 'fiction', 52 types, 8 hapaxes). Note that a high number of COC types in relation to tokens and of hapaxes is characteristic for all categories of written English in the BNC if compared to the spoken text types. These figures, which point at a more frequent exploitation of higher-level COC schemas in written language and are more frequent usage of lowest-level schemas in spoken categories, are in line with the general assumption of shorter activation times for entrenched units on the lexical level, which are stored as a whole: Thus, in conversation, during which language is produced rapidly and in short periods of time, these units can be retrieved quickly and economically (cf. Langacker 1998: 12, 2008: 238; Plag 2003: 49f.). As Langacker (1998: 12) nicely puts it : "[T]he fluency of real-time speech may only be possible owing to the substantial proportion of boilerplate language and standard expressive schemes it employs". Dabrowska (2004: 203), who incorporates basic insights of CL into a discussion of language processing, and more generally, of a psychologically realistic grammar, arrives at a similar conclusion: One of the reasons why language processing is so fast is because it relies to a considerable extent on prefabricated units which need to be modified only minimally in order to be used in communication. (emphasis added, SH) In writing, there being no such time pressure on the progress of production, the slower route of activating a higher-level schema for the creation of infrequent or new (i.e. non-entrenched) forms can be taken, resulting in a relatively high number of types as observed in our analysis of the corpus data. The above examination of the distribution of COCs over different register categories has contributed to a more complete picture of the usage of the construction types. It could be demonstrated that, although the token frequencies of COCs do not correlate with the degree of formality of the respective text categories defined in this study, the type frequencies depend to a large extent on these categories (and, more generally, on the mode of language). The findings once more support the postulation of constructional schemas of different levels of abstraction, with the entrenched lowest level forms (i.e. lexically fixed ones) primarily used in spoken language. Written language, on the other hand, is characterized by the usage of both those entrenched low-level patterns and the exploitation of more abstract <?page no="280"?> 268 schemas. The latter has been shown to be most applicable for the language of imaginative text types, which obviously led scholars to label COCs as stylistic devices or embellished forms of language. Creative use of these patterns is illustrated in the examples quoted in the study by Macfarland (1995: 35f.), who directed attention to the following lines in Shakespeare's Richard II (7a) and Ripley's Scarlet (7b): (7) a. Bolingbroke: […] Here comes his grace in person. My noble uncle. York: Grace me no grace, nor uncle me no uncle. b. But me no buts, Miss Eleanor. Just hand me the biscuits. I'm going. These forms can be readily explained in terms of the activation of abstracted high-level schemas. As described in Chapter 3, constructions come in various forms and sizes: The English language, for example, offers a complex and systematic inventory of NP constructions, VP constructions, argument structure construction, information structure constructions etc. (cf. Goldberg 2006: 10). Consequently, each linguistic utterance must be seen as being made up of a combination of different constructions. With regard to 'clause-level' constructions, the COC can, as demonstrated in an earlier section of this book, combine with the passive construction; moreover, in the examples given in (8), the A-COC schema is obviously paired with the abstract ditransitive schema. Other such combinations of COCs and different construction types are possible, giving rise to highly creative forms of language usage. The following examples give a brief overview of these possible combinations: COC + Ditransitive (8) a. Diana looked thoughtful as Bruce took her hands in his, she smiled him a smile that conveyed all her feelings of love and affection for him. (www.jlaunlimited.com/ eFiction1.1/ vistory. php? action= print able&sid=4628; accessed 01/ 06/ 07) b. Prayed them a prayer, gave them a key I wanted to spread the joy that had overfilled my spirit. (http: / / keshanicole.tripod.com/ ; accessed 05/ 27/ 07) COC + Resultative Construction (9) a. Where can the Level Two take us in this next phase….that is up to all of us to come together and dream the dream into reality. (http: / / www.visionarymusic.com/ newsletters/ 2006/ VMM5- 7-06.html; accessed 04/ 06/ 07) <?page no="281"?> 269 b. But to think this thought to the end it also will have then to include TM content generation. (http: / / www.isotopicmaps.org/ pipermail/ tmql-wg/ 2007- February/ 000337.html; accessed 04/ 06/ 07) COC + Passive (10) a. But the one true R'n'R death was died by the one true original R'n'R star, Elvis, who died on the lavatory of what boiled down to a lethal dose of constipation. (http: / / findarticles.com/ p/ articles/ mi_qn4158/ is_20021017/ ai_n12649630; accessed 02/ 06/ 07) b. […] and fights were fought to a finish when a man was downed and had been battered into insensibility 7 . (BM9: 146) COC + It -Clefting construction: (11) It was a song Mr Malik sang, and she was singing it to what must, surely, be his words. (HR8: 1698) COC + Left-Dislocation of NP (NP-Preposing) (12) Long tales she told, to kill the tedious hour. (AN4: 1090) While the first two categories display an extension of argument structure, 'extending' the monotransitive COC into a ditransitive or complextransitive (resultative) pattern, the latter three exhibit a re-arrangement of information units. The possibility of merging COCs with the ditransitive, resultative or passive construction undoubtedly underlines the status of COs as true objects: All of these constructions provide an argument slot to be filled by a PATIENT (expressed via a direct object), and in the examples given above it is the CO which occurs in these slots. 11.5 Summary and conclusion: A usage-based network of constructions With the discussion of lowand lowest-level schemas of COCs and the distribution and exploitation of different schemas in the various categories of register distinguished in the BNC, the design of the bottom layer of a usage-based constructional network could be spelt out, completing the picture of possible constructional templates provided so far in Chapters 5- 7 Note that example (10b) displays a combination of three different clause-level constructions, integrating the COC, the passive and the resultative construction. <?page no="282"?> 270 10 and refining the multifaceted portrayal of COCs as found in actual language performance. Major steps, findings and insights which have contributed to this detailed depiction are to be briefly summarized again at this point, concluding the usage-based description of the construction: In the course of this comprehensive investigation of COCs found in the BNC claims which are frequently found in the literature with respect to the construction's degree of transitivity, its ability to passivize and its aspectual potential have been put to test and checked against the data compiled in my corpus research. Likewise, general statements on the combinability of the CO with determiners and modifiers and the use of COCs in different types of register have been probed on the basis of naturally occurring language examples. Due to a careful and close scrutiny of the data it could be demonstrated that the majority of commonly made generalizations cannot be maintained in the strong form they are often put forward and thus need to be formulated more subtly. Furthermore, it has become evident that the validity of the specifications stated in those claims varies with the type of COCs. Especially the use of the definite article with a CO noun could be shown to strongly correlate with a particular type of construction: COCs of the A-type occur more frequently with the definite determiner the than EV/ R 1 and R 2 , which take COs denoting created entities. However, even for the core type of COCs, i.e. EV/ R 1 - COCs, most of the claims have turned out to be unfounded: In the majority of cases, the CO-NP behaves similarly to prototypical types of objects, with restrictions on the use of particular determiners, modifiers or passive forms of the construction arising from pragmatic factors and needs of information-packaging. Moreover, the integration of CO-NPs specified by the determiner the or possessive pronouns and unmodified CO-NPs into EV/ R 1 -COCs could be shown to be compatible with the semantics of the construction as depicted in this study (i.e. transitivity, relations within the energetic chain, aspect) and to result in special semantic nuances of the whole pattern in some cases. Next to the discussion of the corpus findings, the information gathered from the close inspection of the data has been integrated into a usage-based network of COCs and related constructions, representing phenomena of usage on different levels of abstraction. Elicitation of frequency data and calculations of frequent co-occurrences led to the identification of those verbs which are most likely to be used in any of the sub-construction and allowed for a specification of COC patterns - ranging from the most abstract representation of the three types of COCs (i.e. EV/ R 1 , R 2 and A) to lexically fully specified phrases. The variety and frequency of patterns found on different levels of schematicity have not only confirmed the existence of entrenched low-level patterns (idioms and strong collocations) but also support the plausibility of entrenched abstract schemas. The activation <?page no="283"?> 271 of either low-level patterns or high-level schemas could be shown to correlate with language register. The overall outcome of this analysis is a highly complex network of the family of COCs, which not only depicts the linkage between subtypes and instantiating forms, but also the embedding of these subtypes into a larger network of argument structure constructions. This form of representing linguistic knowledge reflects Langacker's vision of language, […] one of a massive network in which structures with varying degree of entrenchment, and representing different levels of abstraction, are linked together in relationships of categorization, composition and symbolization. (ibid. 2000: 5) The account of COCs provided here is but one study which demonstrates that aspects of linguistic structure can be reasonably described and clearly arranged along these lines. <?page no="284"?> 272 12 I have a tale to tell: Conclusion and outlook 12.1 Summary of findings Although each of the previous chapters includes a summarizing section, the main findings of my comprehensive analysis of COCs are to be briefly reviewed at this place for the purpose of a coherent presentation of the construction. One of the major goals of the investigation presented in this book was an exploration of COCs as symbolic patterns, i.e. pairings of form and meaning. In the following, the most significant components which were identified as contributing to the status of COCs as constructions in their own right shall be adduced again. Generally, COCs have been depicted as constructions whose content is motivated by form to a considerable degree. That is, in accordance with the principle of isomorphism (i.e. "one form - one meaning"), the construction's schematic form of a monotransitive pattern ([NP SUB V NP CO ]) could be shown to bias the interpretation of the CO-NP as referring to an entity which is involved in the activity denoted by the verb rather than as a constituent describing manner of the action (i.e. adverbial). Rather than postulating one construction, I have set up COCs as a family of three different, but related, constructions. By doing so, the different semantic characteristics of the cognate objects, the different semantic relationships between objects and verbs (i.e. effected vs. affected) and the relationship to alternative constructions (i.e. VACs and LVCs) could be accounted for. I have shown that two of the types, the forms labelled R 2 and A, are particular sub-patterns of well-established transitive constructions with effected and affected objects respectively, their particularity arising from the specific morphological relationship between verb and cognate noun. The third subtype, labelled EV/ R 1 and usually treated as the true/ core type of COCs in relevant studies, could be identified as a metaphorical extension of transitive constructions involving an effected object, with the conceptual metaphor EVENTS ARE OBJECTS rendering actions and processes as concrete, bounded objects which come into existence through the execution of the respective activities (named by the verb) by an AGENT . Starting out from the core semantic and pragmatic features of the COC prototype (i.e. of the type EV/ R 1 ), the following notions were recognized as indispensable for an adequate description of patterns of this type: <?page no="285"?> 273 1. Transitivity: The action is depicted as proceeding unidirectionally from an AGENT to a PATIENT , i.e. the construction profiles an energetic transfer from a source to a sink. 2. Creation: The outcome of the energetic transmission is not an affected, but an effected entity, constituting the sink of the transfer. 3. Argument structure: Taken together, points 1 and 2 determine the argument roles usually associated with the construction: The subject slot combines with an AGENT role, the object slot is associated with a PATIENT participant. 4. Event-Result-Ambiguity: Typically, the CO designates a reified version of the process encoded by the verb. The reified form can in many cases refer both to the event as a whole unit or to the result of the action. That is, the prototypical COC is polysemous, with the two interpretations linked by a PART - FOR - WHOLE metonymy (i.e. PART : result; WHOLE : event) or motivated by an end-point-focus transformation. The EVENT -reading might cause slight changes in the constellation of argument roles of the construction in that the trajector is construed as an AGENTIVE EXPERIENCER and the landmark as a THEME (which was shown to be the most plausible construal of COCs involving the verb die). 5. Temporal boundedness: The CO as a reified process, i.e. construed as having concrete boundaries, imposes a temporal frame within which the described action develops. 6. Intensity / emphasis: The double encoding of conceptual content in the form of verb and CO-NP evokes a higher degree of intensity of the action (following principles of iconicity). 7. Elaborate style: Except for highly frequent COC expressions such as live a …life, sing a …song or tell a …tale, which are commonly used in colloquial and other less formal registers of spoken English, COCs primarily occur in stylistically elaborate forms of language usage, such a fiction writing, academic prose or public speeches. Since COCs constitute a family of constructions which are organized in the form of a prototype category, where subtypes range from prototypical to more peripheral schemas, not all instances of the construction can be captured by means of the features listed above. In R 2 -forms, for which the CO was defined as describing a concrete, material result, the notions of eventresult ambiguity, intensity and elaborate style do not figure prominently or do not apply at all, while the A-types, involving COs which denote affected entities, lack the features of event-result ambiguity, creation, intensity and elaborate style. Clearly, the more peripheral the forms are, the fewer features are applicable for the description of the respective construction types. <?page no="286"?> 274 Due to the semantic properties of the construction and the close interaction between the construction and the verbs occurring therein, two observations concerning the verbs likely to be found in the pattern should be emphasized here again: Firstly, COCs (except for the A-forms) primarily attract dynamic verbs describing actions which lead to some kind of "output" (e.g. verbal: song, cry; non-verbal: smile, frown; cognitive: dream, thought; bodily: sweat, blood; material: rain, record etc.). Secondly, verbs which in their conventional core meaning lack the notion of 'creation' or 'producing output' (e.g. live, die, fight etc.), inherit these from the construction when fused with it. Such changes on the semantic level are closely tied to, i.e. coerced by, the change in the syntactic frame which the fusion of verb and construction might bring about: In the majority of cases of EV/ R 1 constructions this fusion has substantial effects on the argument structure of the verbs, in that those verbs which are conventionally associated with a syntactically intransitive frame inherit one argument slot from the construction. As regards the formal aspects of the construction, it could be shown on the basis of a large set of instances of COCs that the constructional types do not considerably deviate from prototypical monotransitive constructions, especially with respect to passive forms or their combination with other information-structure constructions such as patterns of NP-fronting or clefting. Nor do CO-NPs display a behaviour distinct from that of noncognate objects: Grammatically, CO-NPs seem to allow the same range of structural possibilities as 'normal' NPs. Contrary to the claims frequently found in the literature, CO-NPs can, for example, occur unmodified and with a definite determiner, if the appropriate coand context is given. A description of the construction as pursued in the project and as summarized here quite naturally brings up again the question of the compositionality of linguistic expressions. As clearly pointed out by Langacker (2008: 168), 'composition' cannot be equated with 'summation', i.e. one cannot arrive at an expression's composite meaning by simply "viewing the component meanings collectively". Instead, the constructional schema (itself a meaningful, yet abstract assembly) contributes a substantive part to the full meaning of a certain expression and guides its interpretation or formation. This is much in line with Goldberg's model of argument structure constructions, where she assumes that verbs are only ascribed basic semantic meanings which are then specified by the linguistic context, that is the constructional pattern in which they are used 1 . In fact, Goldberg's 1 It should be of interest, however, to explore in how far the progressing institutionalisation of the use of particular verbs in uncommon syntactic surroundings might cause changes in their so-called "basic" senses. One might, for example, speculate whether linguists involved in Construction Grammar have stored a slightly changed <?page no="287"?> 275 model aims at preserving compositionality (cf. Goldberg 1995: 16): It is the regular fashion in which verb and constructional meaning are assumed to combine in her model which saves aspects of semantic transparency to a certain extent. Thus, if compositionality is viewed as the regular, ordered integration of component meaning and constructional meaning, COC instances such as The moment you relax you start dreaming day dreams or Jacob smiled a seductive smile, expressing the creation of an event, can be considered compositional, even though neither of the two verbs has 'creation' as one of its basic meaning components. The component may be understood as contributed by the construction. Note that besides Goldberg's account of complex constructional meaning on the basis of fusing verb meaning and a separate constructional meaning, an alternative model was suggested by Fauconnier & Turner (1998a, b), who explain the principles of the usage and interpretation of verbs in constructions that do not match with their inherent participant roles on the basis of conceptual blending 2 : The meaning of expressions of the caused-motion type, for example, as in Junior sped the toy car around the Christmas tree (ibid., 1998b: 5), is discussed as the result of a conventional blend of two input spaces which are related to specific verb frames. Input space 1 includes a typical basic form for a fully integrated event (e.g. throw, which inherently signifies caused motion), while Input 2 is an unintegrated causal sequence (such as "Junior presses remote control. Car speeds around tree."). In the blending process, which "motivates the emergence and main properties of such a construction" (ibid.), the participant roles and the corresponding syntax are mapped from Input 1 3 , with the content for these roles then added by Input 2, yielding a blended form in which the scenario described in Input 2 is now 'equipped' with a caused-motion frame. Note that an integral sub-process of blending, which yields new emergent structure, is that of 'composition' (next to 'completion' and 'elaboration'), the importance of which is accounted for by Turner & Fauconnier (1998a: 144, emphasis added, SH) as follows: "blending composes elements from the input spaces, providing relations that do not exist in the separate inputs". Clearly, in doing away with a 'building-block model' of compositionality, meaning of sneeze after having extensively dealt with sentences such He sneezed the napkin off the table… 2 Fauconnier & Turner (1998b: 1) provide a very 'handy' (yet simplified) definition of this conceptual process: "In blending, structure from two input spaces is projected to a separate space, the 'blend'. The blend inherits partial structure from the input spaces, and has emergent structure of its own." See Fauconnier &Turner (1998a, b) for a more detailed description and discussion of 'blending'. 3 As Fauconnier & Turner (1998b: 5) explain, Input 1 includes "an agent role a, an object role o, a role e that subsumes causal action, means, manner of the action, motion of the object, manner of that motion, and a direction role dm". In English, the corresponding syntactic frame is NP V NP PP. <?page no="288"?> 276 in which "each morpheme always represent[s] the same rigid and stable semantic chunk or building-block" (Sweetser 1999: 136), the model of conceptual integration must be considered as an alternative approach towards an explanation of linguistic compositionality in which the meaning of elements (i.e. morpheme, words, syntactic structures) does not simply add up to an expression's overall meaning, but where these serve as a prompt to construct meaning. Fauconnier & Turner (1998b) point out that, while both their account and Goldberg's model rest on the notion of blending and fusion respectively, it is the nature of the blended/ fused elements, which distinguishes the models: In contrast to Goldberg, Fauconnier & Turner do not explicitly propose a separate clause-level construction existing independently of lexemes which instantiate it. Instead, they suggest that the structure and meaning of the resulting construction is motivated by that of a verb prototypically used in the construction (in the case of caused-motion most probably put or throw) (see also Hampe & Schönefeld 2003). For a discussion of COCs in this light, one would have to consider in how far the subtypes of the construction get their meaning from verbs prototypically associated with the corresponding non-cognate monotransitive subtypes. Alternatively, one would have to scrutinize whether they all are motivated by prototypical verbs occurring in the core type of the monotransitive construction (i.e. forms taking an affected object) and a group of conceptual metaphors, creating metaphorical blends which motivate the usage of verbs of creation (build, create) or verbs of facial expressions (smile, frown) in the construction. Such a mechanism is advocated by Fauconnier & Turner (1998b: 21f), for example, for the usage of verbs of communication in the ditransitive construction, which itself receives its constructional schema from the verb give. The blending of the ditransitive structure (provided by give) and communication verbs is supposed to be 'driven' by the conduit metaphor, discussed by Reddy (1979). My attempt of a systematic examination of COCs within the Cognitive Linguistic framework (incorporating assumptions of Construction Grammar) has clearly exemplified the usefulness and the advantages of the descriptive and analytic inventory of the paradigm. Firstly, with the framework being "data friendly" (Janda 2006: 23) and usage-based, the incorporation of findings from a careful corpus analysis allowed for the refinement of the structural description of COCs. Moreover, "naughty" data 4 , which may occur rarely in natural discourse and which, in fact, can only be detected through the scrutiny of large corpora (e.g. the occurrence of passive COCs, in particular involving the verb die), were not ignored but 4 i.e. data which violate pre-established theoretical claims <?page no="289"?> 277 studied as cases which are well-motivated by cognitive or experiential factors, or as cases which form important links/ transitions to other structural categories. Here, the abandonment of the conception of clearly delineated categories in which all members are equally good representatives in favour of prototype categories or continua of forms could be shown to permit a logical and reasonable explanation of certain linguistic phenomena. Besides the depiction of COCs as prototype categories comprising several subcategories, two 'continua' were discussed in relation to COCs in the present study: First, the construction itself covers a particular section on a transitivity continuum, ranging from core-transitive forms (instances of COCs of the A-type) to less transitive instances (e.g. die a…death). Closely related is the continuum between the syntactic categories of argument (object) and adjunct (adverbial): CO-NPs were shown to be borderline (transitional) cases in some instances, especially in expressions involving the verb live in combination with a CO-NP which might as well be interpreted as a temporal adverb (e.g. all PossPr life, as in He was born in Russia and lived all his life there). Approaching COCs as a family of constructions which are arranged around a prototype and which are definable first and foremost against the notion of transitivity quite naturally led to the depiction of a network of argument structure constructions, based on the assumption that certain argument frames correlate with certain force-dynamic event scenarios. This network, which incorporates usage data and which constitutes only a tiny section of a tremendous arrangement of linguistic structure, captures the relations holding between the different constructional schemas in the form of instantiation or extension links which are motivated by metonymy and metaphor. With the latter being grounded in primary bodily experiences, the network illustrates the successful merging of theoretical modelling, experientialism and the study and statistical analysis of linguistic data obtained from natural usage. 12.2 Future perspectives The analysis of COCs presented in this book has aimed at a portrayal of linguistic structures on the basis of converging evidence, i.e. by combining facts, concepts, ideas and data which were obtained and gathered from different resources. While the present study primarily drew on resources such as theoretical description, language corpora and (the author's) intuition, further studies on the construction might incorporate data and findings gained by other means, methods and models. Let me turn to the modelling of constructions first: My investigation of COCs as a family of related constructions rested on the idea of distinguish- <?page no="290"?> 278 ing the different constructional forms on the basis of the semantics of the CO. Each of the subtypes was shown to attract particular semantic verb classes, even though some verbs were found to be unclassifiable or 'odd ones out'. Recently, Croft (2003) has proposed a theory of verb-classspecific or even verb-specific meanings of constructions for the purpose of specifying facts and peculiarities of particular constructions in much greater detail. Reconsidering Goldberg's (1995) analysis of the highly polysemous ditransitive construction, Croft suggests capturing constructions in terms of verb-specific constructions, which are schemas filled with specific lexical content, and verb-class-constructions, which are slightly more abstract but still much less schematic than constructional schemas such as [ SBJ VB OBJ 1 OBJ 2] for the ditransitive pattern (2003: 58f.) 5 . The concepts underlying the constructional schemas and sub-schemas I have worked out for a network of COCs basically parallel Croft's ideas to some degree, but my representations are only a first step towards a much more intricate network. Future studies on COCs should take up Croft's insights and work out more elaborate constructional sub-schemas. A good candidate for a verb specific schema, for example, is die (the only unaccusative verb occurring in a COC in the BNC), whereas verb-class specific schemas are conceivable for verbs of manner of speaking, verbs of manner of motion, verbs of non-verbal communication etc. (cf. Table 5.2, p.131). A second aspect for possible future projects relates to methodological and empirical issues: First of all, with respect to the elicitation of data, questionnaires distributed among and completed by native speakers of English are valuable and insightful sources of information on language usage and linguistic knowledge. Self-reports on preferences of one construction over another (i.e. VAC vs. COC. vs. LVC) or judgements on the assumed marked or peripheral status of forms are, even if intuitively arrived at, a complement to the linguist's (partly) introspective description. Furthermore, speakers' listing of features or properties which describe the meaning of expressions and which they associate with expressions containing COCs will provide clues on how speakers interpret the construction. Alternatively, participants might be given key words, properties or paraphrases relating to COC expressions and their respective VAC counterparts from which they have to choose those items which best capture the meaning of the particular forms (cf. Hasson & Giora, 2007: 314ff.). Possible key 5 Croft (2003: 58) discusses, for example, the verb-specific schemas [[SBJ permit OBJ1 OBJ2]/ [enabling XPoss by permitting]] and [[SBJ allow OBJ1 OBJ2]/ [enabling XPoss by allowing]] for a representation of the fact that of the class of permission verbs, only permit and allow, but not let or enable can occur in the ditransitive construction. Verb class specific constructions of the ditransitive construction take the following form (ibid.: 57): [[S BJ G IVING V ERB O BJ 1 O BJ 2]/ [actual XPoss]] or [[SBJ B ALL .M OT .V ERB O BJ 1 O BJ 2]/ [actual XPoss via ballistic motion]]. <?page no="291"?> 279 phrases should include the following pairs: high/ low intensity, high/ low control over action, high/ low involvement of acting participant, complete/ incomplete action etc., i.e. notions which most probably express the meaning of COCs as assumed in this study. Thirdly, psycholinguistic experiments might allow for the verification of hypotheses on the processing and storage of instances of the construction: Reading tests, for example, can provide insights into the actual processing load of COCs as compared to the syntactically simpler VACs. If conducted in a setting in which the sentences are presented one word at a time, a reading test can help to single out points of difficulties which speakers might experience when they process these sentences. Applied to expressions including COCs, such a procedure might show whether the integration of the CO-NP into the VP (including a conventionally intransitively used verb) poses a problem for the reader. As reading times can be taken as indicators of processing difficulties, it might be of interest to compare the times needed by speakers to process sentences involving COCs and their respective VAC variants. On the one hand, it seems plausible to hypothesize that due to the greater conceptual load the construction carries (cf. Ch. 4), reading times for COCs will be longer than for VACs. On the other hand, as several lowest-level schemas of COCs were postulated as possibly entrenched units, reading times for these particular forms should not be different from those of their simpler VAC forms, or even shorter. As has been demonstrated in psycholinguistic experiments on idiom comprehension, for example, highly familiar idioms are retrieved quickly, sometimes even more quickly than their literal counterparts, due to their being stored as complete, prefabricated units (cf. Gibbs 1980, Blacksburg & Cackier 1994, Blacksburg 2001). What is more, it is generally assumed in psycholinguistic theories that speakers store a great amount of ready-made (even non-idiomatic) phrases, and that these are accessed much more easily and quickly 6 . Hence, the measuring of reading times might corroborate assumptions on the status of selected lexically filled COCs as entrenched units (e.g. die a thousand deaths, live my life, sow the seeds of etc., see Ch. 11.2-3). Fourthly, as regards corpus-linguistic work and the evaluation of corpus data, I wish to add the following comments: Although my analysis of COCs has exploited data which were obtained from an investigation of a huge language corpus, and has incorporated recently developed procedures of inferential statistics for an analysis of these data, the possibilities offered by statistical corpus linguistics have not been completely exhausted in the study. In recent years, highly complex, corpus-based methodologies 6 See Dabrowska (2004: 18ff.) for an overview and discussion of psycholinguistic experiments which aim at demonstrating the psychological reality of completely stored, lexically filled units in the mental lexicon of speakers. <?page no="292"?> 280 and algorithms for the discovery of collocational and collostructional patterns in large language corpora have been advanced for the modelling of language usage. A brief description of two possible methods which might conceivably be applied to the study and description of COCs shall be outlined here: In my investigation of the formal and pragmatic characteristics of the different types of COCs I closely examined factors such as type of determination and modification of the CO-NP, voice, and preferred register of usage in order to find out about correlations between these factors and the respective constructional types. In this respect, I carried out a monofactorial analysis, considering the single factors in isolation. I did, however, not investigate whether there are significant correlations between all of the factors considered and the construction type, i.e. an interrelationship between constructional type, voice, type of determiner and modification, and register. Such a multifactorial analysis, as demonstrated by Gries (2003a, b), may provide a more fine-grained depiction of the syntactic and semantic assembly of constructional prototypes and allow to "predict native speakers' choices in authentic discourse situations" (ibid. 2003a: 8), defining circumstances of linguistic usage more precisely 7 . The second option I would like to consider for an even more extensive corpus study of COCs involves a rather 'large-scale enterprise': A strictly corpus-driven paradigm is currently developed and tested at the IDS Mannheim (Germany) by Keibel and Belica (see ibid. 2007), whose goal is an in-depth profiling of a broad range of linguistic phenomena 8 . The project includes the compilation of collocation profiles of a large number of lexemes, based on collocational clusters which were generated from a 2.2 billion word sub-corpus of the Mannheim German Reference Corpus. By means of their collocation profiles, semantic similarities between words can be measured in fine detail, a procedure which is part and parcel of the subproject "Modelling Semantic Proximity": For given lexemes (focal words), sets of words whose collocational profiles are almost similar to the focal words, are retrieved. The identified words are then arranged on a twodimensional grid showing the similarity relations between their collocational profiles and those of the focal words. Applied to a comparison of, for example, all the verbs which were measured as being attracted to the different types of the COC in my analysis, other possible verbs which have a 7 Note that Gries (2003a) employs a multifactorial analysis in order to work out the contrast between two alternating forms of verb-particle constructions. Thus, this procedure might as well be used for the comparison of COCs and forms which where discussed as alternatives in my study, i.e. LVCs and VACs. 8 I thank Holger Keibel (p.c.) for giving me some insight into the strengths and potentials of the project pursued by his research group. For more information and illustration, see http: / / corpora.ids-mannheim.de/ ccdb/ . <?page no="293"?> 281 similar collocation profile as those attracted verbs might be identified as candidates for the construction. Moreover, a close examination of those identified words which form an intersection (i.e. which are equally similar to the focal words) might point at semantic characteristics of the construction. Likewise, an application of this method to all the modifiers (i.e. adjective-adverb pairs) which were identified as distinctive collexemes in my comparison of COCs and their VAC counterparts (see Ch. 9) could improve findings on the semantic features shared by the respective distinctive collexemes and thereby yield more reliable and verifiable insights on nuances of meanings of the alternating patterns. Such an investigation would, of course, presuppose the computation and compilation of the collocation profiles of a great number of lexemes recorded in the BNC or, if available, even larger sets of data. Due to their very time-consuming nature, which would have taken away from the comprehensiveness of this study, and also because of their recency, the described procedures were not employed in the present project. If, however, further investigations on COCs are to be conducted, the described methods should be taken into consideration. Because of their potential to operate on immense amounts of data, the application of the procedures deployed in the Mannheim project might bring to light new and maybe unexpected facets of structure and meaning of COCs and their relation to other linguistic constructions. <?page no="294"?> 282 13. References Aarts, Flor. 1986. English Grammars and the Dutch Contribution: 1891- 1985. In: Leitner, Gerhard (ed.), 363-386. Aarts, Flor. 1991. Traditional Grammars of English. In: Leitner, Gerhard (ed.), 293-308. Adger, D. 1994. The Licensing of Quasi-Arguments. In: Ackema, P. & M. Schoorlemmer (eds.). ConSole 1 Proceedings. The Hague: Holland Academic Graphics.1-18. Aijmer, Karin & Bengt Altenberg (eds.). 2004. Advances in Corpus Linguistics. Papers from the 23rd International Conference on English Language Research on Computerized Corpora (ICAME 23). Amsterdam/ New York: Rodopi. Aitchison, Jean. 1994. 'Say, say it again Sam': The Treatment of Repetition in Linguistics. In: Fischer, Andreas (ed). Repetition. SPELL 7. Tübingen: Gunter Narr,15-34. Algeo, John. 1991. American English Grammars in the 20th Century. In: Leitner Gerhard (ed.), 113-138. Algeo, John. 1995. Having a look at the expanded predicate. In: Aarts, Bas & Charles F.Meyer. The verb in contemporary English. Cambridge: CUP, 203-217. Allan, Quentin. 1998. Delexical verbs and degrees of desemanticization. Word 49, 1-17. Allerton, David. J. 1982. Valency and the English Verb. London: Academic Press. Asher, R.E. & J.M.Y. Simpson (eds.). 1994. The Encyclopedia of Language and Linguistics. Oxford/ New York etc.: Pergamon Press. Aston, Guy & Lou Bernard. 1998. The BNC Handbook. Exploring the British National Corpus with SARA. Edinburgh: Edinburgh UP. Barcelona, Antonio (ed.). 2000. Metaphor and Metonymy at the Crossroads: A Cognitive Perspective. Berlin/ New York: Mouton de Gruyter. Baron, Naomi. 1971. On Defining Cognate Object. In: Glossa 0017-1271, 71- 98. Barlow, Michael & Suzanne Kemmer (eds.). 2000. Usage Based Models of Language. Stanford, CA: CLSI. Bates, Elizabeth & Brian MacWhinney. 1982. A functionalist approach to grammatical development. In E. Wanner & L. Gleitman (eds.). Language acquisition: The state of the art. Cambridge: Cambridge University Press, 173-218. Biber, D., S. Conrad & R. Reppen. 1998. Corpus Linguistics: Investigating Language Structure and Use. Cambridge: CUP. Biber, Douglas et al. 1999. Longman Grammar of Spoken and Written English. Harlow: Longman. <?page no="295"?> 283 Boas, Hans Christian. 2003. A constructional approach to resultatives. Stanford, Calif.: CSLI Publ. Boberg, C. 2002. Fact or opinion: A Sociolinguistic View of Native-Speaker Intuitions as Evidence in Linguistics. LACUS FORUM XXVIII : 3-13. Boers, Frank. 1996. Spatial prepositions and metaphor: a cognitive-semantic journey along the up-down and front-back dimensions. Tübingen: Narr. Bolinger, Dwight. 1977. Meaning and Form. London: Longman. Bongers, Herman. 1947. The history and principles of vocabulary control : as it affects the teaching of foreign languages in general and of English in particular. Woerden: Wocopi. Boyd, Richard. 1993. Metaphor and theory change: What is metaphor a "metaphor" for? In: Ortony, Andrew (ed.), 481-532. Brinton, Laurel J. 1995. The Aktionsart of Deverbal Nouns in English. In: Bertinetto, Pier Marco et al. (eds.). Temporal Reference, Aspect and Actionality. Vol. 1: Semantic and Syntactic Perspectives. Torino: Rosenberg & Sellier. Broccias, Cristiano. 2006. Cognitive Approaches to Grammar. In: Kristiansen, Gitte, Michel Archard, René Dirven & Francisco J. Ruiz de Menodza Ibánez (eds.), 81-118. Broccias, Cristiano & Willem B. Hollmann. 2007. Do we need summary and sequential scanning in (Cognitive) grammar? Cognitive Linguistics 18-4, 487-522. Brown, Keith & Jim Miller (eds). 1996. Concise Encyclopedia of Syntactic Theories. Oxford: Pergamon. Brugman, Claudia. 1988. The story of over: polysemy, semantics and the structure of the lexicon. New York: Garland Press.[1981. The story of over. Berkeley, CA: UC Berkeley Master's thesis.] Brugman, Claudia. 2001. Light verbs and Polysemy. Language Sciences 23, 551-78. Burnard, Lou (ed.). 1995. Users Reference Guide for the British National Service. Oxford: Oxford University Computing Services. Bussmann, Hadumod. 1998. Routledge Dictionary Of Language and Linguistics. Transl. & ed. by Gregory P. Trauth & Kerstin Kazzazi. London: Routledge. Butt, Miriam. 2003. The Light Verb Jungle. http: / / ling. unikonstanz.de/ pages/ home/ butt/ harvard-work.pdf2. Bybee, Joan. 2003. Cognitive Processes in Grammaticalization. In: Tomasello, Michael (ed.). The New Psychology of Language. Vol.2. Mahwah, N.J. / London: Lawrence Erlbaum, 145-168. Bybee, Joan. 2006. From usage to grammar: The mind's response to repetition. Language 82 (4), 711- 733. Cambridge International Dictionary of Idioms. 1998. Cambridge: CUP. <?page no="296"?> 284 Cappelle, Bert. 2006. Particle Placement and the case for 'allostructions'. In: Constructions. SV1-7/ 2006.urn: nbn: de: 0009-4-6839. www. constructionsonline. de. Casad, Eugen H. (ed).1996. Cognitive Linguistics in the Redwoods. The Expansion of a New Paradigm in Linguistics. Berlin/ New York: Mouton de Gruyter. Cattell, Ray.1984. Composite Predicates in English. North Ride, New South Wales: Academic Press Australia. Chandler, Daniel. 2006. Semiotics: The Basics. Rev. ed. Milton Park: Routledge. Chomsky, Noam. 1957. Syntactic Structure. 's-Gravenhage: Mouton. Chomsky, Noam. 1965. Aspects of the Theory of Syntax. Cambridge, Mass: MIT Press. Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris Publications. Chomsky, Noam. 1995. The Minimalist Program. Cambridge, Mass: MIT Press. Clark, Eve V. & Clark, Herbert H. 1979. When nouns surface as verbs. Language 55, 767- 811. Coleman, Linda & Paul Kay. 1981. Prototype semantics: the English word LIE. Language 57, 26-44. Conradie, C. Jac. 2003. The iconicity of Afrikaans reduplication. In: Müller, Wolfgang G. & Olga Fischer (eds.), 203-223 Croft, William. 1994. Voice: Beyond Control and Affectedness. In: Fox, Barbara & Paul J. Hopper (eds.). Voice - Form and Function. Amsterdam/ Philadelphia: John Benjamins, 89-118. Croft, William. 2001. Radical Construction Grammar. Oxford: OUP. Croft, William. 2003. Lexical rules vs. constructions: A false dichotomy. In: Cuyckens, H., T. Berg, R. Dirven & K.U. Panther (eds.), 49-68. Croft, William & D. Alan Cruse. 2004. Cognitive Linguistics. Cambridge: CUP. Cuyckens, Hubert, Dominiek Sandra & Sally Rice. 1997. Towards an Empirical Semantics. In: Smieja, Birgit & Meike Tasch. Human Contact through language and linguistics. Frankfurt: Peter Lang, 35-54. Cuyckens, H., T. Berg, R. Dirven & K.U. Panther (eds.) 2003. Motivation in Language. Amsterdam / Philadelphia: John Benjamins, Dabrowska, Eva. 2004. Language, Mind and Brain. Some psychological and Neurological Constraints on Theories of Grammar. Edinburgh: Edinburgh UP. Dahl, Östen. 1981. On the definition of the telic-atelic (bounded-nonbounded) distinction. Syntax and Semantics Vol. 14. Tense and aspect. Philip Tedeschi & Zaenen (eds.), New York: Academic Press, 79-90. <?page no="297"?> 285 Dik, Simon C. 1997a. The Theory of Functional Grammar. Part 1: The Structure of the Clause. Berlin / New York: Mouton de Gruyter. Dik, Simon C. 1997b. The Theory of Functional Grammar. Part 2: Complex and Derived Constructions. Berlin / New York: Mouton de Gruyter. Dirven, René. 2005. Major Strands in Cognitive Linguistics. In: Ruiz de Mendoza Ibanez, Francisco & M. Sandra Pena Cervel (eds)., 17-68. Cognitive Linguistics. Internal Dynamics and Interdisciplinary Interaction Dixon, R.M.W. 1991. A New Approach to English Grammar, on Semantic Principles. Oxford: Clarendon. Downing, Angela & Philip Locke. 1992. A University Course in English Grammar. New York: Prentice Hall. Dowty, David. R. 1979. Word meaning and Montague Grammar: The semantics of verbs and times in Generative Semantics and in Montague's PTQ. Dordrecht: Reidel. Dowty, David R. 1991. Thematic Roles and Argument Selection. Language 67, 547- 619. Egan, Thomas. 2008. Non-finite complementation: a usage-based study of infinitive and -ing clauses in English. Amsterdam: Rodopi. Epstein, Richard. 2001. The definite article, accessibility, and the construction of discourse referents. Cognitive Linguistics 12-4, 333-378. Ernst, Thomas. 1981. Grist for the linguistic mill: idioms and 'extra' adjectives. Journal of Linguistic research 1(3), 51-68. Evans, Vyvyan & Melanie Green. 2006. Cognitive Linguistics. An Introduction. Edinburgh: EUP Fauconnier, Gilles & Mark Turner. 1998a. Conceptual Integration Networks.Cognitive Science 22 (2), 133-187. Fauconnier, Gilles & Mark Turner. 1998b. Blending as a Central Process of Grammar (expanded web.version). http: / / markturner.org/ centralprocess WWW/ centralprocess.pdf. Felser, Claudia & Anja Wanner. 2001. The Syntax of Cognate Objects and Other Unselected Objects. In: Dehé, Nicole and Anja Wanner (eds.) Structural Aspects of Semantically Complex Verbs. Frankfurt: Peter Lang, 105-130. Filip, Hana. 1996. Boundedness in temporal and spatial domains. In: Casad, E. H. (ed.), 655-692. Fillmore, Charles J.1968. The case for case. In Bach, Emmon and Robert T. Harms (eds.).Universals in Linguistic Theory. Vol.1. New York: Holt, Rinehart and Winston. Fillmore, Charles. 1978. "On the organization of the semantic information in the lexicon". In: D. Farkas, W. M. Jacobsen, & K. W. Todrys (eds.) Papers from the Parasession on the Lexicon. 148-173. Chicago: Chicago Linguistic Society. <?page no="298"?> 286 Fillmore, Charles. 1977. Topics in lexical semantics. In: R.W.Cole, ed. Current issues in linguistic theory. Bloomington, London: Indiana University Press, 76-138. Fillmore, Charles.1986. Pragmatically controlled zero anaphora. In: BSL 12, 95-107. Fillmore, Charles, Paul Kay & Mary Catherine O'Connor. 1988. Regularity and Idiomaticity in Grammatical Constructions: The Case of Let Alone. Language 64, 501-538. Fischer, Kerstin & Anatol Stefanowitsch. 2006. Konstruktionsgrammatik: Ein Überblick. In: Fischer, Kerstin & Anatol Stefanowitsch (eds.). Konstruktionsgrammatik. Von der Anwendung zur Theorie. Tübingen: Staufenburg, 19-38. Fischer, Olga & Max Nänny. 1999. Introduction: Iconicity as Creative Force in Language Use. In Nänny, Max & Olga Fischer (eds.) Form miming meaning. Iconicity in language and literature. Amsterdam/ Philadelphia: John Benjamins, xvi-xxxiii. Freidin, Robert.1996. Generative Grammar: Principles and Parameters. In: Brown, Keith and Jim Miller (eds.), 119-147. Fried, Mirjam & Jan-Ola Östman. 2004. A thumbnail sketch of Construction Grammar. In: Fried, Mirjam & Jan-Ola Östmann (eds.). Construction Grammar in a Cross-Language Perspective. Amsterdam/ Philadelphia: John Benjamins, 11-86. Fries, Charles C. & Aileen Traver. 1940. English Word Lists. A Study of Their Adaptability for Instruction. Washington: American Council on Education. Fries, Charles C. 1952. The Structure of English. New York etc: Harcourt, Brace and World. Geeraerts, Dirk.1999. Idealist and empiricist tendencies in cognitive semantics. In: Janssen Theo & Gisela Redeker (eds.), 163-193. Geeraerts, Dirk. 2006. Methodology in Cognitive Linguistics. In: Kristiansen, Gitte, Michel Archard, René Dirven & Francisco J. Ruiz de Menodza Ibánez (eds.), 21-50. Geeraerts, Dirk & Hubert Cuyckens (eds.). 2007. The Oxford Handbook of Cognitive Linguistics. New York, N.Y.: Oxford UP. Gibbs, Raymond W. 1980. Spilling the beans on understanding and memory for idioms in conversation. Memory and Cognition 8, 149-156. Gibbs, Raymond W. 1996. What's cognitive about cognitive linguistics? . In: Eugen Casad (ed.), 27-54. Gibbs, Raymond. 2005. The psychological status of image schemas. In: Hampe, Beate (ed.), 114-135. Gibbs, Raymond W. 2007. Why cognitive linguists should care more about empirical methods. In: Gonzales-Marques, Monica et al. (eds.), 2-18. <?page no="299"?> 287 Gibbs, Raymond W. & Marcus Perlman. 2006. The contested impact of cognitive research on the psycholinguistics of metaphor understanding. In: Kristiansen, Gitte, Michel Archard, René Dirven & Francisco J. Ruiz de Menodza Ibánez (eds.), 211-228. Gillete, Marie & Ernst-Jan C. Wit. 1998. What is Linguistic Redundancy? Technical Report. http: / / galton.uchicago.edu/ ~wit/ redundan.html. Givón, Talmy. 1985. Iconicity, Isomorphism and Non-arbitrary Coding in Syntax. In: Haiman, John (ed.), 187-219. Givón, Talmy 1991a. Isomorphism in the grammatical code: Cognitive and Biological Considerations. Studies in Language 15: 85-114. Givón, Talmy.1991b. Markedness in grammar. Distributional, communicative and cognitive correlates of syntactic structure. Studies in Language 15: 335-370 Givón, Talmy. 1998.The Functional Approach to Grammar. In: Tomasello, Micheal (ed.), 41-66. Givón, Talmy. 2001. Syntax: A Functional-Typological Introduction. Vol.1. Rev. ed. Amsterdam: Benjamins. Glucksberg, Sam. 2001. On understanding figurative language. Oxford: OUP. Goldberg, Adele E.1995. Constructions. A construction grammar approach to argument structure. Chicago: Chicago University Press. Goldberg, Adele E. 1997. Construction Grammar. In: Brown, Keith & Jim Miller (eds.), 68-71. Goldberg, Adele E. 1998. Patterns of Experience. In: Tomasello, Michael (ed.), 203-219. Goldberg, Adele. 2006. Constructions at Work: the nature of generalization in language. Oxford: OUP. Gonzalez-Marquez, Monica, I. Mittelberg, S. Coulson & M. J. Spivey. 2007. Methods in Cognitive Linguistics. Amsterdam: John Benjamins. Goossens, Louis. 1994. Transitivity and the treatment of (non)prototypicality in Functional Grammar. In: Engberg-Pedersen, Ingrid, Lisbeth Falster Jakobsen & Lone Schack Rasmussen (eds.). Function and Expression in Functional Grammar. Berlin/ New York: Mouton de Gruyter, 65-80. Grade ak-Erdelji , Tanja. 2004. Periphrasis in the coding of events. In: Lewandowska-Tomaszczyk, Barbara & Alina Kwiatkowska (eds.), 431- 443. Greenberg, Joseph H. 1966. Language Universals. Janua Linguarum Series Minor, 59. The Hague: Mouton. Gries, Stefan. 2003a. Multifactorial Analysis in Corpus Linguistics. A Study of Particle Placement. NewYork / London: Continuum. Gries, Stefan. 2003b. Towards a corpus-based identification of prototypical instances of constructions. Annual Review of Cognitive Linguistics 1, 1-27. Gries, Stefan Th. 2004. Coll.analysis 3. A program for R for Windows 2.x. <?page no="300"?> 288 Gries, Stefan Th. 2006. Introduction. In: Gries, Stefan Th. & Anatol Stefanowitsch (eds). Corpora in Cognitive Linguistics. Corpus-Based Approaches to Syntax and Lexis. Berlin/ New York: Mouton de Gruyter. Gries, Stefan Th. & Anatol Stefanowitsch. 2004a. Extending collostructional analysis: A corpus-based perspectives on 'alternations'. International Journal of Corpus Linguistics 9, 97-129. Gries, Stefan Th. & Anatol Stefanowitsch. 2004b. Co-varying collexemes in the into-causative. In: Achard, Michel & Suzanne Kemmer (eds.). Language, Culture, and Mind. 225-236. Gries, Stefan Th., Beate Hampe & Doris Schönefeld. 2005. Converging evidence: Bringing together experimental and corpus data on the association of verbs and constructions. Cognitive Linguistics 16-4, 635-676. Grimshaw, Jane & Armin Mester. 1988. Light verbs and theta-marking. Linguistic Inquiry 19, 205-232. Haiman, John. 1980. The Iconicity of grammar: isomorphism and motivation. Language 56, 515-540. Haiman, John. 1983. Iconic and economic motivation. Language 59 (1): 781- 819. Haiman, John (ed.). 1985. Iconicity in Syntax. Amsterdam/ Philadelphia: Benjamins. Haiman, John. 1994. Iconicity. In: Asher, R.E. & J.M.Y. Simpson (eds.), 1629- 1633. Halliday, M.K.A. 1967. Notes on Transitivity and Themes in English, Part1. Journal of Linguistics 3, 37-81. Halliday, M.K.A. 1985. An Introduction to Functional Grammar. London: Edward Arnold. Halliday, M.K.A. 2004. An Introduction to Functional Grammar. Rev. ed. London: Edward Arnold. Hampe, Beate (ed.) 2005. From Perception to Meaning. Image Schemas in Cognitive Linguistics. Berlin / New York: Mouton de Gruyter. Hampe, Beate & Doris Schönefeld. 2003. Creative Syntax. Iconic Principles within the Symbolic. In: Müller, Wolfgang G. & O. Fischer (eds.). From Sign to Signing. Iconicity in language and literature 3. Amsterdam/ Philadelphia: John Benjamins Publishing Company, 243-261. Haspelmath, Martin. 1997. From Time to Space: Temporal Adverbials in the World's Languages. München: Lincom Europa. Hasson, Uri & Rachel Giora. 2007. Experimental methods for studying the mental representation of language. In: Gonzalez-Marques, Monica et al. (eds.), 302-322. Hawkins, John A. 1991. On (in)definite articles: Implicatures and (un)grammaticality prediction. Journal of Linguistics 27, 405-442. <?page no="301"?> 289 Herbst, Thomas. 2004. A valency dictionary of English : a corpus based analysis of the complementation patterns of English verbs, nouns and adjectives. Berlin/ New York: Mouton de Gruyter, 2004 Herskovits, Annette. 1986. Language and cognition. An interdisciplinary study of the prepositions in English. Cambridge: CUP. Hopper, Paul and Sandra Thompson. 1980. Transitivity in Grammar and Discourse. Language 56: 251-99 . Hopper, Paul & Sandra Thompson.1985. The iconicity of the universal categories 'noun' and 'verb'. In: Haiman, John (ed.), 151-183. Horita, Yuko. 1996. English Cognate Object Constructions and their Transitivity. English Linguistics 13, 221-247. Hornstein, Norbert, Jairo Nunes & Kleanthes K. Grohmann. 2006. Understanding Minimalism. Cambridge: CUP. Huddleston, Rodney. 1984. Introduction to the Grammar of English. Cambridge: CUP. Huddleston, Rodney & Geoffrey K. Pullum (2002). The Cambridge Grammar of the English Language. Cambridge: CUP. Hurford, J.R. 1994. Redundancy. In: Asher, R. & J.Simpson (eds), Vol.7, 3481-3482. Ikegami, Yoshihiko. 1999. Transitivisation: the contrast between prepositional phrase and noun phrase as a verb compliment. In: Falkner, Wolfgang & Hans-Jörg Schmid (eds.). Words, lexemes, concepts - approaches to the lexicon. Tübingen: Gunter Narr, 175-187. Iwasaki, Shin-ya. 2007. A cognitive analysis of English cognate objects. In: Constructions 1/ 2007. urn: nbn: de: 0009-4-11744. www.constructionsonline.de. 1-40. Jackendoff, Ray. 1991. Parts and Boundaries. Cognition 41, 9-45. Janda, Laura.2002. The Conceptualization of Events and Their Relationship to Time in Russian. http: / / www.seelrc.org/ glossos/ issues/ 2/ janda.pdf. Janda, Laura. 2004. A metaphor in search of a source domain. The categories of Slavic Aspect. Cognitive Linguistics 15-4, 471-527. Janda, Laura. 2006. Cognitive Linguistics. Glossos 8, http: / / www.seelrc.org/ glossos/ . Janda, Laura. 2007. Semantic Motivation for Aspectual Clusters of Russian Verbs. www.unc.edu/ depts/ slavdept/ lajanda/ janda% 20ohrid % 20 article.doc. Janssen, Theo. 2003. Monosemy vs polysemy. In: Cuyckens, Hubert, René Dirven & John R. Taylor (eds.). Cognitive Approaches to Lexical Semantics. Berlin/ NewYork: Mouton de Gruyter. Janssen, Theo & Gisela Redeker (eds.). 1999. Cognitive Linguistics: Foundations, Scope and Methodology. Berlin/ New York: Mouton de Gruyter. <?page no="302"?> 290 Jespersen, Otto.1954 (1927). A Modern English Grammar On Historical Principles, Part 3, Syntax (Second Volume). London: George Allen and Unwin. Jespersen, Otto. 1954 (1942). A Modern English Grammar on Historical Principles. Part 6, Morphology. London: George Allan & Unwin. Jespersen, Otto. 1968 (1924). The Philosophy of Grammar. London: George Allen and Unwin. Johnson, Mark. 1987. The Body in the Mind. Chicago/ London: The University of Chicago Press. Johnson, Mark. 2005. The philosophical significance of image schemas. In: Hampe, Beate (ed.), 15-33. Johnstone, Barbara et al. 1994.Repetition in Discourse: A Dialogue. In: Johnstone, Barbara (ed.), 1-22. Johnstone, Barbara (ed.). 1994. Repetition in Discourse: Interdisciplinary Perspectives. Vol. I. Norwood. N.J.: Ablex Publishing Corporation. Jones, Michael A. 1988. "Cognate objects and the case filter". Journal of Linguistics 24, 89-110. Kay, Paul & Charles Fillmore. 1999. Grammatical Constructions and Linguistic Generalizations: The What's X doing Y ? Construction. Language 75, 1-33. Kearns, Kate. 2002 (1988). Light Verbs in English. Ms. MIT, Cambridge, MA. http: / / www.ling.canterbury.ac.nz/ documents/ lightverbs.pdf . Keibel, Holger & Cyril Belica. 2007. CCDB: A Corpus-Linguistic Research and Development Workbench. In: Proceedings of Corpus Linguistics 2007, Birmingham. Kemmer, Suzanne. 1992. Grammatical Prototypes and Competing Motivations in a Theory of Linguistic Change. In: Davis, Garry & Gregory Iverson (eds). Explanations in Historical Linguistics. Amsterdam / Philadelphia: John Benjamins, 143-166. Kemmer, Suzanne. 2003. Human Cognition and the Elaboration of Events: Some Universal Conceptual Categories. In: Tomasello, Michael.2003. The New psychology of Language. Vol.2. Mahwah, N.J. / London: Lawrence Erlbaum Associates. Kemmer, Suzanne & Michael Barlow. 2000. Introduction: A Usage-Based Conception of Language. In: Barlow, M. & S. Kemmer (eds.), vii - xxviii. Knowles, John. 1979. Lexemic iteration. Linguistics: An Interdisciplinary Journal of the Language Sciences 17, 641-57. Kuhn, Thomas S. 1993. Metaphor in science. In: Ortony, Andrew (ed.), 533- 542. Kuno, Susumu & Ken-ichi Takami. 2004. Functional Constraints in Grammar. On the unergative-unaccusative distinction. Amsterdam/ Philadelphia: John Benjamins. Kövecses, Zoltan. 1986. Metaphors of Anger, Pride and Love: A Lexical Approach to the Structure of Concepts. Philadelphia: John Benjamins. <?page no="303"?> 291 Kövecses, Zoltan. 2005. Metaphors in Culture. Cambridge: CUP. Kövecses, Zoltan & Peter Szabo.1996. Idioms: A View from Cognitive Semantics.Journal of Applied Linguistics 17-3, 326-355. Krifka, Manfred. 1992. Thematic Relations as Links between Nominal Reference and Temporal Constitution. In: Sag, Ivan & Anna Szabolcsi (eds.). Lexical Matters. CSLI Publications, Chicago University Press, 29- 53. Kristiansen, Gitte, Michel Archard, René Dirven & Francisco J. Ruiz de Menodza Ibánez (eds.).2006.Cognitive Linguistics: Current Applications and Future Perspectives. Berlin/ New York: Mouton de Gruyter. Lakoff, George.1987. Women, Fire, and Dangerous Things: What Our categories reveal about the Mind. Chicago: Chicago University Press. Lakoff, George.1993. The contemporary theory of metaphor. In: Ortony, Andrew (ed.), 202-231. Lakoff, George & Mark Johnson. 1980. Metaphors We Live By. Chicago: The University of Chicago Press. Lakoff, George & Mark Johnson. 1999. Philosophy in the Flesh. New York: Basic Books. Lakoff, George & Mark Turner.1989. More than Cool Reason. A Field Guide to Poetic Metaphor. Chicago/ London: University of Chicago Press. Lambrecht, Knud. 1994. Information structure and sentence form. Cambridge: CUP. Langacker, Ronald.1982a. Space grammar, analysability, and the English passive. Language 58 (1): 22-80. Langacker, Ronald. 1982b. Remarks on English Aspect. In: Hopper, Paul (ed.).Tense-Aspect: Between semantics and pragmatics. Amsterdam/ Philadelphia: John Benjamins, 265-304. Langacker, Ronald. 1987a. Foundations of Cognitive Grammar, Volume I: Theoretical Prerequisites. Stanford: SUP. Langacker, Ronald. 1987b. Nouns and verbs. Language 63, 53-94. Langacker, Ronald.1988a.An overview of cognitive grammar. In: Rudzka- Ostyn, Brygida (ed.), 3-48. Langacker, Ronald. 1988b. A view of linguistic semantics. In: Rudzka- Ostyn, Brygida (ed.), 49-90. Langacker, Ronald .1988c. The Nature of Grammatical Valence. In: Rudzka- Ostyn, Brygida (ed.), 91-125. Langacker, Ronald. 1988d. A usage-based model. In: Rudzka-Ostyn, Brygida (ed.), 127- 61. Langacker, Ronald. 1991a. Foundations of Cognitive Grammar, Volume II: Descriptive Application. Stanford: SUP. Langacker, Ronald. 1991b. Concept, Image, and Symbol: The Cognitive Basis of Grammar. Berlin/ New York: Mouton de Gruyter. <?page no="304"?> 292 Langacker, Ronald. 1998. Conceptualization, Symbolization, and Grammar. In: Tomasello, Michael (ed.), 1-40. Langacker, Ronald. 1999a. Grammar and Conceptualization. Berlin/ New York: Mouton de Gruyter. Langacker, Ronald. 1999b. Assessing the cognitive linguistic enterprise. In: Janssen, Theo & Gisela Redeker (eds.), 13-60. Langacker, Ronald. 2000. A Dynamic Usage-Based Model. In: Barlow, Michael & Suzanne Kemmer (eds.), 1-63. Langacker, Ronald W. 2001. Cognitive linguistics, language pedagogy, and the English present tense. In: Pütz, Martin, Susanne Niemeier & René Dirven (eds.). Applied Cognitive Linguistics I: Theory and Language Acquisition. Berlin/ New York: Mouton de Gruyter, 3-40. Langacker, Ronald W. 2003. Constructional Integration, Grammaticization, and Serial Verb Constructions. Language and Linguistics 4.2, 251-278. Langacker, Ronald W. 2004. The first 40 years. In: In: Lewandowska- Tomaszczyk, Barbara & Alina Kwiatkowska (eds.), 31-41. Langacker, Ronald W. 2005. Construction Grammars: cognitive, radical and less so. In: Ruiz de Mendoza Ibanez, Francisco & M. Sandra Pena Cervel (eds.), 101-162. Langacker, Ronald. 2008. Cognitive Grammar. A basic introduction. Oxford: OUP. Langlotz, Andreas. 2007. Idiomatic Creativity. Amsterdam / Philadelphia: John Benjamins. Larson, Richard K. 1988. On the double object construction. Linguistic Inquiry 19, 335-391. Leino, Jaako. 2005. Frames, profiles and constructions. Two collaborating CGs meet the Finnish Permissive Construction. In: Östman, Jan-Ola & Mirjam Fried (eds.), 89-120. Leitner, Gerhard. 1986. English Grammars - Past, Present, Future. In: Leitner, Gerhard (ed)., 409-431. Leitner, Gerhard (ed.). 1986. English Reference Grammar. Tübingen: Max Niemeyer. Leitner, Gerhard (ed.).1991. English traditional grammars: an international perspective. Amsterdam/ Philadelphia: Benjamins. Lemmens, Maarten.1998. Lexical Perspectives on Transitivity and Ergativity. Causative Constructions in English. Amsterdam/ Philadelphia: John Benjamins. Levin, Beth.1993. English verb classes and alternations: a preliminary investigation. Chicago: Univ. of Chicago Press. Levin, Beth. 2000. Aspect, Lexical Semantic Representation, and Argument Expression. BSL 26, 413-430. Levin, Beth & Tova Rappaport.1988. Lexical Subordination. In: MacLeod, Lynn, Gary Larson & Diane Brentari. Papers from the 24 th Regional Meet- <?page no="305"?> 293 ing of the Chicago Linguistic Society. CLS 24, Part 1. Chicago: Chicago Linguistics Society, 275-289. Levin, Beth & Malka Rappaport Hovav.1995. Unaccusativity. At the Syntax- Lexical Semantic Interface. Cambridge(Mass.) / London: MIT. Levin, Beth & Malka Rappaport Hovav. 2005. Argument realization. Cambridge/ New York etc: Cambridge University Press. Lewandowska-Tomaszczyk, Barbara & Alina Kwiatkowska (eds). 2004. Imagery in Language. Festschrift in Honour of Professor Ronald W. Langacker. Frankfurt etc: Peter Lang. Macfarland, Talke.1995. Cognate Objects and the Argument/ Adjunct Distinction in English. Ph.D. diss., Northwestern University, Evanston, IL. Marsen, Richard. Cain’s Face, and Other Problems: The Legacy of the Earliest English Bible Translations. http: / / www. tyndale.org/ Reformation / 1/ rmarsden.html. Massam, Diane.1990. "Cognate Objects as Thematic objects". In: The Canadian Journal Of Linguistics 35, 161-190. Matlock, Teenie. 2004. The conceptual motivation of fictive motion. In: Radden, Günther & Klaus-Uwe Panther (eds.), 221-248. Matsumoto, Masumi.1996. The Syntax and Semantics of the Cognate Object Construction. English Linguistics (Journal of the English Linguistic Society of Japan) 13, 190-220. McGlone, M.S., S. Glucksberg & C. Cacciari. 1994. Semantic productivity and idiom comprehension. Discourse Processes 17, 167-190. Michaelis, Laura. 2004. Type Shifting in Construction Grammar: An Integrated Approach to Aspectual Coercion. Cognitive Linguistics 15, 1-67. Mittwoch, Anita.1998. Cognate Objects as Reflections of Davidsonian Event Arguments. In: Rothstein, Susan (ed.). Events and Grammar. Dordrecht/ Boston/ London: Kluwer Academic Publishers, 309-332. McEnery, Tony & Andrew Wilson. 1996. Corpus Linguistics. Edinburgh: Edinburgh University Press. Moltmann, Friederike.1989. Nominal and Clausal Event Predicates. In: CLS 25, 300- 314. Moravcsik, Edith A. 1978. Reduplicative Constructions. In: Greenberg, Joseph H. (ed.). Universals of Human Language. Stanford: Stanford University Press. Mukherjee, Joybrato. 2004. Corpus data in a usage-based cognitive grammar. In: Aijmer, Karin & Bengt Altenberg (eds.), 85-100. Müller, Wolfgang G. & Olga Fischer (eds.). 2003. From Sign to Signing. Amsterdam/ Philadelphia: John Benjamins. Newman, John. 1996. Give. A Cognitive Linguistic Study. Berlin / New York: Mouton de Gruyter. Nöth, Winfried. 1985. Handbuch der Semiotik. Stuttgart: J.B. Metzler. <?page no="306"?> 294 Norvig, Peter & George Lakoff. 1989. Taking: A Study in Lexical Network Theory. BSL 15, 195 - 206. Oakes, Michael P. 1998. Statistics for Corpus Linguistics. Edinburgh: Edinburgh UP. Östman, Jan-Ola & Mirjam Fried (eds.).2005. Construction Grammars. Cognitive grounding and theoretical extensions. Amsterdam/ Philadelphia: John Benjamins. Olsen, Mari Broman. 1994. The semantics and pragmatics of lexical aspect features. Studies in the linguistic sciences 24, 361-375. Olsen, Mari Broman. 1997. A Semantic and Pragmatic Model of Lexical and Grammatical Aspect. New York: Garland. Olsen, Mari Broman & Talke Macfarland. 1996. Where's Transitivity? Paper presented at the Seventh Annual Formal Linguistic Society of Mid- America conference, May 17-19, The Ohio State University. http: / / www.umiacs.edu/ users/ molsen. Olsen, Mari Broman & Philip Resnik .1997. "Implicit Object Constructions and the (In)transitivity Continuum. Proceedings of Chicago Linguistic Society - 33., 327-336. Ortony, Andrew (ed.). 1993. Metaphor and Thought. Cambridge etc: CUP. Oxford Collocation Dictionary for Students of English. 2002. Oxford: OUP. Oxford Concise Dictionary of Quotations. 4 th ed. 2001. Oxford: OUP. Paprotté, Wolf. 1988. A Discoursive Perspective on Tense and Aspect in Standard Modern Greek and English. In: Rudzka-Ostyn, Brygida (ed.), 447-505. Peirce, Charles Sanders. 1931-1958. Collected Papers of Charles Sanders Peirce. Vol.1-6. Ed. C.Hartsthorne & P. Weiss. Cambridge, MA: Harvard UP. Pinker, Steven. 1989. Learnability and Cognition. The Acquisition of Argument Structure. Cambridge (Mass.)/ London: MIT Press. Plag, Ingo. 2003. Word formation in English. Cambridge: CUP. Pustejovsky, J. 1992. The Syntax of Event Structure. In: Levin, Beth & Steven Pinker (eds.). Lexical and Conceptual Semantics.. Cambridge, MA.: Blackwell, 47-81. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech & Jan Svartvik. 1985. A Comprehensive Grammar of the English Language. London: Longman. Radden, Günter. 1992. The cognitive approach to natural language. In: Martin Pütz (ed.).Thirty years of Linguistic Evolution: Studies in Honour of René Dirven on the occasion of his Sixtieth Birthday. Philadelphia: John Benjamins, 512-541. Radden, Günther & Klaus-Uwe Panther. 2004. Introduction: Reflections on motivation. In: Radden, Günther & Klaus-Uwe Panther (eds.). Studies in Linguistic Motivation. Berlin / New York: Mouton de Gruyter, 1- 46. Radford, Andrew.2004. Minimalist syntax. Exploring the structure of English. Cambridge: CUP. <?page no="307"?> 295 Rapoport, T.R. 1999. Structure, Aspect, and the Predicate. In: Language 75: 653-77 Reddy, Michael J. 1993[1979]. The conduit metaphor: A case of frame conflict in our language about language. In: Ortony, A. (ed.), 164 - 201. Rice, Sally.1987. Towards a Transitive Prototype: Evidence from some atypical English Passives. BLS 13, 422-434. Rice, Sally.1988. Unlikely Lexical Entries. BLS 14, 202-212. Robins, R.H. 1986. The Evolution of English Grammar Books since the Renaissance. In: Leitner, Gerhard (ed.), 292 - 306. Rohdenburg, Günter. 2003. Aspects of Grammatical Iconicity in English. In: Müller, Wolfgang G. & Olga Fischer (eds.), 263-286. Rosch, Eleanor. 1978. Principles of Categorization. In: Rosch, Eleanor & B.B. Lloyd (eds.). Cognition and Categorization. Hillsdale, NJ: Lawrence Erlbaum Associates, 27-48. Rothstein, Susan. 2004. Structuring Events: A Study in the Semantics of Lexical Aspect. Oxford: Blackwell. Round, Anna. 1998. Grammatical constructions and prototype effects in a group of analytic phrases. Papers from the Thirty-fourth Regional Meetings, Chicago Linguistic Society (CSL 34), 333-345. Rudzka-Ostyn, Brygida (ed.). 1988. Topics in Cognitve Linguistics. Amsterdam/ Philadelphia: John Benjamins Publishing Company. Ruiz de Mendoza Ibanez, Francisco & M. Sandra Pena Cervel (eds). 2005. Cognitive Linguistics. Internal Dynamics and Interdisciplinary Interaction. Berlin / New York: Mouton de Gruyter. Sampson, Geoffrey & Diana McCarthy. 2004. Corpus Linguistics. Readings in a widening discipline. London/ New York: Continuum. Sasse, Hans-Jürgen. 2001. Recent Activity in the Theory of Aspect: Accomplishments, Achievements, or just Non-progressive State? Arbeitspapier Nr.40, Universität Köln: Institut für Sprachwissenschaft. Saussure, Ferdinand de.1959 (1916). Course in General Linguistics. Eds. Bally, Charles & Albert Sechehaye. Transl.: Baskin, Wade. New York: Philosophical Library. Schemann, Hans & Paul Knight. 1997. English-German Dictionary of Idioms. London: Routledge. Schmid, Hans-Jörg. 2000. English abstract nouns as conceptual shells: from corpus to cognition. Berlin: Mouton de Gruyter. Schönefeld, Doris. 2006. Constructions. In: Constructions. SV1-1/ 2006. urn: nbn: de: 0009-4-6839. www.constructions-online.de. Shepherd, Susan. 1994. "Grammaticalization and Discourse Function of Repetition". In: Johnstone, Barbara (ed.), 221-229. Shibatani, M. 1985. Passives and related constructions. Language 61, 821-48. Sinclair, John.1998. The lexical item. In: Edda Weigand (ed.). Contrastive Lexical Semantics. Amsterdam/ Philadelphia: John Benjamins, 1-24. <?page no="308"?> 296 Sinclair, John. 2004. Intuition and annotation - the discussion continues. In: Aijmer, Karin & Bengt Altenberg (eds.), 39-60. Siwierska, Anna.1991. Functional Grammar. New York: Routledge. Spears, Richard. 1988. PONS English and American idioms dictionary. Stuttgart: Klett. Stefanowitsch, Anatol & Stefan Th. Gries. 2003. Collostructions: Investigating the interaction between words and constructions. International Journal of Corpus Linguistics 8.2, 209-43. Stefanowitsch, Anatol & Stefan Th. Gries. 2005. Covarying collexemes. Corpus Linguistics and Linguistic Theory 1, 1-43. Stefanowitsch, Anatol (ed.). 2006a. Corpus-Based Approaches to Metaphor and Metonymy. Berlin/ New York: Mouton de Gruyter. Stefanowitsch, Anatol. 2006b. Distinctive collexeme analysis and diachrony: A comment. Corpus Linguistics and Linguistic Theory 2-2, 257-262. Sweet, Henry. 1968 (1891). A New English Grammar. Part I: Introduction, Phonology, And Accidence. Oxford: Clarendon. Sweetser, Eve. 1990. From Etymology to Pragmatics: Metaphorical and Cultural Aspects of Semantic Structure. Cambridge: CUP. Sweetser, Eve. 1999. Compostionality and blending: semantic composition in a cognitively realistic framework. In: Janssen, Theo & Gisela Redeker (eds.), 129 - 162. Takahashi, Hidemitsu. 2000. English Imperatives and Passives. In: Foolen, Ad & F. van der Leek (eds.). Constructions in Cognitive Linguistics. Selected Papers from the Fifth International Cognitive Linguistic Conference, Amsterdam 1997. Amsterdam/ Philadelphia: John Benjamins, 239-258. Talmy, Leonard.1988. The Relation of Grammar to Cognition. In: Rudzka- Ostyn, Brygida (ed), 165-205. Talmy, Leonard. 1996. Fictive motion in language and 'ception'. In: Bloom, Paul, Mary A. Peterson, Lynn Nadel & Merrill F. Garrett (eds.). Language and Space. Cambridge, MA / London: MIT Press/ Bradford. Talmy, Leonard. 2000. Towards a Cognitive Semantics. Vol. I: Concept Structuring Systems. Cambridge, MA / London: MIT Press. Talmy, Leonard. 2007. Foreword. In: Gonzalez-Marquez, Monica et al. (eds.), xi - xxi. Tannen, Deborah. 1989. Talking Voices: Repetition, Dialogue and Imagery in Conversational Discourse. Cambridge: Cambridge UP. Taylor, John R. 1995. Linguistic categorization: prototypes in linguistic theory. 2 nd ed. Oxford: Oxford Univ. Press Taylor, John R. 1996a. On Running and Jogging. Cognitive Linguistics 7, 21- 34. Taylor, John R. 1996b. Possessives in English. Oxford: Clarendon. Taylor, John R. 1998. Syntactic Constructions as Prototype Categories. In: Tomasello, Michael (ed.), 177-202. <?page no="309"?> 297 Taylor, John R. 2002. Cognitive Grammar. Oxford: OUP. Taylor, John R. 2003. Meaning and Context. In: Cuyckens, H., T. Berg, R. Dirven & K.U. Panther (eds.), 27-48. Tenny, Carol. 1987. Grammaticalizing aspect and affectedness, MIT: Ph.D. dissertation. Tenny, Carol L.1994. Aspectual Roles and the syntax-semantics interface. Dordrecht: Kluwer. Tenny, Carol L. & James Pustejovski (eds.).2000. Events as Grammatical Objects. Stanford: CLSI. Thompson, Sandra. 1992. Functional Grammar. In: William Bright (ed.). International Encyclopedia of Linguistics, 37-39. New York etc: OUP. Thompson, Sandra and Y. Koide.1987. Iconicity and indirect objects in English. Journal of Pragmatics 11: 399-406. Tobin, Yishai .1993. Aspect in the English verb. Process and Result in Language. London / New York: Longman. Tomasello, Michael (ed.).1998. The New Psychology of Language.Vol.1. Mahwah, NJ / London: Lawrence Erlbaum Associates. Tomasello, Michael (ed.).2003. The New Psychology of Language.Vol.2. Mahwah, NJ / London: Lawrence Erlbaum Associates. Turner, Mark. 1991. Reading Minds: The Study of English in the Age of Cognitive Science. Princeton (NJ): Princeton UP. Tyler, Andrea & Vyvyan Evans. 2004. The semantics of English prepositions. Spatial scenes, embodied meaning and cognition. Cambridge: CUP. Turner, Mark.1996. The Literary Mind. New York etc.: OUP. Ungerer, Friedrich & Hans-Jörg Schmid. 2006. An Introduction to Cognitive Linguistics. 2 nd rev. ed. Reading, Mass.: Addison Wesley. Vendler, Zeno.1967. Linguistics in philosophy. Ithaca, N.Y.: Cornell University Press. Viola, Patrizia. 2004. Embodiment at the crossroads between cognition and semiosis. Recherches en Communication 19: 199-234. Visser, F. Th. 1963. An Historical Syntax of the English Language. Part I: Syntactical Units with one Verb. Leiden: Brill. Wierzbicka, Anna. 1982. Why can you have a drink when you can't have an eat? In: Language 58, 753-799. Wierzbicka, Anna. 1988. The Semantics of Grammar. Amsterdam/ Philadelphia: John Benjamins. Zandvoort, Reinard Willem. 1972 (1945). A Handbook of English Grammar. 6 th ed. London: Longman. Zubizarreta, Maria Luisa. 1987. Levels of Representation in the Lexicon and in the Syntax. Dordrecht: Forris Publications. Zwicky, Arnold. 1971. In a manner of speaking. Linguistic Inquiry 2, 223- 232. <?page no="310"?> 298 14. Appendix Table 14.1 Overall frequencies of COC in the BNC Verb CO Total COC Type of CO semantic class of verb live life 699 EV/ R 1 existence sing song 466 R 1 performance tell tale 401 R 1 communication smile smile 202 EV/ R 1 non-verbal expression sow seed 198 A putting produce product 141 R 2 creation give gift 128 A change of possession build building 100 R 2 creation die death 87 EV/ R 1 disappearance think thought 78 R 1 mental activity see sight 72 A perception do deed 67 EV/ R 1 execution name name 45 A verbal expression. dream dream 45 EV/ R 1 mental activity weave web 26 R 2 creation smell smell 24 A perception feel feeling 23 A perception drink drink 23 A ingesting feed food 21 A ingesting fight fight 19 EV/ R 1 social-interaction grin grin 18 EV/ R 1 non-verbal expression plant plant 17 A putting sleep sleep 14 EV/ R 1 bodily process dance dance 12 EV/ R 1 performance laugh laugh 10 EV/ R 1 non-verbal expression light light 8 A preparing sigh sigh 8 EV/ R 1 non-verbal expression publish publication 7 A "release" tie tie 7 A attaching pray prayer 7 EV/ R 1 communication paint painting 7 R2 creation (immage) charge charge 6 A measure (bill) fly flight 6 EV/ R 1 manner of motion <?page no="311"?> 299 record record 6 R2 creation farm farm 6 A edit edition 5 A (creation) jump jump 5 EV/ R 1 manner of motion talk talk 5 EV/ R 1 communication shoot shoot 4 EV/ R 1 throw-verbs draw drawing 4 R 2 creation (image) drop drop 4 A putting breathe breath 4 EV/ R 1 bodily process execute execution 3 EV/ R 1 (killing) kick kick 3 EV/ R 1 throw lose loss 3 EV/ R 1 change of state rain rain 3 EV/ R 1 weather walk walk 3 EV/ R 1 manner of motion describe description 3 R 1 characterize attach attachment 2 A attaching find finding 2 A obtain invest investment 2 A change of possession load load 2 A putting mix mix 2 A combining nail nail 2 A attaching taste taste 2 A perception operate operation 2 EV/ R 1 change of state arrange arrangement 2 EV/ R 1 creation bite bite 2 EV/ R 1 bodily impact bow bow 2 EV/ R 1 gestures cough cough 2 EV/ R 1 bodily process cry cry 2 EV/ R 1 manner of speaking decide decision 2 EV/ R 1 frown frown 2 EV/ R 1 non-verbal expression kiss kiss 2 EV/ R 1 social-interaction nod nod 2 EV/ R 1 gestures offer offer 2 EV/ R 1 change of possession present present 2 EV/ R 1 provide run run 2 EV/ R 1 manner of motion scream scream 2 EV/ R 1 manner of speaking serve service 2 EV/ R 1 change of possession yawn yawn 2 EV/ R 1 bodily process copy copy 2 R 2 creation model model 2 R 2 creation <?page no="312"?> 300 button button 1 A attaching dress dress 1 A dress earn earn 1 A obtain experience experience 1 A perception feature feature 1 A fish fish 1 A change of possession herd herd 1 A group existence love love 1 A psych verbs measure measure 1 A measure (register) pursue pursuit 1 A recruit recruit 1 A repeat repetition 1 A share share 1 A separating smoke smoke 1 A ingest spray spray 1 A putting switch switch 1 A transcribe transcribe 1 A answer answer 1 R 1 communication say say 1 R 1 communication shape shape 1 R 2 creation (image) purr purr 1 EV/ R 1 animal sounds attempt attempt 1 EV/ R 1 chatter chatter 1 EV/ R 1 communication (manner of speaking) cut cut 1 EV/ R 1 cut pace pace 1 EV/ R 1 manner of motion stammer stammer 1 EV/ R 1 communication (manner of speaking) step step 1 EV/ R 1 manner of motion stitch stitch 1 EV/ R 1 attaching strike strike 1 EV/ R 1 ? ? ? work work 1 EV/ R 1 creation yell yell 1 EV/ R 1 communication (manner of speaking) breed breeding 1 R 2 (creation) develop development 1 R 2 creation plan plan 1 R 2 creation plot plot 1 R 2 creation shit shit 1 R 2 bodily process 3139 <?page no="313"?> 301 Table 14.2 Collexeme analysis / Collocational strength verbs word freq in BNC obs. freq in COC exp.freq in COC relation coll. strength 1 live 27501 699 8.46 attraction Inf 2 sing 5700 466 1.75 attraction Inf 3 tell 71351 401 21.94 attraction Inf 4 sow 547 198 0.17 attraction Inf 5 smile 8720 202 2.68 attraction 297.9094632 6 produce 26880 141 8.27 attraction 118.8687254 7 build 19670 100 6.05 attraction 83.1064042 8 dream 1331 45 0.41 attraction 74.1631838 9 die 19858 87 6.11 attraction 67.0985713 10 name 4797 45 1.48 attraction 49.3171758 11 weave 816 26 0.25 attraction 42.5363268 12 give 116135 128 35.72 attraction 32.8022754 13 smell 1732 24 0.53 attraction 30.6824942 14 grin 1834 18 0.56 attraction 20.5684481 15 drink 4750 23 1.46 attraction 19.2846741 16 plant 2012 17 0.62 attraction 18.3931186 17 feed 5257 21 1.62 attraction 16.0352252 18 dance 2163 12 0.67 attraction 11.0907643 19 fight 8624 19 2.65 attraction 10.1506220 20 farm 300 6 0.09 attraction 9.1245645 21 sleep 5239 14 1.61 attraction 8.7065932 22 sigh 2227 8 0.68 attraction 6.1915389 23 think 141160 78 43.41 attraction 5.9205229 24 light 2608 8 0.80 attraction 5.6867855 25 pray 2321 7 0.71 attraction 5.0026088 26 tie 3021 7 0.93 attraction 4.2810390 27 edit 1382 5 0.43 attraction 4.0938225 28 paint 3493 7 1.07 attraction 3.8935361 29 laugh 8484 10 2.61 attraction 3.4179565 30 yawn 297 2 0.09 attraction 2.4074298 31 nail 323 2 0.10 attraction 2.3367153 32 cough 329 2 0.10 attraction 2.3212335 33 charge 4979 6 1.53 attraction 2.3103665 34 rain 1224 3 0.38 attraction 2.1735678 35 jump 4273 5 1.31 attraction 1.9547833 36 execute 1622 3 0.50 attraction 1.8454383 <?page no="314"?> 302 37 breathe 3111 4 0.96 attraction 1.7846759 38 shit 54 1 0.02 attraction 1.7832173 39 kick 1992 3 0.61 attraction 1.6135387 40 model 819 2 0.25 attraction 1.5712308 41 fly 7291 6 2.24 attraction 1.5700743 42 see 184991 72 56.89 attraction 1.5425675 43 bow 948 2 0.29 attraction 1.4553275 44 record 7806 6 2.40 attraction 1.4481202 45 herd 126 1 0.04 attraction 1.4200261 46 stammer 161 1 0.05 attraction 1.3158906 47 publish 10625 7 3.27 attraction 1.3120879 48 button 168 1 0.05 attraction 1.2978707 49 taste 1198 2 0.37 attraction 1.2735629 50 purr 181 1 0.06 attraction 1.2663617 51 transcribe 211 1 0.06 attraction 1.2017410 52 load 1360 2 0.42 attraction 1.1772911 53 chatter 228 1 0.07 attraction 1.1692110 54 copy 1464 2 0.45 attraction 1.1221773 55 stitch 273 1 0.08 attraction 1.0939495 56 frown 1686 2 0.52 attraction 1.0184445 57 feeling 55992 23 17.22 attraction 0.9809354 58 shoot 6017 4 1.85 attraction 0.9322547 59 scream 2002 2 0.62 attraction 0.8959434 60 pace 487 1 0.15 attraction 0.8565948 61 spray 513 1 0.16 attraction 0.8356981 62 bite 2356 2 0.72 attraction 0.7841680 63 mix 2670 2 0.82 attraction 0.7015413 64 plot 813 1 0.25 attraction 0.6550777 65 invest 3020 2 0.93 attraction 0.6232968 66 kiss 3066 2 0.94 attraction 0.6139220 67 fish 931 1 0.29 attraction 0.6037466 68 drop 8621 4 2.65 attraction 0.5603724 69 yell 1072 1 0.33 attraction 0.5514346 70 breed 1291 1 0.40 attraction 0.4844441 71 smoke 1320 1 0.41 attraction 0.4766037 72 shape 1421 1 0.44 attraction 0.4508559 73 recruit 1691 1 0.52 attraction 0.3919046 74 attach 4621 2 1.42 attraction 0.3814049 75 cry 5231 2 1.61 attraction 0.3205774 76 nod 5277 2 1.62 attraction 0.3164678 <?page no="315"?> 303 77 arrange 6248 2 1.92 attraction 0.2422409 78 feature 2981 1 0.92 attraction 0.2216155 79 say 315507 1 97.04 repulsion 40.8088888 80 do 534968 67 164.53 repulsion 18.0000159 81 find 90104 2 27.71 repulsion 9.4672782 82 work 54331 1 16.71 repulsion 6.0269432 83 run 34629 2 10.65 repulsion 2.7963248 84 develop 20714 1 6.37 repulsion 1.9016093 85 offer 25032 2 7.70 repulsion 1.7625744 86 decide 21799 2 6.70 repulsion 1.4338407 87 cut 16013 1 4.92 repulsion 1.3674997 88 lose 23656 3 7.28 repulsion 1.1659835 89 describe 22066 3 6.79 repulsion 1.0300209 90 love 12137 1 3.73 repulsion 0.9467163 91 talk 26573 5 8.17 repulsion 0.7556484 92 serve 14253 2 4.38 repulsion 0.7284872 93 walk 17885 3 5.50 repulsion 0.6960773 94 plan 9406 1 2.89 repulsion 0.6664612 95 answer 8649 1 2.66 repulsion 0.5920590 96 share 8347 1 2.57 repulsion 0.5628574 97 present 12196 2 3.75 repulsion 0.5579985 98 draw 18064 4 5.56 repulsion 0.4575971 99 attempt 6774 1 2.08 repulsion 0.4159274 100 operate 9358 2 2.88 repulsion 0.3458986 101 strike 5767 1 1.77 repulsion 0.3273342 102 repeat 5700 1 1.75 repulsion 0.3216200 103 experience 5376 1 1.65 repulsion 0.2943380 104 measure 5319 1 1.64 repulsion 0.2896006 105 earn 5014 1 1.54 repulsion 0.2645850 106 step 4466 1 1.37 repulsion 0.2211596 107 dress 4382 1 1.35 repulsion 0.2146897 108 pursue 4283 1 1.32 repulsion 0.2071321 109 switch 3611 1 1.11 repulsion 0.1579257 <?page no="316"?> 304 Table 14.3 Distinctive collexeme analysis: live+ X adj life vs. live X adv 1 lexemes obs. freq. in COC obs. freq. in VAC exp. freq. in COC exp. freq. in VAC pref. occur. in coll. strength ALONE 6 350 84.15 271.85 VAC 38.4966093 LONGER 3 141 34.04 109.96 VAC 13.5268055 HAPPY 6 117 29.07 93.93 VAC 8.2249921 LOCAL 0 40 9.46 30.54 VAC 4.7577623 LONG 8 90 23.16 74.84 VAC 4.4189712 COMFORTABLE 3 35 8.98 29.02 VAC 1.9617607 DANGEROUS 1 21 5.20 16.80 VAC 1.6982249 CHEAP 0 11 2.60 8.40 VAC 1.2933895 ROUGH 0 8 1.89 6.11 VAC 0.9395378 HARMONIOUS 0 5 1.18 3.82 VAC 0.5865201 MAGICAL 0 5 1.18 3.82 VAC 0.5865201 WILD 0 5 1.18 3.82 VAC 0.5865201 COMMUNAL 0 4 0.95 3.05 VAC 0.4690322 SUCCESSFUL 0 4 0.95 3.05 VAC 0.4690322 VICARIOUS 0 4 0.95 3.05 VAC 0.4690322 COLLECTIVE 0 3 0.71 2.29 VAC 0.3516364 ETERNAL 0 3 0.71 2.29 VAC 0.3516364 FAST 0 3 0.71 2.29 VAC 0.3516364 MERRY 0 3 0.71 2.29 VAC 0.3516364 POSITIVE 0 3 0.71 2.29 VAC 0.3516364 UNEASY 0 3 0.71 2.29 VAC 0.3516364 FRUGAL 1 6 1.65 5.35 VAC 0.3195539 OPEN 1 5 1.42 4.58 VAC 0.2468860 ANONYMOUS 0 2 0.47 1.53 VAC 0.2343325 APPROPRIATE 0 2 0.47 1.53 VAC 0.2343325 INTENSE 0 2 0.47 1.53 VAC 0.2343325 INTIMATE 0 2 0.47 1.53 VAC 0.2343325 JOYFUL 0 2 0.47 1.53 VAC 0.2343325 METAPHYSICAL 0 2 0.47 1.53 VAC 0.2343325 RESPECTABLE 0 2 0.47 1.53 VAC 0.2343325 RIGHTEOUS 0 2 0.47 1.53 VAC 0.2343325 SANE 0 2 0.47 1.53 VAC 0.2343325 1 The table lists all those lexemes which occur at least twice in either of the two constructions. <?page no="317"?> 305 SELF - SUFFCIENT 0 2 0.47 1.53 VAC 0.2343325 NORMAL 34 1 8.27 26.73 COC 20.4155477 FULL 14 5 4.49 14.51 COC 5.3139193 PRIVATE 5 0 1.18 3.82 COC 3.1415916 DIFFERENT 6 2 1.89 6.11 COC 2.5188846 EVIL 4 0 0.95 3.05 COC 2.5113447 ORDINARY 4 0 0.95 3.05 COC 2.5113447 HEALTHY 5 1 1.42 4.58 COC 2.4575394 SOLITARY 5 1 1.42 4.58 COC 2.4575394 DECENT 4 1 1.18 3.82 COC 1.9025148 EXCITING 3 0 0.71 2.29 COC 1.8820655 MARRIED 3 0 0.71 2.29 COC 1.8820655 MISERABLE 3 0 0.71 2.29 COC 1.8820655 RELIGIOUS 3 0 0.71 2.29 COC 1.8820655 AUTHENTIC 2 0 0.47 1.53 COC 1.2537506 CONTENTED 2 0 0.47 1.53 COC 1.2537506 IMPOVERISHED 2 0 0.47 1.53 COC 1.2537506 ISOLATED 2 0 0.47 1.53 COC 1.2537506 NOMADIC 2 0 0.47 1.53 COC 1.2537506 POLITICAL 2 0 0.47 1.53 COC 1.2537506 RURAL 2 0 0.47 1.53 COC 1.2537506 SECRET 2 0 0.47 1.53 COC 1.2537506 SPARTAN 2 0 0.47 1.53 COC 1.2537506 STRANGE 2 0 0.47 1.53 COC 1.2537506 QUIET 14 30 10.40 33.60 COC 0.8757287 HIGH 2 1 0.71 2.29 COC 0.8507448 PURE 2 1 0.71 2.29 COC 0.8507448 USEFUL 2 1 0.71 2.29 COC 0.8507448 SEPARATE 8 15 5.44 17.56 COC 0.8140916 SIMPLE 5 9 3.31 10.69 COC 0.6600404 INDEPENDENT 14 37 12.06 38.94 COC 0.5132911 FREE 2 3 1.18 3.82 COC 0.4705319 GOOD 10 30 9.46 30.54 COC 0.3182514 EASY 1 2 0.71 2.29 COC 0.2557099 PROPER 1 2 0.71 2.29 COC 0.2557099 LAVISH 1 3 0.95 3.05 COC 0.1801930 UNHAPPY 1 3 0.95 3.05 COC 0.1801930 <?page no="318"?> 306 Table 14.4 Distinctive collexeme analysis: die + X adj death vs. die X adv lexemes obs. freq. in COC obs. freq. in VAC exp. freq. in COC exp. freq. in VAC pref. occur. in coll. strength SUDDEN 1 146 13.86 133.14 VAC 5.72745648 QUICK 0 38 3.58 34.42 VAC 1.67589088 TRAGIC 0 24 2.26 21.74 VAC 1.04823538 ALONE 0 23 2.17 20.83 VAC 1.00386884 UNEXPECTED 0 17 1.60 15.40 VAC 0.73894932 IMMEDIATE 0 14 1.32 12.68 VAC 0.60730470 PREMATURE 1 21 2.07 19.93 VAC 0.43329919 NEEDLESS 0 10 0.94 9.06 VAC 0.43261390 EARLY 2 31 3.11 29.89 VAC 0.41794326 EASILY 0 6 0.57 5.43 VAC 0.25886831 MYSTERIOUS 0 6 0.57 5.43 VAC 0.25886831 PEACEFUL 0 6 0.57 5.43 VAC 0.25886831 COMPLETE 0 5 0.47 4.53 VAC 0.21557841 UNFORTUNATE 0 5 0.47 4.53 VAC 0.21557841 FAST 0 4 0.38 3.62 VAC 0.17234679 UNNECCESSARY 0 4 0.38 3.62 VAC 0.17234679 BEAUTIFUL 0 3 0.28 2.72 VAC 0.12917330 PROPER 0 3 0.28 2.72 VAC 0.12917330 RAPID 0 3 0.28 2.72 VAC 0.12917330 WILLING 0 3 0.28 2.72 VAC 0.12917330 NATURAL 16 9 2.36 22.64 COC 11.0813845 HORRIBLE 4 1 0.47 4.53 COC 3.46835183 VIOLENT 5 6 1.04 9.96 COC 2.71681021 TRADITIONAL 2 0 0.19 1.81 COC 2.05663114 UGLY 2 0 0.19 1.81 COC 2.05663114 UNNATURAL 2 0 0.19 1.81 COC 2.05663114 ROMANTIC 2 1 0.28 2.72 COC 1.60698431 CASUAL 1 0 0.09 0.91 COC 1.02563794 CRUEL 1 0 0.09 0.91 COC 1.02563794 IGNOBLE 1 0 0.09 0.91 COC 1.02563794 NASTY 1 0 0.09 0.91 COC 1.02563794 POLITICAL 1 0 0.09 0.91 COC 1.02563794 PREDICTABLE 1 0 0.09 0.91 COC 1.02563794 STRANGE 1 0 0.09 0.91 COC 1.02563794 <?page no="319"?> 307 UNHEROIC 1 0 0.09 0.91 COC 1.02563794 UNJUST 1 0 0.09 0.91 COC 1.02563794 UNSURPRISING 1 0 0.09 0.91 COC 1.02563794 SAD 2 4 0.57 5.43 COC 0.98919181 PAINFUL 2 6 0.75 7.25 COC 0.77116909 HONOURABLE 1 1 0.19 1.81 COC 0.74531281 NOBLE 1 1 0.19 1.81 COC 0.74531281 TERRIBLE 1 1 0.19 1.81 COC 0.74531281 QUIET 2 16 1.70 16.30 COC 0.28500308 Table 14.5 Distinctive collexeme analysis: smile + X adj smile vs. smile + X adv2 Lexemes obs. freq. in COC obs. freq. in VAC exp. freq. in COC exp. freq. in VAC pref. occur. in coll. strength RUEFUL 0 53 5.40 47.60 VAC 2.50194650 POLITE 0 48 4.89 43.11 VAC 2.26326219 SUDDEN 0 47 4.79 42.21 VAC 2.21559260 FAINT 1 63 6.52 57.48 VAC 2.10042181 SWEET 1 58 6.01 52.99 VAC 1.89245451 BROAD 2 73 7.64 67.36 VAC 1.87758339 WEAK 0 38 3.87 34.13 VAC 1.78757047 SLIGHT 0 37 3.77 33.23 VAC 1.74012366 WARM 1 44 4.58 40.42 VAC 1.32702814 NERVOUS 0 24 2.44 21.56 VAC 1.12532447 SAD 1 38 3.97 35.03 VAC 1.09432615 BRIEF 0 23 2.34 20.66 VAC 1.07818607 ENCOURAGING 0 22 2.24 19.76 VAC 1.03106955 THIN 5 80 8.66 76.34 VAC 0.92045453 INWARD 0 19 1.93 17.07 VAC 0.88985105 GENTLE 2 44 4.68 41.32 VAC 0.86328696 BENIGN 0 18 1.83 16.17 VAC 0.84282183 KINDL 0 16 1.63 14.37 VAC 0.74882870 CHEERFUL 0 15 1.53 13.47 VAC 0.70186475 COLD 0 15 1.53 13.47 VAC 0.70186475 2 Note that due to the great number of adverbs which occur in the smile + X adv - construction as recorded in the BNC I only include those in the table which occur at least twice in this pattern. <?page no="320"?> 308 EASY 0 15 1.53 13.47 VAC 0.70186475 INDULGENT 0 15 1.53 13.47 VAC 0.70186475 MERE 0 14 1.43 12.57 VAC 0.65492251 HAPPY 1 26 2.75 24.25 VAC 0.65407687 WICKED 0 12 1.22 10.78 VAC 0.56110310 UNCERTAIN 1 22 2.34 20.66 VAC 0.51760385 RADIANT 0 11 1.12 9.88 VAC 0.51422588 SHEEPISH 0 11 1.12 9.88 VAC 0.51422588 SYMPATHETIC 0 11 1.12 9.88 VAC 0.51422588 WISTFUL 0 11 1.12 9.88 VAC 0.51422588 SHY 2 32 3.46 30.54 VAC 0.50699901 GRATEFUL 0 10 1.02 8.98 VAC 0.46737030 BLEAK 0 9 0.92 8.08 VAC 0.42053633 SOUR 0 9 0.92 8.08 VAC 0.42053633 WRY 5 55 6.11 53.89 VAC 0.38032476 CONTENTED 0 8 0.81 7.19 VAC 0.37372396 FOND 0 8 0.81 7.19 VAC 0.37372396 INNOCENT 0 8 0.81 7.19 VAC 0.37372396 SHAKY 0 8 0.81 7.19 VAC 0.37372396 SOFT 0 8 0.81 7.19 VAC 0.37372396 UNPLEASANT 0 8 0.81 7.19 VAC 0.37372396 APOLOGETIC 1 17 1.83 16.17 VAC 0.35743472 PLEASANT 2 25 2.75 24.25 VAC 0.32708271 REASSURING 2 25 2.75 24.25 VAC 0.32708271 AMIABLE 0 7 0.71 6.29 VAC 0.32693316 APPROVING 0 7 0.71 6.29 VAC 0.32693316 TREMULOUS 0 7 0.71 6.29 VAC 0.32693316 BRAVE 1 15 1.63 14.37 VAC 0.29746713 LAZY 1 15 1.63 14.37 VAC 0.29746713 WAN 4 41 4.58 40.42 VAC 0.29263281 BEATIFIC 0 6 0.61 5.39 VAC 0.28016392 BLAND 0 6 0.61 5.39 VAC 0.28016392 CALM 0 6 0.61 5.39 VAC 0.28016392 CONFIDENT 0 6 0.61 5.39 VAC 0.28016392 DAZZLING 0 6 0.61 5.39 VAC 0.28016392 HOPEFUL 0 6 0.61 5.39 VAC 0.28016392 INGRATIATING 0 6 0.61 5.39 VAC 0.28016392 SERENE 0 6 0.61 5.39 VAC 0.28016392 TOLERANT 0 6 0.61 5.39 VAC 0.28016392 WIDE 2 22 2.44 21.56 VAC 0.25831597 <?page no="321"?> 309 BRILLIANT 0 5 0.51 4.49 VAC 0.23341621 BRISK 0 5 0.51 4.49 VAC 0.23341621 FOOLISH 0 5 0.51 4.49 VAC 0.23341621 GENIAL 0 5 0.51 4.49 VAC 0.23341621 GRACIOUS 0 5 0.51 4.49 VAC 0.23341621 INVOLUNTARY 0 5 0.51 4.49 VAC 0.23341621 MISCHIEVOUS 0 5 0.51 4.49 VAC 0.23341621 MOCKING 0 5 0.51 4.49 VAC 0.23341621 PROUD 0 5 0.51 4.49 VAC 0.23341621 RELUCTANT 0 5 0.51 4.49 VAC 0.23341621 REMINISCENT 0 5 0.51 4.49 VAC 0.23341621 THOUGHTFUL 0 5 0.51 4.49 VAC 0.23341621 UNEASY 0 5 0.51 4.49 VAC 0.23341621 WARY 0 5 0.51 4.49 VAC 0.23341621 GRIM 5 45 5.09 44.91 VAC 0.22239984 FALSE 1 11 1.22 10.78 VAC 0.18676642 TRIUMPHANT 1 11 1.22 10.78 VAC 0.18676642 ABSENT 0 4 0.41 3.59 VAC 0.18669002 AFFABLE 0 4 0.41 3.59 VAC 0.18669002 AUTOMATIC 0 4 0.41 3.59 VAC 0.18669002 BLANK 0 4 0.41 3.59 VAC 0.18669002 CONSPIRATO- RY 0 4 0.41 3.59 VAC 0.18669002 CRUEL 0 4 0.41 3.59 VAC 0.18669002 LOVING 0 4 0.41 3.59 VAC 0.18669002 PATRONISING 0 4 0.41 3.59 VAC 0.18669002 SCORNFUL 0 4 0.41 3.59 VAC 0.18669002 SHARP 0 4 0.41 3.59 VAC 0.18669002 QUICK 2 18 2.04 17.96 VAC 0.17615017 TIGHT 2 18 2.04 17.96 VAC 0.17615017 COOL 1 10 1.12 9.88 VAC 0.16143297 ACID 0 3 0.31 2.69 VAC 0.13998532 CAUTIOUS 0 3 0.31 2.69 VAC 0.13998532 COMPLACENT 0 3 0.31 2.69 VAC 0.13998532 CURIOUS 0 3 0.31 2.69 VAC 0.13998532 DELIGHTED 0 3 0.31 2.69 VAC 0.13998532 DEMURE 0 3 0.31 2.69 VAC 0.13998532 EVIL 0 3 0.31 2.69 VAC 0.13998532 FIERCE 0 3 0.31 2.69 VAC 0.13998532 HUMOURLESS 0 3 0.31 2.69 VAC 0.13998532 <?page no="322"?> 310 ICY 0 3 0.31 2.69 VAC 0.13998532 INANE 0 3 0.31 2.69 VAC 0.13998532 MODEST 0 3 0.31 2.69 VAC 0.13998532 PAINFUL 0 3 0.31 2.69 VAC 0.13998532 PLACATING 0 3 0.31 2.69 VAC 0.13998532 SELDOM 0 3 0.31 2.69 VAC 0.13998532 STIFF 0 3 0.31 2.69 VAC 0.13998532 TACTFUL 0 3 0.31 2.69 VAC 0.13998532 UNDERSTAN- DING 0 3 0.31 2.69 VAC 0.13998532 SLOWLY 16 22 3.87 34.13 COC 6.63264257 SMALL 7 1 0.81 7.19 COC 6.11660335 PREDATORY 4 0 0.41 3.59 COC 3.97859385 BEST 3 0 0.31 2.69 COC 2.98142314 BRIGHT 3 0 0.31 2.69 COC 2.98142314 HARSH 3 0 0.31 2.69 COC 2.98142314 TWISTED 3 0 0.31 2.69 COC 2.98142314 LOP - SIDED 5 5 1.02 8.98 COC 2.76384567 AMUSED 3 1 0.41 3.59 COC 2.41345163 FRIENDLY 2 0 0.20 1.80 COC 1.98593924 INNER 2 0 0.20 1.80 COC 1.98593924 LOVELY 2 0 0.20 1.80 COC 1.98593924 MISHAPPEN 2 0 0.20 1.80 COC 1.98593924 MOCK - STERN 2 0 0.20 1.80 COC 1.98593924 REAL 2 0 0.20 1.80 COC 1.98593924 SUPERIOR 2 0 0.20 1.80 COC 1.98593924 WRECKED 2 0 0.20 1.80 COC 1.98593924 DISARMING 2 1 0.31 2.69 COC 1.53910559 FLEETING 3 5 0.81 7.19 COC 1.40183186 SECRETE 3 5 0.81 7.19 COC 1.40183186 INFURIATING 2 2 0.41 3.59 COC 1.26808995 SEDUCTIVE 2 2 0.41 3.59 COC 1.26808995 BITTER 4 11 1.53 13.47 COC 1.23559965 CROOKED 4 11 1.53 13.47 COC 1.23559965 SMUG 4 12 1.63 14.37 COC 1.14545351 STRANGE 2 3 0.51 4.49 COC 1.07597832 CHARMING 3 8 1.12 9.88 COC 1.03137897 ENIGMATICAL 3 10 1.32 11.68 COC 0.85736855 KNOWING 2 6 0.81 7.19 COC 0.71633038 MIRTHLESS 2 6 0.81 7.19 COC 0.71633038 <?page no="323"?> 311 RARE 2 8 1.02 8.98 COC 0.56716315 BOYISH 1 2 0.31 2.69 COC 0.55981605 CHEERY 1 2 0.31 2.69 COC 0.55981605 SARCASTIC 1 2 0.31 2.69 COC 0.55981605 SIDEWAYS 1 2 0.31 2.69 COC 0.55981605 TIRED 1 2 0.31 2.69 COC 0.55981605 TOOTHLESS 1 2 0.31 2.69 COC 0.55981605 ARCHY 2 9 1.12 9.88 COC 0.50801115 QUIET 2 9 1.12 9.88 COC 0.50801115 BENEVOLENT 1 3 0.41 3.59 COC 0.45666951 IRONICAL 1 3 0.41 3.59 COC 0.45666951 NATURAL 1 3 0.41 3.59 COC 0.45666951 OPEN 1 3 0.41 3.59 COC 0.45666951 READY 1 3 0.41 3.59 COC 0.45666951 SECERETIVE 1 3 0.41 3.59 COC 0.45666951 VAGUE 2 11 1.32 11.68 COC 0.41137056 IMPISH 1 4 0.51 4.49 COC 0.38114676 APPRECIATIVE 1 5 0.61 5.39 COC 0.32294942 PATIENT 1 5 0.61 5.39 COC 0.32294942 DRY 1 7 0.81 7.19 COC 0.23877703 WINNING 1 7 0.81 7.19 COC 0.23877703 CYNICAL 1 8 0.92 8.08 COC 0.20741234 SARDONIC 1 8 0.92 8.08 COC 0.20741234 <?page no="324"?>