Marion Schulte The Semantics of Derivational Morphology A Synchronic and Diachronic Investigation of the Suffixes -age and -ery in English Language in Performance LiP The Semantics of Derivational Morphology 49 Edited by Rainer Schulze Advisory Board: Anita Fetzer (Augsburg), Thomas Herbst (Erlangen), Andreas Jucker (Zürich), Manfred Krug (Bamberg), Christian Mair (Freiburg i.Br.), Ute Römer (Atlanta, GA, USA), Andrea Sand (Trier), Hans-Jörg Schmid (München), Josef Schmied (Chemnitz) and Edgar W. Schneider (Regensburg) Marion Schulte The Semantics of Derivational Morphology A Synchronic and Diachronic Investigation of the Suffixes -age and -ery in English Bibliografische Information der Deutschen Nationalbibliothek Die Deutsche Nationalbibliothek verzeichnet diese Publikation in der Deutschen Nationalbibliografie; detaillierte bibliografische Daten sind im Internet über http: / / dnb.dnb.de abrufbar. Dissertation, Universität Bielefeld, D 361 © 2015 · 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 Printed in Germany ISSN 0939-9399 ISBN 978-3-8233-6963-9 5 Contents Tables .......................................................................................... 8 Figures ......................................................................................... 8 Acknowledgements ................................................................ 10 1 Introduction......................................................................... 11 2 Affixes and Meaning ......................................................... 15 2.1 Morpheme-based Approaches to Morphology................ 15 2.1.1 Morpheme-based Accounts and Affixes ............................... 17 2.2 Word-based Approaches to Morphology ......................... 21 2.3 Complex Words in the Mental Lexicon............................. 22 2.4 Summary ............................................................................... 27 3 Accounting for the Semantics of Grammatical Categories ............................................................................ 29 3.1 Previous Approaches........................................................... 29 3.2 Lieber's Skeleton and Body Model .................................... 31 3.3 Semantic Maps...................................................................... 34 3.4 Summary ............................................................................... 38 4 Adapted Semantic Map Model........................................ 39 4.1 Readings and Ontological Categories ............................... 39 4.2 Construction of Semantic Maps ......................................... 48 4.3 Semantic Structure ............................................................... 51 4.4 Productivity........................................................................... 58 4.5 Summary: Using Semantic Maps to Account for Semantic Structure .................................................... 65 5 Dictionary Investigation of -age and -ery Derivatives........................................................................... 67 5.1 Data Source and Methodology........................................... 67 6 5.2 The Suffix -age in the OED .................................................. 71 5.2.1 Previous Descriptions .............................................................. 71 5.2.2 Middle English Neologisms .................................................... 72 5.2.3 Present Day English Neologisms ........................................... 94 5.2.4 Comparison of ME and PDE Neologisms ........................... 103 5.3 The Suffix -ery in the OED................................................. 106 5.3.1 Previous Descriptions ............................................................ 106 5.3.2 Middle English Neologisms .................................................. 107 5.3.3. Present Day English Neologisms ......................................... 126 5.3.4 Comparison of ME and PDE Neologisms ........................... 135 5.4 Summary: The Semantic Structure of -age and -ery Neologisms in the OED ................................................ 137 6 Corpus Investigation of -age and -ery Derivatives .... 139 6.1 Data Source and Methodology......................................... 139 6.2 The Suffix -age in the BNC ................................................ 142 6.2.1 CONCRETE Derivatives ............................................................ 143 6.2.2 ABSTRACT Derivatives ............................................................. 147 6.2.3 EVENT Derivatives ................................................................... 150 6.2.4 STATE Derivatives.................................................................... 152 6.2.5 Semantic Maps of -age Derivatives....................................... 153 6.2.6 Comparison to Dictionary-based Analysis ......................... 157 6.3 The Suffix -ery in the BNC................................................. 161 6.3.1 CONCRETE Derivatives ............................................................ 162 6.3.2 ABSTRACT Derivatives ............................................................. 166 6.3.3 EVENT Derivatives ................................................................... 168 6.3.4 STATE Derivatives.................................................................... 170 6.3.5 Semantic Maps of -ery Derivatives ....................................... 172 6.3.6 Comparison to Dictionary-based Analysis ......................... 176 6.4 Summary: The Semantic Structure of -age and -ery Derivatives in the BNC................................................. 180 7 Conclusion: The Semantic Structure of -age and -ery Derivatives ......................................................... 183 7.1 Comparison of -age and -ery Derivatives ........................ 183 7.2 Evaluation of the Adapted Semantic Map Model ......... 187 7 9 Appendix............................................................................ 197 8 References .......................................................................... 191 Primary Sources .......................................................................... 191 Secondary Sources...................................................................... 191 8 Tables Table 1: Calculation of box sizes on semantic maps ............................. 49 Table 2: Thickness of connecting lines on semantic maps ................... 50 Table 3: Ontological categories of -age neologisms (ME) ..................... 73 Table 4: CONCRETE readings of -age neologisms (ME) .......................... 75 Table 5: ABSTRACT readings of -age neologisms (ME) ........................... 80 Table 6: STATE readings of -age neologisms (ME) .................................. 88 Table 7: Ontological categories of -age neologisms (PDE) ................... 94 Table 8: CONCRETE readings of -age neologisms (PDE)......................... 95 Table 9: ABSTRACT readings of -age neologisms (PDE).......................... 97 Table 10: Ontological categories of -ery neologisms (ME) ................... 109 Table 11: CONCRETE readings of -ery neologisms (ME)......................... 110 Table 12: ABSTRACT readings of -ery neologisms (ME) ......................... 116 Table 13: STATE readings of -ery neologisms (ME) ................................ 120 Table 14: Ontological categories of -ery neologisms (PDE) ................. 126 Table 15: CONCRETE readings of -ery neologisms (PDE) ....................... 127 Table 16: ABSTRACT readings of -ery neologisms (PDE)........................ 130 Table 17: Ontological categories of -age derivatives (BNC) ................. 142 Table 18: CONCRETE readings of -age derivatives (BNC) ...................... 143 Table 19: ABSTRACT readings of -age derivatives (BNC) ....................... 147 Table 20: EVENT readings of -age derivatives (BNC) ............................. 151 Table 21: STATE readings of -age derivatives (BNC) .............................. 152 Table 22: Ontological categories of -ery derivatives (BNC) ................. 162 Table 23: CONCRETE readings of -ery derivatives (BNC)....................... 162 Table 24: ABSTRACT readings of -ery derivatives (BNC) ....................... 166 Table 25: EVENT readings of -ery derivatives (BNC) ............................. 168 Table 26: STATE readings of -ery derivatives (BNC) .............................. 170 Figures Figure 1: Word formation schema -ing suffixation ................................ 22 Figure 2: First attestation dates of -age neologisms (ME)....................... 73 Figure 3: Number of OED entries in 50 year periods over the course of the ME period............................................................. 74 Figure 4: Semantic map of CONCRETE readings of -age neologisms (ME) ......................................................................... 79 Figure 5: Semantic map of ABSTRACT readings of -age neologisms (ME) ......................................................................... 86 9 Figure 6: Semantic map of STATE readings of -age neologisms (ME).... 90 Figure 7: Semantic map of all readings of -age neologisms (ME) ......... 93 Figure 8: Semantic map of CONCRETE readings of -age neologisms (PDE) ....................................................................... 97 Figure 9: Semantic map of ABSTRACT readings of -age neologisms (PDE) ....................................................................... 99 Figure 10: Semantic map of all readings of -age neologisms (PDE)...... 102 Figure 11: Semantic map of all readings of -age neologisms (ME) ....... 104 Figure 12: First attestation dates of -ery neologisms (ME)..................... 108 Figure 13: Number of OED entries in 50 year periods over the course of the ME period........................................................... 108 Figure 14: Semantic map of CONCRETE readings of -ery neologisms (ME) ....................................................................... 115 Figure 15: Semantic map of ABSTRACT readings of -ery neologisms (ME) ....................................................................... 118 Figure 16: Semantic map of STATE readings of -ery neologisms (ME) .. 123 Figure 17: Semantic map of all readings of -ery neologisms (ME) ....... 125 Figure 18: Semantic map of CONCRETE readings of -ery neologisms (PDE) ..................................................................... 130 Figure 19: Semantic map of ABSTRACT readings of -ery neologisms (PDE) ..................................................................... 131 Figure 20: Semantic map of all reading of -ery neologisms (PDE) ....... 134 Figure 21: Semantic map of all reading of -ery neologisms (ME) ......... 136 Figure 22: Semantic map of all readings of -age derivatives (BNC) based on token frequency........................................................ 154 Figure 23: Semantic map of all readings of -age derivatives (BNC) based on type frequency.......................................................... 156 Figure 24: Semantic map of all readings of ME -age neologisms (OED)..................................................................... 159 Figure 25: Semantic map of all readings of PDE -age neologisms (OED)..................................................................... 160 Figure 26: Semantic map of all readings of -ery derivatives (BNC) based on token frequency........................................................ 174 Figure 27: Semantic map of all readings of -ery derivatives (BNC) based on type frequency.......................................................... 175 Figure 28: Semantic map of all readings of ME -ery neologisms (OED)..................................................................... 178 Figure 29: Semantic map of all readings of PDE -ery neologisms (OED)..................................................................... 179 10 Acknowledgements This book is a revised version of my PhD thesis, which was accepted and defended at Bielefeld University in 2014. During this project, my supervisors Anne Schröder and Ingo Plag offered great advice and helped me to develop and improve my research. I would like to thank them for their continuous encouragement and guidance. I am also lucky to have wonderful colleagues in Bielefeld and elsewhere who provided helpful comments on my work and shared all kinds of highs and lows with me: Steve Gramley, Vivian Gramley, Jens Thomas, Sabine Arndt-Lappe, Till Meister, Christoph Haase, Kathrin Höppner, Jenny Krome, Petra Peschke thank you! This book couldn’t have been written without my family’s endless support. I am extremely grateful for that. Bielefeld, May 2015 Marion Schulte 11 1 Introduction The present study is situated at the interface of morphology and lexical semantics. It investigates the semantics of two derivational English suffixes: -age and -ery. Although some recent investigations have started to study the semantics of affixation systematically (e.g. Lieber 2004), there are, on the whole, very few thorough accounts of the semantics of derivational morphology. This leads Lieber (2012: 2108) to claim that “the most neglected area of morphological theory in the last three decades has been derivational semantics”. One of the reasons for the lack of research in this area may be the challenges that word formation processes pose for semantic description. Semantic description is, of course, never easy or straightforward. Even for word meaning, a phenomenon that is much better researched and understood than the semantics of derivation, there are multiple theories and approaches that vary greatly with regard to the aspects of word meaning they formalise and the ways in which they do so (e.g. Wierzbicka 1996, Geeraerts 1997, Pustejovsky 1995). Apart from strengths and weaknesses that are particular to the varying approaches, the problems encountered in the area of lexical semantics are similar to those of derivational semantics. One of the main differences is that aspects like polyfunctionality or synonymy seem to be much more pronounced for derivational semantics and that, especially regarding affixation, there is the added difficulty of trying to account for the meaning of elements that never occur in isolation and are always dependent on other morphological elements. Polyfunctionality and synonymy in particular are often mentioned as problematic with regard to affixation (e.g. Beard & Volpe 2005). The derivatives of one and the same affix can have very different readings. Consider, for example, the following derivatives of -age: schoolage 'a fee paid for tuition at a school' 1 , patronage 'the right of presenting a member of the clergy to a particular ecclesiastical benefice or living', screenage 'screens collectively', hourage 'the aggregate number of hours spent in working or travelling', bondage 'the tenure of a bonde', riffage 'riffing, esp. on a guitar', victimage 'the condition of being a victim', parachutage 'a drop site', or spillage 'that which is spilt'. At first glance it certainly seems unlikely that one and the same affix can give rise to derivatives with so many highly different readings. On the other hand, the derivatives of different affixes can have very similar readings. Both -age and -ery derivatives, for example, often refer to collectives such as blossomry 'blossoms collectively' or twiggage 'twigs collectively'. Sometimes doublets with identical semantics, at least according to the para- 1 Where possible, the semantic paraphrases used in this study are based on the semantic paraphrases provided in the Oxford English Dictionary online (OED). 12 phrases in the OED, are recorded, e.g. vassalage 'a body or assemblage of vassals' and vassalry '=vassalage'. Such amounts of polyfunctionality on the one hand and synonymy on the other are not common in the realm of lexical semantics but seem to be the norm for derivational semantics. In light of this situation, the present study has two aims. First, it investigates the semantic structure of the two derivational suffixes -age and -ery in Middle English (ME) and Present Day English (PDE) in two different data sources - a dictionary (Oxford English Dictionary online (OED)) and a corpus (British National Corpus (BNC)). Although a number of researchers provide overviews of the semantics of English affixes (e.g. Marchand 1969, Dalton- Puffer 1996, Bauer, Lieber & Plag 2013), such detailed accounts of individual affixes are rare. The present study aims to find out which readings are expressed by derivatives at which point in time, whether there is semantic change from ME to PDE, and which readings are used more often than others. This goes beyond a mere list of possible readings, because the connections and dependencies between these readings, and, thus, the semantic structure of each morphological category is exposed. Such an account can also shed light on the development of polysemy within a morphological category - do polysemous structures, for example, arise slowly over time, or are they already found in the earliest usage period of derivatives? The second aim of this study is the introduction of a new way of representing the semantic structure of grammatical categories such as derivational affixes: an adapted semantic map model. This model is based on implicational semantic maps as spelled out in Haspelmath (2003). The adapted semantic map model is envisaged as a pre-theoretical tool that can help to expose and compare the semantic structure of morphological categories, allowing for generalisations and abstractions based on the attested data. These can then, in a later step, form the basis of theoretical approaches to derivational semantics. The theoretical background to this investigation is discussed in chapters 2 to 4. Chapter 2 explores the issue of affixes and meaning. Because of problems in accounting for the semantics of derivation like the ones already mentioned above, it is not clear whether it should be assumed that affixes make an independent semantic contribution to derivatives. There are two fundamentally different approaches to morphology: one, the morpheme-based approach, assumes that morphemes, including affixes, have independent meaning, and another, the word-based approach, claims that only words themselves have meaning. These two opposing views will be discussed in chapter 2 before the stance taken in the present study is described. Chapter 3 then discusses previous approaches to the semantics of grammatical categories. Two accounts in particular, Lieber's skeleton and body model (Lieber 2004) and the semantic map approach (Haspelmath 2003 among others), have informed the approach taken in the present investigation and will be discussed in some detail. Chapter 4 introduces the adapted semantic map 13 approach that will be used in this study. Details on the individual elements of the semantic maps and on their exact construction are provided here, and it is also explained how the semantic structure of grammatical categories is described. Chapters 5 and 6 form the empirical part of this investigation. Chapter 5 discusses the neologisms of -age and -ery attested in two different time periods in the OED, and is thus a diachronic analysis, while chapter 6 is based on corpus data from the BNC, which represents late 20 th century language usage. The different data sources provide information on the range of readings expressed by the derivatives of both affixes, and the incorporation of corpus data allows us to investigate which readings are frequently expressed and which might be rare occurrences. Semantic maps are used to discuss the semantics of the derivatives in both of these chapters. A conclusion in chapter 7 describes the semantic structure of -age and -ery derivatives and compares the results of the different data sources. It also assesses the suitability of the adapted semantic map model in accounting for the semantics of derivational affixation. 15 2 Affixes and Meaning As a study that is concerned with the semantics of affixation, the present investigation has to come to terms with some fundamental issues regarding the semantics of affixes. There is a major divide in morphological theory concerning the analysis of affixed words such as unhappy or beautiful. Some approaches assume that such words are concatenated from a number of morphemes, e.g. unand happy. These are called morpheme-based approaches. A different way of analysing words like unhappy is to assume that they are part of a set of words that contain the word-initial phonological string / ʌn/ and express a similar meaning to unhappy, e.g. unlawful or unbelievable. These are called word-based approaches. The difference between these two ways of analysing words lies in the concept of the morpheme. Morpheme-based approaches consider the morpheme as a unit of form and meaning and the smallest meaningful building block of words. Morphemes can be added to one another according to certain word formation rules to form larger structures, namely morphologically complex words. Wordbased approaches look at the paradigmatic structure of sets of words and do not have to use a concept like the morpheme to find regularities between words. Affixes are naturally viewed very differently by these two approaches. While morpheme-based theories consider affixes to have meaning and to make a semantic contribution to their derivatives, word-based approaches do not make reference to affixes as independently meaningful elements. This chapter reviews the main tenets of both approaches as well as some of their strengths and weaknesses. Evidence on the processing of morphologically complex words is also presented. Different processing models are closely related to the theoretical approaches to morphology, and the psycholinguistic data will shed some light on how complex words are analysed. 2.1 Morpheme-based Approaches to Morphology Morpheme-based approaches to morphology assume that the morpheme is a unit of form and meaning, and the “smallest meaningful unit” in language (Plag et al. 2007: 66, but compare also Bauer 1983, Plag 2003, Katamba & Stonham 2006). Such a definition is intuitively appealing, as many examples quickly come to mind to support this position: free morphemes such as tree or chair are also lexemes, and these clearly have meaning. But bound morphemes such as affixes are also assumed to be meaningful in a morphemebased approach. The meaning of such elements is not as intuitively obvious as that of lexemes, but it can still be discerned: unmeans 'not' in unhappy, unlucky and unimportant; -ness means 'state of being X' in happiness, laziness 16 and blueness; -ing means 'action of doing X' in drawing, flattening and questioning. The complex words that are derived from a base and an affix in the above examples combine bases and affixes in a straightforward way not dissimilar to how words are combined to form a sentence. A similarity to syntax is indeed postulated in many morpheme-based approaches, for example when bases and affixes are combined according to word formation rules that are influenced by syntactic phrase structure rules. Both bases and affixes have a specified phonological form and a meaning that can be combined to render a composite phonological form and a composite meaning. Such an analysis seems to account well for the examples given above. It also has the advantage of explaining why affixation, i.e. the concatenation of morphemes, is so pervasive in the world's languages. Haspelmath and Sims (2010: 43) claim that “morpheme concatenation is the most common kind of morphological pattern cross-linguistically. By treating concatenation as the fundamental (or only) type of morphological rule, the morpheme-based model provides a natural explanation for this fact”. In spite of their theoretical elegance, morpheme-based approaches have problems explaining other processes of word formation than affixation. The most obvious problems with the definition of the morpheme as a unit of form and meaning are a morpheme with a form, which is usually called morph, but without meaning, as well as a morpheme without a morph, but with meaning. Both of these cases are attested in English. Some suffixes, for example, are clearly recognisable, but do not seem to make a semantic contribution to the supposed derivative. Consider, for example, the synonymous adjectives syntactic and syntactical, whose parallel existence and identical semantics show that -al does not add any meaning to the complex word in this case. The suffix -al is, in spite of its lack of meaning, still recognisable as a distinct element, as it also occurs in other adjectives, for example magical, where it derives adjectives from nouns, and thus clearly makes a semantic contribution. One might conclude that only magical contains the suffix -al, while the string <al> as the final element in syntactical does not represent a suffix. However, the structural and functional properties of the two adjectives are so similar that one hesitates to accept this explanation. Let us turn to the second possible problem with the traditional definition. A morpheme without a phonological realisation, but with meaning, is usually called a zero morph. This somewhat abstract construct is used to explain the otherwise puzzling occurrence of conversion, a word formation process which is, in English, often used to transpose nouns into verbs or verbs into nouns. The noun cook, for example, is derived from the verb to cook, and the verb to google is derived from the noun google without any overt marking. To explain this phenomenon without disposing of the morpheme, the morph is said to be invisible, while the meaning change from verb to noun or vice versa represents the conceptual side of the morpheme. 17 Further problems for the notion of the morpheme as the smallest meaningful unit in language are generated by non-concatenative word formation processes, i.e. processes that do not consist of the addition of meaningful morphemes to other morphemes in a linear way. These processes clearly have meaning, but their form is often difficult to establish - it may be invisible (conversion), consist at least partly of the deletion of material (truncation, backformation, -y diminutives), alternate the existing material (vowel change), or might be expressed by more than one form (extended exponence). Even affixational processes raise questions as to their supposed unity of form and meaning, as the following discussion will show. It is therefore fair to conclude that the morpheme as the smallest meaningful unit of language is a problematic notion. Carstairs-McCarthy (2005: 22) even says that it is an understandable position if some linguists “conclude that the term 'morpheme' has hindered rather than helped our understanding of how morphology works”. According to him there have been different reactions to this problem in recent years: some scholars continue to use the term morpheme, but “some or all morphemes are explicitly not regarded as Saussurean signs” (Carstairs-McCarty 2005: 20), others continue to use it without “much theoretical weight being attached to it” (ibid.), and a third group does no longer use the term at all. In this study, the term morpheme will be avoided, as it seems to raise more questions than it answers. If it occurs, it will not have any theoretical value, i.e. I will not use it to comment on the sign-like status of morphological elements. 2.1.1 Morpheme-based Accounts and Affixes Although the traditional definition of the morpheme is at the very least problematic, one could argue that affixes may still be considered form-andmeaning units, as most of the problems concerning the morpheme are posed by non-affixational word formation processes. One of the issues mentioned above that also pertains to affixes is their polyfunctionality. The same affix can often be found in derivatives with a range of meanings, but different affixes also seem to give rise to very similar derivatives. The suffix -age, for example, can be found in derivatives denoting locations (orphanage), collectives (baggage), actions (creepage), amounts (mileage), or states (victimage). Derivatives of -ery have a similar range of meanings and refer to locations (eatery), collectives (blossomry), actions (milksoppery), or states (smuggery). Polyfunctionality is only a problem if one presupposes a one-to-one relationship between form and meaning of an affix, a principle also known as isomorphism. According to this principle, a given form should only have one meaning, which poses problems for polysemous forms. It is, however, “a truth universally acknowledged that natural languages do not exhibit an absolute one-to-one correspondence between meaning and form” (Booij 1986: 503). This is true for both lexemes and affix- 18 es. Löbner claims that there is hardly a lexeme that is not polysemous: “If one consults a more comprehensive monolingual dictionary one will hardly find a word with just one meaning given” (2002: 42). If the existence of polysemy does not challenge the form-and-meaning unity of lexemes it should hardly be used to challenge a form-and-meaning unity of affixes. The meaning expressed by affixes, however, has often been said to be different from the meaning shown by lexemes, as affixes only express limited semantic functions, which are rather abstract and general. Lehrer, for example, says that they “encode only a limited set of meanings” and many of them “are semantically general” (2000: 152). Beard and Volpe go even further and claim that “[m]orphological modifications with multiple meanings select from a single pool of meaning, always grammatical functions” (2005: 190). This is a strong claim, and there is evidence to suggest that it is indeed too strong a claim. Mithun has shown that affixes in Bella Coola, a Salishan language spoken in Canada, “show characteristics that are considerably more root-like than most affixes“ (1997: 369). She therefore calls these affixes lexical affixes. It is certainly correct, however, that most affixes usually express limited meanings, and thus differ from lexemes, as the meanings that may be expressed by lexemes are potentially infinite. However, this seems to be a difference of degree rather than a difference in kind. Lehrer claims that “[i]n looking at all the cases we see a cline rather than a clear-cut division” (2000: 152). Such an approach would allow some affixes to have a very limited meaning, which is restricted to grammatical functions, while other affixes could express more lexical meanings. It is generally assumed that English derivational affixes have rather restricted grammatical meanings, but the results of this study suggest that at least some English affixes, i.e. the suffix -age, which often refers to charges and taxes in Middle English coinages, also have more lexical meanings in addition to grammatical ones. It thus seems better to assume a cline from more grammatical to more lexical meaning in affixes rather than a clear-cut division between affix and lexeme meanings. Synonymy between different affixes is another point that is often made to argue that affixes should not be considered linguistic signs, i.e. a unit of form and meaning (Beard 1995, Beard & Volpe 2005, Raffelsiefen 2010). Indeed, a number of different affixes derive complex words with similar functions. Action nouns, for example, are attested with all of the following suffixes, but may also be derived by conversion: -ing, -age, -ment, -ure, -th, -ation, and -al. The same is true for collective nouns, which may contain, among others, the suffixes -age, -ery, -ship, -hood, or -dom. Synonymy is of course also attested for lexemes, which is often illustrated by the pair couch and sofa. Semanticists have, however, argued successfully and repeatedly that total synonymy is extremely rare (e.g. Löbner 2002), as a word pair would need to correspond on all levels of meaning: descriptive, social and expressive. A 19 more common phenomenon is partial synonymy, which exists when “[t]wo lexemes [...] have one meaning variant in common” (Löbner 2002: 46). It is puzzling that total synonymy is proven to be very uncommon in lexemes - because it is so rare for two words to correspond on all levels of meaning - but is at the same time assumed to be the normal state of affairs for affixes. A good example for presumably synonymous affixes are the suffixes -hood, -dom and -ship, which have been described as semantically similar, e.g. by Plag (2003). All three form nouns with state readings, e.g. subjecthood 'the state or condition of being a subject', coupledom 'the state or condition of being (part of) a couple', recruitship 'the state of being a recruit'. A striking example for the similarity of -hood, -dom and -ship are the attested doublets and triplets. In these cases, the same base is attested with two or even all three of these suffixes with no detectable meaning change - at least not according to the paraphrases in the OED. Compare, for example, the triplet pariahship '=pariahdom', pariahhood '=pariahdom', pariahdom 'the state or condition of a pariah or outcast'. Here, the suffixes seem to be virtually interchangeable without any meaning change. The extreme similarity in this particular case might be due to the limited information available in the dictionary paraphrases. But even if pariahship, pariahhood, and pariahdom were completely synonymous, this would not prove that -ship, -hood, and -dom would be completely synonymous as well. They would only be synonymous in this particular instance but would not necessarily be interchangeable in all other contexts, as the following short overview of 20 th century -hood, -dom, and -ship neologisms as attested in the OED will show. While -hood gives rise to only a handful of derivatives in the 20 th century, nearly all of them with a state reading, -dom and -ship give rise to 29 and 40 neologisms respectively. A substantial part of these do not refer to a state or condition. In the case of -dom, 18 derivatives have a state reading, but 15 also refer to a collectivity, and another 15 to a realm or sphere. Many derivatives are polysemous between these readings. Only 18 out of 40 -ship derivatives have a state reading, and the most common meaning of -ship neologisms is 'skill' or 'practice'. This reading is extremely frequent, especially on the basis of agent nouns ending in -man, for instance buymanship, cocksmanship or upmanship. In some of these cases the supposed base noun ending in -man is not attested - the OED does not list *buyman or *upman independently - so that the amalgamation of -man and -ship into the new suffix -manship is the most likely source for these creations. This shows how common the 'skill' or 'practice' readings are in -ship derivatives. This suffix should therefore not be considered synonymous with -dom and -hood, as their meaning ranges differ considerably. While the meanings of -hood, -dom and -ship derivatives are strikingly similar in the area of state nouns, they differ considerably in other respects. Partial synonymy that is restricted to certain semantic functions is a more accurate description than complete synonymy. 20 In sum, affixes commonly exhibit polysemy, and synonymy that is limited to certain functions of the affixes involved is also frequently observed. With regard to polysemy, affixes do not behave differently from lexemes, as this kind of ambiguity is widespread in lexemes as well. This issue can therefore not be used to support an argument against the position that affixes are a unit of form and meaning. Synonymy seems more pronounced in affixes than lexemes, however. Although Riddle (1985) claims to have shown for -ness and -ity that subtle meaning differences between semantically very similar affixes can exist (cf. also Baeskow 2012 for a similar result), Arndt- Lappe (2014) is able to model the distribution of these two suffixes more convincingly as a result of the phonological properties of the bases. It is certainly the case that sets of affixes do show marked similarity, which looks like synonymy in some cases, regarding at least some of their functions. This rarely occurs in lexemes, and such a difference between affixes and lexemes should not be ignored. Still, there does not seem to be a qualitative difference between affixes and lexemes: both exhibit polysemy and synonymy, albeit to different extents. Lehrer can therefore claim that “[a]lthough the concepts represented by English affixes are limited, they show polysemy that is similar to that found in lexemes” (2003: 229). The strongest arguments against a morpheme-based theory of morphology therefore do not pertain to affixes and affixation. From a theoretical point of view it is thus not impossible for affixes to be perceived as formand-meaning units similar to lexemes. It is far from clear how the meaning of affixes can be established, however. The problem is that affixes, by definition, occur only in combination with other morphemes and are thus never attested independently. If one wants to establish the meaning of an affix, one has to assume that the meaning of the derivative can be interpreted as the combination of the meanings of each morpheme, so that the affix meaning can be 'subtracted' from the derivative's meaning. This premise is called Compositionality Principle. The principle of compositionality is a semantic principle which originates in syntactic theory, but is also often referred to in other areas of linguistics, including word formation. It is here given in Pelletier's words: “The Principle of Semantic Compositionality is the principle that the meaning of an expression is a function of, and only of, the meanings of its parts together with the method by which those parts are combined“ (1994: 11; italics in original, MS). This is quite a strong claim to make, and it does seem to be too strong for the semantics of word formation processes. Even in the area of compounding, which is sometimes called composition, most researchers would probably agree that the different parts contribute to the meaning of the derivative, but that it is very difficult to predict the semantics of an isolated compound only from the meanings of its parts and the way they are combined (Klos 2011, Plag 2003, Bauer 1983). A weakened version of the compositionality principle is widely accepted in the literature (e.g. also Lehmann 2007; Costel- 21 lo & Keane 2005), even if the different scholars have different theoretical standpoints on this issue. Other linguists, however, believe that compositionality is not crucial in interpreting novel compounds. Dunbar, for example, finds that “[t]hese findings challenge the notion that the meaning of NN compounds can be reliably composed from the meanings of the constituent concepts” (2005: 223ff). However, even Dunbar admits that at least some, if by no means all, novel compounds may be interpreted compositionally (2005: 227). For him, compositionality seems to be an existing and useful notion, even if it is not always necessary in understanding new compounds. If a strict version of the compositionality principle is unlikely to hold for compounding, it is at the very least unlikely to hold for derivation. A weakened form would imply that both affix and base contribute to the semantics of the derivative, but other factors, for example context, may play a role in interpretation as well. This would make it extremely difficult, if not impossible, to establish the recurring semantic contribution of an affix from its derivatives. Morpheme-based approaches provide an elegant theoretical account of affixation and they also explain why concatenating morphological processes are so prevalent cross-linguistically. However, they have serious difficulties in accounting for non-concatenating morphological processes. The following part of this chapter will review a different approach to morphology, namely word-based theories. 2.2 Word-based Approaches to Morphology A different theoretical position is taken by word-based approaches to morphology. These do not analyse words into their constituent parts but compare them paradigmatically, i.e. to other words. The only form-and-meaning unit in these approaches is the word itself, not the morpheme. For example, a paradigmatic approach to morphology would see the semantic and formal similarities between the words laughing, barking, and cooking, but it would not assume, like morpheme-based approaches, that the -ing in the above words means 'action of doing something'. This meaning would be assigned to the word as a whole, and the paradigm could be extended by creating similar formations, e.g. sleeping or bathing. Abstractions over sets of similar words are possible here as well. A word-based analysis of these formations could for example say that a word with the structure Xing would refer to an action. Such relations can be formalised in schemas that contain information on the phonology, word class and semantics of words, for example as shown in figure 1. 22 Figure 1: Word formation schema -ing suffixation This schema shows that verbs with the form / X/ and the meaning 'do X' relate directly to nouns with the form / Xing/ and the meaning 'the action of doing X'. The bidirectional arrow illustrates a relationship that goes both ways. This has the great advantage that not only the nouns on the righthand side of the schema can be seen as a result of the addition of -ing to the verb form, which would be the analysis in a morpheme-based approach, but also that the verbs on the left-hand side could be the result of the 'subtraction' of that element from the noun. Such an analysis would naturally account for word formation processes such as backformation, e.g. to babysit < babysitter, which are more problematic for morpheme-based accounts. It is indeed one of the major advantages of word-based accounts of morphology that they are able to describe non-concatenative processes just as easily as concatenative processes. A word-based approach to morphology thus accounts for all attested morphological processes, but it is far less restrictive than a morpheme-based approach. Word-based approaches, for example, do not explain why concatenative processes are so common in various languages. Haspelmath and Sims thus conclude their discussion of these two approaches by saying that “[t]he word-based model is more empirically adequate, but at a cost of lost restrictiveness. It is therefore not the case that one approach is inherently superior to the other” (2010: 53). The two approaches have theoretical strengths and weaknesses, but is there any evidence on how complex words are analysed in practice? Do speakers perform a morpheme-based analysis, where individual morphemes would have independent meaning and therefore their own lexical entries, or do speakers analyse morphologically complex words paradigmatically without such independent entries in the mental lexicon? These questions are related to and have implications for the theoretical approaches discussed so far. The last part of this chapter will thus review a number of processing models and present some evidence from psycholinguistic investigations. 2.3 Complex Words in the Mental Lexicon Similarly to the different theoretical views on morphology, there are different models of morphological processing. Two opposing models assume that words with an internal structure, e.g. affixed words, may be either decomposed into their various elements or accessed holistically, i.e. as whole words. Both of these models make reference to an entity called the mental / X/ verb 'do X' / Xing/ noun 'the action of doing X' 23 lexicon in which the accessed parts - either morphemes or whole words - are stored. Before we can discuss the advantages and disadvantages of the processing models, the notion of the mental lexicon should be clarified. The mental lexicon is not to be confused with a real-world lexicon or dictionary. The latter are manufactured and very different in size and organisation to the mental lexicon - the mental lexicon being both much larger and more complex than any book (cf. Aitchison 2012: 11ff). The question that has concerned much of psycholinguistic study is what kind of information is stored in the mental lexicon - does it contain whole words or (also) parts of words? This is difficult to answer, as information on the mental lexicon can only be inferred and not accessed directly. Before psycholinguistic experiments could be carried out, conceptions of the mental lexicon could not be tested at all and linguists had to rely on what they assumed was happening in the brain when language was processed. Bloomfield, for example, believed that “[t]he lexicon is really an appendix of the grammar, a list of basic irregularities” (1933/ 1969: 274). In his view, this list of irregularities was supplemented by the grammar, which would produce regular forms. Although similar statements have been made after Bloomfield as well, more recent theories of the mental lexicon assume a more complex system than just a list of irregularities. These assumptions are based on a great number of psycholinguistic experiments (see, for example, McQueen & Cutler (2001) for a discussion of some of these experiments with results supporting different theories of the mental lexicon). Although some recent models of linguistic processing do not refer to a mental lexicon in the traditional sense, as they do not make reference to discrete entries for morphemes (e.g. Baayen et al. 2011, Baayen 2014: 113-116), most theories still use the concept of the mental lexicon in one shape or another. With regard to complex words, it is a central question, as already outlined, whether they are stored and accessed as whole words or whether they are decomposed. Two kinds of models have been proposed to answer this question: symbolic and subsymbolic ones. The former rely on entries in the mental lexicon and rules to combine different elements, the latter reject such combinatorial rules. To illustrate the strengths and weaknesses of both kinds of models examples from both groups will be introduced in the following, starting with the symbolic view. The argument has been made that the mental lexicon contains the building blocks of words - morphemes - and the rules by which these are combined into words. Additionally, irregular complex words are stored as wholes (cf. Günther 2005: 1767). These may be, for example, verbs that have past tense forms created by ablaut, as in sing-sang-sung, lexicalised forms like business, or words that contain relatively unproductive affixes like growth. A system like this would store only absolutely necessary information - it would be minimally redundant and therefore relatively small. Such a model would account for the storage of both regular and idiosyncratic complex 24 words in a logical fashion. However, Sandra points out that “the existence of a logical system in the language does not necessarily imply that the mental structures that develop for dealing with this system will reflect this logic” (1994: 245). He stresses the difference between the model we make of morphological processing and the actual processes that happen in the brain. Another problem with this economic storage model is that minimal redundancy is not a common feature of natural languages and that “the way words are stored in the human brain is not totally economical” (Plag 2003: 48). A different issue is an inherent disadvantage of economical storage, because the small size of the mental lexicon in this model would come at a cost: processing time. If every regular complex word has to be built or decomposed on-line while speaking or listening, the processing costs, and therefore the time it takes to access words, will be quite high. Processing time would be considerably reduced if all words had their own representations, but such a model would have the disadvantage of being much larger than the one described above. It could also not account for novel formations, which have to be decomposed into their parts in order to be understood. So which of these models is to be preferred then? It has in fact been argued that speakers make use of both models. Aitchison suggests that complex words are usually accessed as whole words, but that they may also be decomposed. “The primary purpose of this ability [decomposition, MS], therefore, seems to be to enable speakers to make up new words of their own and to comprehend the novelties coined by others. Its secondary purpose may be as a memory aid to enable people to link up words containing similar morphemes” (2012: 206f.). She distinguishes between inflectional and derivational affixation and also between derivational prefixation and suffixation. While inflected words are usually decomposed into their components, “a few words used commonly in their inflected form, such as peas, eyes, happened, needed [...]” (Aitchison 2012: 150; italics in original, MS) may well be accessed as whole words. For derivational affixation, the situation seems to be slightly different. Aitchison claims that prefixes are “normally attached to stems” (2012: 152), and that there is some empirical evidence that supports holistic access to suffixed words as well (cf. ibid. 153). But she also says that decomposition of these derivatives is possible as speakers might indeed use this strategy as a “back-up procedure in order to construct a complex word if their normal memory for the word fails them […]” (2012: 154). Decomposition as a back-up plan may even be carried out at the same time as whole-word access (ibid.: 155). Her remarks are supported by recent studies. Bozic et al. (2013), for example, have found clear differences in the processing of inflected and derived words in auditory processing: whole-word access is generally favoured for derived forms, whereas inflected words are decomposed. There are subtle differences between various categories of derivatives, however, and traces of morphological structure seem to be present for semantically transparent 25 derivatives that contain productive suffixes. Evidence for decomposition is also provided by priming experiments for various languages (e.g. Rueckl & Aicher 2008, Smolka et al. 2014). It thus seems most likely that decomposition and holistic access are both employed in natural language processing. Models that incorporate both strategies are therefore favoured in many current models of morphological processing. Reid and Marslen-Wilson (2003: 288), for example, claim that “[t]he most prominent approaches over the last decade […] have been those that have taken the middle ground, arguing for a mix of morphemic and full-form representations”. Although many models share the general assumption that complex words may be either decomposed or accessed as whole words, they differ on finer points. Caramazza and colleagues, for example, argue for a model called Augmented Addressed Morphology (Caramazza et al. 1988). According to this approach, holistic access to word representations is favoured, because it is always faster than decomposition. Words are only successfully decomposed if they are new and do not yet have whole-word representations in the mental lexicon. Other models, for example the Morphological Race Model (e.g. Frauenfelder & Schreuder 1992), assume that decomposition and holistic access are more equal. According to this approach, complex words are both decomposed, via the decomposition route, and looked up as whole words, via the whole-word route. The fastest route then accesses the relevant entry or entries. A completely different approach to processing is taken by connectionist and emergentist models. Although models that fall under this rather general heading differ in various points, the so-called network model developed by Bybee will be taken as an example for how these approaches differ from the ones described above. According to Bybee, complex words are not strictly speaking decomposed, as they are “not broken up into their constituent morphemes” (1995: 428). However, their internal structure is recognised, because connections are made between the complex word in question and other words with similar phonological and semantic features. The “morphological structure emerges from the connections they make with other words in the lexicon” (ibid.: 428f). Here, bound morphemes such as affixes do not have independent representations, but are always parts of words - a crucial difference to the models outlined above. This makes this model easily compatible with a word-based approach to morphology. The network model does, however, not account for the decomposition of complex words, a strategy that is employed at least for inflected words (Bozic et al. 2013), and is predicted by symbolic models. Another important difference to symbolic models is the gradual structure of morphology that is predicted by the network model. The connections between different words, for example derived words and their bases, or between different sets of derived words, can be relatively strong or relatively weak and may vary between speakers or situations. Dual route models, on 26 the other hand, assume that a word is either decomposed or accessed holistically, but there is now evidence that the nature of morphology is gradual rather than categorical in this respect (e.g. Hay & Baayen 2005, Ali & Ingleby 2010). In spite of the abundant evidence for the importance of paradigmatic relations, Hay and Baayen still point out that “[s]tems and affixes may well develop their own representations” (2005: 343). These representations would, however, depend on “continuing probabilistic support received from paradigmatic analogy” (2005: 344). This means that affixes may develop independent representations, but these can only continue to exist if other complex words with a similar internal structure are contained, and perceived as such, in the lexicon. Once the connections between a word and other words with similar parts weaken, the independent representations of the parts weaken as well, and may even disappear at some point. This shows that gradual structure and decomposition do not have to contradict each other. In a gradual model of morphological structure the strategies assumed by dual route models could represent the poles on a cline from fully decomposable to completely holistic access. Some factors that influence whether a given word is decomposed or accessed holistically have been established. The most important of these are probably frequency effects - both the overall frequency of the derived word and the frequency of the derivative relative to its base. Aitchison already hints at frequency as a factor when she says that words, which are “used commonly in their inflected form” (2012: 150) might have whole-word representations, although inflected words are usually decomposed. It is indeed a commonly held view in the literature on morphological processing that high frequency forms are accessed as whole words, while low frequency forms are decomposed. Plag, for example, claims that “[i]nfrequent complex words have a strong tendency to be decomposed. By contrast, highly frequent forms, be they completely regular or not, tend to be stored as whole words in the lexicon” (2003: 51). Hay (2003) has, however, shown that the absolute frequency of the derived word is not the only frequency effect important for morphological processing. The experiments she has conducted lead her to conclude that complex words are likely to be decomposed if they are less frequent than their bases, but tend to be accessed holistically if they are more frequent than their bases (ibid.: 88). Even low frequency complex words might then be accessed as whole words, at least if they are more frequent than their bases, and high frequency words can be decomposed if they are less frequent than their bases. This result is contrary to the results predicted by a model which relies solely on absolute frequency. Thus, the relative frequencies of the components of a morphologically complex word play an important role in decomposition (see also Burani & Thornton 2003). Relative frequency also has an effect on semantic transparency, which is traditionally assumed to be due to the absolute frequency of derived words as well. For example, high frequency complex words such as business or 27 government are usually thought to be less compositional than lower frequency words like calmness. But such an account does not explain why happiness is compositional and at the same time a high frequency word. According to Hay, semantic drift, and therefore low compositionality of the derived form, is connected to relative frequency (cf. 2003: 113f). Additionally, she also finds effects of absolute frequency in this area: highly frequent words often acquire additional meanings and become polysemous. “As long as the base is higher frequency [sic! ] than the derived form, however, these forms do not lose their semantically transparent meanings” (Hay 2003: 117; italics in original, MS). This shows that both absolute and relative frequency are important factors that influence polysemy and semantic transparency. Frequency is, of course, not the only factor that plays a role in morphological processing, the phonotactics of a complex word are also crucial. If a complex word exhibits phonotactics at the affix boundary which do not occur in simple words, the derivative is relatively likely to be decomposed. Hay uses the -ful derivatives bowlful and pipeful as examples (cf. 2003: 16): bowlful is less likely to be decomposed than pipeful, because the transition at the affix boundary is also attested in simplex words like dolphin, while [pf] in pipeful does not occur in simple English words. This tendency is independent of frequency effects. Both frequency and phonotactics influence the productivity of an affix. Hay and Baayen summarise this as follows: Affixes represented by more words which are infrequent relative to their bases, and which contain low probability phonotactics, are not only the most likely to be more highly segmentable and to develop stronger independent representations, they are also more readily available for use in new words; that is, they tend to be more PRODUCTIVE. (Hay & Baayen 2005: 345; emphasis in original, MS) 2.4 Summary The evidence from studies on morphological processing suggests that models that account for gradual structure describe the data best. While there is evidence for words with internal structure being both accessed as whole words and being decomposed, it is not yet clear that morphemes that are not words have independent representations in the mental lexicon (cf. Hay & Baayen 2005). This has implications on the debate concerning morphemebased and word-based morphological theories. We have already seen that word-based theories account more easily for the morphological processes attested in the world's languages. Although such theories are less restrictive than morpheme-based accounts, and restrictiveness is usually seen as an important aspect of a theory, they are more in line with current research on 28 morphological processing. But the evidence is not compelling enough to dismiss morpheme-based theories completely. Fortunately, a definitive answer as to which theory of morphology is the 'right' one does not have to be given in the context of the present investigation of the semantic structure of -age and -ery derivatives. My results can be interpreted in both approaches: a morpheme-based account would assign meaning to the suffixes themselves, while a word-based account would assume that the meaning is expressed by the derivatives. Psycholinguistic studies have shown that morphologically complex words are decomposed under certain conditions. Their internal structure is thus clearly recognised by speakers, independently of the question whether morphemes are represented by independent lexical entries in the mental lexicon. Especially Hay's findings regarding the importance of relative frequency and her points on the semantic development of morphologically complex words (cf. Hay 2003) will play a role in the empirical part of this study. The relative frequency of derivatives and their bases in particular will inform transparency judgements in the corpus-based part of this investigation. 29 3 Accounting for the Semantics of Grammatical Categories This chapter introduces some prominent approaches to the meaning of elements other than content words. This includes approaches to the semantics of inflectional affixes, e.g. Haspelmath (2003), derivational affixes, e.g. Lieber (2004), and function words, e.g. Tyler and Evans (2001). Although these accounts are situated in very different theoretical frameworks they deal with similar issues, as the phenomenon of lexical meaning is on the whole far better understood than that of grammatical meaning or the semantics of sublexical elements. Some of these issues will be discussed in the following section, which also introduces other approaches that have influenced the model developed in the present work. Lieber's skeleton and body model (Lieber 2004) and the semantic map approach (Haspelmath 2003 among others) will be presented in more detail in the remainder of this chapter, as they have had the greatest influence on the adapted semantic map model introduced in chapter 4. They can also serve as exemplary investigations into the semantics of non-lexical entities. 3.1 Previous Approaches It has traditionally been assumed that affixes express a different meaning than lexemes, if they are assumed to have meaning at all (see chapter 2). According to this view, affixes are semantically bleached and thus rather abstract. Mithun says with regard to this traditional position that “[w]e expect roots to carry such meanings as 'rock', 'mouth', or 'catch', and affixes to indicate such features as causation, tense, or gender” (Mithun 1997: 357). Derivatives of unand -ness will illustrate this position. If the prefix unis attached to verbs it has a reversative function, e.g. in undo 'to unfasten', unzip 'to unfasten a zip', or unthink 'to remove from thought'. The semantic contribution of the prefix to the derivatives is quite general and abstract. This is similar for the suffix -ness in the derivatives kindness 'the quality of being kind' or happiness 'the quality or condition of being happy'. While this may be true for most or even all English affixes, the situation is different in other languages. Mithun (1997: 357) describes a number of so-called lexical affixes in the Salishan language Bella Coola that seem to have “root-like functions”, e.g. -lst 'rock' or -łp 'tree'. These are formally clearly affixes, but their semantics cannot be said to be general and abstract. Lieber thus states that “lexical affixes indeed present a challenge to the notion that derivation covers a rather abstract, fixed set of categories” (2012: 2099). The meanings of affixes 30 thus seem to range from more abstract and general to more concrete and specific (cf. also Lehrer 2000). A model developed to account for the semantics of lexemes should therefore in principle be able to account for the semantics of affixes or affixed words as well. There are many such models, as the notion of word meaning has been the focus of a whole linguistic discipline: lexical semantics. This issue has been approached from various theoretical backgrounds. To name only a few examples, Jackendoff's Conceptual Semantics (e.g. Jackendoff 1991) approaches the problem of meaning from a generative perspective, while Geeraerts (1997) works in the tradition of prototype theory, and Wierzbicka (1996) develops a metalanguage based on semantic primitives. None of these models has been developed with the semantics of affixation in mind, however, and Lieber (2004: 6) explains that they do not have “all the characteristics that I believe are necessary” to account for the semantics of derivational affixes. According to Lieber then, approaches to lexical semantics that are suitable for lexemes may not be suitable for affixes, although she assumes a general similarity between word and affix meaning as she “claims that the circumscription in the semantic range of derivation follows […] from the semantic categories that distinguish classes and subclasses of simplex lexical items from one another” (Lieber 2012: 2109). Lieber clearly supports a morpheme-based view on morphology, as she believes affixes to make independent semantic contributions to their derivatives and that these contributions can be expressed by semantic features. Many approaches that can be applied to the semantics of grammatical categories differ substantially from Lieber (2004) in that they are not decompositional (e.g. Haspelmath 2003, Tyler & Evans 2001, Lehrer 2003). They do not break complex words down into their parts and assign meaning to the individual elements, but consider only the meaning of the derivative. Such approaches are therefore more easily applied to a word-based theory of morphology than to a morpheme-based account. One example for such an approach in the cognitive tradition is Tyler and Evans (2001), who study the semantics of the spatial preposition over in English. They aim to find out how the different senses of this highly polysemous word are connected and provide criteria to inform further research in this area. As polysemy is the main topic of their study, their work will be discussed in chapter 4.3 of the present work, which also deals with polysemy. Another cognitive approach can be found in Haselow (2011), an investigation of the meaning and change of Old and early Middle English word formation patterns. Haselow groups derivatives into five ontological categories: PERSON , OBJECT , LOCATION , AB- STRACT , and ACTION . By comparing the developments of each of these groups over various periods of the English language, he can show how the inventory of word formation patterns has changed over time. Haselow investigates the typological change of English towards a predominantly analytic language, so the information provided for a single process is limited to 31 the somewhat general properties expressed by the relevant ontological category and a short discussion of the more specific readings of a suffix's derivative. If the categories were more specific, an approach like this would be usable to investigate the semantics of a single word formation process, and the ontological categories employed by Haselow seem particularly suitable to compare the semantics of different word formation processes. 3.2 Lieber's Skeleton and Body Model Lieber (2004) develops a model for the description of affix semantics. She acknowledges the influences of a number of scholars working on lexical semantics on her work, e.g. Jackendoff, Wierzbicka, Pustejovsky and Szymanek. Her model can be situated in the tradition of structuralist linguistics. Lieber says that “[a]lthough each of these systems has some attractive characteristics, none of them has all the characteristics that I believe are necessary to the task at hand” (2004: 6). The above mentioned linguists have developed componential approaches to describe the meaning of lexemes, but Lieber is the first who tries to build a specialised method to account for the meaning of morphemes. To do so, she studies a number of English affixes, among them both prefixes and suffixes. Lieber's decompositional metalanguage utilises semantic features such as [dynamic] or [material] and employs them both in a privative and in an equipolent way (cf. 2004: 23). These features have what she believes to be the right “grain size” (Lieber 2004: 10) to describe the semantic contribution made by affixes. They are used across different ontological categories (cf. chapter 4.1). Lieber's model is constructed around an anatomical metaphor: the two parts of the meaning of morphemes are called skeleton and body. It is only the skeletal part of an affix's meaning that is formalisable: it contains “those aspects of meaning which have consequences for the syntax” (Lieber 2004: 10). This part is described using semantic features. The body, on the other hand, is holistic and cannot be decomposed. It also differs from speaker to speaker and is prone to diachronic change. The skeleton though is “less amenable to change” (2004: 10), and it is also the part that “allows us to extend the lexicon through various word-formation processes” (ibid.). A feature-based approach like the one proposed here adheres to the principle of isomorphism. Lieber uses different features to describe an affix's semantics, but the result is always a single meaning. This instantly raises the issue of polysemy, which is frequently observed in affixes. To illustrate how Lieber accounts for polysemy, let us have a look at her discussion of the suffixes -age and -ery (cf. Lieber 2004: 148ff). Both of these have the features [+B, +CI], which refer to quantity. [B] stands for 'Bounded', its positive value [+B] indicating that the item “is limited spatially or temporally” (Lieber 32 2004: 136). [CI] means 'Composed of Individuals', so “[i]f an item is [+CI], it is conceived of as being composed of separable similar internal units” (ibid.). An -age or -ery derivative does then refer to a collectivity of something relating to the base. Such a paraphrase accounts for a number of derivatives, e.g. for certain senses of words like mileage 'extent or distance in miles; an aggregate number of miles travelled or covered' or jewellery 'jewels collectively'. Many other -age and -ery derivatives do not have collective readings, however. Lieber has, of course, also noticed the multiplicity of meanings that some of these derivatives show. In her book, she discusses -ery derivatives with the readings 'collective', 'location' and 'behaviour', and -age derivatives referring to 'collective', 'location' and 'condition of being/ behaviour of'. As both suffixes mainly seem to refer to collectives, she assumes that the semantic representations contained in the skeletons of these suffixes are identical. Both groups of derivatives may also have a behavioural aspect, as shown by derivatives like milksoppery 'the characteristics or behaviour of a milksop'. This is, according to Lieber, no accident, but the result of a regular sense extension. As the 'behaviour' derivatives take nouns denoting a person as their bases, their reading follows from the interaction with these base nouns: “So bases like buffoon, midwife, or brigand are taken to stand for 'what buffoons, midwives or brigand do'” (Lieber 2004: 149, italics in original, MS). The location readings are also explained as sense extensions. Lieber points out that the sense extension from place name to collectivity is quite common, and she suggests that this extension is “in fact bidirectional” (Lieber 2004: 150). She illustrates this by explaining that from a word referring to a collectivity of animals kept in a certain place it is only a small step to denoting that place itself. Lieber's model provides a means to account for the polysemy of derivational affixes. Affixes have a single core meaning, which is expressed by semantic features. Additional readings that may be expressed by derivatives are assumed to be the results of sense extensions, which arise due to paradigmatic pressure when a language does not provide a speaker with a suitable derivational process for the concept they want to express. This approach offers a new way of accounting for affix semantics in a structured and logical fashion. Her model is open to the addition of new features, or primitives, if they prove useful, and also aims to explain the often encountered polysemy of affixes by postulating a single core sense and sense extensions. Like all semantic metalanguages, it enables comparison not only between different affixes in a single language, but also crosslinguistically. The skeleton/ body model has already proven very influential, as it has been utilised by a number of linguists working on other affixes than those discussed by Lieber herself. Trips (2009), for example, uses it to describe the nominalising English suffixes -hood, -dom, and -ship. But this approach has also been applied to non-English affixes, notably by Uth (2011), 33 who discusses the semantics of French -ment and -age, or Melloni (2011), who applies it to Italian event and result nominalisations. In spite of all these advantages, Lieber's model also has some drawbacks. As she acknowledges herself, the use of semantic features has been heavily criticised in the past, but she decides to make use of such features nevertheless. Some fundamental points of criticism of classical categorisation, e.g. the circularity of describing categories with features that can only be acquired because the category is already known, are spelled out in Taylor (2003). Other issues are related to the fuzziness of word meaning. Semantic features imply a clear-cut categorisation of linguistic entities, but this is not usually found in language. Instead, words and affixes are polysemous and their meaning(s) may also be flexible in certain contexts. Lieber addresses some of these issues in her model, for example by postulating a core sense, which is the only sense of an affix described by semantic features, and sense extensions. In this way she can incorporate polysemy into her model, while keeping a feature-based description. But one of the major problems in applying this model to different data is the lack of criteria to distinguish a core sense from a sense extension. Lieber merely postulates a core sense for the affixes she discusses, but she does not explain how she arrived at that conclusion. Some of her remarks regarding the nature of sense extensions are also questionable. For example, she claims that the locative reading of -ery and -age derivatives is an extension of the collective reading that these affixes also express. The reverse extension, from location to collectivity, is quite common, so Lieber claims that “the sense extension involved with this equivalency (location = collectivity, MS) in in fact bidirectional” (2004: 150). Directionality of meaning change is a much discussed topic in grammaticalisation studies, and, to the best of my knowledge, this bidirectional sense extension has not been found anywhere else. Bidirectionality in language change is in fact rare (Haspelmath 2004). This does not mean that this particular polysemy of the suffixes -ery and -age can not have arisen in the way described by Lieber, but it shows that she is perhaps too quick to claim a natural relation between senses. Within her model, it is necessary to keep a single reading as the core sense, so she has to try and relate all readings to this core sense. If the derivatives were analysed outside of such a rigid model, one might arrive at a different conclusion. Another disadvantage is the model's inability to account for language change. This is also related to the categorisation principles connected with the use of semantic features. Lieber claims that the skeletal part of affix meaning is unlikely to change dramatically over time (cf. 2004: 10). This view is also supported by some studies employing her model, e.g. by Uth, who claims “[...] dass sich Nominalisierungssuffixe im Speziellen und Derivationsaffixe im Allgemeinen durch eine extrem hohe diachrone Stabilität 34 auszeichnen” (2011: 4) 2 . The -age and -ery data from the OED, however, suggests extensive diachronic change regarding the semantics of derivatives, which is difficult to account for by a model that assumes a rather rigid and unchanging skeletal meaning from the start. Because of the nature of this model, a number of a priori assumptions regarding the structure of a polysemous category have to be made. These are necessary, but possibly inaccurate, and are the main reason why this model will not be used in the present study. 3.3 Semantic Maps There is no unified semantic map approach, but rather several slightly different versions with many similarities. However, Haspelmath (2003: 213) provides a description of the semantic maps he uses that seems general enough to be applicable to the various approaches: “A semantic map is a geometrical representation of functions in “conceptual/ semantic space” that are linked by connecting lines and thus constitute a network”. An early semantic map approach 3 is described in Anderson (1982). In this article, Anderson compares the grammatical category Perfect in various different languages. To do so, he organises the different meanings of this category in a two-dimensional space, taking inspiration from Berlin and Kay (1969), who mapped colour terms, and Labov (1973), who mapped the properties of drinking implements. Anderson points out that, in contrast to those studies, there are no objective criteria for arranging the meanings of a grammatical category, but he explains how one should be able to arrive at a layout inductively. The first step is the comparison of different languages: Anderson considers meanings and the forms by which these are expressed. “If two particular meanings are often expressed by the same surface form (across a random sample of languages), then we can assume that the two meanings are “similar” to the human mind” (1982: 227). His example for such a similarity is the fact that “a person receiving a gift and a person psychologically affected are often marked in the same ways” (ibid.). Once such similarities are found, one can arrange their meanings spatially, with similar meanings being grouped closely together, and “non-similar meanings being farther apart” (ibid.: 228), although he makes no attempt to quantify these distances. If one has successfully constructed a semantic map in such a way, the meanings of a single word or grammatical category should occupy a contiguous area on the map. Anderson illustrates this with a multitude of 2 “[...] that nominalising suffixes in particular and derivational affixes in general are characterised by extremely high diachronic stability” (my translation, MS) 3 Anderson himself refers to his maps as mental maps. But as the term semantic map seems to have been established in the literature for this kind of approach, it will be used here instead. 35 maps comparing the different meanings of the Perfect in English, Turkish and Mandarin Chinese. In an interim summary, he stresses the universal nature of semantic maps. Comparing his work to Berlin and Kay's (1969) results, he is not sure whether “human experience does provide such universal foci for grammar” (Anderson 1982: 242), but further research should throw light on this question. But he certainly finds universal aspects in the relationships of the different meanings. Apart from the universality of meaning relationships, Anderson addresses other advantages of semantic maps, such as the possibility of accounting for diachronic changes. He also briefly touches on the notion of common meaning, or Gesamtbedeutung. He clarifies that his maps show the “normal USES of each category” (1982: 232; emphasis in original, MS), but they do not define the common meaning of a category. In fact, there might not even be such a singular meaning for each category. A more recent approach to semantic maps is Haspelmath (2003). He presents semantic maps as a way for accounting for polysemy, or, as he calls it, multifunctionality of grammatical morphemes (grams), i.e. function words and affixal categories. Haspelmath does not want to distinguish between senses and uses, i.e. between conventionalised readings (polysemy) and readings elicited by the context (vagueness). He uses the term functions instead, in order “to be neutral between these two interpretations” (2003: 212). Although he, too, arranges the meanings of grammatical categories in twodimensional space, his model is notably different from Anderson's. Unlike Anderson, Haspelmath expresses similarity not only by close proximity, but also by connecting lines between the different functions. But like Anderson, he makes no attempt at quantifying the degree of similarity between these meanings. Haspelmath also constructs his maps inductively based on crosslinguistic comparison of surface forms and their meaning(s). He claims that “it is generally sufficient to look at a dozen genealogically diverse languages to arrive at a stable map” (217). On a finished map, a grammatical category then occupies a contiguous space, just like in Anderson's approach. Although the different functions are usually selected and arranged based on cross-linguistic investigations, Haspelmath points out that the comparison of grammatical morphemes with “overlapping distribution” (2003: 218) in a single language may also lead to the successful construction of a semantic map. Cysouw (2007) proposes changes to the terminology used in the field. Haspelmath's functions become analytical primitives and these are used as a tertium comparationis to compare language-specific categories. But, more crucially, he also introduces a concept that has not played a role in previous semantic map approaches: frequency of occurrence. In the building of a semantic map the researcher has to establish which functions, or analytical primitives, are expressed by a given category. Traditionally, this matter is a yes/ no question - a function is either attested, in which case it needs to be 36 accounted for by the semantic map, or unattested. The frequency of occurrence is of no consequence, so that a rare function has the same weight as a very common one. Such rare, but attested, functions also seem much more important than unattested ones, although new, currently unattested, functions might be added if the sample was widened. The distinction between rare and unattested functions thus seems rather artificial - a result of the discrete separation of attested vs. unattested meanings - and it might merely be due to the particular language sample used for comparison (cf. Cysouw 2007: 232). He points out another problem with the received view of semantic maps, namely their imbalance between coverage and accuracy. They often predict a rather high number of categories that are not actually attested in order to cover all the attested categories. Cysouw uses a semantic map of indefinite pronouns from Haspelmath (1997) and claims that this map “has a coverage of 100%, but it actually predicts the existence of 105 different categories of which only 39 are attested” (Cysouw 2007: 234). Cysouw aims to improve the semantic map approach by incorporating quantitative notions. One possibility he mentions is to draw the connecting lines between functions in varying degrees of thickness depending on the frequency of co-occurrence of these functions. While this may be a good visualisation in some cases, the maps become rather messy and therefore difficult to interpret when more functions are added. Another method is multidimensional scaling (MDS), which is often used in recent approaches to semantic maps (e.g. Wälchli 2010, Wälchli & Cysouw 2012). Multidimensional scaling works on the basis of similarity semantics, and positions similar objects close to each other in the graphical representation. This is similar to the organisation of implicational semantic maps like Haspelmath's, which express similarity through connecting lines as well as through spatial proximity. But unlike in traditional semantic maps, these distances are quantifiable - the closer two objects are, the more similar they are. This allows some interesting conclusions concerning the relation of form and meaning, e.g. that “more similar meanings are more likely to be expressed by the same form” (cf. Wälchli & Cysouw 2012: 675). But Cysouw points out himself that semantic maps generated with this technique are not necessarily better than traditional ones, as they have to reduce the actual variability that is found in the data to two dimensions (cf. 2007: 237). They therefore do “not suffice as a model for the cross-linguistic variation, because there is an arbitrary cut-off point of data-reduction as determined by the dimensionality of display” (ibid.: 237). The work done by Cysouw (2007) and Wälchli & Cysouw (2012) shows that the incorporation of quantitative information can greatly improve traditional semantic maps. There are a number of different ways to add this kind of information, but, unfortunately, all of them have weaknesses. Which method is the most suitable depends on the amount of data considered and the type of study carried out - there is no one ideal method that fits all pur- 37 poses. A number of recent publications utilise semantic maps not only to account for specific linguistic phenomena in a variety of languages, but they have also sparked new discussions about principles of linguistic categorisation, and future interdisciplinary research between different linguistic fields (cf. the articles and comments in Cysouw et al. 2010). Semantic maps have many advantages in accounting for the meanings of grammatical categories. It should have become clear from the weight both Anderson and Haspelmath attach to cross-linguistic comparison that semantic maps are a particularly suitable tool for this. Cross-linguistic comparability can of course also be achieved by other means of meaning representation, for example by a semantic metalanguage such as Lieber's feature-based model. There are major differences in the ways semantic maps and semantic metalanguages represent meaning, however. The meanings, or functions, expressed on semantic maps are relatively concrete and can “easily be discussed, improved on, or proven wrong” (Haspelmath 2003: 231). This is not to say that the categories or descriptions used by Lieber, or in other semantic metalanguages, cannot be adapted or improved, but because the inventory of such approaches is restrictive, the addition of a new feature is more difficult than the addition of a new function on a semantic map. The features also have to be rather abstract, because the limited number of features should be relevant to describe a number of readings. The semantic map approach, on the other hand, does not impose conditions on the number of functions employed. If new readings are encountered, a new function can be easily incorporated into the semantic map. So although Lieber works with a core meaning and sense extensions and only describes the core meaning with the help of features, her descriptions remain much more abstract than a semantic map of the same grammatical category. A network of different semantic functions also avoids a number of problems that other methods of meaning representation might have regarding sense disambiguation. There is, for example, no need to distinguish between polysemy and vagueness on a semantic map such as Haspelmath's (cf. Haspelmath 2003: 231). Furthermore, semantic maps do not make a priori judgements regarding the way that meanings are structured. It has often been found that word meanings, including function words, are organised around a prototype (Lakoff 1987, Tyler & Evans 2001). In such an approach, the prototype may be the most basic use, sense or function of a word or grammatical category, or an abstraction that contains only the most prominent meaning components. Other, less salient senses are clustered around that prototypical sense. There are also models that do not use the notion of prototype, but that still display a similar layout. Lieber, for example, works with a single core sense and sense extensions that are directly related to the core sense. Such radial structures account quite well for some words, but, as Haspelmath points out, it probably “is not a good strategy to look for one 38 single central sense in all cases” (2003: 232). As semantic maps do not make a priori judgements to this nature, they can account both for categories that have a single core sense and additional sense extensions, and for those that have a different structure. Another advantage that is related to the cross-linguistic application of semantic maps is their usefulness in generating implicational universals. The layout of the maps is claimed to be the same in all languages, different grammatical categories in different languages merely occupy a certain part on a map. Implicational universals then “emerge as an automatic side effect of the construction of a map that allows the representation of cross-linguistic similarities and differences” (Haspelmath 2003: 233). Connected with implicational universals, and extremely important for the purposes of this study is the ability of semantic maps to represent diachronic change. If the functions of a category change over time, this category will shift the space it occupies on the map, which can easily be represented. If a category's functions do not change, it simply does not move on the map. This means that a semantic map can account for change if it happens. But semantic maps can do more than merely represent diachronic change after it has happened - they can also predict the path along which this change will progress. It is assumed that the meanings of grammatical categories progress from one function to an adjacent one, but they cannot skip functions. So, similarly to the implicational universals generated by the maps, they allow to predict which meanings a category will likely express next if it is subject to change. Semantic maps can thus show the directionality of language change. 3.4 Summary This section has introduced a number of very different approaches to the semantics of grammatical categories. It was shown that Lieber's skeleton and body model, although it addresses the issue of derivational affixation and semantics in a systematic way, has some disadvantages. The semantic map approach, on the other hand, is more flexible and offers a relatively unbiased way of modelling language data. Semantic maps thus seem to offer a good way of modelling the semantics of complex words. However, semantic maps are usually used to account for the functions of inflectional categories across a large variety of languages. This is substantially different from the aim of the present study, namely to describe the semantic structure of the derivatives of only two affixes in a single language. The semantic map approach thus has to be adapted considerably in order to be applicable to this type of research. These adaptations are introduced in the next chapter. 39 4 Adapted Semantic Map Model The semantic maps used in this study are based on the model described in Haspelmath (2003). A number of adaptations had to be made to that approach, mainly because of the different kind of data analysed in this study. Haspelmath builds his semantic maps on the basis of a wealth of data from different languages, while I study two suffixes, -age and -ery, in one and the same language. This is the equivalent of only two categories in Haspelmath's terminology, and would not be sufficient to create a semantic map with the methods he uses. Another reason for adaptations is the lack of frequencybased information in the original approach. Cysouw (2007) and Wälchli and Cysouw (2012) have found ways of incorporating such information into their semantic maps, but they also base these on large amounts of crosslinguistic data. The challenge here lies in finding a way to take advantage of the strong points of semantic maps, while adapting their use to a completely different dataset. This chapter gives a detailed account of the adapted semantic map model that is used in this investigation. It describes the assumptions behind the adaptations that are made to similar previous approaches and explains how the semantic maps found in the following empirical part are constructed. It also provides information on how the readings of derivatives are captured on the maps, and how the semantic structure of the derivatives can be analysed with the help of this method. 4.1 Readings and Ontological Categories The main components of a semantic map are the functions (Haspelmath 2003) or analytical primitives (Cysouw 2007). These primitives are labels that convey information on the function or semantics of the members of a particular morphological category. In order to arrive at such labels, the semantics of the derivatives has to be analysed and categorised. It is not intuitively apparent, however, what such a semantic categorisation should look like. The first part of this chapter reviews some approaches to semantic categorisation before introducing the system that is used in the present investigation. A candidate for a universally accepted classification are ontological categories, as these are used in semantic analyses from very different theoretical backgrounds. Feature-based approaches, for example, which originate in the structuralist tradition, make reference to these categories. This has recently been done by Lieber (2004), who proposes a distinction between SITUATION on the one hand, and SUBSTANCE / THING / ESSENCE on the other. To describe 40 members of these categories, she uses well-known features like [material] or [dynamic]. Scholars working with other theoretical frameworks have also made reference to such higher-level categories, e.g. Haselow (2011) with a cognitive approach to word formation. He uses conceptual categories like ACTION or THING , which are remarkably similar to the ontological categories employed elsewhere. A reason for this widespread use of ontological categories is that they are at least partly “independent of the structure of particular languages” (Lyons 1977: 449). Unfortunately, they are not used in a uniform fashion. Nearly every study that makes reference to these categories uses them in a slightly different way, sometimes consolidating a number of them into a larger category, sometimes making finer distinctions. This means that there is no ontological system that is universally accepted. What adds to this puzzling condition is the fact that ontology is not only a concept used in linguistics, but also in philosophy, psychology, computer science, and other fields. Naturally, there is some discussion about the exact number and types of categories, and many different systems of classification currently exist. A good example for this situation is the category type EVENT , which is often named as an ontological category, and is much discussed in linguistics and other fields. This category is quite complex and contains a number of subcategories. Casati and Varzi (2006) mention many different approaches to the exact categorisation of this group, which shows quite clearly the extent of disagreement between different scholars. According to Casati and Varzi (2006), four different types are regularly mentioned: activities, accomplishments, achievements, and states. Some scholars propose this four-way distinction, others classify accomplishments and achievements together, or see achievements as different from the three other types, which are then grouped together. A few examples of ontological categorisation in linguistics will further illustrate their diverse application. Dixon (1991) assumes that “words can be grouped together in a natural way into large classes that have a common meaning component” (ibid.: 6). He calls these classes “semantic types” (ibid.). These semantic types are different from syntactic categories such as noun or verb, but they sometimes correlate with one another. He asserts that the noun class always contains “words with [ CONCRETE ] reference” (ibid.: 9). An interesting point regarding membership in the semantic types is raised slightly later. According to Dixon, “central representatives of a type […] have unequivocal membership” (Dixon 1991: 75), but other words may combine elements of different semantic types. He thus seems to believe in a prototypical organisation of semantic types. Most of his monograph deals with verb semantics, but a short section also introduces semantic types of nouns. He proposes five different semantic types: 1. CONCRETE , 2. ABSTRACT , 3. STATES and PROPERTIES , 4. ACTIVITIES , and 5. SPEECH ACTS (ibid.: 76). Most of these types can be further divided. The CONCRETE type, for example, contains the subtypes HUMAN , other ANIMATE , (body and other) PARTS , and IN- 41 ANIMATE , which may again have subtypes themselves. A few example words for each type are given, but a definition of the semantics of each type is not provided. This makes it difficult to apply this categorisation in practice. Also, it is not explained what makes these semantic types a natural distinction, which is a claim Dixon makes explicitly. To have a main type comprised of SPEECH ACTS , for example, seems to be a language-centred perspective, and a distinction probably not made by non-linguists. Lyons (1977) offers another way of classifying nouns and nominals based on ontological considerations. He proposes the use of first, second, and third order entities. First order entities are physical objects, including persons. Lyons offers a set of criteria to define membership to this class: “it is characteristic of all first-order entities (persons, animals and things) that, under normal conditions, they are relatively constant as to their perceptual properties; that they are located, at any point in time, in what is, psychologically at least, a three-dimensional space; and that they are publicly observable” (Lyons 1977: 443). This class seems to contain lexical units that are sometimes described as concrete, while the other two classes contain lexical units with what is traditionally labeled abstract reference. There is, however, a difference between these last two classes in Lyons' terminology. Second order entities “mean events, processes, states-of-affairs, etc. […]; and by [thirdorder entities] we shall mean such abstract entities as propositions, which are outside space and time” (ibid.). But even though Lyons develops such an elaborate system based on ontological considerations, he does not include all ontological categories that would be relevant to describe language: “It must be emphasized, however, that our threefold classification is not intended to be exhaustive. Nothing has been said, for example, about the ontological status of numbers, sets, etc. [...]” (ibid.: 446). A more recent approach is Haselow (2011), who uses five conceptual categories based on cognitive principles to analyse complex nouns in English. These are PERSON , OBJECT , LOCATION , ACTION , and ABSTRACT . While the first three labels are quite intuitive, the latter two warrant some explanation. Haselow classifies those words as ACTIONS that “denote a concept that implies a temporal contour, thus having a more or less definable starting point and a more or less definable end point” (2011: 67). ABSTRACT words “are usually mental constructs, which become materialized either through particular actions which they are related to, or through objects or persons for which a particular abstract quality is characteristic” (2011: 68). These few examples illustrate the observation made earlier about the diverse application of ontological categories. However, they also show that certain categories appear quite often with similar definitions. Dixon's classification corresponds to Haselow's, as well as Lieber's in many ways, and Lyons also uses similar categories, although he groups them differently. These similarities are even more striking since these scholars are from different theoretical backgrounds. So although there is no definitive set of onto- 42 logical categories, some categories appear regularly in the literature and can be considered canonical. Murphy (2010: 138) gives some examples: THING , SITUATION with the subcategories EVENT and STATE , PROPERTY , MANNER , PLACE , TIME , DIRECTION , AMOUNT . With the exception of the subcategories of SITUATION , she does not specify how these categories relate to each other - whether they are on the same level, or whether they can be grouped together into more general categories. A number of categories that are particularly well-described are those surrounding actions and events. Most studies concerned with words or sentences describing things that happen work with the category SITUATION , which can be divided into EVENT and STATE , with possible additional categories and subcategories. The most canonical division is that of SITUATION into EVENTS and STATES though. In his seminal work on aspect, Comrie (1976) distinguishes two main situation types, stative and dynamic. He defines states as situations that are naturally continuous: “unless something happens to change the state, then the state will continue” (1976: 49). A dynamic situation is different in that it “will only continue if it is continually subject to a new input of energy” (ibid.). The two types are similar, as they may both involve change, and are thus subcategories of a higher-level category. This differentiation is generally recognised in the literature and will be accepted here as well. The morphological category of derived nouns, which often have readings that fall into the SITUATION category, is also well researched (e.g. Chomsky 1970, Grimshaw 1994, Melloni 2011). Such nominalisations often have event or action readings, but various result interpretations are also common. The suffixes analysed in the present study attach mostly to nouns and verbs to form nouns, so while some derivatives are deverbal nominalisations, others are non-transpositional. Unfortunately, a well-described classification of the semantics of non-eventive nouns comparable to that for different SITUATION types is not available as far as I am aware. The eventive nouns in my data can thus be classified according to criteria that have been proven to be useful, but the other formations cannot. It is unlikely that all the different categories that have been proposed in the literature are going to be applicable to the derivatives of -age and -ery, as these represent only a small part of all nouns, and it is not the aim of this study to find a universal system to classify all possible noun meanings. I therefore approach the problem of categorisation partly in a top-down, and partly in a bottom-up fashion. This means that some categories are specified a priori. These are rather broad ontological categories, such as EVENT and STATE , which have been identified as important for the meaning of nouns in the literature. In particular, Lieber's broad distinction between SITUATION and SUB- STANCE / THING / ESSENCE seems useful here (Lieber 2004). She uses Comrie's criteria to differentiate between the subcategories STATE and EVENT in her ontological category SITUATION . A further distinction in SUB- 43 STANCE / THING / ESSENCE is that between ABSTRACT and CONCRETE , for which she uses her features [+ material] and [material]. This differentiation is well-known in lexical semantics, and is proven to be a useful one. It will thus be adopted in this study as well. Lieber adds an additional level at this point and introduces the [dynamic] feature to separate concrete and abstract words with a situational flavour from those that do not have one. This is the result of her aim to use cross-categorial features. She does not introduce the [material] feature into the SITUATION category, however. Her reasoning for this is the following: “What I seem to be saying is that in some sense [ SUB- STANCE / THING / ESSENCE ] have ontological priority over [ SITUATIONS ]. Put informally, things can be processual, but processes, events, and states can't be “thingy” without, of course, ultimately being things” (Lieber 2004: 27). Lieber argues that all SITUATIONS involve “participants or arguments”, which belong to the SUBSTANCE / THING / ESSENCE category, “but [ SUBSTANC- ES / THINGS / ESSENCES ] do not presuppose situations” (ibid.). Lieber's broad distinctions SUBSTANCE / THING / ESSENCE and SITUATION with their two subcategories each are used for this investigation. All derivatives are classified into one of the four categories CONCRETE , ABSTRACT , EVENT , or STATE . If the words are polysemous across different categories, they are recorded in all applicable groups. There are, however, some differences to Lieber's manner of categorisation: Lieber classifies all nouns exclusively as members of the SUBSTANCE / THING / ESSENCE category, but the present study also classifies them as SITUATIONS . This ontological system is remarkably similar to the categorisation found in Haselow (2011), who distinguishes between PERSON , OBJECT , LOCATION , ACTION , and ABSTRACT . The first three of these are subsumed by my category CONCRETE , Haselow's ABSTRACT and my ABSTRACT would be fairly similar, and his ACTION seems to be more or less equivalent to my EVENT . As mentioned above, this account is based on cognitive linguistics, and thus comes from a different theoretical background than Lieber's model. The fact that these two classifications are so similar again speaks for the validity of the categories. The four lower level ontological categories used in this investigation with example readings and derivatives are given below. CONCRETE : This category contains all words that are material. This includes persons, animals, and objects, but also locations and collectives. Examples are pelfry 'a pilfered or plundered item', pastry 'a stiff but malleable mixture of flour moistened with water or milk and kneaded to make dough', marriage 'a spouse', glovery 'a place in which gloves are made or sold', and twiggage 'twigs collectively'. ABSTRACT : This category contains derivatives that are not material. These derivatives refer to, among others, mental concepts, charges, and rights. 44 Examples are ancestry 'distinguished or ancient descent', gluttonry 'gluttony', plankage 'a charge levied for the use of planks at landing places', and coinage 'the right of coining money'. EVENT : A SITUATION that will only continue if energy is continuously put into it. Otherwise, it will change. This group contains practices, behaviours, and actions. Examples from my data are avauntry 'boasting', banditry 'the practices of bandits', madcappery 'the behaviour of a madcap; impulsive or reckless conduct', and scourage 'the act of scouting or skirmishing'. STATE : A SITUATION that will continue if nothing happens to change it. This category contains conditions and positions. Examples are problemage 'the state or condition of being a problem', and portmanry 'the position or rank of a portman'. These categories are obviously rather broad. While such rough distinctions suffice for some studies, for example those comparing derivatives of many different word formation processes, like Haselow (2011), a more fine-grained approach is necessary here. Previous discussions of the investigated suffixes have shown that the semantics of their derivatives often cluster around certain readings. For example, both -age and -ery derivatives often refer to collectivities, but -ery formations also denote locations, and -age is known to have a tax reading in ME (e.g. Dalton-Puffer 1996, Fleischman 1977). Such interpretations are not on the same level as the ontological categories, but they can be placed within these larger categories. If a derivative denotes a collectivity of objects, it can be placed within the CONCRETE group; if it refers to a collectivity of abstract concepts, it can be considered an ABSTRACT derivative and so forth. Within the ontological categories the various readings are grouped together, allowing for all CONCRETE collectives to be analysed and compared to all the CONCRETE derivatives denoting an object, for example. But the CONCRETE collectives can then also be compared to the ABSTRACT collectives. In this way, common properties between similar readings in different ontological categories but also within a single category can be found, and differences between the various groups of derivatives can also be discerned. In contrast to the ontological categories, the readings are more specific, and also often particular to the suffixes discussed here. As previous discussions of the suffixes exist, a number of readings were expected to occur, yet a reading is only established if a number of derivatives with that interpretation are attested in the data. The readings are based on the formulations of the semantic paraphrases and quotations found in the OED. If a paraphrase, for example, mentions an 'action of doing something', the derivative is classified as having an ACTION reading, and if the paraphrase includes a formulation like 'X collectively' the derivative is assumed to have a COLLECTIVE 45 reading. Most of the paraphrases are composed in a way that makes the semantic classification relatively straightforward, but a small number of formulations require a certain amount of interpretation because they do not contain any obvious markers like the ones mentioned above. Examples for this are lovage 'praise, honour', or victorage 'victory'. Such difficult formulations turned out to make especially the distinction between ABSTRACT and STATE readings difficult. In order to be consistent, the remarks Lieber makes about her ontological categories are used as a guideline in this regard. She lists properties and conditions as belonging to the STATE category, and her examples for ABSTRACT include such words as love. The above examples lovage and victorage are thus classified as ABSTRACTS , while only conditions and positions are assumed to be STATE readings. These words refer to SITUA- TIONS , e.g. barnage 'childhood', because beginning and end points of a state are inherent to the conditions they describe. This makes them different from words like victorage, which refer to more general concepts without necessarily making reference to a SITUATION . Another problem lies in the inconsistency of the paraphrases provided in the OED. Similar words are sometimes described very differently, but it is doubtful whether this signifies a marked difference in their semantics. The diverging formulations are instead probably due to different lexicographers creating them. These paraphrases are not composed with a study like the present one in mind, so they are not always ideal for my purposes. In order to keep the results as objective as possible, such unexpected differences in the paraphrases are noted, but they do not affect the categorisation. For that, each paraphrase is considered on its own, and each derivative is classified accordingly. The advantages of using the OED do, however, far outweigh such drawbacks. Even if not all of the paraphrases are ideally formulated, they still provide an objective basis for this investigation, exactly because they have been created by trained lexicographers independently of the present work. With such an approach, there is only a small chance of a bias towards particular interpretations according to previously held assumptions or expectations. It is important to point out that the readings used here are abstractions. They are given mnemonic labels expressing a semantic property that all of these derivatives have in common. An example from the data will illustrate this point: Middle English -age derivatives often have a locative reading, e.g. hermitage 'the dwelling of a hermit', thanage 'land held by a thane', or reclusage 'a place of seclusion'. While all of them denote some kind of location, the individual locations they refer to are actually quite different, and include houses, limited areas outside, or even large parts of a country. So the locative interpretation that all of these words share does not take the same form in each derivative, but is an abstraction of the individual instances. It does not provide precise information on the exact kind of location denoted by the derivative, but is only meant to show which properties these derivatives 46 share. The reading assigned to the formations hermitage, thanage, and reclusage is LOCATION . New mnemonic labels can be created if a number of derivatives have semantics that do not fit into any of the already established groups. As a general rule, the number of groups is kept relatively small. It is of course possible to make even more fine-grained distinctions by positing a high number of more specific readings, but this is not necessary for the present study. In order to compare the readings of two morphological categories, in this case two suffixes, with one another, it seemed advisable to use more general readings rather than too specific ones. This is not a theoretical limitation, and the question of how homogeneous the reading groups should be depends on the overall purposes of the investigation. The readings arrived at through this bottom-up procedure are, as already pointed out, very similar to those proposed for the two suffixes in the literature, e.g. in Marchand (1969) or Dalton-Puffer (1996). This is hardly surprising, because these descriptions are based on the careful consideration of language data as well. But this similarity justifies the procedure used in the present study, and also makes the results arrived at here comparable to previous work. The following readings are encountered in both the dictionary and the corpus for both -age and -ery derivatives. As readings that are part of the same ontological category are grouped together on the semantic map the list is organised in the same way. CONCRETE Readings OBJECT : Derivatives with an OBJECT reading denote physical objects, both countable and mass. This includes animals. Examples are septage 'waste matter', poultry 'a bird', and wormery 'a container in which worms are kept'. PERSON : This reading is assigned to derivatives that denote a human being as well as imaginary beings, e.g. devilry 'a demon', lavendry 'a laundress', or personage 'a person of high rank'. LOCATION : This group contains formations that denote places such as houses, but also larger areas of land, e.g. baronage 'the domain of a baron', teacherage 'a house or lodgings provided for a teacher by a school' or plushery 'a luxurious or high-class restaurant'. COLLECTIVE : Derivatives with a COLLECTIVE reading denote a collectivity of any of the previous three CONCRETE readings, e.g. blossomry 'blossoms collectively', cousinage 'kinsfolk collectively', or signage 'signs collectively'. This reading has much in common with AMOUNT in the ABSTRACT category, but unlike AMOUNT it denotes exclusively CONCRETE entities. ABSTRACT Readings GENERAL ABSTRACT : Formations with this reading refer to quite general abstract concepts such as gluttonry 'gluttony', desidery 'desire, wish', or lovage 'praise, honour'. 47 CHARGE : Derivatives with a CHARGE reading refer to taxes, payments, benefices etc. Examples are ferriage 'the fare or price paid for the use of a ferry', tonnage 'a tax or duty formerly levied upon wine imported in tuns or casks', or canonry 'the benefice of a canon'. TENURE : This reading is assigned to formations that refer to tenures on land or property such as sergeantry 'a form of feudal tenure on condition of rendering some specified personal service to the king', villeinage 'the tenure by which a feudal villein held or occupied his land', or bondage 'the tenure of a bonde'. RIGHT : Derivatives with a RIGHT reading refer to feudal or religious rights or privileges e.g. coinage 'the right of coining money', pickage 'the right to collect a toll paid for breaking the ground in setting up a booth, stall, tent etc. at a fair or market', or advowry 'the right to present a member of the clergy to a particular benefice or living'. AMOUNT : The formations in this group refer to a collectivity of abstract entities or a number or share of abstract entities, e.g. lollardry 'the tenets of the Lollards', headage 'the number of animals', or wattage 'an amount of electrical power'. This reading is therefore parallel to COLLECTIVE in the CONCRETE group. EVENT Readings ACTION : The distinction between ACTION and STATE is taken from Comrie (1976). Derivatives with an ACTION reading thus refer to events that require an input of energy in order to continue. Examples are arrivage 'the act of coming to shore or into port', mockage 'the action of mocking', or skulkery 'the practice of skulking'. STATE Readings CONDITION : This reading is expressed by derivatives that refer to the state or condition of being or a skill, i.e. a situation that continues even without a further input of energy once it has started. Examples are beggary 'the state or condition of a beggar', problemage 'the state or condition of being a problem', or martial artistry 'achievement or skill in a martial art'. POSITION : Words in this group refer to an office or a position within society, e.g. constablery 'the office of a constable', portmanry 'the position or rank of a portman', or bondage 'the position … of a serf or slave'. The nature of the readings established in this way is very diverse. Some, like OBJECT or ACTION , express more general entities that can also be expected to occur in a number of other other words, both simplex and complex. Others, like CHARGE or TENURE , are much more specific and thus unlikely to occur in derivatives of other affixes. Especially the more specific readings could probably be incorporated into more general readings. For example, it can be argued that houses, which are classified as LOCATION in this study, could 48 also be seen as OBJECTS , or that OBJECTS and PERSONS could be grouped together as THINGS . However, the readings described above provide the basis for an interesting comparison of the semantics of -age and -ery derivatives, and its empirical usefulness probably justifies the classification into these particular readings. It is not seen as a disadvantage of this approach that the readings are so diverse because it is exactly this mixture of more general and more specific readings that is unique to each of the two morphological categories. 4.2 Construction of Semantic Maps Each reading established in the manner described above is represented by a separate box on a semantic map. These boxes differ in size depending on the number of derivatives with that particular reading. The more derivatives express a certain reading, the larger the corresponding box is going to be. But as different affixes differ enormously regarding the overall number of derivatives they give rise to, the size of the boxes reflects the relative frequency of a reading in a morphological category rather than the absolute number of derivatives with that reading. For example, affix A gives rise to 200 derivatives, 50 of which express reading A', and affix B gives rise to 800 derivatives, 200 of which express reading B'. Although the absolute number of derivatives with the readings A' and B' is different, the boxes reflecting them would have the same size, because both readings account for a quarter of the derivatives in their respective morphological categories. This ensures direct comparability between different morphological categories. In the dictionary investigation the box sizes are based on type frequency, i.e. the number of derivatives attested with each reading. The corpus investigation contains maps based on type frequency, but also maps based on token frequency. When assembling the semantic maps the boxes are organised according to ontological category, so that readings in the same ontological category are grouped together. This reflects the fact that these readings are broadly similar. The exact distances between the different readings does not reflect the degree of difference or similarity between different readings, however. The box sizes are calculated according to the specifications spelled out in table 1 below. For example, if a reading in the dictionary investigation accounts for a share of 13% of all types, the box representing that reading measures 1.5 cm by 0.75 cm on the semantic map. As a general rule, a bigger box represents a more frequent reading. 49 % of types or tokens box size (cm) 0 - 4.9 0.5 x 0.25 5 - 9.9 1 x 0.5 10 - 14.9 1.5 x 0.75 15 - 19.9 2 x 1 20 - 24.9 2.5 x 1.25 25 - 29.9 3 x 1.5 30 - 34.9 3.5 x 1.75 35 - 39.9 4 x 2 40 - 44.9 4.5 x 2.25 45 - 49.9 5 x 2.5 50 - 54.9 5.5 x 2.75 55 - 59.9 6 x 3 Table 1: Calculation of box sizes on semantic maps The boxes can be connected by lines. These lines represent overlaps in meaning: if a single derivative has multiple readings, a connecting line is drawn between the two boxes representing these readings. This shows how interconnected the readings of a morphological category are: a high number of connecting lines, for example, suggests a polysemous category, while many isolated readings indicate a number of homonymous affixes. But the connecting lines also vary in thickness: The more derivatives express a particular polysemy, the thicker the connecting line is going to be between the corresponding boxes. By making the lines sensitive to frequency as well, it is easier to see which readings are particularly closely connected, and which may be more isolated. Note that the connecting lines are based on absolute figures, not on relative ones like the box sizes. If many individual derivatives often have the same polysemies, it is possible that one reading is a sense extension of the other (see also the discussion in section 4.3). A sense extension is here understood as an “extension of the range of meanings of a word through conceptual mechanisms such as metaphor and metonymy” (Booij 2007: 357). Sense extensions are senses that are directly related to a sense from which they originate, which is established by using three different criteria. First, the two readings have to be connected through polysemous derivatives with both interpretations. The other two criteria differ for corpus and dictionary results. As the dictionarybased investigation is diachronic, the criterion that applies here is diachronic, too. If a reading occurs consistently later than a connected reading in those derivatives that have both interpretations it is considered a sense extension. The corpus investigation is synchronic, so this criterion cannot be applied there. But the corpus study allows us to look at token frequencies, 50 which is exploited to establish sense extensions. If a reading occurs consistently less often than a connected reading in derivatives that show that overlap, a sense extension originating in the more common reading is established. These criteria are sensitive to frequency and express the tendencies of co-occurring readings to be patterned in a certain way. That means that there may be individual exceptions to this behaviour. Arrow heads pointing towards the sense extension are added to the connecting lines between two readings when a sense extension is established. As a sense extension is only established when there is a clear tendency, they are the exception. Most readings are only connected by lines, which merely signify that two readings co-occur in individual derivatives. Table 2 shows the thickness of the connecting lines as dependent on the number of derivatives with a particular semantic overlap. Seven types that refer to the same two readings account for a continuous connecting line of 0.1 cm thickness, for example. no. of overlaps thickness of connecting line (cm) 1 - 2 Dashed 0.0 3 - 5 Continuous 0.0 6 - 10 Continuous 0.1 11 - 15 Continuous 0.2 16 - 20 Continuous 0.3 21 - 25 Continuous 0.4 26 - 30 Continuous 0.5 31 - 35 Continuous 0.6 36 - 40 Continuous 0.7 41 - 45 Continuous 0.8 46 - 50 Continuous 0.9 51 - 55 Continuous 1.0 56 - 60 Continuous 1.1 Table 2: Thickness of connecting lines on semantic maps The semantic maps that are built on this basis retain the advantages of traditional semantic map models. They do not make a priori assumptions regarding the structure of a word formation process's semantics. Judgements on this aspect can be made later with the help of the finished maps, however. If changes in meaning occur, they can be displayed with ease. Such semantic maps provide a good visualisation of the different readings of a morphological category. But apart from giving a detailed picture of the synchronic state or the diachronic development of a single word formation process, they can also be used in a straightforward way to compare two or more such process- 51 es. They could therefore contribute to judgements on the similarity or difference of the semantics of different affixes, for example. The two suffixes investigated in this study are often claimed to be similar, because they can both refer to a collectivity of things or people. Comparing the semantic maps of the suffixes shows whether they refer to a collectivity with similar frequency, whether their additional readings are also expressed by the other suffix, and if so to which extent that is the case, or whether the sense extensions or connections between the different readings are similar. It would be interesting to find out whether affixes with (some) similar or even identical readings show similar sense extensions, or whether connections between readings are unique to a particular morphological category. An extension of the data to a larger number of derivational affixes whose meanings are modelled in such a way could help answer questions about the number and extent of the different readings of polysemous affixes, the way these readings are structured and connected, and the extent to which derivational meaning changes over time. Especially the last point is quite controversial, as some scholars claim that derivational affixes are semantically stable over time. Uth, for example, speaks about the “diachrone Konsistenz” 4 (2011: 273) of the derivational suffixes she investigates. She acknowledges changes in the readings of derivatives, but claims that these do not affect the abstract semantic features of the suffixes, which are diachronically more stable. The results of the present study, however, suggest that significant semantic change can be found in affixes. For example, some previously highly productive readings like the CHARGE reading of -age derivatives (see section 5.2) are virtually extinct in recent neologisms. But although this study has found evidence for the loss of certain readings, there are no new readings in neologisms either. The change that is there is limited to a change in productivity of already attested readings. It would be interesting to see whether this is the case for other derivational affixes as well. After all, the meaning change attested for -age and -ery might be rather unusual. But it is also possible that the semantics of other affixes changes even more dramatically and semantic maps could help to shed light on this issue. 4.3 Semantic Structure One of the central issues of this investigation, apart from establishing which readings are expressed by the derivatives of -age and -ery, is to find out if and how these readings relate to one another. The relations of different readings are much better explored with regard to lexemes than with regard to affixation. As it has been shown that the relations of different readings in lexemes and derivational affixes are similar (Lehrer 2003), the following 4 “diachronic consistency” (my translation, MS) 52 discussion will make frequent reference to lexeme semantics as well as derivational semantics. A term that is often used in the lexical semantics literature instead of reading is sense. Murphy (2010: 36) defines a sense as an “abstract representation of what the referents of a word have in common”. What exactly that abstract representation looks like varies between different approaches. In this study, the readings established according to the principles outlined above will serve as such abstractions or senses. Lexemes with a single sense are monosemous, while those with more than one sense may be either polysemous or homonymous. In the case of polysemy, a single word is seen as having more than one sense, while homonymy means that one is really dealing with separate lexemes that have the same form by accident (cf. Murphy 2010: 84). The distinction between polysemy and homonymy relies on the perceived relatedness of the different senses. A good example for polysemy are the different senses of the word foot. One sense of this word could be paraphrased as 'the body part that humans usually use to stand and walk on', while another sense is 'the base of a lamp'. These readings can be seen as related because of their similarities - both denote the lowest part of a physical object, and that part is utilised by both objects, the human body and the lamp, in a similar way, namely to stand on. Homonymy is, for example, shown by the different senses of the word bank. This word may either refer to an institution that deals with money, to the land directly beside a river, or to a bench. These senses are not as obviously related to one another as the senses of foot, which is why they are usually seen as the unrelated senses of three separate words bank: bank 1 , bank 2 , and bank 3 . Another factor that plays a role in this distinction is the etymology of the words or senses: While the similarity between the two senses of foot allows us to see the 'base of a lamp' sense as a somewhat specialised version of the 'human body part' sense, such a relation is harder to see between the different senses of bank. Ultimately, even these three words seem to be related however, as they all originate in a single Proto-Germanic word (cf. the OED entries bank 1 , bank 2 , and bank 3 for more information). The key to the distinction between homonymy and polysemy thus is the perceived difference between senses, and not necessarily the etymological (un)relatedness of different words. A notion that is linked to polysemy is that of a core sense and sense extension. In the above example, the 'base of a lamp' sense of foot can be seen as a sense extension of the 'human body part' sense. This core sense is the earliest sense of the polysemous word, and the sense extension is a later modification, specialised version or figurative use of the core sense. Some modifications are so common that they can be called regular polysemy. Murphy (2010: 90), for example, mentions the CONTAINER - CONTENT polysemy, which describes the fact that words for containers are also regularly used to refer to the contents of those containers, e.g. in Would you like to drink another glass? Regular polysemy is also well-known in complex words, and 53 the often cited AGENT - INSTRUMENT polysemy expressed by derivatives of the suffix -er, e.g. printer, cooker, or teacher (cf. the discussion of Dutch -er in Booij 1986) is a good example. Some of these derivatives refer to agents, others to instruments, and some can be interpreted as both. For example, teacher 'someone who teaches' denotes a person carrying out an action, while printer can denote both a person ('a person who prints') and an instrument ('a device used for printing'). The following investigation aims to find out if and, if yes, how the different readings of derivatives in the same morphological category, i.e. those that are derived by the same affix, are related. The first question to answer is then whether the readings found in the data are the related senses of a polysemous morphological category or the unrelated senses of more than one homonymous morphological category. The traditional notions of polysemy and homonymy in words thus have to be extended to the data considered here: All encountered derivatives of each of the two suffixes -age and -ery form a morphological category, and all readings expressed by these derivatives are the senses of these categories. For this purpose, the categories themselves occupy the same position as lexemes in traditional accounts, so they can be either monosemous, polysemous, or homonymous. While it is theoretically possible that -age and -ery derivatives express a single sense, this seems an unlikely option, because we already know from previous accounts of these derivatives that the readings expressed by them are quite diverse. Lieber (2004: 148), for example, calls the range of meanings shown by -age and -ery derivatives “challenging”. A single sense that accounts for all derivatives would have to be so abstract and general that it would not be useful to distinguish the derivatives of one affix from those of any other. This leaves the second option of multiple senses per affix. There are no conceptual limitations on the number of senses per morphological category, and it is to be expected that the number of senses varies from affix to affix. Whether these multiple senses are related or unrelated can be detected on the semantic maps. The connecting lines on the maps symbolise individual derivatives with more than one reading, and if two readings are often expressed by individual derivatives, the connecting line between these two readings is particularly thick. A large amount of connecting lines represents a large number of derivatives with more than one sense, which would suggest a polysemous rather than a homonymous category. Readings that are completely unrelated would probably not be expressed by the same derivatives with any regularity. An analysis of recurrent overlaps similar to an analysis of the senses of polyfunctional lexemes can also show whether the senses are related. An example from the data will illustrate this point. A relatively high number of Middle English neologisms of -age have both AC- TION and LOCATION readings, e.g. passage 'a journey by water' ( ACTION ) and 'a place at which a river may be crossed' ( LOCATION ). As these two senses cooccur much more often than others in the same derivative, the senses can be 54 seen as related rather than as coincidental co-occurrences. Apart from the fact that this overlap occurs often, the senses are also similar. Both make reference to a body of water and the action of travelling across that body. The LOCATION reading denotes a specific place that has to do with such a journey, while the ACTION reading refers to a journey on water more generally. The amount of connecting lines between the readings expressed by derivatives can therefore help to expose polysemous or homonymous relationships between these readings. If many connecting lines link all readings to each another, the morphological category is polysemous. But if the readings are unconnected, they are homonymous. Of course, the semantic maps may also expose a combination of polysemy and homonymy. There could, for example, be a number of islands on the maps that show a large amount of connectivity within themselves, and would thus be polysemous, but are not connected to other clusters of readings. This would signify a number of homonymous categories that are each polysemous. If readings are found to be related, a number of possible relations between the polysemous senses of a category are conceivable. By far the most common structure that has been found both in polysemous lexemes and polysemous grammatical categories is a radial structure. Here, a central sense is found and additional senses are analysed as extensions of the central sense. Studies conducted by Tyler and Evans (2001) and Geeraerts (1997) will serve to illustrate this structure. Tyler and Evans (2001) provide a detailed and careful investigation of the polysemies of the English preposition over. They first suggest two criteria in order to distinguish between separate senses in a more objective fashion than previous accounts of the same preposition have done. They assume that a separate sense must involve a meaning that is not purely spatial in nature and/ or in which the spatial configuration between the TR [trajectory, MS] and LM [landmark, MS] is changed vis-à-vis the other senses associated with a particular preposition. Second, there must be instances of the sense that are contextindependent, instances in which the distinct sense could not be inferred from another sense and the context in which it occurs. (Tyler & Evans 2001: 731f) But Tyler and Evans do not just claim that there are various separate senses for a preposition, they also assume that there is a single sense that all other senses of a polysemous category are derived from. They call this sense the “primary sense” (Tyler & Evans 2001: 733f.). Based on similar work by Langacker (1987), they formulate four criteria that, taken together, identify the primary sense. These are “(1) earliest attested meaning, (2) predominance in the semantic network, (3) relations to other prepositions, and (4) grammatical predictions” (Tyler & Evans 2001: 734). A further level of abstraction is added by positing a protoscene, which represents the primary sense but is an abstract concept and not a concrete instantiation of the primary sense. 55 These criteria are particularly well suited to account for the semantics of spatial prepositions. Also, they are built on the assumption that a primary sense exists, and that the primary sense of prepositions in general “is a particular spatial relation between a TR and an LM” (Tyler & Evans 2001: 731). Similar a priori assumptions cannot be made for other words or affixes, so that Tyler & Evan's criteria cannot be used for other categories than spatial prepositions without extensive adaptations. Geeraerts (1997) describes the semantic structure and development of the polysemous Dutch word legging. We do not need to go into too much detail regarding this particular investigation, the point that is important here is that Geeraerts assumes a core sense of this word from which other senses originate. His investigation is grounded in prototype theory, so he equates the core sense with the prototypical sense: “the semasiological structure of the category is characterized by a dominant core (the prototypical instantiation of the category), surrounded by peripheral instantiations that deviate in one or more features from the central cases” (ibid.: 39). Closely related to a prototypical account of polysemy such as Geeraerts's is the notion of family resemblance (cf. Wittgenstein 1968). In a category with a family resemblance structure, polysemous senses would be related to each other, but they would not necessarily all originate in a core sense. Rather, they would form a chain in which sense A is related to sense B, sense B is related to sense C and so forth. This would mean that senses A and D are neither directly linked nor are they both connected to the same sense. Wittgenstein has famously postulated such a structure for the category GAME (1968: 31ff.). Lehrer has also found such a “chain-like structure” (2003: 218) for the different senses of the English suffix -ist, thus showing that this relationship can also be expressed by derivational affixes. Mixtures of these structures are, of course, possible as well. Even if a category is established to be polysemous rather than homonymous one might find multiple core senses and sense extensions. At the same time, some readings might be directly linked to a core sense while others are linked via a family resemblance structure. It is also possible that there is no core sense in a category, but rather equally central senses that are connected to each other. Many different accounts of polysemy from various theoretical backgrounds use terms like core sense, primary sense, and sense extension (e.g. Brugman & Lakoff 2006, Lieber 2004). The terminology and assumptions are thus not always employed in a uniform manner, although there are similarities. Sense extension, for example, is a term used by Copestake and Briscoe (1995) only for predictable, regular relations between senses, but such restrictions on this expression are not always imposed. The most commonly used term of the above is core sense, and the different views on what exactly constitutes a core sense are not always as clearly spelled out by criteria and discussions as explicit as those proposed by Tyler and Evans (2001). Blank (2003: 16), for example, claims the following: “[t]he corresponding core sense 56 comprises a bundle of properties, some of which can be suppressed [...]”. His core sense is similar to Tyler and Evan's primary sense in that it can have concrete instantiations, but he does not abstract further to a protoscene. He also introduces the notion of a collectivity of semantic properties to define the core sense. If all of these properties are expressed, that particular use of a word is prototypical, but if some properties are suppressed, the word is used in a less typical way. His core sense can probably be equated with a prototypical usage instance of a word. A different approach is taken by Lieber (2004), who postulates a rather abstract core sense of affixes, which is described by a combination of the semantic features she employs. This core sense is neither a prototypical usage instance nor is it necessarily comprised of a collectivity of features, but it is the centre of a radial structure of senses similar to the core senses proposed by Tyler and Evans (2001), for example. What all of these approaches, with the exception of Wittgenstein's family resemblance relations, have in common is that they assume a single core sense at the heart of a polysemous category. The additional senses are then either directly related to the core sense, or indirectly through other readings which are connected to the core sense itself. Such an assumption has certain appeals, and is not without its reasons. In the case of spatial prepositions, for example, it seems intuitively likely to assume a single original reading, which still forms the most basic sense of the word, as this is unlikely to change much over time. Additional senses of words often arise through metaphor or metonymy, and would thus clearly be related to previously attested senses. This would also allow one to retain a form of isomorphism even for polysemous categories, which makes for a neat description of language. But languages are known to have numerous idiosyncrasies and irregularities, so that such theoretical neatness might be artificial. A priori assumptions like these could limit the discussion of language data in fundamental ways and one might be prejudiced to analyse the data in ways that suit the expectations. At the same time, it is also evident that not all senses have the same importance within a polysemous category. Some senses are used frequently, others only sporadically, for example. With the help of historical data it is possible to distinguish between senses that have been used for a long time, and senses that have arisen more recently. Also, patterns of regular polysemy have been established that explain some senses as the consequence of other already established ones. Words denoting a container, for example, can often be used to refer to the contents of the container as well (cf. Copestake & Briscoe 1995 for a discussion of such metonymies). These regular extensions are wide-spread and can usually be found in a number of languages. Such differences between senses should be taken into account, so that some senses might well be more prominent or more central than others. The notion of core sense and sense extension seems a useful way to describe such differences and dependencies. 57 In the present study, a slightly different approach to this terminology will be taken. It will not be assumed that each morphological category has to have exactly one core sense and that the other senses are either directly or indirectly related to this. A core sense is also not understood as the most basic or oldest sense of a category, but rather as a particularly productive sense, and a morphological category might well have more than one of these productive senses. A productive word formation process is a process that is likely to give rise to new formations. In analogy to this, a productive reading of a morphological category is a reading one is likely to encounter in neologisms. This means one should be able to find a number of newly coined derivatives of a particular affix that express this reading. The notion of productivity and the criteria used to determine the productivity of a reading are going to be discussed in section 4.4. A sense extension is not necessarily the outcome of regular or predictable polysemy, and also does not have be the result of semantic widening or narrowing. Instead, a sense extension is assumed to be a sense that can be linked directionally to another sense. This may be due to a process of regular polysemy, e.g. the well-known resultative interpretations of action nouns like product or agent (cf. Bauer et al. 2013: 209), but it may also be a more local pattern that cannot be found in other polysemous categories. A prerequisite for a sense extension is the existence of polysemous derivatives that show both readings under consideration. If two readings are unconnected by polysemous formations, one cannot be the sense extension of the other. If this condition is met, the diachrony of the readings is an important indicator. A reading that is attested considerably later than another reading, or a reading that is regularly expressed later than another by a number of polysemous derivatives may be a sense extension of the earlier reading. The readings ACTION and LOCATION of ME -age neologisms illustrate this relation: eight derivatives have both readings, e.g. pilgrimage 'a journey (usually of a long distance) made to a sacred place as an act of religious devotion' ( AC- TION , c1275) and 'a shrine, holy city, etc., to which pilgrims travel; a sacred place' ( LOCATION , ? a1425). The ACTION reading of these words is usually attested before the LOCATION reading. But it is not only the diachronic sequence of senses that plays a role - the token frequency distribution of different readings may also suggest the directionality of a sense extension. If a reading is expressed much more often than another it may be the origin of a sense extension. This criterion is employed only in the corpus investigation, because token frequency cannot be established in the dictionary. An example for the application of this criterion is the sense extension ACTION → COL- LECTIVE found in -age derivatives in the BNC. Types that show this polysemy usually have an ACTION reading much more often than a COLLECTIVE reading, so the token frequency of ACTION is higher. 90% of all drainage tokens, for example, can be read as ACTION , but only 13% have a COLLECTIVE interpretation. On the semantic maps, sense extensions are represented by con- 58 necting lines with arrow heads, which indicate the direction of that extension. Core senses, then, are particularly productive readings. A sense extension is a sense that either occurs later or is expressed less often than another sense directly connected with it. But not all senses have to be either core senses or sense extensions - a sense may not be directionally linked to another one, but exist relatively independently and is then called an additional sense. Such a classification allows a detailed analysis of a category's structure using clear criteria. The use of the terms productive and productivity in this context warrants some explanation, as these words evoke theoretical assumptions that are perhaps not usually made in the context of polysemy. 4.4 Productivity Productivity is an important issue for any study on word formation. The present study is not concerned with the productivity of different word formation processes, but with the productivity of different readings or senses of a single word formation process. The aspects of this notion that are relevant for the present study will be outlined in this section. I do not aim to give a definitive overview of the topic here, as such attempts have already filled whole books. For more thorough discussions of morphological productivity see for example Plag (1999) or Schröder (2011). Productivity refers to the readiness of a word formation process to give rise to new complex words (cf. Plag 2003: 44). Consider the behaviour of two suffixes to illustrate this: Some affixes are more often used to coin new words than others, the former are productive, the latter unproductive. The suffix -ness, for example, is very productive in Present Day English. It creates nouns, often on the basis of adjectives, and the derived nouns usually refer to a state or condition. Recent examples are geekiness, masslessness and nanniness, all of which are first attested in the OED after 1950. Other suffixes are considerably less productive. Nominalising -th is one of the classic examples for an unproductive suffix. It is attested in a number of English words, among them depth, growth and health, but it is generally not used to coin new words similar to these. Obviously, the suffix was at one point in time used to create these nouns, so it must have been productive then. This shows that productivity is subject to change over time - a process which is productive now might not be productive in the future, or might not have been productive in the past, but a now unproductive process might well have been productive, indeed it must have been to leave traces in the language, and might even become productive again. Despite the importance of productivity for the study of word formation and the numerous works that have been devoted to it in the past decades, 59 there is still no generally agreed upon exact definition, let alone a measurement. Although a loose definition like the one given above would probably be accepted as the lowest common denominator of many different research traditions in the field, it is notoriously vague. What exactly is meant by 'readiness'? Which new complex words should be taken into account - only those created unintentionally by a regular process or also those created intentionally by analogy? Can productivity be measured? If so, how should it be measured? These are among the issues that are discussed, and often disagreed upon, in the literature. When exactly a word formation process can be said to be productive is a matter of debate. Indeed, much of the controversy surrounding the notion of productivity is connected to this question. Some linguists take productivity to be a yes-or-no question, with a process being either productive or not. In this view, productivity is a qualitative notion. Bauer, for example, claims that “a morphological process is available if it can be used to produce new words as they become necessary” (2001: 205). A process is therefore either available or unavailable, without any middle ground. But he also distinguishes between two aspects of morphological productivity: availability and profitability. Bauer takes into account different factors that may limit productivity and believes that they are a question of availability if they are absolute, but belong to profitability if they are merely preferences. He, for example, mentions the tendency in English to avoid colour terms with three or more syllables (cf. 2001: 207). The competition between rival word formation processes with similar absolute constraints, and therefore similar availability, and pragmatic considerations are also examples of profitability. Profitability is then, unlike availability, not a categorical yes-or-no question, but rather “reflects the extent to which its availability is exploited in language use, and may be subject unpredictably to extra-systemic factors” (2001: 211). According to Bauer, availability and profitability interact to determine the productivity of a morphological process. As these interactions are very complex, Bauer does not believe that productivity can be measured in any straightforward way (cf. 2001: 162), despite the fact that many different approaches and measurements, some of them also highly complex, have been proposed. With regard to qualitative criteria for productivity, for example those proposed by Bauer, Plag concludes that “this notion boils down to the property of a given word formation process or affix to be used to derive a new word in a systematic fashion” (1999: 22), but an additional qualification is not useful. Many linguists therefore disagree with such a qualitative approach and prefer a quantitative account. They understand productivity as a scalar concept, a continuum between fully productive processes and unproductive processes as the extremes of a scale. Such an account would also predict processes which give rise to new formations but are not fully productive. In principle, all word formation processes of a language could be 60 arranged on such a scale and their productivity levels could be compared to one another. In my view, a gradual approach is preferable to a qualitative one because it mirrors the insight that morphology as a whole is a gradual phenomenon, which has already been discussed in relation to morphological processing above (cf. chapter 2). Another advantage of such a definition is the comparability of different word formation processes that follows from it. If processes, or readings in the context of this study, are not merely productive or unproductive, one would expect more or less subtle differences between them. In order to arrange them on a scale, however, ways of measuring their productivity have to be found. Various productivity measurements have indeed been proposed over the years, but many of them remain problematic. Some of those relevant for the present study will be discussed in the following. An apparently easy way of measuring productivity is the counting of derivatives of a given process: To investigate the productivity of a suffix, one counts all the derivatives this suffix has given rise to. This is a measurement of type frequency. Such a method is relatively straightforward to carry out, which ensures a certain amount of popularity. Unfortunately, it has some serious drawbacks. A large number of the counted types are probably old words, i.e. words that have been used in the language for a long time. With regard to productivity, these words are not particularly interesting, as one wants to find out how many, or how few, new words are probably going to be coined by a word formation process in the future. But such a count of established words tells us more about the past productivity of a process than about its current productivity. This is, of course, a problem that all productivity accounts that are based on previously produced utterances have to deal with. This is true for dictionary-based methods, but also for corpusbased accounts. Elicitation tests suffer somewhat less from this disadvantage, because in them subjects are asked to actively produce words, meaning paraphrases, explanations, or to judge the acceptability of a given input. Perhaps surprisingly, this type of measurement has not often been employed in productivity studies. One of the few exceptions is Schröder (2011), who investigates the productivity of English prefix verbs with the help of all three accounts: dictionaries, corpora, and elicitation tests. Although elicitation tests are a useful tool, they are also problematic, as Schröder points out (cf. 2011: 58f). She therefore employs them as a supplement to the more traditional dictionary and corpus-based accounts and not independently of these. As simple word counts can clearly not tell us much about the productivity of a word formation process, more refined measurements have been proposed. Some of the most influential recent methods are based on word frequency and were developed by Harald Baayen and his colleagues (cf. for example Baayen 1993, Baayen & Renouf 1996). As they rely on frequency, these measurements can only be used with corpora; dictionary-based studies 61 require the use of different approaches, which will be discussed below. For many of these frequency-based methods, the number of hapax legomena, hapaxes for short, is crucial. Hapaxes are word types which occur only once in a corpus - they have a token frequency of one. Baayen assumes that the number of hapaxes correlates with the number of neologisms, “as it is primarily among the hapax legomena that novel words are expected to appear” (Baayen & Renouf 1996: 75). This is only true for sufficiently large corpora, however. If the corpus is small, one will find mostly established words among the hapaxes (cf. ibid.). A measurement that relies heavily on hapaxes is 'productivity in the narrow sense', or P. P is calculated as follows: P = n 1 / N (cf. for example Baayen & Lieber 1991: 809). The numerator n 1 represents the amount of hapaxes and the denominator N the number of all tokens that are derivatives of a given word formation process. Baayen and Lieber explain: “P estimates the probability of coming across new, unobserved types, given that the size of the sample of relevant observed types equals N” (Baayen & Lieber 1991: 809f., italics in original, MS). Another measurement that relies on the importance of hapax legomena is, for example, global productivity (P*) (cf. Baayen & Lieber 1991: 818f for more information). All methods relying on word frequency have, however, been subject to criticism. Baayen's formulae in particular have been criticised by van Marle, who doubts that both word frequency and the number of hapax legomena can tell much about the productivity of a word formation process (cf. van Marle 1992). Baayen himself has acknowledged the validity of some of these criticisms and has developed new measurements in response. The different methods now “have the great advantage that they make certain intuitive aspects of morphological productivity explicit and calculable” (Plag 1999: 33). But even so, they still heavily depend on hapax legomena, and there is no one measurement of productivity that captures all aspects of it. An example from the BNC data used for this study will illustrate the difficulties that especially productivity in the narrow sense, the most wellknown of the measurements, runs into. A major problem is that there is no one-to-one relation between low frequency and productivity and high frequency and non-productiveness. There are bound to be low frequency types that are rare, and not necessarily new words, and there are also high frequency words that have been coined recently and represent productive word formation processes. Derivatives of -age that refer to a charge or tax are fairly rare in the BNC: only 48 tokens with such a reading can be found. 19 different types are responsible for this token frequency, and seven of these are hapaxes. This leads to a P value of 0,146, the highest of all readings. This should indicate a productive pattern, but further analysis shows that the opposite is true. The low frequency types with this reading are old formations that are only rarely used. Many of them are medieval terms that are only found in academic literature about the middle ages. These words are 62 clearly not in current use. Such a distortion may be due to the small size of the reading groups, the largest contains only 127 types for -age and 211 types for -ery, and most of them are much smaller than that. In any case, the value of P does not seem to be a useful indicator of the productivity of different reading groups and will therefore not be used in the empirical part of this study. Productivity is a complex phenomenon, and the number of hapaxes is only one part of this, so in order to find which readings are productive other indicators have to be used here. One of these indicators is the type-token ratio. Although P is only of limited value in the context of this study, the assumption that a word formation process is more productive if it contains many low-frequency types and few high-frequency types is, of course, still valid. The type-token ratio indicates this relationship, but, unlike P, it does not only consider hapaxes and the overall token frequency. If a reading contains many dislegomena or other low-frequency types, this influences the type-token ratio as well. The highest possible type-token ratio is 1, and the higher, i.e. the closer to 1, this ratio is, the more likely it is that a process or reading is productive. But this measure has to be used with caution and should not be the sole indicator of productivity. While a high type-token ratio may indicate a process or reading that gives rise to a high number of types with a low token frequency, which would be expected of a productive process, it may also be the result of a process or reading that contains only a very small number of types with a low token frequency. Such a process or reading could not be seen as overly productive. A good example for this is the PERSON reading of -ery derivatives in the BNC (cf. chapter 6). A mere six types are represented by 316 tokens, and the analysis of these formations shows that this reading is unproductive. The type-token ratio (0.018), however, is higher than that of some more productive readings. This method can also only be used to assess the productivity of the corpus data, because the token frequency of different types can only be determined here. A different way of establishing productivity involves hybrid formations. Hybrid formations are complex words which consist of parts from different typological origins, one of them native, the other non-native. A good example for a Middle English hybrid is thanage, with a base of native origin and a non-native Romance suffix. Hybrids with non-native bases and native affixes also exist, and are indeed more common, consider e.g. secretiveness (Romance base + English suffix). Dalton-Puffer points out that the borrowing of a derivational affix is a sign for closer language contact than the mere borrowing of a word which may function as a base for further derivation (cf. 1996: 211). When a borrowed affix is used on borrowed bases, it can be argued that the resulting derivative is not perceived as decomposable in the borrowing language, but might just have been borrowed as a whole. If such a borrowed affix is also used on native bases, however, it can arguably be perceived as an active element in word formation. And if, finally, such hy- 63 brid formations with native base and foreign affix occur frequently, the affix may be said to be productive. The important question to ask here is, of course, at which point these constructions occur often enough to consider them productive. One or two, or even a handful of occurrences are probably not enough, but such a judgement also depends on the overall type frequency of derivatives containing the affix in question: If a substantial amount of all derivatives of that affix are hybrids, even if the overall number is not very high, one may consider that process relatively productive. The analysis of hybrid formations is especially useful for this study, as it focuses on two suffixes of Romance origin, so all derivatives of -age and -ery that contain Germanic bases are classified as hybrid formations. One should probably speak of an indicator for productivity rather than a measurement here because mathematical formulas and exact figures are not postulated. The number of hybrid formations as a different way of establishing productivity seems a useful addition to other procedures, which arguably do not provide definite results either, presented above. Note that this indicator is only used to assess the productivity of Middle English neologisms. It is assumed here that 20 th century speakers do not distinguish between native Germanic and borrowed and established French words as a source for new coinages. While the type-token ration can only be calculated for the corpus data, the number and kind of hybrid formations can also be assessed for the dictionary-based investigation. A few short remarks on dictionary-based accounts of productivity seem advisable here, as such studies have sometimes been criticised (e.g. by Baayen & Renouf 1996). They have, however, been shown to be a profitable source if used with care (see Plag 1999: 96ff for a discussion of the advantages and disadvantages of dictionaries for productivity studies). Dictionary-based accounts, especially when they rely on very large dictionaries like the OED, have some particular strengths. They contain a large number of types due to the fact that an extremely wide range of texts was sampled in the compilation process. The OED, as a historical dictionary, also allows searches restricted to certain periods. Because of this, one can exclude words from the search which are no longer used, or only include words which have been first attested in a given period. This is particularly important, as one can be reasonably sure that these words are neologisms of that period. Another crucial point is the inclusion of semantic paraphrases compiled by professional lexicographers in dictionaries. Especially for a study that aims to analyse the semantic structure of a word formation process, this is extremely useful (see section 4.1 for more information on the use of the paraphrases in the present study). These advantages only hold for very large historical dictionaries, however, and do not apply to regular desk dictionaries. Also, even the OED faces the problem that its compilers might have overlooked some words. Especially productively coined neologisms are prone to this, as they often go unnoticed (cf. Plag 1999: 98). It is therefore 64 not advisable to rely exclusively on dictionary data for a productivity study, but its inclusion surely is valuable. The third possible data source, elicitation tests, is not going to be explored in the present study, although this was shown to be a valuable addition to the productivity study conducted in Schröder (2011). The main aim of the present study is the exploration of the semantic structure of a morphological category; it is not an investigation into the morphological productivity of certain word formation processes. Schröder (2011), however, compares the productivity of different word formation processes, namely prefix verbs and particle verbs. The dictionary and corpus data used in the empirical part of this investigation is seen as sufficient to answer these research questions. It will be assumed that morphological productivity is a gradual notion which can, in principle, be measured. As there is yet no accepted measurement that takes all the different aspects of productivity into account, different data sources and measurements will be combined in this study. The productivity of individual readings will be assessed based on neologisms attested in the OED and derivatives found in the BNC. For the dictionarybased investigation, the number of neologisms is the most important indicator of productivity: the more new formations in each time period are attested, the more likely it is that this reading is productive. Another indicator for a productive reading is a high number of hybrid formations in ME. Finally, a reading should only be considered productive if it is transparent. A reading is assumed to be transparent when its derivatives form a homogeneous group. This can mean that the derivatives have a highly similar structure, for example when they contain similar bases. Recurring polysemies also contribute to the uniformity of a reading. An example for a transparent group are PDE -age neologisms with an AMOUNT reading. Most of these derivatives are based on nouns that denote units of measurement, e.g. minutage or wattage. In the OED, these formations are paraphrased as 'the number or amount of X', where X stands for the base. A more opaque group are ME -age neologisms with a GENERAL ABSTRACT reading. The bases in this group differ enormously, and the derivatives have such different paraphrases as 'praise, honour' (lovage), 'alliance, confederation' (alliage), or 'pre-eminence, supremacy' (vassalage). Slightly different criteria are used in the corpus-based investigation: a high type-token ratio indicates a productive reading, because it shows that a high number of low frequency types are attested. The transparency criterion employed in the dictionary part also plays a role here. In addition to the homogeneity of the formations with a particular reading, the level of decomposability will be taken into account in the corpus investigation, because the relative frequency of base and derivative can be established. If a base is substantially more frequent than its derivative, decomposition of the formation is unlikely, but if a base is less frequent than its derivative, decomposition is more likely (cf. Hay 2003 and the discussion in section 2.3). 65 Readings that contain a high number of decomposable derivatives are assumed to be transparent. Exact figures and rankings of productivity do not seem advisable in the context of this study. It will, however, be determined which readings are the most and least productive readings in a given morphological category. The most important criteria for productivity are the number of neologisms in the dictionary data and the type-token ratio for the corpus data. Other indicators like the number of hybrid formations and the transparency of derivatives contribute to the overall productivity judgements by either supporting or qualifying the initial assessment. The readings attested in neologisms, i.e. in the dictionary analysis, are, of course, all somewhat productive, simply because they are found in new formations. But an assessment of the relative productivity of each reading is still possible here, as a reading that is expressed by many neologisms can be assumed to be more productive than a reading that is only found rarely. The most productive readings in a given time period or data source are assumed to represent core senses of that category. 4.5 Summary: Using Semantic Maps to Account for Semantic Structure The adapted semantic map model introduced in this chapter promises to be a good method for linguistic comparison, both synchronically and diachronically. It retains many of the advantages of previous semantic map models such as Haspelmath's (e.g. Haspelmath 2003), but it is also particularly suitable to account for the semantics of a single derivational affix. It is an accurate account of the attested formations and provides information on the frequency of occurrence of individual readings and on the structure of a morphological category. However, a detailed analysis of individual readings is still necessary in order to obtain information on the productivity of individual readings, and thus their status within a morphological category, as well as the exact formal and semantic make-up of derivatives in each reading group. The readings that are found to be the most productive are seen as the core senses of the morphological category under investigation. The structure within each category is formed by the different levels of productivity of the readings and the relations between them. If the readings are connected by many polysemous derivatives, the category can be assumed to be polysemous; if there are only few connections and the individual readings are isolated from each other, one is probably dealing with separate homonymous categories. If two readings are linked by a large number of polysemous derivatives one of them may be a sense extension of the other. A sense extension is postulated when a reading is either attested later or has a lower token 66 frequency in the same type than a reading that is directly connected with it. This is represented by arrowheads on the connecting lines pointing towards the sense extension. The semantic map approach is applied to dictionary and corpus data in the following part of this work. This account is supplemented with a careful analysis of the structural and semantic properties of the derivatives. 67 5 Dictionary Investigation of -age and -ery Derivatives This chapter presents a dictionary-based account of -age and -ery neologisms from the Middle English (ME) and Present Day English (PDE) periods. This investigation aims to find out which readings are expressed by the neologisms and whether these have changed from ME to PDE. In order to compare the new coinages from both periods, not only the readings themselves, but also the formal make-up of the derivatives that express them, e.g. partof-speech and semantics of the bases, are taken into consideration. Apart from providing an account of the attested readings of derivatives, this investigation also aims to analyse the structure of these readings. How are they related to each other? Are there many polysemous derivatives which connect the readings, or are these relatively separate - in other words, is there evidence for a polysemous morphological category or do we find multiple homonymous suffixes? If the readings are connected, how are they linked? Is there, for example, a single core sense which gives rise to sense extensions, are the individual readings linked by a chain of sense extensions, or is there a completely different structure? This investigation thus aims to provide detailed information on the kind of readings attested in new derivatives of -age and -ery, and also on diachronic change in these readings if that occurs. It also aims to account for the organisation of these readings and wants to shed light on the semantic structure of the two morphological categories in question. This chapter provides information on the dictionary that is used for this analysis and the methodology employed to classify derivatives before it discusses the derivatives of -age and -ery in turn. Each part contains a discussion of influential previous descriptions of the suffixes and then proceeds to analyse the Middle English and Present Day English neologisms. Semantic maps are created for the separate ontological categories and the two time periods as a whole. The maps of ME and PDE coinages are then compared to each other before a summary reviews the main findings of this investigation. 5.1 Data Source and Methodology The data for this investigation comes from the Oxford English Dictionary online (OED), which is accessible at http: / / www.oed.com. The OED is an extremely large dictionary - according to its own website it contains some 600,000 words (“about” 2013). These include not only recently coined words or words in current use, but also old words that may not be used any more. What makes the OED particularly appealing for a study like the present one 68 is the information it provides on first attestation dates of these words. The dictionary entries give semantic paraphrases of different senses, which are followed by quotes from various sources using the word in question. Crucially, these quotes are dated. The available sources do not necessarily represent the first time a particular word was used, but they are usually the first written proof of the existence of a word, which makes them a good approximation of the time period in which a word was coined. The OED provides a search function that allows us to filter the results in a number of ways, e.g. according to certain time periods or parts-of-speech. However, it does not allow us to search for derivatives of a certain affix, but one can only search for a string of letters as part of a word. For this investigation, the search was limited to nominal headwords ending in <age> and <ry> whose first attestation dates lie within two different time periods: 1100- 1499 for Middle English (ME), and 1900-2013 for Present Day English (PDE) 5 . The elements used in the OED to modify attestation dates, such as ? , a, or c in front of dates are disregarded. Only those entries or senses are considered for which the OED provides both a semantic paraphrase and a quotation. The results obtained with these search parameters still have to be cleaned up manually, because many of the words are not derivatives of the two suffixes -age and -ery. In order to be kept in the result file, a word has to contain a base that is either attested independently in the OED or that is attested as part of another derivative with similar semantics. Similarity is, of course, a somewhat subjective notion, and other researchers may classify the relations between derivatives and their potential bases slightly differently. In doubtful cases, please refer to the tables in the appendix (Appendix A - D), which list all formations analysed as -age or -ery derivatives and include the bases that are assumed to be sufficiently similar to their derivatives as well as the semantic classifications. Examples of derivatives that are based on independently attested words are altarage 'the revenue arising from an altar' and foxery 'the character, manners, or behaviour of a fox'; an example for a derivative that is based on a bound stem is curtilage 'a small court, yard, …', which is related to curtiler 'a gardener'. A semantic connection between base and derivative can be made for all these formations, as the bases occur elsewhere with a similar meaning. Such an approach is also taken by Plag (1999), who includes such formations as baptise in his data, because of the existence of other, semantically and formally similar derivatives like baptism (cf. ibid.: 107). 5 The OED updates its entries on a regular basis. The search results that this investigation is based on were obtained on 20 March 2013 (nominal headwords ending in <age> 1050-1499), 10 June 2013 (nominal headwords ending in <age> 1900-2013), 18 October 2013 (nominal headwords ending in <ry> 1050-1499), and 27 November 2013 (nominal headwords ending in <ry> 1900-2013). 69 To make sure that the derivatives were analysable and therefore at least potentially transparent at the time of coinage, i.e. that speakers could perceive them as complex words and derivatives of one of the suffixes, the independent words or bound bases had to be attested in English sources either before or at around the same time as the derivative. Some words had to be excluded from the result file because of this criterion, e.g. village, which is first attested in c1386, while its putative base villa is only recorded in 1611. Note that this derivative is included as an analysable formation in the corpus investigation because the corpus data comes from the late 20 th century, and by that point the base villa is attested independently in English. The general policy regarding which formations are kept in the result file and which are not considered is quite restrictive. The criteria described above ensure that a word is only kept if it is certain that it is a derivative of one of the investigated suffixes. Although a more inclusive approach would be possible, such a strategy might lead to data pollution, which would be highly problematic for the kind of study envisaged here. The present investigation is concerned with the semantics of -age and -ery derivatives, so if formations that are not derivatives of these suffixes are analysed, the results would not be valid. The risk of data pollution is particularly high for the suffix -ery, which is why all formations that are not clearly derivatives of this suffix are excluded from the analysis. The central issue here is to distinguish -ery from the suffix -y, which is also used as a nominalising suffix whose derivatives refer to actions or states (s.v. -y, suffix 3 in the OED). The semantic and formal similarity of the derivatives of -ery and -y arises because the suffixes have a similar origin: -y derives from the Latin suffix -ia and its French form -ie, and -ery combines the two suffixes -er, forming agentive nouns, and the above mentioned -y. An example will illustrate the difficulties in distinguishing the two from each other. It is probably uncontroversial to assume that a formation like Tartary 'the country of the Tartars' < Tartar n. 'a native inhabitant of the region of central Asia' is a derivative of -y. But what distinguishes this formation from cutlery 'articles made or sold by cutlers' < cutler n. 'one who makes, deals in, or repairs knives', which is usually assumed to be a derivative of -ery (cf. Marchand 1969)? In both cases, only a <y> is added to the base noun, both bases denote persons, and both bases are morphologically simplex. In order to understand this situation better, a few points concerning the morphology and phonology of -ery formations have to be made before the criteria used to distinguish between -ery and -y derivatives are presented. The first point is the known allomorphy of -ery: the forms <ry> and <ery> are both representations of the suffix -ery. According to Bauer et al. (2013: 178), the distribution of these two allomorphs “is prosodically governed, with -ry attaching to disyllabic feet, and -ery to monosyllabic feet”, although a few counterexamples to this rule exist as well. Secondly, two 70 phonological processes shape the appearance of -ery derivatives: elision and haplology. Derivatives like almonry 'a place were alms are … distributed' can be considered -ery derivatives although the agentive ending of their bases, in this case almoner 'a person who gives alms to the poor', does not surface in the derivative, but is substituted by -ry. This is due to elision. Haplology, a process that leads to the “omission of some of the sounds occurring in a sequence of similar [articulations]” (Crystal 2008: 224) affects derivatives that are based on words ending in / ər/ such as cutler or butcher. The suffix also contains this sound string, and the expected derivatives *cutlerery and *butcherery are not attested. Otherwise, these derivatives fit well into a pattern made by completely uncontroversial -ery derivatives like merchandry 'merchandise' < merchant n. - they are based on nouns denoting persons, and they have similar semantics. These points lead to the following criteria. If the base of a putative derivative is attested independently or within another, semantically and formally similar, derivative, and <ery> or <ry> are added to that base, the word is kept in the result file. The important point here is that the semantic relationship between base and derivative has to be plausible at the time of coinage, which makes this criterion somewhat subjective. Other researches may thus come to different results regarding the semantic connections between a formation and a possible base, but such variation would only affect a very small number of words so that no significant differences in the outcome of this investigation would have to be expected. If a headword is given with a spelling ending in <ary> or <ory> it is only kept if it is also attested with an <ery> or <ry> spelling in the OED. Formations like commissary or procuratory are therefore omitted, to exclude derivatives of the suffixes -ary or -ory. If a form of -ery substitutes the agentive suffix -er, as in almonry, the resulting formation is included in the analysis. If a formation is based on a noun that ends in / ər/ and denotes a person, such as cutlery or plumbery, these formations are assumed to be derivatives of -ery and are therefore kept in the result file. In all other cases when only <y> is added to the base form, e.g. Tartary or Barbary, the formations are excluded, as it is not certain whether they are derivatives of -ery. The criteria for -age are, of course, identical. A base for a putative derivative has to be attested independently or within a semantically and formally similar derivative before or at around the same time as the derivative. But as there are no noteworthy spelling variations, allomorphy, or phonological effects that influence the surface form of the suffix -age, it is relatively straightforward to decide whether a given formation is a derivative of -age or not. After applying these criteria the semantics of -age and -ery derivatives are analysed. The semantic paraphrases and quotations provided in the OED are the basis for this classification and these are quoted in the following, where possible, to give information on the semantics of derivatives. For details on 71 this process, please refer to chapter 4.1, which describes the ontological categories and readings into which the derivatives are grouped. For each of the derivatives the origin of its base word is determined. The OED provides etymological information, which can be used to find out whether a base word is of native Germanic origin or was loaned in the ME period from a Romance language. For the purpose of the present study, a more finegrained distinction than that between Germanic and Romance languages is not necessary. A derivative is classified as a hybrid formation if it contains a Germanic base that is not attested in a Romance language as well. Due to the relatedness of Germanic and Romance and prolonged language contact between these two families, many of the bases of the -age and -ery derivatives in ME are attested in both language families. These are not classified as hybrid formations. The number of hybrid formations is only used as an indicator of productivity for the ME data (cf. chapter 4.4). Semantic maps are then created for each ontological category and for each time period as a whole (cf. chapter 4.2). 5.2 The Suffix -age in the OED 6 5.2.1 Previous Descriptions According to Marchand (1969), -age takes verbal or nominal bases and forms nouns. He lists a number of readings for the derivatives; among them are 'condition, state, rank, office of -', 'act, fact, mode of -', 'place', 'collectivity of -'. The collective reading has a variant that forms “words denoting the total of measure units” (ibid.: 235). Elsewhere, this type is usually called 'amount'. Although most of the other readings have equivalents in French coinages, this reading is English in origin. A group of rather unusual readings is labelled 'right, liberty, toll'. Marchand explicitly says that this type became “established” (ibid.: 235) early in ME, and gives a number of examples. He also mentions the extended polysemy of derivatives with this suffix, as “[m]ost words belong to several sense groups” (ibid.: 236). Dalton-Puffer (1996) analyses the Middle English -age derivatives found in the Helsinki Corpus. She believes that “denominal AGE became analysable before deverbal AGE, as in several of the deverbal derivatives the verb appears later than the derived noun” (ibid.: 99). Dalton-Puffer discusses the unusual semantics of the derivatives. Some of the words she found have readings already described by Marchand (1969), but especially the ones with person nouns as their bases seem to behave in a predictable fashion. Deverbal formations, for example, have action readings only rarely, and the often mentioned reading 'collectivity' can only be found in non-analysable items like beverage (cf. Dalton-Puffer 1996: 100f). Her solution is to cluster the 6 A previous version of parts of this chapter has appeared as Schulte (2014). 72 readings around referential fields rather than to use more traditional senselike descriptions. Another detailed description of the suffix -age is Fleischman (1977). She discusses the suffix in various Romance languages, but also includes a section on -age in English. In addition to denominal and deverbal derivatives, which are often difficult to tell apart, Fleischman finds a few deadjectival formations. She also notices that -age shows a significant amount of polysemy particularly in English: “Although multiplicity of function has to some extent characterized -age nouns in all the languages here surveyed, it is nowhere as striking as in English” (1977: 406). She then provides a list of derivatives with these various readings. Sometimes the formation of these words is discussed in terms of the semantics of the base word, but this is not always the case. Also, the semantic groups she forms are not easy to compare to other accounts. There is, for example, one group for 'tax designations', one for 'taxes and rights', one for 'taxes and action nouns', and another for 'tax, action noun and result of action' (ibid.: 410ff.). The most recent large-scale discussion is Bauer et al. (2013). Here, affixes are grouped together and discussed according to their semantic output. The authors mention the mostly deverbal and denominal derivation of -age formations, and list examples for the different readings. This work provides an overview of this word formation pattern including its formation, phonological considerations, and resulting semantics, but it does not attempt a semantic analysis of this suffix. None of these descriptions offer a comprehensive analysis of the formation and semantic structure of -age derivatives although many important points are made in all of them. The different readings they mention are not put in relation with each other, apart from the relatively heterogeneous large clusters of senses that Marchand proposes and the various overlapping groups found in Fleischman. But the connection between these groups is not explicitly discussed anywhere. Dalton-Puffer's solution of using referential fields instead of more typical derivational meaning descriptions is a different approach, but it does not bring us closer to a solution to the puzzling semantics of this suffix. Regarding the formation, she merely asserts that deverbal and denominal patterns “cut across these semantic groups” (Dalton-Puffer 1996: 101), but does not offer a more detailed analysis. 5.2.2 Middle English Neologisms Altogether 142 words that can be analysed as -age derivatives in the OED are first attested between 1100 and 1499. These derivatives have a number of different readings across all four ontological categories (see chapter 4). The table below shows the number of types in each ontological category with examples for each group. Polysemous derivatives are counted once per onto- 73 logical category they occur in, which is why the total number of types adds up to more than 142. ontological category types examples CONCRETE 52 hermitage, skevinage, cordage ABSTRACT 74 arrearage, lovage, stallage EVENT 59 leakage, steerage, bondage STATE 22 tillage, marriage, thanage Table 3: Ontological categories of -age neologisms (ME) These categories differ enormously not only in size, but also in various other aspects. Each category will therefore be analysed separately below, but the following paragraphs discuss some general tendencies that apply to all -age derivatives. The earliest formation, hidage 'a tax payable to the royal exchequer' (a1195), is first attested in the late 12 th century. Only a small number of -age derivatives are attested this early, but the number of new coinages increases considerably as time goes on. Figure 2 illustrates this development. Figure 2: First attestation dates of -age neologisms (ME) More than one fifth of all formations are first attested before 1350, and thus in early Middle English. From the earliest attestation in the late 12 th century the number of neologisms rises continuously until the 14 th century. A plat- 1100-1149 1150-1199 1200-1249 1250-1299 1300-1349 1350-1399 1400-1449 1450-1499 0 10 20 30 40 50 60 year of first attestation number of neologisms 74 eau is reached here, before the number of new formations increases more significantly in the 15 th century, so that more than 60% of all coinages are attested between 1400 and 1499. The significant increase of the number of -age neologisms over the course of the Middle English period seems to suggest a clear increase in the productivity of this suffix. This development has to be seen in relation to the data source, however. The number of entries in the OED is not constant over time. According to the information provided on the OED website (“timelines” 2013), the dictionary contains very few entries for words that are first attested in early ME. Only 168 entries are attested for the 50 years between 1100 and 1149, for example. This number increases subsequently, but it does not do so evenly, as can be seen in figure 3. Figure 3: Number of OED entries in 50 year periods over the course of the ME period Although there are differences between the graphs representing the number of -age neologisms and the number of OED entries, the general development is similar. The number of new entries in the OED rises substantially in the late 14 th century and remains at a high level throughout the 15 th century. This is also the period during which we find most -age neologisms, so that this increase should not be misunderstood as an obvious sign for a rising productivity of this suffix. The majority of 60% of the -age derivatives is based on words of Romance origin. Another 20% contain bases that are attested in both Germanic and Romance languages. 25 words, or 18%, are hybrid formations on exclusively Germanic bases, e.g. swannage 'payment for the right to keep swans', or cartage 'the process of conveying by cart; the price paid for this'. These hybrids 1100-1149 1150-1199 1200-1249 1250-1299 1300-1349 1350-1399 1400-1449 1450-1499 0 2000 4000 6000 8000 10000 12000 year of first attestation number of entries 75 can be used as indicators for the transparency of a word formation process. Once a foreign affix is used on native elements, this affix must be perceived as a transparent and available means of word formation by speakers. The resulting derivatives are also proof that an affixational process is productive at a given time, because such formations can usually not be considered loanwords. The number of hybrid formations in a reading group can thus be used as evidence to show whether that particular reading is transparent (cf. also the discussion in chapter 4 on this point). Finally, three derivatives are based on words of unknown origin. Just under 40% of all formations contain bases that are attested as members of different word classes. The vast majority of these may either be deverbal or denominal, e.g. forceage 'the action of forcing', which is either based on force n. 'strength' (a1300) or on force v. 'to apply force' (a1300), although a few deadjectival coinages and one based on a combining form are also possible. A slightly larger group consists of exclusively denominal formations, such as villeinage 'the tenure by which a feudal villein held or occupied his land' from villein n. 'one of the class of serfs in the feudal system', or putage 'harlotry, prostitution' from pute n. 'a prostitute'. These base nouns almost always denote a concrete entity, most commonly a person or object. The remaining derivatives are based on verbs (15%), bound bases (4%), or adjectives (0.7 %). The majority of -age derivatives are thus probably based on nouns, but verbal bases are also quite common. Other bases occur only sporadically. The semantics of the resulting derivatives can be described by a variety of different readings across all four ontological categories. The following section will analyse each reading separately to find out which of them can be seen as a core reading of -age in Middle English. 5.2.2.1 CONCRETE Derivatives More than one third of all derivatives, 52 formations, have a CONCRETE reading. As table 4 shows, these denote objects, persons, locations, or a collectivity of these. reading types examples OBJECT 11 fardellage, altarage, vintage LOCATION 24 reclusage, thanage, eremitage PERSON 4 hostage, marriage, personage COLLECTIVE 19 baggage, plumage, vicarage Table 4: CONCRETE readings of -age neologisms (ME) 76 Eleven derivatives are first attested with a CONCRETE reading in early ME, i.e. before 1350. The earliest of these is pottage 'a thick soup or stew' in ? c1225. Nearly all the derivatives in this category are based on words of Romance origin. Only five formations, or 10%, are hybrid formations on Germanic bases. These are burgage 'a freehold property in a borough', cottage 'a dwelling-house of a small size', herbryage 'entertainment, lodging', lastage 'the ballast of a ship', and thanage 'the land held by a thane'. Another 13 words contain bases that are attested both in Germanic and Romance languages. The share of hybrid formations in this group is thus significantly below the average over all semantic categories, while exclusively Romance-based derivatives are overrepresented. Also, all of the hybrids are attested only in late ME. The existence of these hybrid formations suggests that -age suffixation with a CONCRETE reading is transparent and also somewhat productive by late ME, but probably not before that. More than half of the bases are clearly nouns. These mostly denote a concrete entity themselves; persons, e.g. cousin 'a collateral relative' and objects, e.g. herb 'a plant', are the most common. Some bases are exclusively verbal, e.g. marry or pill 'to rob', and three derivatives, companage 'whatever is eaten along with bread', curtilage 'a small court, yard, garth', and vintage 'the produce or yield of the vine', are formed on bound bases. For the remaining words a number of possible bases are attested, these are mostly verbs and nouns, but some adjectives are also possible. OBJECT Eleven derivatives denote an object of some kind. This includes countable single objects like fardellage 'a package', but also mass nouns referring to objects or materials, e.g. presserage 'something extracted or squeezed out by pressure', and natural phenomena like umbrage 'shade, shadow'. About half of the formations with an OBJECT reading are monosemous, but when additional readings are attested those are usually earlier than OBJECT . All of the bases already have CONCRETE interpretations themselves, so the suffixation does not add a CONCRETE reading, but merely preserves it. If derivatives are polysemous, the OBJECT reading is often very similar to the meaning of the base, e.g. in altarage 'the revenue arising from an altar' (? c1430) and 'an altar maintained for a priest' (1446), which is based on the noun altar; or murage 'a toll or tax levied for the building or repairing of town walls' (1424) and 'the building of walls. Also: a system of defensive walls' (1450) based on mure n. 'wall' or mure v. 'to wall in'. These OBJECT readings are distinct from their bases - an altar maintained for a priest is different from just any altar, and a system of defensive walls is different from any wall - but they are still quite similar, and it is likely that this reading is merely an additional reading expressed by polysemous derivatives that primarily refer to something else. This conclusion is also supported by the fact that 77 OBJECT is a later, additional interpretation when the derivatives are polysemous. The strongest links of polysemous derivatives exist with the readings CHARGE , RIGHT , and ACTION , but none of these is systematic enough to suggest a sense extension. The derivatives in this group are quite varied with regard to the kind of object they denote. In the monosemous formations this is virtually identical with the object denoted by the base, as in altarage and murage, and in the polysemous formations it seems to be a rare additional interpretation of other readings. This lack of systematicity as well as the small number of neologisms with an OBJECT reading strongly suggest that this reading is not productive. PERSON Only four derivatives denote a person: hostage 'a person … given and held in pledge', lineage 'an individual descendant', marriage 'a spouse', and personage 'a person of high rank'. All of these four words have a number of additional readings, which are mostly earlier than the PERSON interpretation. The only exception is personage, but here the base word already refers to a person, which weakens the importance of the suffix in this regard. Due to the infrequency of words with this reading it is certainly not productive. LOCATION The largest individual group among the CONCRETE category is LOCATION , containing 24 derivatives. This equals nearly half of the CONCRETE words, and 17% of all derivatives. But this group does not just stand out in terms of size, it also contains four of the five hybrid formations attested among CON- CRETE formations, and some of the earliest coinages like curtilage (1206) and pilgrimage (c1275). In spite of this, most derivatives only acquire their LOCA- TION readings in late ME. More than a quarter of the derivatives are monosemous and only refer to a location, for example cottage 'a dwelling-house', eremitage 'the dwelling of a hermit', and skevinage 'a district under the jurisdiction of a local magistrate'. All of the monosemous derivatives apart from skevinage, and also some of the polysemous ones, refer to a house or other kind of dwelling place, so a specific kind of location. These words are probably denominal and are mostly based on words denoting persons, e.g. parsonage, hermitage, or reclusage. This group of derivatives is thus highly uniform and can be considered transparent. The rest of the LOCATION group refers to larger entities, mostly areas of jurisdiction like baronage 'the domain of a baron' or thanage 'the land held by a thane'. Many of these words are also based on person nouns, but they do not refer to the dwelling place of the person. Most of these derivatives are polysemous and refer to an action as well as a location, e.g. rivage 'a coast' and 'landing on a shore', or passage 'the action of going or moving onward' 78 and 'a place at which a river may be crossed'. The ACTION reading is then usually earlier than LOCATION , so that LOCATION can be seen as a sense extension of ACTION in these cases. Such a sense extension from an action to the place in which the action happens is one of a number of common sense extensions in English eventive nominalisations (cf. Bauer et al. 2013: 209), so it is not surprising that one finds this polysemy here. Overlaps with other categories are much less frequent than the ACTION - LOCATION polysemy, but LOCATION is connected to a number of other readings nevertheless. The relation with TENURE should be mentioned, as the LOCATION reading of the words with that polysemy is probably due to a sense extension TENURE → LOCATION . There are three words with this overlap, socage, thanage, and villeinage, which is more than half of all words that refer to a tenure. All three have a TENURE reading before they also denote a LOCATION . LOCATION is a frequent reading especially in late ME neologisms. The formation of derivatives is systematic, and some of the polysemies follow a regular pattern as well. Although some early derivatives are attested with this reading, it probably only becomes productive by late ME. COLLECTIVE Derivatives denoting a collectivity of concrete entities also form a substantial group among CONCRETE derivatives: 19 words have this interpretation. Most of them refer to a collectivity of objects, e.g. baggage, or cordage 'cords or ropes collectively', but there are also some that denote a collectivity of persons, e.g. baronage 'the body of barons collectively', or cousinage 'kinsfolk collectively'. The COLLECTIVE interpretation is both frequent and systematic, as these derivatives often refer to a collectivity of the base word. If the base denotes a person, as in baronage 'the body of barons collectively', cousinage 'kinsfolk collectively', or vicarage 'a college of vicars', the derivative refers to a collectivity of these persons. Virtually all of these derivatives are based on nouns denoting a person or a collectivity of persons. If the base denotes an object, the derivative is usually understood as the collectivity of these objects, as in coinage 'coins collectively', or baggage. But not all derivatives that denote a collectivity of objects are based on such nouns. A small number of bases referring to abstract concepts, locations, and verbs are also attested. In these cases, the derivatives denote a collectivity of objects that is clearly connected to the denotation of the base: pesage 'cargo' is based on either peise v. 'to weigh' or peise n. 'weight', portage 'cargo, freight, baggage' is based on port 'a harbour', and pillage 'booty, plunder, spoils' comes from the verb pill 'to rob'. Most derivatives in this group have additional readings in other categories. The deverbal derivative pillage, for example, can also mean 'an action or an act of plundering', and lastage 'the ballast of a ship' also refers to a charge 'a toll payable by traders attending fairs and markets'. Overlaps with the readings CHARGE and ACTION are especially common. This is probably due to 79 the fact that CHARGE is very productive and takes similar bases as COLLEC- TIVE . The COLLECTIVE reading is often the first one attested, so it cannot be seen as a sense extension of the more frequent CHARGE reading. Most of the derivatives that show an overlap with ACTION are highly polysemous and have many other interpretations as well, so that it is not clear that COLLEC- TIVE is a direct sense extension of ACTION . Although the polysemies of derivatives are somewhat systematic and a number of neologisms are attested with a COLLECTIVE reading, this is probably not a productive interpretation of -age derivatives until late ME. COLLEC- TIVE is very uncommon to be encountered in early ME, and the only hybrid formation with this reading, lastage 'the ballast of a ship', refers to a charge before it is attested with a COLLECTIVE reading. However, monosemous formations with a COLLECTIVE interpretation start to occur towards late ME, and neologisms become much more frequent both as additional and primary readings in polysemous words as well. The productivity of this reading thus seems to increase towards the end of the ME period. Semantic Map of CONCRETE Derivatives The four different CONCRETE readings discussed above can be plotted on a semantic map. The different sizes of the individual boxes reflect the difference in frequency - the larger the box, the more frequent the reading (see chapter 4.2 for more details on the construction of semantic maps). The connecting lines illustrate the overlaps between the different CONCRETE readings. Dashed lines stand for one or two derivatives with readings in both groups, continuous lines for three to five polysemous formations. Figure 4: Semantic map of CONCRETE readings of -age neologisms (ME) The map reflects the differences in frequency in the various groups. LOCA- TION is clearly the most frequent reading, COLLECTIVE occurs less often, and OBJECT and PERSON contain only very few derivatives. The connecting lines show that many polysemous derivatives have readings in more than one COLLECTIVE OBJECT LOCATION PERSON 80 group, but also illustrate that not all of the readings are connected to each other in this way. The four PERSON formations, all of them highly polysemous, have one overlap with OBJECT , and one with COLLECTIVE , but none with LOCATION . Although half of the words in this group have additional CONCRETE readings, most of their additional interpretations are in the AB- STRACT category. The strongest connection within the CONCRETE category exists between COLLECTIVE and LOCATION , with the three derivatives baronage, boscage, and ménage having readings in both groups. But the derivatives in both groups have significantly more overlaps with non- CONCRETE readings. So although the different readings in this ontological category are connected by polysemous derivatives, significant regularities in this area cannot be detected. 5.2.2.2 ABSTRACT Derivatives This group contains the highest number of derivatives. As many as 74 words, about half of all derivatives, are included in this category. reading types examples GENERAL ABSTRACT 20 alliage, falsage, testimonage CHARGE 48 butlerage, plankage, stallage RIGHT 10 coinage, patronage, tronage TENURE 5 socage, thanage, villeinage AMOUNT 7 portage, superplusage, usage Table 5: ABSTRACT readings of -age neologisms (ME) These formations belong to five different reading groups. The most frequent one is CHARGE , e.g. in stallage 'a tax or toll levied for the liberty of erecting a stall in a fair or market'. Formations with a GENERAL ABSTRACT reading, e.g. in alliage 'alliance, confederation', occur less often, but are still quite common, while the other three readings RIGHT , TENURE , and AMOUNT are only rarely found in neologisms. Each of these groups will be discussed below in detail. Only 16% of all ABSTRACT derivatives are attested in early ME, which is well below the average over all semantic categories. Many of these early coinages have CHARGE readings, but TENURE is also quite common at this point. ABSTRACT readings in general, with the exception of CHARGE and TEN- URE , are not among the earliest interpretations of -age derivatives, but AB- STRACT readings become more frequent towards the end of the ME period. Hybrid formations are slightly more common in this category than in others. More than 20% of ABSTRACT derivatives are based on Germanic 81 words, compared to the average of 18%. Two of these, hidage 'a tax payable to the royal exchequer, assessed at a certain quota for each hide of land' and thanage 'the tenure by which lands were held by a thane', are already attested in early ME but the vast majority of formations is first recorded in the 15 th century. Just below a quarter of derivatives contain bases that are attested in both Romance and Germanic languages. The origin of one verb, prime 'to fill, charge, load', which forms the base of primage 'a customary payment to the master and crew of a ship for loading and taking care of the cargo', is unknown, but the remaining formations, equalling more than half of all AB- STRACT derivatives, are based on Romance words. There are no major discrepancies between this distribution and the average, which is not astonishing, given that this category is so large and therefore influences the average to a considerable degree. The same is true when it comes to the part of speech of the bases and their semantics. Slightly more than 40% of the derivatives are denominal, with the majority of bases denoting CONCRETE concepts like persons or objects. Some bases can also be qualified as AB- STRACTS , but these are in the minority. 14% of the derivatives are clearly deverbal, but more than a third have multiple possible bases. Deadjectival formations and those built on bound bases are very rare. There are, however, significant differences concerning all of these points between the various readings that make up this category, as will become clear in the individual discussions below. GENERAL ABSTRACT This group contains derivatives that have less specific readings than the other derivatives in the ABSTRACT category, e.g. testimonage 'testimony' or lovage 'praise, honour'. As a result, this group is semantically less homogeneous than the others. Twenty words are included here, so GENERAL AB- STRACT accounts for more than a quarter of all ABSTRACT derivatives, and 14% of all -age formations. About a third of these words are monosemous, but many derivatives with a GENERAL ABSTRACT reading are highly polysemous and express three or more readings each. Marriage, for example, can have an ACTION , a CONDI- TION , a GENERAL ABSTRACT , an AMOUNT , a CHARGE , a PERSON , and a RIGHT reading. Such words are not very informative when one tries to find regularities with regard to their polysemy, because they have so many possible interpretations. There seems to be a regular overlap of GENERAL ABSTRACT and ACTION , however. In the six words with this polysemy, the ACTION reading is usually earlier than the GENERAL ABSTRACT interpretation. A sense extension ACTION → GENERAL ABSTRACT can, however, not be established on the basis of this data, because these words have so many other readings that a direct relationship between those two interpretations in particular is not certain. This is also true for other polysemies like RIGHT or COLLECTIVE . As regularities regarding additional readings are hard to detect, this reading 82 does not seem particularly transparent. Even the monosemous derivatives are quite diverse. Compare, for example, the interpretations of the monosemous derivatives disadvantage 'detriment, loss, or injury to interest', victorage 'victory', lovage 'praise, honour', and alliage 'alliance'. In sum, a GENERAL ABSTRACT reading seems to arise mainly as an additional reading in words that are already highly polysemous. Due to the variety of readings expressed by such derivatives, regularities in the semantic overlaps cannot be established and it is thus not clear that GENERAL ABSTRACT is a sense extension of one or more other readings. Even the monosemous derivatives do not contribute to the semantic homogeneity of this reading, as they refer to a variety of concepts. This reading is overall quite opaque. As transparency is a requirement for productivity, a GENERAL ABSTRACT reading cannot be considered productive, in spite of its relatively high frequency. CHARGE The derivatives in this group refer to taxes, charges or other payments that have to be made. Derivatives with this reading are quite numerous in ME, as one third of all derivatives, not only the ABSTRACT ones, expresses this reading. Such a high number is certainly striking, given that this is a highly specialised reading that is fairly uncommon for derivational affixes in English. But these formations are not only frequent, they are also among the oldest -age derivatives in English. The earliest -age derivative hidage 'a tax payable to the royal exchequer' (a1195), and several other early ME coinages have a CHARGE interpretation. In spite of this, the share of early ME formations with a CHARGE reading lies at only 14%, which is below the average of 20% for all -age derivatives. The share of hybrid formations in the CHARGE group is, however, higher than the average: 21% of all derivatives with a CHARGE reading are coined on Germanic bases, compared to 18% of all -age derivatives. Almost all of these hybrids are first attested in late ME, which proves that -age derivatives with CHARGE readings are transparent derivatives by late ME at the latest. 42% of the derivatives are denominal, and the same amount contains bases that are attested as both nouns and verbs. Exclusively deverbal derivatives are quite rare. Interestingly, the derivatives that are based on verbs belong to the most highly polysemous formations with CHARGE readings, and these derivatives refer to a charge only after other readings, predominantly ACTION readings, are expressed. The CHARGE readings of most deverbal formations thus seem to be additional readings, and possibly sense extensions from previous ACTION readings. The denominal derivatives are very different in that they are often monosemous. The bases themselves usually denote concrete entities: Physical objects like swan or crane are common, but locations, e.g. pont 'a bridge' or quay, and persons, e.g. vicar or parson, are also found. Only few bases have other readings, among them are trewe 'tribute' and gavel 'payment to a supe- 83 rior', which already denote some kind of payment, or pound, which denotes a unit of measurement, and sene 'a synod, a meeting of clergy'. So overall there is a clear pattern discernable: the derivatives in this group refer to a charge or payment made to a person if the base denotes a person, or a charge for the use of an object or location if the bases denote objects or locations. Good examples for this are butlerage 'a duty formerly payable to the king's butler', pontage 'a toll for the use of a bridge', and cranage 'dues paid for the use of a crane'. Especially the objects and locations that form the bases of derivatives with a CHARGE reading are thus closely connected to actions carried out with these objects or at these locations. This connection to actions is also evident in those derivatives that may be either denominal or deverbal. If these are interpreted as deverbal nouns, the tax denoted by the derivative is paid for an action that is carried out - compare, for example stowage 'a duty levied on goods stowed' from stow v. 'to put in a certain place'. If they are interpreted as denominal formations, the bases usually denote physical objects, and the reading is identical to the exclusively denominal derivatives, as in plankage 'a charge levied for the use of planks at landing places', which is either based on plank n. or plank v. 'to provide with planks'. Unsurprisingly then, many of the polysemous CHARGE derivatives have additional ACTION readings. Although polysemous items are overall not very common in this group, as more than half the derivatives only have a CHARGE reading, two thirds of the items that are polysemous can also be interpreted as actions. Examples are ferriage 'the action or business of ferrying' and 'the fare or price paid for the use of a ferry', and cartage 'the process of conveying by cart' and 'the price paid for this'. Even the monosemous items often make implicit reference to an action, because they refer to a charge that is paid for an action, e.g. pickage 'a fee paid for breaking the ground and setting up a booth, stall, tent …' or quayage 'a tax levied on goods landed or loaded at a quay'. The overlap of CHARGE and ACTION does not only occur as the result of possible deverbal derivation, as exclusively denominal derivatives also sometimes show this polysemy. Both readings are often attested in the same semantic paraphrase, so it is impossible to say which of them is the earliest, and a sense extension cannot be established. Another recurring additional reading of derivatives in the CHARGE group is RIGHT . One of the six derivatives with this reading is pickage 'a toll paid for breaking the ground in setting up a booth, stall, tent, etc., at a fair or market' and '(also) the right to collect such a toll'. All of these factors, the frequency of derivatives with a CHARGE reading, the high number of monosemous formations, and the regularity of additional readings lead to the conclusion that this reading is highly productive in ME and should be considered a core reading of -age derivatives. 84 RIGHT Derivatives with this reading form a relatively small group with only ten members. All of these words are first attested in late ME or acquire this reading only at that point. The structure of these formations is slightly unusual, as most of them are either clearly denominal or clearly deverbal, and only three have possible nominal and verbal bases. Most of the nominal bases denote physical objects, e.g. altar or mure 'a wall', but this regularity does not lead to a transparent word formation process. Only one derivative is monosemous, and the remaining words have many additional readings. Most of these words have an additional reading in one of the CONCRETE groups, and six derivatives, mostly the deverbal ones, also refer to an ACTION . Another six formations, more than half of this group, have an additional CHARGE reading. Both readings, CHARGE and RIGHT , are normally attested within the same paraphrase, so that it is not clear which of them is the earlier interpretation. A sense extension can therefore not be established. In sum, these words are clearly highly polysemous, most of them having at least three different interpretations, and this is not a transparent group of derivatives. The complete lack of hybrid formations also supports this assessment, as does the fact that the RIGHT reading is usually a later addition to a derivative coined with another reading. This reading is not productive. TENURE Five derivatives denote a form of tenure, for example villeinage 'the tenure by which a feudal villein held or occupied his land' or thanage 'the tenure by which lands were held by a thane'. The small number of items with this reading suggests that this is a rather marginal group. Two of these words are first attested with a TENURE reading in early ME, so even if this is not a frequent reading, it is certainly an early one. This group also contains two hybrid formations, bondage 'the tenure of a bonde' and thanage 'the tenure by which lands were held by a thane', which leads to a very high share of hybrids at 40%. Because of the small number of words altogether, this high share should not be over-interpreted. Remarkable is the lack of verbal bases in this group. Four derivatives are denominal, and one is denominal or deadjectival. The semantics of these bases is also quite uniform: four out of five bases denote persons. This regularity suggests a transparent word formation pattern. Regularities are also found in the polysemies of these derivatives. All of them have additional readings: LOCATION and POSITION are the most common, but ACTION , CONDI- TION , and GENERAL ABSTRACT also occur. The overlaps with LOCATION probably occur because TENURE is already obviously connected with LOCATION , as the derivatives refer to a tenure that is held on a property or place. More than half of all derivatives in this group have an additional LOCATION reading, which is always later than their TENURE interpretation. This systematici- 85 ty suggests a sense extension TENURE → LOCATION . The overlap of TENURE and POSITION is probably due to similar bases selected by both readings rather than to a connection between the two readings themselves. Person nouns dominate both groups, and the derivatives may denote a tenure held by a person or a position or office held by that person. The fact remains that the derivatives are highly polysemous and these polysemies are not entirely predictable. This reading is therefore not completely opaque, but it should not be considered a productive reading, as the number of neologisms is very low. AMOUNT Seven derivatives have an AMOUNT reading, e.g. arrearage 'an amount overdue' or vantage 'a greater amount of something'. Not all derivatives that are included here have such a straightforward reference to an amount in their semantic paraphrases. Some of the derivatives refer to an amount of money, but others, like superplusage 'a surplus amount' do not specifically refer to money, and usage 'the body of rules or principles followed by a particular group of persons' refers to a collectivity of abstract concepts. Most of the derivatives in this group are polysemous. The most common semantic overlap exists with CONDITION (three polysemous formations), and GENERAL AB- STRACT and ACTION are each found in two of the seven derivatives. Most of the polysemous derivatives only acquire their AMOUNT reading after other readings, but a clear sense extension cannot be established due to the small number of words in this group, their heterogeneity, and the high amount of additional readings across all ontological categories. This reading thus forms a relatively heterogeneous group, and this heterogeneity does not only apply to the interpretations of the derivatives. The structure of these words is also quite diverse, with exclusively nominal, exclusively verbal, and verbal and nominal bases being attested in the very small group of derivatives. Monosemous words are in the minority in this group, and some of the most highly polysemous words like marriage or advantage are found here. All of this limits the transparency of these derivatives. These words are also not particularly frequent, so that AMOUNT cannot be considered a productive reading. Semantic Map of ABSTRACT Derivatives The semantic map visualises the significance of derivatives with a CHARGE reading. It is by far the largest group in this category. The next most frequent reading, GENERAL ABSTRACT , is already far less common, and the other readings in this category, especially TENURE , account for only a small number of derivatives. Apart from TENURE , which is fairly isolated within this category, the readings are highly interconnected through polysemous derivatives. The closest link exists between CHARGE and RIGHT , as most derivatives with 86 RIGHT readings also refer to a CHARGE . GENERAL ABSTRACT words also often have a CHARGE reading, although this connection is less pronounced. AMOUNT and CHARGE on the other hand overlap only rarely with one another. GENERAL ABSTRACT is the only reading that is connected to all four other groups in this category, which is probably due to its fairly heterogeneous make-up. Figure 5: Semantic map of ABSTRACT readings of-age neologisms (ME) 5.2.2.3 EVENT Derivatives This category contains more than 40% of all derivatives. These derivatives are not subdivided into different readings, but can all be classified as AC- TION . Good examples are arrivage 'the act of coming to shore', towage 'the action or process of towing or being towed', and wharfage 'the stowage of goods on, or loading or unloading at, a wharf'. Many of these are first attested in early ME, but the share of coinages from before 1350 is lower than the average of early ME coinages of all -age derivatives. New formations with an ACTION reading become especially numerous towards the second half of the 15 th century. The share of hybrid formations in the ACTION group at 25% lies significantly above the average of 18%. None of these formations is attested with an ACTION reading in early ME, the earliest hybrids are bondage 'the service rendered by a bonde' (1381) and stowage 'the action or operation of stowing cargo on board ship' (1390). The vast majority of hybrids in this group is GENERAL ABSTRACT RIGHT AMOUNT TENURE CHARGE 87 only attested in the 15 th century, and the high amount of hybid formations proves that an ACTION reading is certainly transparent and productive in late ME. A striking difference to derivatives in the other ontological categories is the prevalence of verbal bases: nearly a quarter of the coinages are clearly deverbal. The same percentage is clearly denominal, and about half of the formations in this group are either deverbal or denominal. The contrast to the average across all categories, where more than 40% of -age derivatives are denominal, and only 13% deverbal, is striking. This difference is probably due to the semantics of the derivatives. Action nouns in general, not only derivatives of -age, are often deverbal, so it is not surprising to find a large amount of verbal bases in this group. But the equally high number of clearly denominal formations also shows that -age suffixation with an ACTION reading is not always a transpositional process. The high frequency of derivatives with an ACTION reading already suggests that this reading is productive. But frequent neologisms are only one criterion for a productive reading - transparency is another. In order to be transparent, derivatives should conform to a clearly discernible pattern regarding the nature of the bases and the semantics of the resulting suffixed forms. Let us first investigate the denominal formations. All of these words are built on bases denoting concrete entities. These are mostly persons, as in putage 'prostitution' from pute 'prostitute', but a small number of objects, e.g. crane, and locations, e.g. wharf, are also attested. Most of the derivatives on these bases are polysemous and often denote a charge or tax connected with an action. A good example of this polysemy is cranage 'the use of a crane to hoist goods; dues paid for the use of a crane'. Fourteen derivatives have both an ACTION and a CHARGE reading, making this the most frequent polysemy of all ME -age neologisms. A sense extension from one of these readings to the other cannot be established, as most of the derivatives with this overlap have both readings attested in the same paraphrase, so it is not clear which is the earlier interpretation. The paraphrases usually mention ACTION readings first, e.g. in cartage 'the process of conveying by cart; the price paid for this', but this probably only indicates a logical relationship between the two interpretations, namely that charges are paid for actions. This logical relationship is so strong that actions are implicitly mentioned in paraphrases of CHARGE derivatives that do not have an overt ACTION reading. Good examples for this are pesage 'a duty imposed for the service of weighing commodities', pavage 'a tax or toll towards the paving of highways or streets', or guidage 'a fee or tax paid for guidance'. When the two readings are given in separate paraphrases, there is no clear tendency regarding their chronology. The deverbal derivatives show a different pattern from the denominal ones. There are also some polysemous derivatives among them, but most of these coinages are monosemous and only denote an action. Examples are coupage 'the cutting up or carving of meat at the table' or abritrage 'the pro- 88 cess of arbitration'. The polysemous derivatives marriage, passage, and repassage, have many additional readings, which is probably a result of extensive usage, and does not indicate the opaqueness of this pattern. They should rather be seen as isolated incidents, as deverbal derivation with this reading is, in general, highly transparent. The group of derivatives with possible verbal and nominal bases also gives rise to many monosemous formations. The most common overlaps occur with CHARGE , for the same reasons as described above for the denominal formations, and LOCATION . Additional readings in the CONCRETE category can often be seen as the result of an action, which is a well-known phenomenon (cf. Bauer et al. 2013: 209). The locations described by the polysemous derivatives in this group are places where an action happens or denote the endpoints of a movement or action described by the base, e.g. rivage 'a coast, a shore' and 'landing on a shore', or passage 'the action of going or moving onward' and 'a place at which a river may be crossed'. The derivatives that can be connected to both nominal and verbal bases have a predictable interpretation and regular additional readings, so they can be seen as transparent. To summarise, -age derivatives with ACTION readings are frequently based on verbal bases, and many of the deverbal derivatives can be interpreted as transpositional nouns. However, a number of denominal formations constitute a regular pattern. The nominal bases usually denote a concrete entity, and the resulting actions either describe an action performed by someone, with the help of an object, or at a particular location. This reading is not common in early ME, but becomes frequent and highly transparent by late ME. Derivatives with an ACTION reading are also very frequent, and this reading should be considered highly productive. It is one of the most productive readings in ME and can therefore be seen as a core reading. 5.2.2.4 STATE Derivatives This ontological category contains two readings: CONDITION and POSITION . This is the smallest ontological category with only 22 words in total. reading types examples CONDITION 19 dotage, tillage, sorage POSITION 5 parage, parentage, thanage Table 6: STATE readings of -age neologisms (ME) Nine of the 22 derivatives have a STATE reading already in early ME, e.g. servage 'servitude, bondage' (c1290) or arrearage 'the state or condition of being behind … with the payment of what is due' (1330), but most, e.g. 89 cousinage 'the condition of being cousins' (c1430), acquire such a reading only in late ME. The share of early ME derivatives in this group lies at 40% and is thus significantly above the average. STATE readings are therefore more common among early ME coinages than among late ME neologisms. More than half the derivatives are clearly denominal, slightly below a fifth are deverbal and more than a quarter could be based on nouns, verbs, or even adjectives. The share of deverbal derivatives is thus relatively high at the expense of the doubly connected formations, but one has to keep the small number of derivatives in mind here and should not over-interpret this distribution. The percentage of hybrid formations at 22% is higher than the average, which suggests a certain amount of productivity of STATE readings. CONDITION Nearly all of the derivatives in the STATE category have a CONDITION reading, e.g. dotage 'the state of one who dotes' or marriage 'the condition of being a husband or wife'. The general characteristics of the whole category as discussed above are therefore true for this reading as well. Five derivatives are monosemous: corsage 'bodily condition as to size and shapeliness', maritage 'marriage', tillage 'the state or condition of being tilled or cultivated', sorage 'the first year of a hawk', and barnage 'childhood'. But most of these derivatives have multiple readings, some, for example marriage, as many as six. Such high numbers of additional readings are not very informative, as they only occur in individual cases, presumably when a particular derivative is used very frequently. Regularities regarding common sense extensions or general developments of semantic change can not be detected on the basis of such words. A tendency of derivatives referring to CONDITIONS seems to be an additional ACTION reading, e.g. in concubinage 'the cohabiting of a man and a woman who are not legally married; the practice of having a concubine' and 'the state of being a concubine'. Seven derivatives show this overlap, which is nearly a third of all CONDITION formations. Derivatives with such a polysemy sometimes show the CONDITION reading before the ACTION reading, even in the case of the only deverbal derivative among them, marriage. But the ACTION readings of these words should not be seen as a sense extension, because this tendency is not strong enough. CONDITION itself is not uncommon - after all, more than 10% of all derivatives have this interpretation - and the structure of these derivatives is quite regular. It is clearly not as productive as other readings, in particular ACTION and CHARGE , and should therefore not be considered a core reading of -age derivatives. POSITION This group contains five derivatives that refer to social positions and offices, e.g. bondage 'the position or condition of a serf' or thanage 'the rank or office of a thane'. Most of these derivatives have additional readings that are earli- 90 er than POSITION . Although the derivation of these words is highly systematic - four out of five words are based on person nouns - and this group contains two hybrid formations, POSITION is clearly not a productive reading of Middle English -age derivatives due to the lack of neologisms alone. Semantic Map of STATE Derivatives Figure 6: Semantic map of STATE readings of -age neologisms (ME) It is obvious from the semantic map that POSITION occurs only rarely. CONDI- TION is more common but compared to readings in other ontological categories also only moderately frequent. Overlaps within this group are sparse, which is shown by the dashed line. The overlaps between CONDITION and ACTION are shown on the unified semantic map in figure 7, as they cut across two different ontological categories. 5.2.2.5 Semantic Map of ME Neologisms The combined semantic map includes all readings and overlaps. It is immediately apparent on the map that there are enormous differences in frequency between the various readings. ACTION and CHARGE clearly stand out among the other readings in this regard. Others, like LOCATION , GENERAL ABSTRACT , or COLLECTIVE are moderately frequent, but most readings only give rise to a handful of derivatives. Although even these low-frequency readings show a certain amount of productivity - after all, they are attested in some neologisms - they cannot be considered central readings. The core readings of this morphological category are ACTION and CHARGE . Another productive reading is LOCATION , but this gives rise to much fewer derivatives than ACTION and CHARGE and should therefore not be considered a core reading. Some of the other readings may become productive in a later period of the language, as they are highly transparent. The COLLECTIVE reading is an example for such a potentially productive interpretation. Another important feature of the map concerns the connecting lines. This morphological category is clearly highly interconnected. Each reading is linked to multiple others, and even a small group like TENURE , which contains only five derivatives, has overlaps with five other readings. This is CONDITION POSITION 91 strong evidence against the homonymy of different suffixes, but supports the polysemy hypothesis. Apart from supporting the notion of a polysemous morphological category, the high number of overlaps also reveals the amount of polysemy of single derivatives. Many of the dashed lines, which represent overlaps of one or two polysemous derivatives, are the result of a small number of highly polysemous derivatives. These formations exhibit basically every reading an -age derivative can have. Words like advantage or marriage, which are two of the most highly polysemous items in the data, thus contribute to the interconnectedness of the morphological category, but they are not useful to determine any regularities. They are clearly extraordinary occurrences. But not all of the dashed lines are due to these highly polysemous words. Take, for example, the five derivatives in the TENURE group. Many of the connecting lines originating from this reading group represent only one or two overlaps. These are the result of all words in this group being polysemous, but they do not have a high amount of additional readings. The high number of single connections thus shows that there is not much regularity regarding additional readings in this case. An exception to this is the overlap with LOCATION , which, as the arrow shows, represents a sense extension TENURE → LOCATION . Such diversity and irregularity regarding additional readings is not found in all small reading groups, however. The clearest example for a small reading group with highly regular additional interpretations in RIGHT . This group contains ten derivatives and is connected with multiple other readings. But it has a very strong predictable connection with CHARGE and ACTION , exemplified by the thick continuous connecting lines. Nearly two thirds of the derivatives in the RIGHT group also have a CHARGE reading, and more than half also refer to an action. This shows that even reading groups with relatively few members can express regular overlaps with other groups. Most readings are connected by up to five polysemous derivatives, as represented by dashed and thin continuous lines, but a few show a higher number of formations with recurrent overlaps. All of these are connected to one of the core readings ACTION or CHARGE , but the strongest connection exists between ACTION and CHARGE themselves. The map also shows that both of these large readings are closely connected to a number of smaller groups as well. Even relatively small groups like RIGHT can show such regularities, as this reading is connected with CHARGE as often as the much larger group LOCATION is connected with ACTION . These frequent connections on the semantic map illustrate another characteristic of the RIGHT reading: most of the derivatives in this group have to be polysemous to sustain such a high number of overlaps. Directional sense extensions, represented by the arrows in the semantic map, could only be found for few readings. Only two such cases are supported by the data: TENURE → LOCATION , and ACTION → LOCATION . Such 92 connections depend on the diachronic evidence provided in the OED. If a certain reading usually occurs after another in a polysemous derivative, a directional sense extension can be postulated. But in most cases no such regularity could be detected. 93 Figure 7: Semantic map of all readings of -age neologisms (ME) 94 5.2.3 Present Day English Neologisms Since 1900, 78 new -age formations are attested in the OED. Similarly to the Middle English neologisms, these have readings across all ontological categories, but some significant differences between the two periods can be found as well. ontological category types examples CONCRETE 25 pre-package, teacherage, twiggage ABSTRACT 35 gallonage, minutage, tuneage EVENT 40 creepage, dressage, preshrinkage STATE 3 plaçage, problemage, victimage Table 7: Ontological categories of -age neologisms (PDE) The majority of all neologisms are first attested between 1900 and 1949, only one fifth are recorded as later coinages. The most recent of these is riffage 'riffing, espec. on a guitar' (1991). Such a drop in the number of neologisms may indicate a decreasing productivity of the suffix. Compared to the development of OED entries in general, this drop seems less significant, however. The entries between 1900-1949 make up nearly two thirds of all entries after 1900 (“timelines” 2013). It is thus not surprising that the number of new -age derivatives drops in the second half of the 20 th century. But it does decrease more markedly than the number of overall entries, which could signify a small decrease in the productivity of this suffix. Another factor that might play a role in this context concerns the perception of productive formations by language users. Language users are less likely to notice regular formations, and Plag has pointed out that “even the OED lexicographers fall victim to the unavoidable tendency to include the more salient idiosyncratic forms and neglect the listing of regular derivatives” (1999: 98; italics in original, MS). Such a bias would explain at least part of the drop in numbers since 1950. Almost half of all PDE neologisms are denominal formations such as turbinage or victimage, and for another third multiple possible bases, usually nouns and verbs, are attested. Exclusively deverbal derivatives are quite rare, they make up less than 10% of all formations. The remaining words are based on bound bases, e.g. intertillage or matrilineage, for which the adjectives intertilled and matrilineal are attested. The share of stem-based derivatives is slightly higher here than in ME, while deverbal derivatives are more common in ME than in PDE. Apart from these small differences, the structure of derivatives is basically identical in both time periods. This structural similarity is remarkable given the differences in the semantics of the derivatives, which will become apparent in the following. 95 5.2.3.1 CONCRETE Derivatives The 25 CONCRETE derivatives can be further subdivided into three different readings. Eleven of them denote objects, six refer to a location, and eleven denote a collectivity. reading types examples OBJECT 11 montage, pre-package, spillage LOCATION 6 coverage, parachutage, teacherage COLLECTIVE 11 screenage, twiggage, voidage Table 8: CONCRETE readings of -age neologisms (PDE) OBJECT and COLLECTIVE are the most frequent readings in this category, and LOCATION is less common. PERSON is already a marginal category in ME, and has disappeared entirely in PDE. OBJECT The most numerous CONCRETE reading denotes physical objects, both countable and mass. Examples are pre-package 'a prepackaged item' and septage 'waste matter or sewage contained in a septic tank'. Five derivatives are monosemous. These are pinotage and meritage, which both refer to a variety of wine, septage 'waste matter', fusilage 'the elongated body of an aeroplane', and stillage 'residue remaining in a still after fermentation'. All remaining derivatives have overlaps with ACTION , e.g. spillage 'the action or fact of spilling; that which spills or is spilt'. This derivative is exemplary for this overlap, as such derivatives usually denote an action or process and the result of this, which is often a physical object. Some derivatives also have semantic overlaps with GENERAL ABSTRACT and COLLECTIVE , but these are less common and less systematic than the OBJECT - ACTION polysemy. The monosemous derivatives denote different kinds of objects; some of them are countable, some are mass nouns. This variation does not make for a particularly transparent reading. This reading is also not particularly frequent compared to readings outside of the CONCRETE category and should thus not be seen as productive. LOCATION These six derivatives denote different kinds of locations from outdoor spaces like parachutage 'a drop site' to teacherage 'a house or lodgings provided for a teacher'. Teacherage and sondage 'a deep trench dug to investigate the stratigraphy of a site' are the only monosemous derivatives in this group. The other four 96 all show overlaps with ACTION , e.g. in coverage 'the act or fact of covering' and 'the area … covered'. Most of these derivatives denote the place where an action takes place. The only exception is plottage 'land in the form of a plot or plots', but the base of this word already refers to a location. Due to the very small number of new derivatives with this reading LOCA- TION is not a particularly productive reading of -age in PDE. COLLECTIVE The eleven formations with a COLLECTIVE reading denote collectivities of objects, e.g. twiggage 'twigs collectively', collectivities of locations such as voidage 'voids collectively', or persons collectively, e.g. matrilineage 'matrilineal lineage'. Most derivatives refer to a collection of physical objects, however. Four of these words, briquetage 'objects fashioned of burnt clay', empennage 'an arrangement of stabilizing planes...', strewage 'a layer or bed of strewed material', and twiggage 'twigs collectively' are monosemous. The remaining seven derivatives often have overlaps with GENERAL ABSTRACT , e.g. in signage 'signs collectively' and 'the design and arrangement of these'. Other words have overlaps mainly with ACTION , e.g. in bricolage 'construction … from a diverse range of materials or sources' and 'a miscellaneous collection'. The COLLECTIVE reading of bricolage and montage probably is the result of an action that happens to involve a collection of physical objects. This result interpretation also explains the polysemy of COLLECTIVE and OB- JECT , as many formations in these two groups can both be understood as results of an action. Although the part of speech of the bases in this group is diverse, they often denote physical objects if they have a nominal form. The derivatives then denote a collectivity of these objects, which makes this reading quite transparent. But it cannot be considered highly productive as it only gives rise to a relatively small number of derivatives. Semantic Map of CONCRETE Derivatives The semantic map below shows the lack of overlaps between the different readings in this ontological category. Only OBJECT and COLLECTIVE are connected by three polysemous derivatives. But this connection is due to another reading of these words, namely ACTION . All three derivatives have CON- CRETE readings because they express the result of an action, which is in these cases sometimes a collection, and sometimes a single object. All CONCRETE readings are connected with additional groups outside of this ontological category, however. 97 Figure 8: Semantic map of CONCRETE readings of -age neologisms (PDE) 5.2.3.2 ABSTRACT Derivatives The numerous derivatives in this category either refer to an amount or number of concrete or abstract entities or actions, e.g. trippage 'the number of journeys made', or gallonage 'an amount in gallons', to a charge in warehouseage 'the cost of warehousing', or to more general abstract entities, e.g. tuneage 'music', or air mileage 'rate or efficiency of travel through the air'. GENERAL ABSTRACT and AMOUNT are of an equally large size, as illustrated in table 9, but CHARGE contains only one derivative. reading types examples GENERAL ABSTRACT 20 air mileage, frottage, tuneage CHARGE 1 Warehouseage AMOUNT 20 hourage, minutage, trippage Table 9: ABSTRACT readings of -age neologisms (PDE) Only 20% of these derivatives are first attested after 1950, thus mirroring the average share. Regarding the syntactic structure of derivatives, this ontological category is dominated by denominal formations, which make up more than half of all derivatives, at the expense of exclusively deverbal ones. Only one derivative, frottage, is clearly deverbal. Two words are based on stems, and both of these have a GENERAL ABSTRACT reading. GENERAL ABSTRACT The derivatives in this group refer to different kinds of abstract entities from tuneage 'music' to sex-linkage 'the occurrence of the gene controlling a particular characteristic on a sex chromosome'. This group is thus, like its ME equivalent, quite heterogeneous. In spite of the semantic diversity of derivatives, the creation of more homogeneous groups with fewer members does OBJECT LOCATION COLLECTIVE 98 not seem advisable, as hardly two words have similar enough readings to be put in the same group. 80% of the twenty words in this group are polysemous. Many of these have overlaps with AMOUNT , referring to an amount of some unit and also a related abstract notion. A good example for this is wattage 'an amount of electrical power' and 'electricity'. In these cases, AMOUNT is mentioned first in the semantic paraphrase, but the two readings usually do not have separate entries. It can thus not be determined which is the earlier interpretation. An even more common overlap exists with ACTION . These derivatives denote an action first, but also refer to an abstract outcome of this action, e.g. in bricolage 'construction from a diverse range of materials or sources' and 'a concept so created'. The four monosemous items sex-linkage 'the occurrence of the gene controlling a particular characteristic on a sex chromosome', beamage 'a deduction for loss of weight by evaporation in cooling', tuneage 'music', and frettage 'damage suffered by two metal surfaces...' are diverse in their semantics and do not contribute to the transparency of this reading. To summarise, the neologisms with a GENERAL ABSTRACT reading constitute a diverse group. The small number of varied monosemous formations shows that this is not a productive reading. CHARGE This group contains only one derivative, warehouseage 'the cost of warehousing'. A CHARGE reading cannot be considered productive simply due to the lack of neologisms. AMOUNT The twenty derivatives in this group are semantically much more homogeneous than other ABSTRACT readings. They refer to an amount or number of objects or measurement units, e.g. metreage 'amount estimated or measured in metres' or spindlage 'the number of spindles employed in a particular mill', or, in the case of coverage, 'the aggregate of risks covered by an insurance policy', a collection of abstract entities. This group also contains a high share of monosemous derivatives: 60% of its members only have a single reading. Nearly all of these are based on units of measurement such as milliampere or minute, or on countable physical objects like spindle. The polysemous derivatives show some regularity as well, and overlaps with GENERAL ABSTRACT are common. Other overlaps occur as well, but are less systematic. More than two thirds of the bases of the AMOUNT derivatives are clearly denominal, e.g. minutage 'the amount of time for which a commercial television company is permitted to broadcast advertisements' or spindlage 'the number of spindles employed in a particular mill, district, trade, etc.'. The remaining formations have multiple possible bases, but they can all be interpreted as denominal. This pattern is therefore highly regular and very 99 transparent. As it is also a frequent reading of -age neologisms in PDE it should be considered a highly productive reading and a core sense. Semantic Map of ABSTRACT Derivatives Similarly to the previous category, not all ABSTRACT readings are connected through polysemous derivatives. This is due to CHARGE only containing a single monosemous derivative. The connection between AMOUNT and GEN- ERAL ABSTRACT , however, is a strong one. Figure 9: Semantic map of ABSTRACT readings of -age neologisms (PDE) 5.2.3.3 EVENT Derivatives All derivatives in this ontological category refer to actions such as creepage 'gradual movement', petrolage 'the covering of water with a film of oil', or scrappage 'the action of sending to the scrap-heap'. Only a small share of these neologisms, 15%, is attested after 1950. The structural make-up differs somewhat from the norm as well. Exclusively nominal bases are less common, although they still account for nearly a third of derivatives, and verbal bases are more common at around 15%. The most numerous group are those derivatives for which both nominal and verbal bases are attested. A higher share of verbal bases in this than in other semantic categories is not surprising, as morphologically complex action nouns often select such bases. In spite of this, the share of nominal bases exceeds that of verbal ones by far, which shows that -age derivatives with an ACTION reading cannot all be considered transpositional nouns. More than half of the ACTION derivatives are monosemous. Only three of these derivatives are exclusively deverbal, and many of them are derived from nominal bases, e.g. portalage 'the construction of portals' or mud pilotage 'the work of a river pilot'. They either refer to an action that is carried out by the person denoted by the base, or an action in which the object denoted by the base is involved or is the result of. The most common overlaps in the polysemous derivatives occur with GENERAL ABSTRACT and OBJECT . Such GENERAL ABSTRACT AMOUNT CHARGE 100 derivatives denote an action and the result of this action, which may be either abstract or concrete. Good examples are spillage 'the action or fact of spilling; that which spills or is spilt', and bricolage 'construction … from a diverse range of materials or sources' and 'a concept so created'. Such result readings more commonly denote physical objects than abstract concepts. The less frequent overlap between ACTION and COLLECTIVE is similar. Polysemous ACTION words with additional locative readings mostly denote an action and the place at which an action takes place, as in parachutage 'a drop of supplies' and 'a drop site'. The polysemous derivatives are also sometimes based on nouns, although bases that are attested as both nouns and verbs are more common here. The remarks regarding the semantics of the monosemous formations made above show that -age suffixation does not merely transpose a verbal meaning into a nominal form. New derivatives with ACTION readings are extremely frequent in PDE. The discussion above has also shown that this reading is very transparent, and that it is certainly productive and a core reading of -age derivatives. 5.2.3.4 STATE Derivatives This category contains only three derivatives and is by far the smallest. All three words refer to a state or condition. Only one derivative, problemage 'the state or condition of being a problem', is monosemous, the other two have overlaps with ACTION . Both denote a practice or custom as well as a condition: plaçage 'In 18th and 19th cent. Louisiana: the custom among many white men of setting up a black or mixed-race woman in her own household in addition to or in place of a wife. Also: such a relationship or situation', and victimage 'the condition of being a victim; also, the practice of seeking out a victim'. From these two cases it is not clear though whether CONDITION is a sense extension of ACTION . The small number of derivatives in this category shows that this reading cannot be considered productive. 5.2.3.5 Semantic Map of PDE Neologisms The extremely high frequency of neologisms with an ACTION reading becomes immediately apparent on the semantic map. This reading contains by far the most derivatives. AMOUNT and GENERAL ABSTRACT are also quite frequent, even if less so than ACTION . The three CONCRETE readings LOCATION , OBJECT , and COLLECTIVE contain a comparable, noticeably smaller amount of neologisms. Finally, CONDITION and CHARGE are only peripheral readings of neologisms at this point. The previous discussion has shown that not all of the frequent readings are equally productive, and the core readings of -age coinages are ACTION and AMOUNT . 101 Most of the readings are connected with others, only CHARGE is isolated, as it contains a single monosemous derivative. Some groups are connected by one or two polysemous words, represented by the dashed lines, but higher numbers of overlapping formations are more common. This is partly a consequence of the high number of formations in the relatively large reading groups, but some smaller groups, e.g. OBJECT , also show significant numbers of overlaps with other readings. Although most of the readings are connected to other readings, the number of connections is quite small. Most groups have overlaps with three or four others, only ACTION , the most numerous reading, is connected to all reading groups that contain polysemous derivatives. 102 Figure 10: Semantic map of all readings of -age neologisms (PDE) 103 5.2.4 Comparison of ME and PDE Neologisms A comparison of the two semantic maps (the map of ME neologisms is reproduced below as figure 11) immediately reveals the differences between the neologisms of these two periods. The most obvious change concerns the number of readings themselves. Four of originally twelve readings, POSI- TION , RIGHT , TENURE , and PERSON , have disappeared completely and are not attested at all in PDE neologisms. The remaining readings have undergone some major developments from ME to PDE as well. Some groups like CONDITION and LOCATION have shrunk in size, but none so much as CHARGE . A third of all ME neologisms express a CHARGE reading, but only a single word in PDE has this interpretation. Other groups like ACTION , OBJECT , and GENERAL ABSTRACT have increased their share of neologisms, but the rise of AMOUNT readings is particularly striking: their share has tripled from about 8% in ME to 25% in PDE. It is not only the number of readings that has decreased from ME to PDE - the amount of connecting lines has also declined considerably. In ME, every reading is connected to multiple other readings, but such relationships are less common in PDE. At that point, only ACTION and GENERAL ABSTRACT are connected to most other readings, while the other groups show significantly fewer overlaps. Much of the interconnectedness in ME is due to one or two derivatives with a particular polysemy, but most of the lines in PDE are continuous, representing at least three polysemous derivatives. This concentration on a smaller number of readings and fewer, but more regular, overlaps between these readings leaves a much clearer picture in PDE than in ME. All of these developments show that the semantics of -age has been subject to considerable change since the suffix was first used in English. This change does not only concern the frequency of certain readings, but also the polysemies and sense extensions of derivatives. 104 Figure 11: Semantic map of all readings of -age neologisms (ME) 105 While such changes are represented on the semantic maps, other developments, such as the absolute number of neologisms, the structure of these formations, and the productivity and transparency of particular readings, are not shown on the maps. The absolute number of -age neologisms has dropped sharply from ME to PDE. 142 coinages are attested in ME, but the suffix only gives rise to 78 new formations in PDE. The amount of OED entries between 1100 and 1499 (39,476) and after 1900 (36,356) certainly differs, but the figures are comparable. Such a decrease then suggests a decline in productivity of this suffix. Most of the readings have undergone some kind of change since ME. As can be seen on the semantic maps, five have disappeared completely, and another three are virtually non-existent in PDE neologisms. But the transparency and productivity of the remaining readings has changed considerably as well. LOCATION , which can be considered productive in late ME, is not productive in PDE, as it is neither transparent nor frequent. The most striking drop in productivity is exemplified by CHARGE , however. This reading is highly productive in ME, but only gives rise to a single coinage in PDE. The reverse can be observed for AMOUNT , which is neither frequent nor transparent in ME, but is certainly productive in PDE. This change is particularly interesting, because these two readings have many similarities. In ME, derivatives with an AMOUNT reading often refer to sums of money, which is naturally close to the CHARGE reading. ACTION , although already productive in ME, has become even more frequent in PDE, it has thus increased its productivity. COLLECTIVE , on the other hand, gives rise to a similar share of neologisms in both periods and can be considered somewhat productive in both ME and PDE. The previous paragraphs have highlighted the changes that have occurred since the Middle English period, but there are some similarities between ME and PDE neologisms as well. One of them is their structure: the syntactic categories and semantics of the bases that are selected has remained relatively stable. Exclusively denominal formations are slightly more common in PDE than in ME, as the share of deverbal formations and formations with multiple possible bases has dropped. Although stem-based derivatives are twice as likely to occur in PDE as in ME, the structure of -age derivatives is overall very similar in both time periods. Another similarity concerns the semantics of the derivatives. Although a number of readings have disappeared and there is significant change in the frequencies of the remaining interpretations, there is some continuity in this area as well. As no new readings occur in PDE, the morphological category seems to have concentrated on a subset of already attested interpretations. Whether this concentration is a general tendency or particular to -age cannot be answered at this point, but a comparison to the development of -ery may shed some light on this question. 106 5.3 The Suffix -ery in the OED 5.3.1 Previous Descriptions Marchand describes -ery and -ry as allomorphic suffixes of French origin forming both concrete and abstract nouns (1969: 282ff). The suffix was borrowed with French words, and because their bases were also borrowed into English, it became an analysable part. Marchand claims that most types are French loans, although some coinages had to be English - he mentions swannery, which is based on a Germanic noun - or were possibly English. The continuous borrowing of French words makes it often “impossible to decide whether a word is an English derivative or a loan from French” (ibid.: 282). Regarding the semantics of the derivatives, he identifies four major semantic groups in PDE: “1) a collectivity of persons (type yeomanry), 2) things taken collectively (type jewelry), 3) acting, behavior (especially undesirable), characteristic of - (type treachery), 4) place which is connect with - (types swannery/ printery)” (ibid.: 282, emphases in original, MS). These semantic groups favour different kinds of base words. One thing he points out with regard to derivatives like swannery is the possible nominal or verbal interpretation of many English words, which makes it sometimes difficult to decide on which base exactly a derivative is built. The result is a certain amount of flexibility of this word formation pattern with regard to denominal and deverbal derivation, although the original denominal formation can still be found in most derivatives. Marchand says of many derivatives that they are “doubly connected” (ibid.: 284) with more than one base word. One reason for this is a possible nominal and verbal interpretation, another is the common coinage of agentive nouns by the suffix -er, which results in the independent attestations of putative base verbs and agentive nouns. Marchand judges locative derivatives like printery to be particularly common in PDE: “Old and very strong today, especially in American English, is the type printery” (ibid.: 284, emphasis in original, MS). He does not comment on the productivity of any other semantic type, however. Dalton-Puffer (1996: 104-106) discusses only the Middle English properties of -ery. Her results are, in contrast to Marchand, who studied the OED, based on the Helsinki Corpus. She concludes that -ery forms nouns based on other nouns or verbs, but also finds many derivatives that contain only a stem rather than an independently attested word. Dalton-Puffer detects four derivatives coined on a Germanic base, where Marchand finds only one. But her semantic classification is similar to Marchand's, although his is formulated for PDE. She also postulates four semantic groups: 1) state, condition, practice of - N or V, 2) (undesirable) behaviour typical of N, 3) collectivity of persons/ things denoted by base, and 4) place to do with base (ibid.: 105f). Interestingly, she evaluates the relations between these groups as well. Ac- 107 cording to Dalton-Puffer, group 1) is the basic meaning of the suffix, while 2) is secondary to this. Groups 3) and 4) are also perceived as additions to the basic meaning. Bauer et al. (2013: 250) describe -ery as a noun-forming suffix that refers to “collectives, locations, and nouns denoting aspects of behaviour”. In their corpus study they frequently find nominal and verbal bases, but occasionally -ery is also attested with adjectives, compounds, phrases, or bound bases. Bauer et al. comment on the distribution of readings and claim that -ery, like -age, is “most frequently used to form collective and location nouns” (2013: 262). However, -ery has another domain, as it is also “sometimes” (ibid.) used to refer to a “kind of behaviour associated with a collective or group” (ibid.). The authors assume that the core reading is collective, and claim that location and behaviour are a “logical extension” (ibid.: 264) of that central sense. 5.3.2 Middle English Neologisms Altogether 226 words could be analysed as -ery derivatives coined before 1499. The earliest of these is gluttonry 'gluttony' (c1175). Another 18 formations are first attested in the 13 th century, e.g. butlery 'a butler's room or pantry' (1297) or Jewry 'the district inhabited by Jews in a town' (? c1225). The number of new coinages rises substantially in the 14 th century, as can be seen in figure 12. Almost twice as many derivatives as before are already recorded in the first half of the 14 th century. This means that a quarter of all ME -ery coinages are first attested in early ME. The number of new formations increases further until it reaches its peak in the first half of the 15 th century, and slightly decreases towards the end of the ME period. However, this rise in the number of -ery neologisms towards late ME should not be taken as proof for the increasing productivity of this suffix, because figure 13 shows that the number of entries in the OED also rises during this period. 108 Figure 12: First attestation dates of -ery neologisms (ME) Figure 13: Number of OED entries in 50 year periods over the course of the ME period Only 13% of all -ery formations are based on Germanic words, and 16% contain bases that are attested in both Germanic and Romance. The overwhelming majority of bases is thus of Romance origin. Some of the hybrid formations, i.e. those with Germanic bases, are already recorded in early ME, 1100-1149 1150-1199 1200-1249 1250-1299 1300-1349 1350-1399 1400-1449 1450-1499 0 10 20 30 40 50 60 70 year of first attestation number of neologisms 1100-1149 1150-1199 1200-1249 1250-1299 1300-1349 1350-1399 1400-1449 1450-1499 0 2000 4000 6000 8000 10000 12000 year of first attestation number of entries 109 e.g. dairy 'a shop in which milk, cream, etc. are sold' (c1290) < dey n. 'a woman having charge of a dairy' (a1000), or husbandry 'the administration of a household' (c1290) < husband n. 'the master of a house' (c1000). The existence of these early coinages proves that -ery suffixation must have already been perceived as a transparent means of word formation at that point. Most derivatives are based on nouns, although many of them, about one third, have bases that are attested as both nouns and verbs. Only nine derivatives, which equals a share of just under 4%, are based on verbs, and even fewer formations are deadjectival. More than 50% of the denominal derivatives are solely based on person nouns, and the remaining part denotes objects or other concrete entities. Suffixation with -ery in ME clearly favours nominal bases denoting persons over other nouns. Although purely deverbal derivatives are rare, a relatively large percentage of bases is attested as nouns and verbs. If interpreted as denominal derivatives, these formations fit the pattern just described: Their bases mostly denote persons, but other concrete entities can also be found. This suggests that these formations are also denominal, and that verbal bases are unusual for -ery neologisms in ME. The derivatives have a number of different readings across all four ontological categories. Table 10 shows the number of derivatives within each category as well as examples. The distribution across the different categories is heterogeneous, with CONCRETE and EVENT containing most derivatives, while far fewer derivatives have readings that fall under the STATE and AB- STRACT headings. Most of the ontological categories contain a number of different readings, and these will be discussed in detail in the following section. ontological category types examples CONCRETE 122 bachelry, barony, diapery ABSTRACT 33 arbitry, chantry, lollardy EVENT 109 enchantery, revelry, treachery STATE 62 beggary, deaconry, portmanry Table 10: Ontological categories of -ery neologisms (ME) 5.3.2.1 CONCRETE Derivatives The CONCRETE category is the largest ontological category, and contains more than half of all derivatives. The 122 derivatives with a CONCRETE reading can be split up into four different subgroups: OBJECT , PERSON , LOCATION , and COLLECTIVE . Table 11 shows that the most frequent of these readings is COLLECTIVE , closely followed by LOCATION . An OBJECT reading is expressed 110 by a much smaller number of derivatives, and a PERSON reading is even less common. reading types examples OBJECT 24 checkery, pastry, tapestry PERSON 10 advowry, falsary, lavendry LOCATION 59 baronry, glovery, pottery COLLECTIVE 67 grocery, jewellry, vassalry Table 11: CONCRETE readings of -ery neologisms (ME) The earliest derivatives with a CONCRETE reading, Jewry 'the district inhabited by Jews in a town' (? c1225), is first attested in the early 13 th century. This category contains a slightly higher share of early ME formations than the average over all categories, namely just under 30%. The share of hybrid formations, however, is lower than the average. Less than 10% of all CONCRETE derivatives contain a Germanic base, but 70% are built on Romance words. A few hybrid formations are already attested in early ME though, e.g. dairy 'a shop in which milk, cream, etc. are sold' (c1290), or ropery 'a long stretch of ground where ropes are made' (1329). Even at this early stage, some -ery derivatives with a CONCRETE reading are already coined on the basis of native vocabulary, which suggests a certain amount of productivity of this pattern. Nearly two thirds of the derivatives in this category are based on nouns, e.g. priory 'a monastery or nunnery governed by a prior...' < prior 'a superior officer of a religious house or order', or catery 'the office concerned with the supply of the provisions of the royal household' < cater 'a buyer of provisions'. Only four formations, avowry 'patronage, protection', vestry 'a room or part of a church ... in which vestments ... are kept', checkery 'checked cloth', and revestry 'the vestry or sacristy of a church', are clearly based on verbs. Adjectives are similarly uncommon as bases, but are attested in such derivatives as novelry 'a new or novel thing', for example. Most of the remaining bases are attested as both verbs and nouns. OBJECT 24 derivatives denote an object. These either refer to countable objects like poultry 'a bird' or pilfery 'an article of stolen or pilfered property', or denote mass objects or material in general, e.g. pastry. But apart from the fact that these formations denote objects they do not have too much in common. Both the countable and the uncountable objects are very diverse, ranging from books (registery) over a litter (saumbury) to checked fabric (checkery). 111 A third of the words in this group are monosemous. The remaining derivatives have additional readings in a number of groups, but ACTION and COLLECTIVE are most common. There is no clear tendency regarding the diachronic distribution of these overlaps, and a sense extension cannot be established. Most of the words with this interpretation are late ME coinages, so OBJECT seems to be a slightly later reading than others. The bases are mostly clearly nominal and denote concrete entities. Some bases, e.g. paste 'a mixture of ingredients or components' as the base for pastry 'a stiff but malleable mixture of flour …', or saumbu 'a saddle-cloth' as the base for saumbury 'a litter' already denote objects. When the base denotes a person, the derivative usually denotes an object the person works on or with; compare, for example, masonry 'stonework, brickwork' from mason 'a builder and worker in stone'. Other regularities cannot be established. Due to the diversity of objects that are denoted by -ery derivatives this pattern cannot be seen as transparent, and can therefore not be considered productive. PERSON PERSON is the smallest CONCRETE group with only ten derivatives. Almost all of the derivatives in this group are derived from bases that already denote a person. These words have additional readings across all ontological categories, e.g. advowry, which may refer to a RIGHT , CONDITION , PERSON , CHARGE , or POSITION . A PERSON reading thus seems to be an occasional additional reading of derivatives that primarily refer to something else. All of this suggests that this reading is not a particularly transparent interpretation of -ery derivatives, and the small number of derivatives also shows that this is not a productive reading. LOCATION The 59 derivatives with a locative interpretation make this group the second largest in the CONCRETE category, and also one of the largest reading groups of all -ery neologisms. The derivatives in this group mostly denote single rooms or parts of houses that are dedicated to certain activities, e.g. plumbery 'a plumber's workshop', butchery 'a slaughter-house, a butcher's shop or stall', or butlery 'a butler's room or pantry'. Larger locations, e.g. baronry 'the domain of a baron', are also attested, but these are rare. A third of these derivatives are already attested in early ME. This share is thus slightly higher than the average over all categories. Some of these early coinages are also derived from Germanic bases, e.g. dairy 'a shop in which milk, cream, etc. are sold' (c1290), or ropery 'a ... long stretch of ground where ropes are made' (1329). -ery formations with a locative interpretation are thus already coined in English, and not merely borrowed from French, at a relatively early point in time. The percentage of hybrid formations in this group is identical to the overall share of derivatives based on Germanic words. 112 There is a clear pattern regarding the kind of base words that combine with the suffix in this group. The majority of bases are nouns that denote persons, e.g. in the derivatives baronry 'the domain of a baron', bailiery 'the jurisdiction of a bailie', or subdeanery 'the residence of a subdean'. The derivatives then denote a location that is connected with the person denoted by the base. A few other formations take bases that denote physical objects, e.g. pomary 'a fruit garden' < pome 'an apple' or buttery 'a place for storing liquor' < butt n. 'a cask for wine or ale'. Very few derivatives contain bases that already denote a location, like bordelry 'a brothel' < bordel 'a brothel'. Only two derivatives, vestry 'a room or part of a church ... in which the vestments ... are kept' and revestry 'the vestry or sacristy of a church', are solely based on verbs, although there are a number of words for which both verbs and nouns are possible bases. Given the general preference for nouns denoting persons as bases of the derivatives in this group, it can be assumed that these doubly connected formations are based on the noun rather than the verb, as these nouns mostly denote persons and therefore fit well into the pattern formed by unambiguous denominal formations. Even in the case of vestry and revestry objects that are part of the actions referred to by the base verbs are picked out by -ery rather than the actions themselves. The derivatives do not denote a location where an action takes place, which would be possible and is done by other suffixes, among them -age, but select objects that are connected to that action and denote the location where these objects are kept. A third of the derivatives with a LOCATION reading are monosemous. Most of the remaining derivatives have overlaps with COLLECTIVE (see the semantic map in figure 14 for an illustration of this relation). Significantly fewer overlaps exist with ACTION and a number of other readings across different ontological categories. It is interesting that not more derivatives refer to an ACTION as well as a LOCATION . A sense extension from ACTION to LOCATION is quite common (cf. Bauer et al. 2013: 209), and it can be found for derivatives of -age, a suffix with a very similar semantic range. An explanation for this could be the fact that the locations denoted by -ery derivatives do not usually denote a location in which an action takes place, but a location in which objects, which may be the results of actions, are kept. This is, as was already mentioned, very different from -age, for which a sense extension ACTION → LOCATION could be established. The derivatives of these two very similar suffixes therefore show subtle differences regarding the relations between readings and the mechanism of the derivation itself. More than a third of all LOCATION derivatives, 26 words in total, have additional COLLECTIVE readings. This also ties in with the observation that the location denoted by -ery derivatives usually is a place where objects are kept. The fact that the formations that denote a location often also refer to a collectivity of said objects is then not surprising. Most of the derivatives with a LOCATION and a COLLECTIVE reading refer to a collectivity of people, com- 113 monly a department or office in a large institution, e.g. the royal household. Good examples for this are almonry 'a place where alms are distributed' and 'an office responsible for the distribution of alms', and saucery 'the department of a household entrusted with the preparation of sauces' and ' that part of a house in which sauces were prepared'. A few of the derivatives with this overlap denote a location and a collectivity of objects that are produced or commonly used in that location, e.g. lormery 'the small ironware produced by lorimers' and 'a place where such ironwork was made or sold' or plumbery 'a plumber's workshop' and 'plumbing appliances collectively'. Polysemous derivatives that denote a location and a collectivity of locations also exist, but these are much rarer than the other two cases, and far less systematic. Given the large amount of person-denoting base nouns in this group as well as in the COLLECTIVE group, and the kind of locations denoted by -ery derivatives, a systematic overlap with person-denoting collectives is not surprising. The derivatives that show both readings, LOCATION and COLLECTIVE , are usually attested with both interpretations at the same time. Only four formations have COLLECTIVE as the earlier reading, and in eight cases LOCATION is attested before COLLECTIVE . Because more than half of the 26 words with this semantic overlap give both readings at the same time a sense extension from one to the other cannot be established. Derivatives with a LOCATION reading are both frequent and transparent. A high percentage of these are monosemous, and the polysemous derivatives have regular overlaps with other readings, especially COLLECTIVE . This reading is also attested frequently in early ME, and there are hybrid formations already at that point. LOCATION can thus be considered a productive reading and also a core sense of -ery derivatives in ME. COLLECTIVE The COLLECTIVE reading is the largest among the CONCRETE category. Derivatives with this interpretation refer either to a collectivity of persons, e.g. Jewry 'Jews collectively', a collectivity of objects, e.g. robbery 'the proceeds of robbery', or a collectivity of locations, e.g. nunnery 'a group of buildings in which nuns live'. The earliest derivative is bachelry 'young knights as a class or body' (1297). A quarter of all derivatives with a COLLECTIVE reading are first attested with that reading in early ME, which is comparable to the average of all -ery derivatives. This reading is therefore neither particularly early, nor is it a late reading to occur. The amount of hybrid derivatives in this group is also similar to the overall share, but hybrid formations with a COLLECTIVE reading are only coined from late ME onwards. Those that are attested earlier, like dairy 'milch cows on a farm collectively' or husbandry 'the body of husbandmen on an estate', only acquire their COLLECTIVE interpretations in late ME. Although the COLLECTIVE -ery derivatives may have been transpar- 114 ent at an earlier time, the coinage of hybrids shows that this reading is certainly transparent and productive by late ME. The bases of the derivatives in this group are overwhelmingly nouns, although about one fifth could also be based on verbs, as possible verbs and nouns are both attested. Most of the bases denote persons, e.g. butcher, vassal, or archer, but some objects and collective objects can also be found. More than a quarter of derivatives are monosemous. Slightly more than half of these denote a collectivity of persons, the others refer to a collectivity of objects. In most cases, the derivative denotes a collectivity of whatever the base denotes, usually either person or object. There are, however, a few exceptions to this general tendency. Sometimes the derivatives pick out objects that are connected to the person denoted by the base and refer to a collectivity of these, e.g. in grocery 'the goods sold by a grocer'. And this also works in the opposite direction: wax chandlery 'the department of a royal household concerned with the provision and storage of wax candles', for example, denotes a collectivity of people although its base refers to an object. Most of the polysemous words have additional LOCATION readings (for more details on this overlap see the discussion of LOCATION above). The derivatives with COLLECTIVE and LOCATION readings are often based on person nouns, so the overlap between these two readings may, at least partly, be connected with the similarity of the bases. This would also explain the high number of overlaps that both LOCATION and COLLECTIVE have with POSITION : POSITION derivatives are also mostly based on person nouns, and an overlap between these three groups can thus be expected. Another systematic overlap exists between COLLECTIVE and ACTION , as a quarter of all collective words show that reading as well. In these cases, the ACTION reading is attested earlier than the COLLECTIVE reading more often than the other way around. This tendency seems strong enough to establish a sense extension ACTION → COLLECTIVE . This implies that most of the derivatives with this polysemy denote a collectivity of objects that is the result of an action or a collectivity of persons who carry out an action. The connection between ACTION and those COLLECTIVE formations that denote a collectivity of objects is especially strong, however. COLLECTIVE is a productive reading in ME, as the derivatives with this interpretation are frequent and transparent. This group contains an average amount of hybrid formations, and, in particular towards late ME, has many monosemous derivatives and systematic overlaps with other readings. The selected bases also show that this pattern is regular and thus transparent. Due to this COLLECTIVE can be considered a core sense of -ery derivatives in ME. Semantic Map of CONCRETE Derivatives The semantic map in figure 14 shows that all the CONCRETE readings are connected to each other by polysemous derivatives. PERSON is only connect- 115 ed by dashed lines, representing one or two polysemous formations, but this group is very small and contains only ten derivatives. There is a stronger relationship between COLLECTIVE and OBJECT , as seven words show both interpretations. Unsurprisingly, six of these denote an object and a collectivity of objects, only the seventh denotes a collectivity of persons as well as an object. The strongest link exists between LOCATION and COLLECTIVE . These two are of course the largest CONCRETE readings, but 40% of their derivatives have this overlap, which is remarkable. Some of these derivatives denote a location and a number of people who work there, but most refer to a location and a collectivity of objects that are connected with that location. The link between the two readings is probably due to the prevalence of persondenoting nouns as bases in both groups and to the kinds of locations denoted by these derivatives. Figure 14: Semantic map of CONCRETE readings of -ery neologisms (ME) 5.3.2.2 ABSTRACT Derivatives ABSTRACT is the smallest ontological category. The 33 derivatives in this category are divided into five different reading groups: GENERAL ABSTRACT , which is by far the largest of these groups, CHARGE , RIGHT , TENURE , and AMOUNT . The number of derivatives in each of these groups is given in table 12 below. OBJECT LOCATION COLLECTIVE PERSON 116 reading types examples GENERAL ABSTRACT 24 desidery, Mahometry, meselry CHARGE 7 advowry, canonry, prependry RIGHT 4 advowry, provendry, seigniory TENURE 1 sergeantry AMOUNT 4 lollardly, mammetry, trumpery Table 12: ABSTRACT readings of -ery neologisms (ME) Early derivatives with a reading in this category are gluttonry 'gluttony' (c1175), and mastery 'superior force or power' (c1225). A quarter of all AB- STRACT derivatives are attested in early ME, which is comparable to the overall share of early ME -ery coinages. The share of hybrid formations is also similar to the overall average, but these coinages occur only in late ME or acquire their ABSTRACT reading at that point. ABSTRACT derivatives are overwhelmingly based on nouns denoting persons, although a few verbal bases are attested too. In this regard as well, this category is very similar to -ery derivatives in general. GENERAL ABSTRACT This is the largest reading group within the ABSTRACT category with 24 members. These derivatives refer to some rather general abstract concepts, e.g. desidery 'desire, wish', or Mahometry 'the Muslim religion'. As a result, this group is something of a mixed bag, as the concepts are quite diverse. It is also sometimes difficult to distinguish between a GENERAL ABSTRACT and a CONDITION reading, as the OED paraphrases are not always ideally phrased for this differentiation. In the ontological system used for this investigation, CONDITION is a STATE reading, and therefore part of the SITUATION category, while GENERAL ABSTRACT is part of SUBSTANCE / THING / ESSENCE . The basic distinction between situations, i.e. nouns that have a temporal structure, and things, i.e. nouns that do not make reference to duration, was employed to discriminate between GENERAL ABSTRACT and CONDITION : the derivatives with a GENERAL ABSTRACT reading do not refer to a concept that lasts for an amount of time, but to abstract concepts in general. The earliest derivative with a GENERAL ABSTRACT reading is gluttonry 'gluttony' (c1175). Formations with this reading are not very frequent in early ME - only one fifth of the derivatives in this group are attested before 1350. The share of hybrid formations at 13% is slightly higher than the average, however. But as most of the hybrids are polysemous, and GENERAL AB- STRACT is not their first interpretation, this does not prove that this reading is transparent. 117 As for the previously discussed readings of -ery derivatives, most of the bases in this group are nouns or are attested as nouns and verbs. Only three derivatives, desidery 'desire', dowry 'a gift or talent with which anyone is endowed by nature or fortune', and fardry 'an effect produced by painting the face', are based on purely verbal forms. One third of the derivatives in this group are monosemous. These formations are often based on person nouns, e.g. lepry 'leprosy' < leper. A small group of words refers to a religion: Mahometry 'the Muslim religion', mammetry 'the Muslim religion', and Jewry 'the Jewish religion'. Overlaps in polysemous derivatives exist with a number of readings, but those with ACTION (nine polysemous derivatives), COLLECTIVE (nine polysemous derivatives), and LOCATION (seven polysemous derivatives) are the most common. In all of these cases, GENERAL ABSTRACT is rarely the earliest reading. As many of the polysemous words have more than two or three readings it is not clear whether GENERAL ABSTRACT is a sense extension of ACTION , COLLECTIVE , or LOCATION . In spite of the relatively high share of monosemous derivatives and the number of hybrid formations in this group, a GENERAL ABSTRACT reading cannot be considered transparent, as the derivatives are too heterogeneous. Also, the number of derivatives with this reading is much higher than that of other readings in the ABSTRACT category, but compared to readings in other ontological categories it is still quite low, and this reading should therefore not be considered productive. CHARGE CHARGE is the second most frequent reading in the ABSTRACT category, but its type frequency is still very low. All seven words refer to a benefice, e.g. canonry 'the benefice of a canon'. All derivatives with this reading are first recorded in late ME, and all of them are polysemous. The pattern is quite systematic, as most of the bases are nouns that denote persons and the derivatives refer to a benefice that is awarded to these persons. Almost all derivatives with a CHARGE reading also refer to an office or position that is connected with the person denoted by the base. The two interpretations CHARGE and POSITION are usually given in the same paraphrase, so it is not clear which of them is the earlier reading. A sense extension can therefore not be established. Given the infrequency of derivatives with a CHARGE reading and their extensive polysemy, CHARGE should not be considered a productive reading of -ery derivatives. RIGHT Four derivatives refer to a right, e.g. avowry 'a … right of presentation to a benefice'. None of these words show a RIGHT reading in early ME, and none are based on Germanic bases. The bases are verbs or nouns with diverse semantics, so that the derivation of these words is unsystematic, and all 118 derivatives are highly polysemous. This reading is neither transparent nor productive. TENURE Only one word, sergeantry 'a form of feudal tenure ...', refers to a tenure. Due to the lack of formations this reading can be seen as unproductive. AMOUNT Only four derivatives, which is less than 2% of all -ery derivatives, denote a collectivity of abstract entities. The low type frequency already shows that this reading cannot be considered a productive interpretation of -ery in ME. The fact that three of the four derivatives are polysemous, and the diversity of the bases also suggest that this pattern is not transparent. Semantic Map of ABSTRACT Derivatives The map in figure 15 shows the readings of ABSTRACT derivatives and the connections between these readings. GENERAL ABSTRACT , the largest reading group, is connected to most other readings in this group by polysemous derivatives. With the exception of an overlap between CHARGE and RIGHT , links among the other readings are not attested. TENURE is even completely isolated, as it shares no polysemous derivatives with other readings in this category. All the connections are tenuous, only one or two words with a particular polysemy are attested in all of these cases. Overlaps with other ontological categories exist for all readings, however. These are shown on the semantic map for all Middle English -ery coinages (figure 17). Figure 15: Semantic map of ABSTRACT readings of -ery neologisms (ME) 5.3.2.3 EVENT Derivatives EVENT is the second largest ontological category with 109 derivatives. There are no subcategories here, all of the derivatives have an ACTION reading. GENERAL ABSTRACT RIGHT CHARGE TENURE AMOUNT 119 The earliest of these derivatives are from the first half of the 13 th century, e.g. treachery (? c1225), ministry 'the action or an act of religious ministration' (a1225), or robbery (a1225). Slightly fewer than a quarter of all ACTION derivatives are attested before 1350. This means that derivatives with an ACTION reading are neither particularly early nor very late compared to -ery coinages in general. The share of hybrid formations is only slightly higher than the average as well. Some of these are attested in early ME, e.g. reavery 'robbery' (c1325) or husbandry 'the administration and management of a household' (c1290), so this word formation pattern is already somewhat transparent at that point. But most of the hybrids are first attested after 1350: -ery suffixation with an ACTION reading is therefore certainly transparent and productive in late ME. Six derivatives with an ACTION reading are based on verbs, so the share of deverbal derivatives is very low. However, this reading group contains most of the verbal bases that are attested in Middle English -ery neologisms. Half of all bases in this group are attested as nouns and verbs. Denominal derivatives constitute a slightly smaller share of nearly 40%. A number of the nominal and potentially nominal bases already refer to actions, e.g. barrat 'deception', but most of the bases denote persons. The derivatives then refer to actions carried out by these persons. Deverbal derivation plays a slightly more important role for derivatives with this reading than for -ery formations in general, but due to the large share of nominal bases even in this group, -ery derivatives with an ACTION reading cannot be interpreted as purely transpositional nouns. A very large share of the derivatives, more than 50%, are monosemous. These refer to various kinds of actions, e.g. michery 'pilfering' or revelry 'the action or an act of revelling'. Interestingly, many of the derivatives have negative connotations - they refer to undesirable actions like robbery, guilery 'trickery' or mockery 'derision, ridicule'. The large amount of monosemous words shows that this pattern is highly transparent. The polysemous derivatives have overlaps with all other readings. ACTION is unusual in this way, as it is one of only two readings, together with COLLECTIVE , that is connected to every single reading group. The most regular overlap occurs between AC- TION and CONDITION : 22 derivatives have both of these readings. This overlap is significant, as it affects most of the derivatives with a CONDITION reading. For most of these words, both readings are given at the same time, but when separate dates are provided, ACTION is expressed before CONDITION . This suggests a sense extension ACTION → CONDITION . The next most frequent polysemies are ACTION - COLLECTIVE (17 formations with both readings), and ACTION - LOCATION (12 formations with both readings). As was already noted in the discussion above, ACTION tends to be the earlier reading in these cases and can be seen as the origin of a sense extension ACTION → COLLECTIVE . 120 ACTION is connected to all readings through polysemous derivatives, and some of these overlaps are systematic. The high share of monosemous words with an ACTION interpretation, and the large number of hybrid formations shows that this reading is certainly transparent. As it is the largest single reading in the data, it is certainly also productive and therefore a core sense of -ery in ME. 5.3.2.4 STATE Derivatives The 62 derivatives in the STATE category have two different readings: They refer to a CONDITION , e.g. beggary 'the state or condition of a beggar', or a POSITION , e.g. wardenry 'the office or position of a warden'. The latter group often refers to the societal position of a person and/ or an office held by a person. Table 13 shows the type frequencies and examples for these two groups. reading types examples CONDITION 37 beggary, foxery, idiotry POSITION 29 deaconry, portmanry, wardenry Table 13: STATE readings of -ery neologisms (ME) The earliest derivative with a STATE reading is mastery 'the state or condition of being master' (c1225). Although a few formations in this category are attested before 1350, the share of early ME coinages lies below the average at 18%. Derivatives with STATE readings become more frequent only towards late ME. The share of hybrid formations is comparable to the overall average, but derivatives on Germanic bases mostly occur in late ME. The only exception is portmanry 'the position or rank of a portman' (1346-7). The hybrid formations show that STATE readings are transparent in late ME, but as these derivatives are rare in early ME, STATE readings probably only become transparent after 1350. Most derivatives in the STATE category are based on nouns. A quarter can be derived from multiple bases, mostly nouns or verbs, and 6% are based on adjectives. This group of derivatives thus has a relatively high share of adjectival bases compared to all -ery derivatives. The base nouns refer almost exclusively to persons. CONDITION Neologisms with a CONDITION reading are slightly more frequent than those with a POSITION interpretation. The semantic paraphrases of these words often mention 'the condition or state of being', e.g. for beggary 'the state or condition of a beggar'. Such a formulation is not always used in the defini- 121 tions; descriptions solely relying on other complex words are also common, for example nigonry 'avarice; niggardliness' or novelry 'novelty, newness; strangeness'. A look at the paraphrases for the words used in the definitions often reveals that these have a clear CONDITION reading, e.g. niggardliness 'the state or quality of being niggardly', so such an interpretation is assumed for the -ery derivative as well. But the inconsistency of the semantic paraphrases makes especially the delineation between CONDITION and GENERAL ABSTRACT very difficult, and a certain amount of interpretation, which potentially limits the reliability of the classification procedure, is unavoidable. As was already mentioned during the discussion of derivatives with a GENERAL AB- STRACT reading above, the distinction between these two readings is based on them belonging to two different ontological categories: CONDITION refers to a type of SITUATION , which has a certain duration, and GENERAL ABSTRACT refers to a SUBSTANCE / THING / ESSENCE , which does not have a starting or end point. The proportion of early ME coinages in this group is comparable to that of the whole STATE category, so it is lower than the average amount of early coinages over all -ery neologisms. The share of hybrid formations is also similar, and all of the hybrid coinages are first attested in late ME. All four deadjectival derivatives found in the STATE category have CON- DITION readings, so this reading contains more than half of all adjectival bases found in -ery derivatives. If an adjective is the base for -ery suffixation, CONDITION is therefore the most likely interpretation of the derivative. Out of the 18 derivatives that are clearly denominal 17 are based on person nouns, although alternative interpretations are also available for a few of these. The only exception is foxery 'wiliness, cunning', which is based on a noun denoting an animal. The bases of formations for which both deverbal and denominal derivation is possible also overwhelmingly denote persons if they are interpreted as nouns. A person noun base is therefore extremely likely to be found in -ery derivatives with a CONDITION reading. Fewer than one fifth of the derivatives in this group are monosemous, which is a low share compared to the other reading groups. Almost two thirds of all CONDITION derivatives also refer to an ACTION . These two readings are mostly given in the same paraphrase in the OED, but when separate dates are given for them, ACTION is the earlier reading. This suggests a sense extension ACTION → CONDITION . Although derivatives with a CONDITION reading are not infrequent, this reading can probably not be considered productive due to the semantic heterogeneity of the derivatives in this group. This reading does, however, become transparent towards the end of the ME period and may thus become productive at a later stage. 122 POSITION 29 derivatives refer to an office or a position, e.g. portmanry 'the position or rank of a portman' or constablery 'the office of a constable'. A quarter of the words in this group are attested with a POSITION reading before 1350, a figure that is comparable to the overall share of early ME formations. The earliest derivative is deanery 'office or position of a dean' (1292). The share of hybrid formations lies just over 10%, which is slightly below the average. Only one of these, portmanry, is an early ME coinage. This group contains two very similar deadjectival derivatives: gentlery and gentry. Both refer to a 'rank by birth' and are early ME coinages. The rarity of such bases and the similarity of these two derivatives shows that deadjectival derivation with a POSITION reading is clearly the exception to an otherwise systematic pattern. All other bases can be interpreted as nouns, although a few have alternative interpretations as verbs or adjectives as well. With the exception of provend 'a stipend, revenue, or estate', which is the base for provendry 'the office … of a prebendary', all of these nominal bases can denote persons, although some have additional interpretations. Slightly under a quarter of all POSITION derivatives are monosemous. This share lies well above that of monosemous CONDITION formations, but still below that of other productive readings groups like COLLECTIVE or LOCA- TION . The polysemous derivatives have a number of frequent overlaps - additional ACTION , LOCATION , and COLLECTIVE readings are the most common. The overlaps with LOCATION and COLLECTIVE were already mentioned in the pertinent discussions above. They are probably due to similar bases in these groups, as all three, LOCATION , COLLECTIVE , and POSITION , have a clear preference for persons nouns as bases. The derivatives that have additional ACTION readings refer to actions carried out by the person denoted by the base as well as their office or position. This overlap is also unsurprising, as ACTION derivatives are often based on person nouns as well. Derivatives with a POSITION reading follow a systematic pattern, as they are almost exclusively based on nouns denoting persons. This interpretation is also relatively frequent. It is, however, not clear that this reading is productive. The share of monosemous formations is quite low, and the overlaps of polysemous derivatives with many different readings are probably due to the selection of similar bases in the various reading groups. Also, given the prevalence for person nouns bases for -ery suffixation in general, POSITION is not very frequent, i.e. it does not exploit the available bases to a large degree. The systematicity of this pattern suggests that this reading is at least transparent, if not overly productive. Semantic Map of STATE Derivatives The semantic map in figure 16 below shows the two readings in the STATE category. CONDITION and POSITION are connected by four polysemous derivatives. 123 Figure 16: Semantic map of STATE readings of -ery neologisms (ME) 5.3.2.5 Semantic Map of ME Neologisms The semantic map in figure 17 shows all attested readings and the overlaps of polysemous derivatives. Like the map of ME -age derivatives (see section 5.2.2.5), this category is highly interconnected: all readings are connected to others via polysemous formations, and links to four or five different reading groups are the norm. This map thus also, like the maps of -age derivatives, suggests a polysemous category, and not multiple homonymous affixes. The differences in type frequency of the individual readings are immediately apparent. ACTION is the largest and most productive reading, containing almost half of all -ery coinages. The next most frequent readings are far less common. COLLECTIVE and LOCATION each include a quarter of derivatives, and these two readings can also be seen as productive core readings of this morphological category. There are also a few readings of mid-range frequency, e.g. CONDITION , POSITION , or GENERAL ABSTRACT . The smallest groups are mostly found in the ABSTRACT category: TENURE , CHARGE , RIGHT , AMOUNT , and PERSON each contain less than 5% of formations. This distribution shows that a few readings, especially ACTION , dominate this category in terms of type frequency. The smaller reading groups are outliers that are unlikely to be encountered in neologisms. In spite of their low frequencies, all of the smaller groups are connected to multiple others. The four formations with a RIGHT reading, for example, have overlaps with eight other readings. All of these links are relatively tenuous though. They are only found in one or two derivatives each, as can be seen by the dashed lines on the map. But a few regularities exist even in these small groups. For example, three out of four AMOUNT formations also refer to an ACTION . The most systematic of the small groups with regard to recurrent polysemies is CHARGE . This group has a highly regular overlap with POSITION , as six out of eight derivatives show that polysemy. The fact that all the small reading groups are either connected by systematic overlaps or a high number of less predictable polysemies shows that most, if not all, the derivatives in these groups are polysemous. CONDITION POSITION 124 But not only the infrequent readings have a high number of overlaps with other groups: ACTION and COLLECTIVE are both connected with every other reading, and LOCATION has links to nine other readings. The more frequent groups are often connected to one another by thicker lines, which represent a higher number of polysemous derivatives. This has to be expected, as these groups contain more derivatives than the smaller ones. The two most frequent overlaps are ACTION - CONDITION and LOCATION - COL- LECTIVE . The first of these affects 22 derivatives, which is almost 60% of the CONDITION group, but only 20% of the ACTION group. It is therefore quite likely for a CONDITION derivative to also express an ACTION reading. The arrow in the map shows that this relation is a sense extension ACTION → CONDITION , because ACTION is usually expressed earlier in polysemous derivatives with separate entries for both senses. The link between COLLECTIVE and LOCATION , also strong, does not represent a sense extension. There are 26 derivatives with both of these readings, which is about 40% of both groups. As the groups are of virtually equal size, this overlap does not affect one of them more than the other. Also, there is no clear tendency as to which reading is expressed earlier in polysemous derivatives, and the readings are similarly productive. The staggering amount of connecting lines shows that this category is already highly polysemous in ME, which is also the time of the earliest attestation of the borrowed suffix -ery in English. This polysemy does therefore not build up over time, but both -ery and -age express it from an early point. A comparison of ME -age derivatives and medieval French -age derivatives (the readings of French -age formations are taken from Uth 2011) reveals that the readings of these two categories are very similar. For French -erie I am not aware of a similar study to the present one, so it is unclear whether the readings expressed in Middle English are similar to the ones attested in French derivatives at the same time. It seems likely, however. 125 Figure 17: Semantic map of all readings of -ery neologisms (ME) 126 5.3.3. Present Day English Neologisms 87 derivatives can be found in the OED with attestation dates after 1900. Nearly three quarters of these formations are first attested in the first half of the 20 th century, and the remaining words are recorded in the second half. The most recent derivative is micromachinery 'micromachines and their components' (1986), so the OED does not contain any -ery coinages from the 21 st century. Almost half of all derivatives are based on nouns, and nearly all of the remaining formations may be based on nouns but also have other possible bases. These are mostly verbs, but some possibly adjectival bases can be found as well. There is only one derivative, pressure cookery < pressure cook v., that is exclusively based on a verb. Many of the nominal bases denote persons, e.g. madcap or martial artist, but there is a substantial part with other meanings. Physical objects, including animals, are very common, e.g. worm or corset, and a few bases refer to an abstract concept or an action. In PDE verbal bases are even less important than in ME, but -ery is also less focused on person nouns as possible bases. Nouns denoting physical objects, and in particular animals, form a substantial group of bases as well. With regard to the formal make-up of the bases, there is also a slight change from ME until PDE. In 20 th century neologisms, one finds a number of names and compounds as bases, e.g. Peter Pan, mumbo-jumbo, nothing-but, or hot dog. The extension of the word formation pattern to include such bases may suggest an increased flexibility, and maybe even an increased productivity. The readings of 20 th century neologisms are distributed over all four ontological categories. The largest of these groups is EVENT , which contains half of all derivatives. CONCRETE follows closely with 35 words, but STATE and ABSTRACT readings are less often expressed by new derivatives. Table 14 contains some examples for each category as well as the exact figures. ontological category types examples CONCRETE 35 corsetry, Reichschancellery, puffinry ABSTRACT 15 beat poetry, Mormonry, quippery EVENT 44 air gunnery, mopery, snoopery STATE 25 nitwittery, Proustery, smuggery Table 14: Ontological categories of -ery neologisms (PDE) 5.3.3.1 CONCRETE Derivatives The derivatives in this category express three different readings: OBJECT , LOCATION , and COLLECTIVE . Table 15 shows the type frequency for each reading and example derivatives. Only four formations denote an object. LOCA- 127 TION and COLLECTIVE are of equal frequency, and contain most words with a CONCRETE reading. This distribution is similar to the one in ME, where LO- CATION and COLLECTIVE are also the largest groups. In ME, OBJECT is also the third most frequent reading, but the importance of this interpretation has clearly diminished. Only 11% of CONCRETE derivatives denote an object in PDE, whereas almost a quarter of CONCRETE words in ME have that interpretation. PERSON is already the smallest group in ME and is non-existent in PDE neologisms. reading types examples OBJECT 4 cokery, pargetry, wormery LOCATION 17 bootery, eatery, snackery COLLECTIVE 17 air gunnery, blossomry, tartanry Table 15: CONCRETE readings of -ery neologisms (PDE) One third of the formations in this ontological category are first attested after 1950. The most recent one is micromachinery (1986). Not surprisingly given the prevalence of nominal bases for -ery neologisms in general, most words in this category are based on nouns. More than half are purely denominal, and almost all remaining ones are either denominal and deverbal, or denominal and deadjectival. Somewhat unusually though, the bases mostly denote physical objects. Almost three quarters of all derivatives in this group are based on object-denoting nouns, e.g. corsetry 'corsets collectively' < corset n., or mulletry 'a place where mullets are bred' < mullet n. 'any of various edible, mainly marine fishes …'. Only one third of the derivatives are based on nouns that denote persons, e.g. noshery 'a restaurant; a snack bar' < nosher 'a person who is fond of snacking'. Bases that denote physical objects are clearly very common in this category, and person nouns are less important as bases. OBJECT An OBJECT reading can only be found in four PDE derivatives: cokery 'a cokefurnace', noshery 'food, a snack', pargetry 'concr. plaster or plasterwork', and wormery 'a container in which worms are kept'. The objects denoted by these formations are clearly quite varied, and a transparent pattern cannot be established. The small number of derivatives also indicates that this reading cannot be a productive interpretation in PDE. LOCATION This group is one of the largest in the CONCRETE category. The 17 derivatives represent 20% of all PDE neologisms. 12 of the 17 derivatives with this read- 128 ing are based on nouns denoting objects, sometimes in addition to other interpretations. Only six bases can denote a person, and one base, nite, refers to an abstract concept. Six of the object bases refer to an animal and give rise to a derivative that denotes the place where such animals are kept. Examples for this pattern are gannetry 'a breeding-place for gannets' and puffinry '(a place occupied by) a breeding colony of puffins'. If the base denotes a physical object, e.g. boot, the derivative refers to a place where such objects are handled or can be found. A similar pattern can be observed for the personnoun bases, as their derivatives denote a place where certain persons work or can be found, e.g. Reichschancellery or microbrewery. Almost all derivatives in this group are monosemous. Only four words have additional readings, two with OBJECT , one with COLLECTIVE , and one with ACTION . Derivatives with this reading are created by a regular and transparent pattern. The bases fall into three main groups: they either denote animals, physical objects, or persons, and the derivatives refer to locations associated with these. The high share of monosemous words contributes to the transparency of this reading, and the relatively high number of neologisms indicates a productive reading. This reading can be considered a core sense of -ery neologisms in PDE. COLLECTIVE COLLECTIVE is one of the largest groups in this ontological category. Almost half of the 17 derivatives are coined after 1950, the share of relatively recent coinages is therefore particularly high in this group. All derivatives are potentially based on nouns, although nearly half of them have other possible interpretations as well. Most of these base nouns denote physical objects like twig or corset, and the derivatives refer to such objects collectively. Person nouns are less common as base words, but are possible for six derivatives. These derivatives can usually be interpreted as a collectivity of persons denoted by the bases, e.g. brigandry 'brigands collectively'. Seven derivatives are monosemous, and all of these refer to a collectivity of objects. The remaining derivatives in this group are polysemous. Overlaps occur with a number of readings, but the most frequent of these are with GENERAL ABSTRACT , ACTION , and CONDITION . With regard to the bases, COLLECTIVE neologisms in PDE are different from those in ME. ME coinages are mostly based on nouns denoting persons, while PDE ones are mostly built on nouns denoting physical objects. The derivatives in PDE also usually denote the collectivity of persons or objects denoted by the base, while the ME words are slightly less predictable in this regard. They sometimes denote a collectivity of objects even if the base is a person noun, because they select an object that is commonly associated with that person and refer to a collectivity of such objects, e.g. in grocery 'the goods sold by a grocer'. The opposite, i.e. a base denoting an object, but the derivative denoting people dealing with that object collectively, is also 129 attested, e.g. in wax chandlery 'the department of a royal household concerned with the provision and storage of wax candles' < wax candle. The PDE neologisms with a COLLECTIVE reading are therefore more predictable than their ME counterparts. New derivatives with a COLLECTIVE reading are quite frequent, and the pattern is also transparent. Compared to LOCATION , which has a similar frequency, COLLECTIVE is somewhat less transparent, because of the more varied semantic overlaps. In spite of this, COLLECTIVE can be seen as a productive reading and a core sense of -ery neologisms in PDE. Semantic Map of CONCRETE Derivatives The semantic map in figure 18 represents the frequency of the CONCRETE readings relative to all PDE neologisms, and the number of overlaps within this ontological category. New derivatives with an OBJECT reading are clearly far less frequent than those with a COLLECTIVE or LOCATION interpretation. Although COLLECTIVE contains a high number of polysemous formations, additional readings within the CONCRETE category are rare. Only one COL- LECTIVE word also refers to a location, and two of the four OBJECT derivatives also have a locative interpretation. There are no derivatives with both OBJECT and COLLECTIVE readings. The change in the structure of this category from ME to PDE is immediately apparent. Apart from the now missing reading PERSON , OBJECT is far less common than before, and the readings are linked by far fewer polysemous derivatives. In ME, overlaps exist between all the CONCRETE categories, and some of these, in particular the overlap between LOCATION and COLLEC- TIVE are extremely strong. In PDE, only LOCATION is connected to both other readings, and these overlaps are quite weak. The weaker connection between COLLECTIVE and LOCATION is probably due to a change in the bases found in neologisms with these readings. Both LOCATION and COLLECTIVE are mostly based on person nouns in ME, but this preference has changed in PDE, where object nouns are more common. In ME, the derivatives based on person nouns are the ones that mostly show an overlap between the two readings. The smaller number of such bases in both reading groups seems to lead to a smaller number of derivatives that show this particular polysemy. Although both COLLECTIVE and LOCATION are productive, their share among -ery neologisms has decreased substantially from ME to PDE. One is thus less likely to encounter either reading in PDE neologisms than in ME coinages. 130 Figure 18: Semantic map of CONCRETE readings of -ery neologisms (PDE) 5.3.3.2 ABSTRACT Derivatives The ABSTRACT category is the smallest with only 15 derivatives. These words refer either to a GENERAL ABSTRACT concept, e.g. nothing-buttery 'an oversimplistic approach to the explanation of a phenomenon, which excludes complicating factors; reductionism', or an AMOUNT , e.g. geekery 'the bizarre or grotesque acts performed by a carnival or circus geek, regarded collectively'. Table 16 shows the type frequency of both readings with example derivatives. GENERAL ABSTRACT is much more frequent than AMOUNT , but compared to readings in other ontological categories, in particular ACTION or CONDI- TION , neither of the two ABSTRACT interpretations are common. reading types examples GENERAL ABSTRACT 12 astrochemistry, crookery, nothingbuttery AMOUNT 4 beat poetry, Mormonry, quippery Table 16: ABSTRACT readings of -ery neologisms (PDE) All 15 derivatives can be interpreted as denominal formations, although a third of them may also be seen as deverbal or deadjectival. The bases are usually person nouns, e.g. neurochemist or geek. Only two bases denote physical objects, e.g. microcircuit, and four already refer to an abstract concept, e.g. quip. The clear preference for person nouns is a major difference to the CONCRETE derivatives, which also often have object-denoting bases. GENERAL ABSTRACT The twelve words with this reading either refer to an abstract concept in general, e.g. neoslavery 'a new or modern form of slavery', or a branch of LOCATION COLLECTIVE OBJECT 131 science, e.g. microcircuitry 'the branch of electronics that deals with microcircuits'. This group contains both base nouns denoting objects, and most of the ones referring to an abstract concept. The remaining bases denote persons. Nearly half of the GENERAL ABSTRACT formations are monosemous. Three of these words refer to a branch of science, so that this rather specialised reading seems to establish a pattern. The polysemous derivatives often overlap with COLLECTIVE , e.g. in microcircuitry 'microcircuits collectively; (also) the branch of electronics that deals with microcircuits'. Additional readings in the ACTION , CONDITION , and AMOUNT groups are less frequent. Apart from those derivatives referring to a branch of science, the formations in this group do not show a high amount of regularity. This reading cannot be considered productive. AMOUNT Only four PDE neologisms have an AMOUNT reading. All of them refer to a collectivity of abstract entities, e.g. quippery 'quips collectively'. Only one derivative, beat poetry 'literary work produced by beat poets', is monosemous, and the others often have an additional ACTION reading. These words refer to an action or behaviour and the collectivity of these actions. AMOUNT cannot be considered a productive reading of -ery neologisms. Semantic Map of ABSTRACT Derivatives The semantic map in figure 19 shows the two ABSTRACT readings. GENERAL ABSTRACT is clearly much more common than AMOUNT , but still not particularly frequent compared to other readings of -ery neologisms outside of this ontological category. The connection between these two groups formed by polysemous derivatives is weak - the dashed line represents a single formation with both interpretations. Figure 19: Semantic map of ABSTRACT readings of -ery neologisms (PDE) 5.3.3.3 EVENT Derivatives 44 derivatives, half of all -ery neologisms, have a reading in this ontological category. All of these refer to an action or behaviour, e.g. do-goodery 'the action or fact of doing good', targetry 'the practice of establishing or aiming at (esp. economic) targets', or madcappery 'the behaviour of a madcap'. The distinction between actions, behaviours, and practices that is made in the paraphrases does not seem to reflect differences in the semantics of the derivatives. All of these words could be described with other paraphrases without altering the meaning substantially: targetry could easily be ex- GENERAL ABSTRACT AMOUNT 132 plained as 'the action of establishing or aiming at targets', and madcappery could also be 'action or practice associated with a madcap'. A single reading, ACTION , is therefore assigned to all such derivatives, and this is the only reading that can be found within the EVENT category. Fewer than 20% of the derivatives are first attested after 1950, which is a very low share compared to other categories and readings. Regarding the part-of-speech of the bases, the words in this group are comparable to the overall picture. Slightly fewer than half of all bases with an ACTION reading are based solely on nouns, and half of them have other possible interpretations as well, usually as verbs. Most of the base nouns denote a person, but object-denoting nouns also occur, although these are much rarer. Only a few bases can be interpreted as actions or abstract concepts. A typical -ery neologism with an ACTION reading is thus based on a noun denoting a person and refers to an action carried out by that person or a behaviour that is characteristic of that person. Good examples are hot doggery 'behaviour characteristic of a hot dog; showing off, hot-dogging' < hot dog 'a flashy, ostentatiously successful person; a show-off', mumpery 'begging' < mumper 'begger' (or mump v. 'to beg'), or banditry 'the practices of bandits'. The derivatives based on an object refer to an action that is carried out in association with that object, e.g. air gunnery 'the skill or practice of aiming and firing guns from on board a military aircraft' < air gun. Half of the derivatives with this reading are monosemous, which is not a particularly high share. However, the other half show a highly regular additional reading, namely CONDITION . 17 derivatives that refer to an ACTION also refer to a CONDITION . Other overlaps, e.g. with AMOUNT or LOCATION , are also attested, but these occur only rarely. Only the overlap between ACTION and CONDITION is highly systematic. Derivatives that have both readings are, for example, smuggery 'the quality or condition of being smug, or an instance of this', or Peter-Pannery 'immaturity, childishness; behaviour characteristic of a Peter Pan'. Unfortunately, the OED does usually not provide information as to which of these two readings is expressed first by derivatives. Derivatives with an ACTION reading are frequent and transparent. They are mostly created on nouns denoting persons, but some verbs or adjectives are also possible as bases. Formations that are not monosemous mostly show highly regular and predictable semantic overlaps. This reading is clearly a highly productive reading of -ery neologisms in PDE and can be considered a core sense. 5.3.3.4 STATE Derivatives This ontological category contains 25 derivatives that refer to a condition of being or a characteristic that is better described as a SITUATION than SUB- STANCE / THING / ESSENCE . As in the EVENT category, there is only a single reading, CONDITION , that describes all STATE derivatives. Examples are mar- 133 tial artistry 'achievement or skill in a martial art', Proustery 'Proustian manner or style', or teenagery 'the period or state of being a teenager'. More than half of all bases can be interpreted as nouns and verbs, or nouns and adjectives. Most of the remaining bases are clearly nouns. The nominal bases predominantly denote persons, e.g. martial artist or crook. Only a small percentage, one fifth, of the derivatives in this group are monosemous. The majority of formations has additional readings, which sometimes occur with GENERAL ABSTRACT , AMOUNT , and COLLECTIVE , but are mostly with CONDITION . 17 derivatives, which equals 70% of all CONDITION formations, also have an ACTION reading. A similarly strong overlap between these two readings can already be found in ME, where we find a sense extension ACTION → CONDITION . As both interpretations are given in the same paraphrase in PDE, such a relationship cannot be established here. Derivatives with a CONDITION reading are quite frequent, and their derivation is also systematic. The polysemies shown by these formations are regular as well. This suggests a somewhat productive reading. 5.3.3.5 Semantic Map of PDE Neologisms The semantic map in figure 20 shows all attested readings of PDE neologisms and the number of polysemous derivatives. ACTION dominates the map, as half of all derivatives can be found in this group. The next most frequent reading is CONDITION . LOCATION and COLLECTIVE , although smaller than CONDITION , are also quite common. They are also productive readings of -ery in PDE and are, like ACTION , core readings of this morphological category. GENERAL ABSTRACT is still less frequent than LOCATION and COLLECTIVE , and AMOUNT and TENURE occur only rarely in neologisms. None of the last three readings can be considered productive. Although none of the readings is isolated on the map and most groups are connected with multiple others, the links between the readings through polysemous derivatives are rather weak. The dashed lines on the map represent one or two polysemous derivatives, and the continuous lines indicate that three to five words show a particular polysemy. The only exception to this is the connection between ACTION and CONDITION , which is very strong and systematic. A sense extension could, however, not be established, because the diachronic distribution of these two readings does not become clear based on the OED data. 134 Figure 20: Semantic map of all reading of -ery neologisms (PDE) 135 5.3.4 Comparison of ME and PDE Neologisms A number of differences are immediately apparent when comparing the semantic maps of PDE and ME derivatives, which is reproduced for convenience as figure 21. Perhaps the most obvious of these differences is the smaller number of readings in PDE. Compared to ME, a number of readings have disappeared completely: TENURE , RIGHT , CHARGE , PERSON , and POSITION . The first four are already rare occurrences in ME, so it is only a small step from being scarcely expressed in neologisms to not being expressed at all. POSITION , however, is quite frequent in ME, and can also be considered transparent at that point. The complete disappearance of this reading is thus rather surprising. This development may be connected with the change in the bases selected by derivatives with a COLLECTIVE and LOCATION reading. In ME, these two groups as well as POSITION select mostly person nouns as bases. But in PDE, object-denoting base nouns are more common than person nouns in these groups. It was pointed out that POSITION is a systematic pattern in ME, as it mostly selects person nouns to denote a position in society or an office held by that person, but given the high number of person nouns selected as bases for other readings of -ery, it does not exploit this pattern to a great extent. The number of person-noun bases has decreased substantially since ME, not only in neologisms with COLLECTIVE and LOCATION readings. While two thirds of all bases denote a person in ME neologisms, only half of all new coinages in PDE have such a base. This significant difference in the derivational pattern from ME to PDE shows again that the semantics of derivational suffixes are not necessarily diachronically stable contrary to the results of some previous investigations. The POSITION reading seems to have vanished completely at the same time as person noun bases have become less common. Another noticeable difference in terms of frequency is the increased size of ACTION , which is already the largest group in ME, but contains an even higher share of PDE neologisms. CONDITION has also increased its share of neologisms. The CONCRETE readings OBJECT , LOCATION and COLLECTIVE , however, contain a smaller percentage of derivatives in PDE than in ME. As some readings have disappeared completely, new coinages seem to concentrate on a smaller number of readings that account for a greater percentage of neologisms. 136 Figure 21: Semantic map of all reading of -ery neologisms (ME) 137 Regarding the connecting lines, the ME and PDE maps are quite different as well. The ME map contains a staggering amount of connections, both weak links and strong ones, and even the smallest groups are linked to four or five other readings through polysemous derivatives. The suffix was borrowed from French in this period, so these are the earliest attestations of -ery in English. The complexity of the semantic map at such an early point shows that this borrowed suffix already expresses a high level of polysemy from the beginning, and does not develop a more complex polysemy network over time. This also proves that -ery is a single polysemous suffix and not multiple homonymous ones. The PDE map is of course somewhat clearer, partly because it contains fewer readings. It is striking though that most of the connecting lines attested in PDE are quite weak. The most frequently attested overlap apart from ACTION - CONDITION is that between COLLECTIVE and GENERAL ABSTRACT , which occurs only four times. The weaker linkage may be partly due to the fact that there are fewer PDE than ME neologisms, which should of course be kept in mind when comparing absolute figures. The numbers of derivatives in the individual readings in PDE would allow more polysemous derivatives though. Both the reduction of readings and the decrease in the number and thickness of the connecting lines lead to a PDE map that seems clearer and simpler than the ME one. This development is equivalent to that of -age. The dictionary investigation has shown that -ery derivatives form a single polysemous category from the earliest point of the suffix's attestation in English. But this category is subject to change over time. Some readings disappear from neologisms, and the connections between readings are not the same in both time periods investigated here. Another noticeable difference concerns the bases that are selected by the suffix: in ME these are mainly nouns denoting persons, but in PDE such bases have become less frequent and have made room for object-denoting nouns. On the other hand, the core senses in both periods have not changed. ACTION is the most productive reading both in ME and PDE and it is also connected to almost all other readings. But COLLECTIVE is also productive and well-connected and can thus be regarded as a core sense as well. LOCATION is somewhat less strongly linked to other readings, especially in PDE, but it is still a productive reading and can therefore be considered a core sense. So in spite of the changes in the semantic structure and formal make-up of -ery neologisms, there is also a certain consistency in this category. 5.4 Summary: The Semantic Structure of -age and -ery Neologisms in the OED The -age and -ery neologisms attested in the OED have many similarities, but also exhibit important differences. They show a similar range of readings, 138 with ACTION being a core reading of both morphological categories in both time periods investigated, but they do differ with regard to the other core readings they express - CHARGE and AMOUNT for -age derivatives and COL- LECTIVE and LOCATION for -ery. The dictionary investigation has shown that there are some significant changes in the semantic structure and the structural properties of the derivatives. The core readings of -age derivatives have changed from ACTION and CHARGE in ME to ACTION and AMOUNT in PDE. While the core readings of -ery derivatives have remained stable from ME to PDE, their structure has changed. In ME, -ery neologisms are coined mostly on nouns denoting persons, but this tendency is much weaker in PDE, where nouns denoting objects are much more common. The PDE pattern also incorporates a large number of compounds as bases, which shows that the derivation of -ery formations has changed quite substantially. Semantic maps are a particularly suitable tool to reveal the overlaps of polysemous derivatives, and these prove to be distinct in each morphological category. Middle English derivatives of -ery show a strong overlap of COLLECTIVE and LOCATION , which is not attested for -age derivatives to the same extent. This overlap seems to be due to the structure of locative -ery formations, which denote places in which various objects are kept, e.g. glovery 'a place in which gloves are made or sold' or pottery 'a factory where porcelain earthenware, etc. is manufactured'. They also often refer to such objects collectively, which explains the COLLECTIVE - LOCATION overlap. Such a relation is not visible for -age derivatives, as these show a closer connection between ACTION and LOCATION than their -ery counterparts. Locative readings of -age formations can be understood as sense extensions of ACTION , and they denote a location at which an action takes place. A good example for this is rivage 'a coast, a shore' ( LOCATION ) and 'landing on a shore' ( ACTION ). So although both -age and -ery derivatives have ACTION , COLLECTIVE , and LOCATION readings, the relations and exact nature of these readings are quite different in the two groups. It therefore seems likely that other affixes, even those with a similar semantic range, would have similarly local patterns that are unique to a particular word formation process. In spite of the many similarities between the derivatives of -age and -ery, there are pronounced differences in the details of their semantic structure, and these two suffixes should not be considered semantically identical. 139 6 Corpus Investigation of -age and -ery Derivatives The corpus analysis discussed in this chapter provides information on the usage of -age and -ery derivatives in natural language. The data comes from the British National Corpus (Davies 2004-; and the freely available BNC Simple Search provided by the British Library at http: / / www.natcorp.ox.ac.uk/ ) 7 , so this investigation is purely synchronic and considers only the current, or, to be more precise, late 20 th century, usage of -age and -ery derivatives. Both type and token frequency are taken into account in this investigation, which is a particular advantage of corpora in comparison to dictionaries. This analysis can thus show whether derivatives are predominantly used to express a certain reading, or whether the different readings of polysemous derivatives are used with similar frequencies. An analysis of the derivatives in each reading group will also show which readings are currently productive, and whether there is evidence for regular sense extensions from certain readings to others. The question of semantic change will not be a concern in this chapter, as the data source is purely synchronic. An additional investigation based on Middle English corpora would, of course, be conceivable, and one could then find out whether the semantic structure of the suffixes has changed since ME. Unfortunately, the semantic classification of -age and -ery derivatives in Middle English texts is very difficult, and I do not feel that my judgements in this regard are reliable enough to include such an investigation of ME corpora in this study. Semantic maps for -age and -ery derivatives attested in the BNC will be created, and the findings will then be compared to the conclusions drawn on the basis of the dictionary data in the previous chapter. 6.1 Data Source and Methodology The British National Corpus (BNC) provides the data source for this part of the investigation. It contains about 100 million words from different genres and registers, both spoken and written. The corpus is accessible online (http: / / corpus.byu.edu/ bnc/ ) free of charge, and can be searched with the help of a relatively simple syntax. Similarly to the OED online, one can only search for a string of letters and not for affixes, although the corpus is tagged for various grammatical categories. In order to find nominal derivatives of -age and -ery, one needs to clean up the results obtained by searching for 7 The rights of the texts cited from the BNC are reserved. 140 nouns ending in <age> and <ry> 8 . Only words that are potentially transparent derivatives of -age and -ery are investigated, so possible bases for these formations need to be attested independently or within another derivative with similar semantics in the BNC or in the OED. Words such as age, language or message are therefore excluded. The criteria employed to distinguish between transparent derivatives on the one hand, and opaque derivatives and words that are no derivatives of -age or -ery at all on the other hand are the same as in the dictionary-based investigation (see section 5.1). But corpora and dictionaries are very different sources, the former usually contain natural language samples, the latter processed information about different aspects of words used in a certain language. Many advantages and disadvantages are connected with both of these sources, which is, of course, the reason for including both in this investigation. Some of the problems that are particular to corpus-based research are discussed in the remainder of this section. One of these issues concerns alternative spellings or misspellings of words, e.g. re-marriage and remarriage. Words with non-standard spelling are included, and the tokens of the alternative spelling are added to the token count of the type with standard spelling. Another problem not encountered when working with a dictionary is the occurrence of non-English words. If a word appears within a foreign language expression or clause, e.g. dommage in c'est dommage! , it is excluded from the result file. In spite of limiting the search to nominal results, the tagging of the corpus is not completely accurate. Verbs and adjectives that are sometimes found by the search query are excluded from the result file. The corpus contains some derivatives with multiple affixes that pose potential problems. These are kept as independent types if possible bases are attested independently in either the BNC or the OED. This leads to the establishment of independent types such as ill-usage < ill-use and water-storage < water-store. Some of these can also be understood as prefixed words that already contain -age or -ery, or as compounds with an -age or -ery derivative as their second element, but as possible bases for the suffixation are attested for all of these types, they are certainly possible derivatives of one of the two suffixes. However, this only affects a small number of types, and not much hinges on this decision, because these words are certainly derivatives of -age or -ery, even if the bracketing of these formations is debatable. The most prominent problem in cleaning up the results is names. Names are clearly words and may even be derivatives of derivational affixes. But this study investigates the lexical meaning, i.e. the sense, of complex words, and names are usually considered to not have a sense (cf. Murphy 2010: 41). The BNC contains a large number of names, all of which are excluded from 8 The search strings used for this investigation are *age.[n*] and *ry.[n*], which search for all words tagged as nouns ending in <age> and <ry>. 141 this analysis. This concerns personal names, but also place names, company names, and names of products. Even if some company or product names sometimes look like -age derivatives, it is not clear what they denote - apart from the company or product, of course. The cosmetics product immunage in the sentence “We recommend Elizabeth Arden's Immunage UV Defence Lotion” (HA6 3755), for example, could be a derivative of -age, but it could also be a compound with the noun age as its second element. Even if it is assumed to be an -age derivative, it is doubtful what exactly this word means. In a study that is concerned with the semantics of complex words, such coinages can thus not be considered 9 . Apart from names, words with unclear meaning are also excluded from the analysis, because, similarly to names, they cannot provide the kind of information necessary for this investigation. An example is flowage in the following context: “Anyway, I'm sure that those people you know who can skate with their legs tied behind their backs are definitely worthy of flowage” (ARM 747). The semantics of the putatively transparent -age and -ery derivatives established in this way are classified according to the same principles as the OED data in the previous chapters (cf. section 4.1 for more information). The four broad ontological categories CONCRETE , ABSTRACT , EVENT , and STATE are subdivided into the reading groups already established for the dictionarybased investigation. These readings account for all the derivatives found in the BNC. However, the utterances in the corpus present specific problems in this area as well, as the distinctions between different readings are not easy to draw in every instance. The discrimination between an action and a condition, and also the one between an abstract entity and an amount is especially difficult at times. Consider, for example, the following sentence: This work has involved the laying of a' spine' of fibre cable across the campus -to which ultimately all faculty and school computer networks will connect -and the installation of satellite networks in all the Schools of the Engineering Faculty together with the provision of hardware bridges to allow their linkage to the main' spine' (G31 391) Linkage could either refer to the action of linking or to the condition of being linked in this example. In many similar cases clues from the context can help to identify a single reading for a derivative. Verbs like stop or occur, for example in “[l]eakage occurred with equal frequency in both groups” (HU3 587), suggest an ACTION reading, and expressions that invoke a number or extent, like the measurement unit in “the off-load d.c. output voltage was a shade over 15V” (C92 1861), suggest an AMOUNT interpretation. Expressions that indicate origin or destination, e.g. from or into, regularly indicate an ACTION reading, as in “there is leakage into surrounding skin tissue” (CDR 9 As a consequence of this procedure, a small number of types that are quite clearly derivatives of -age were excluded, e.g. the company names Eco Emballage and Mercantile Lighterage Ltd. 142 1936). A CONDITION reading is often signalled by prepositions such as in, as in “the most onerous requirement is that he must spend six months in pupillage” (FRA 1413). In spite of this, ambiguities occur regularly and have to be dealt with in a consistent way. Thus, whenever a derivative cannot be clearly assigned to a single reading group, it is assigned to all readings that are possible interpretations. So linkage in the above example is counted once as ACTION , and once as CONDITION . Note that the adapted token count of some types is higher than that given in the BNC due to these double or triple classifications of some derivatives. A reading is assigned to every token of each type for those with a token frequency of less than 100. For types with higher token frequencies, the BNC sample function is used to show 100 examples out of the total token count. The frequency of each reading for these words can be seen as a percentage: If 94 out of 100 samples refer to an action, it is assumed that 94% have an AC- TION reading. This percentage is then extrapolated to the whole token frequency. The token frequency of pilgrimage, for example, is 477. This is multiplied with 0.94, as 94 of the sampled tokens refer to an action. 448 tokens are thus assumed to have an ACTION reading. If tokens have to be excluded for one of the reasons discussed above the total token count is reduced in the same way. The token frequency of village, for example, is given as 10,948 in the BNC, but as six of the 100 samples had to be excluded, the adapted token frequency is reduced by 6% to 10,291. The results in all of these cases are rounded to the closest integer. A list of all types that are analysed as derivatives of -age and -ery together with their readings and the adapted token frequency can be found in the appendix (Appendix E and Appendix F). 6.2 The Suffix -age in the BNC After applying the methods mentioned above, 247 words can be analysed as -age derivatives. Table 17 gives the type and token frequencies of the different ontological categories. ontological category types tokens examples CONCRETE 115 24,132 baronage, cottage, orphanage ABSTRACT 89 28,001 acreage, advantage, mileage EVENT 126 16,111 breakage, re-marriage, shrinkage STATE 38 6,256 cousinage, vassalage, vicarage Table 17: Ontological categories of -age derivatives (BNC) 143 A quarter of these formations are hapax legomena, i.e. they are attested only once in the corpus, and about half of them have frequencies of less than ten. Words with extremely high frequencies are much less common than those with low frequencies - only a quarter occur more than 100 times, and 15 more than 1,000 times. Such a distribution of rank and frequency is in fact quite common and has been described many times before (cf. Zipf 1949/ 1972). Nearly 10% of the 247 derivatives found in the BNC are not recorded in the OED. Unsurprisingly, most of these are hapax legomena, dislegomena, or trislegomena, e.g. grindage in “Astin should definitely chill a little on the grindage” (CGC 442). Many of these new formations could be interpreted as compounds or derivatives with a second element that is an -age derivative, e.g. steam-haulage or un-marriage, but as putative bases are also attested independently, steam-hauled and un-marry in these cases, they can be perceived as direct -age derivatives as well. A few derivatives with slightly higher frequencies are also not attested in the OED: finnage 'the fins of fish collectively' (19 tokens), hectarage 'the amount of hectares' (eleven tokens), and driveage 'a driveway' (ten tokens). As semantic paraphrases for most of these words are not available anywhere, it is sometimes extremely difficult to establish their meaning on the basis of a single sentence. Grindage, for example, could well refer to an action, but a web search strongly suggests the meaning 'food' for this word, which also fits the context. The semantics of these presumably new formations will be analysed in detail together with all derivatives in the following sections. 6.2.1 CONCRETE Derivatives All four CONCRETE readings found in the dictionary are attested in the BNC as well. Table 18 gives their respective type and token frequencies and examples for each group. reading types tokens examples OBJECT 45 4,557 bandage, carriage, sewage PERSON 5 521 appendage, hostage, personage LOCATION 30 13,894 cottage, orphanage, village COLLECTIVE 59 5,160 assemblage, baggage, signage Table 18: CONCRETE readings of -age derivatives (BNC) The differences in frequency are substantial. PERSON readings are very rare, regarding both types and tokens. The token frequencies of OBJECT and COL- LECTIVE are moderate and very similar, but the number of types in these two 144 groups is different with COLLECTIVE being the substantially larger group. LOCATION is very frequent with regard to tokens, it even exhibits the second highest token frequency of all reading groups, but it contains only half the number of types as COLLECTIVE . This suggests that most of the derivatives with a LOCATION reading are high-frequency words, which could indicate an unproductive pattern. The situation is very different for COLLECTIVE , which contains a large number of words with relatively low token frequencies. This reading is therefore more likely to be productive. Each of the readings will be discussed separately below to show whether the initial impression regarding productivity is correct. OBJECT 45 types are responsible for the roughly 4,500 tokens with an OBJECT reading. The most frequent type of these by far is carriage (1,515 tokens). Other highfrequency items include sewage (711 tokens), package (385 tokens), fuselage (244 tokens), and bandage (213 tokens). Altogether ten types have frequencies over 100. 12% of all OBJECT words are hapax legomena, which is significantly below the overall percentage of hapaxes, and many of the types with OBJECT readings are lexicalised formations that are semantically not completely transparent. Carriage, for example, denotes a very specific means of transportation, something that cannot be predicted from the meaning of the base and the suffix alone. Furthermore, even the low frequency items are often polysemous and have much higher token frequencies in other reading groups. Especially the amount of overlaps between OBJECT and ACTION suggests that this reading is likely to be a sense extension of ACTION . More than half of all OBJECT types also show an ACTION reading. This overlap is even more pronounced here than in the dictionary data, as the semantic maps in figures 22 and 23 show. These derivatives can be interpreted as the concrete results of an action, but the same formation may also refer to the action itself. For example leakage refers to the action of leaking in 96% of the tokens in the BNC, but it may also denote the object, the thing that has leaked, in 5% of all tokens. However, one of the hapaxes, grindage, is not attested in the OED. A web search suggests that grindage is a Californian expression for food, and this derivative resembles some established formations like spoilage or seepage. These are also based on verbs, as grindage presumably is, and denote a physical object that is the result of an action. This derivative therefore shows that the OBJECT reading of -age derivatives is not completely unproductive, although the lack of monosemous low-frequency formations and the prevalence of high frequency types in this group makes it clear that this is not a particularly productive reading, but is probably a sense extension of ACTION . 145 PERSON Only five different types denote a person. These are appendage, baggage 'a worthless woman', hostage, ex-hostage, and personage. Hostage is by far the most frequent of these and gives rise to 84% of all PERSON tokens. This reading is clearly not productive, as only a small number of words show it, and most of them are lexicalised formations. This finding is in line with the results obtained from the dictionary data, where no neologism with a PERSON reading is attested in PDE. LOCATION LOCATION is the most frequent of the CONCRETE readings with regard to token frequency, and also one of the most frequent readings in total. Only 30 types are responsible for such a high token frequency, which suggests a limited productivity of this reading. This group contains some of the most frequent types, among them village (9,963 tokens), and cottage (2,432 tokens). These two already account for nearly 90% of all LOCATION tokens. Only three items, bailliage, cellarage and harbourage, are hapaxes, and all of these are Early Modern English coinages. So in spite of the extremely high token frequency of this reading, it does not seem to be particularly productive. But this group also contains two types that are not listed in the OED: curbage 'parking space' and driveage 'ground to drive on'. So new, transparent derivatives are clearly coined with this reading. This is unexpected, given the prevalence of high-frequency formations with this reading. A factor that might explain the existence of some neologisms is the relative frequency of derivative and base word. Hay (2003) has found that relative frequency plays a major role in the decomposition of complex words. She argues that if “the derived form is more frequent than the base it contains, it is more difficult to decompose”, but if “the base base is more frequent than the whole, the word is easily and readily decomposable” (Hay 2003: 88). Many of the LOCATION derivatives, e.g. vicarage (derived form: 178 tokens, base: 868 tokens), and anchorage (derived form: 61 tokens, base: 584 tokens), indeed have much lower token frequencies than their bases. These words are thus easily decomposable. The degree of decomposability also has an affect on the semantic transparency of the derivative, however. Complex words that are likely to be decomposed, like vicarage and anchorage, are also likely to express a reading that is closely connected to the reading of the base (cf. Hay 2003: 117). Such formations are thus semantically transparent. So in spite of the existence of many high-frequency forms among the locative -age derivatives, this reading is transparent, and may thus give rise to new coinages, which it clearly does. Almost half of all LOCATION derivatives also have an ACTION reading, e.g. passage 'the action of moving' and 'a place where there is a way through', or anchorage 'the action of anchoring' and 'a place for anchoring'. This clearly shows the close link between the two readings, which has also become evi- 146 dent in the dictionary data. LOCATION has been shown to be a sense extension of ACTION in ME OED neologisms, but the corpus data do not allow us to draw the same conclusion. The overlap is quite frequent, but the evidence is not conclusive with regards to directionality. While some polysemous derivatives denote an action in the majority of cases - 90% of all tokens of storage, for example, refer to an action and only 6% denote a location - this is not the rule. There are counterexamples like anchorage, where 45% of tokens have a LOCATION , and 25% an ACTION reading, and many words for which the number of tokens with both of these readings is very similar, e.g. pasturage (5% LOCATION , 6% ACTION ), or cooperage (three tokens LOCATION , three tokens ACTION ). So some derivatives, e.g. storage, can be interpreted as the location at which an action takes place, and their LOCATION reading is probably a sense extension of a more prominent ACTION reading. But this is not true for all or even for the majority of formations with a locative interpretation, and LOCATION should therefore not be seen as a sense extension of AC- TION . This group contains many established formations that sometimes give rise to thousands of tokens. However, many of these words are decomposable, because the frequency of the derived form is lower than that of its base. This transparency may encourage some new coinages with this reading. LOCATION is thus not completely unproductive. COLLECTIVE The token frequency of COLLECTIVE is mid-range compared to other readings, but the type frequency of 59 is quite high. This means that the number of tokens per type is low, which suggests a productive reading. This assessment is further supported by the fact that only a small number of highfrequency types like heritage (736 tokens), foliage (706 tokens), or luggage (585 tokens) are responsible for a large share of the tokens. However, only seven hapax legomena have COLLECTIVE readings, and only one derivative, sublineage, is not attested in the OED. The analogy of this formation to established coinages like lineage or matrilineage is evident, and it can be interpreted as the result of sub- + lineage as well as sub-line + -age. The existence of this formation should thus not be used as an argument for a high productivity of this reading. Of the 59 COLLECTIVE types, slightly more than one third, including the hapaxes, are monosemous. Another third have additional ACTION readings, which is illustrated on the semantic maps in figures 22 and 23. Most of these formations have far more tokens in the ACTION than in the COLLECTIVE group, e.g. sewerage (86% ACTION , 51% COLLECTIVE ) or drainage (90% ACTION , 13% COLLECTIVE ). But there are a few words for which this relation is reversed, such as coinage, which refers to a collection (94% of tokens) far more often than to an action (3% of tokens). In spite of those few exceptions, COL- LECTIVE readings can be considered sense extensions of ACTION , and are in 147 this regard very similar to many of the OBJECT derivatives discussed above. COLLECTIVE seems to be a particularly productive sense extension of ACTION , as this group contains a high number of low-frequency types. A concrete result of an action is therefore more likely to have a COLLECTIVE than an OB- JECT reading. 6.2.2 ABSTRACT Derivatives Five different ABSTRACT readings are attested in the BNC, but three of them are extremely rare. Table 19 shows the type and token frequencies and examples for each reading. reading types tokens examples GENERAL ABSTRACT 22 12,576 advantage, marriage, passage AMOUNT 49 15,314 acreage, mileage, percentage CHARGE 19 48 corkage, costage, stallage RIGHT 4 56 herbage, pasturage, patronage TENURE 2 7 burgage, socage Table 19: ABSTRACT readings of -age derivatives (BNC) There are hardly any instances of a TENURE reading in the corpus, regarding both type and token frequency. RIGHT gives rise to only marginally more types, but these have higher token frequencies than the TENURE formations. The token frequency of CHARGE is comparable to that of RIGHT , but the type frequency of CHARGE is relatively high, which suggests a productive reading. AMOUNT and GENERAL ABSTRACT are very different from the other three readings in this ontological category, as their token frequencies are very high. But there is an important difference between these two groups as well: GEN- ERAL ABSTRACT does not contain many types while the type frequency of AMOUNT is also high, suggesting that AMOUNT is a productive interpretation. GENERAL ABSTRACT The 12,576 tokens with a GENERAL ABSTRACT reading are due to only 22 types. Some of these are staggeringly frequent, in particular advantage (7,121 tokens), and only very few derivatives are low-frequency words. This group contains only one hapax: sufferage 'permission, approval', a 17 th century coinage. Furthermore, no new, i.e. not attested in the OED, types have a GENERAL ABSTRACT reading. The words in this group are established formations that are often semantically opaque as well. Consider for example advantage: While it is theoretically possible to link this derivative to the verbal base advance, this decomposition is unlikely to be made by speakers. The slight 148 difference in phonology aside, the derived word has a much higher token frequency than the base (854 tokens in the BNC), which discourages decomposition of the derived form (cf. Hay 2003: 88), and can also lead to semantic changes in the derivative. Hay (2003: 117) finds that “[r]elative frequency affects semantic transparency - derived forms which are more frequent than their bases tend to drift away from the meaning of the base. They also proliferate in meaning, and so tend to be more polysemous than derived forms which are less frequent than their bases”. Other derivatives in this group, e.g. vintage, are based on stems rather than independent words, which also impedes decomposition. All of this points to the fact that GENERAL ABSTRACT is not a productive reading of -age derivatives. Some formations have a lower frequency than their bases, however. Leverage, for example, gives rise to 212 tokens with a GENERAL ABSTRACT reading, and to 239 tokens in total, but its base lever is attested 629 times. This should mean that leverage is easily decomposed and therefore also semantically closely related to its base. This is indeed the case for those tokens of leverage that refer to the action of a lever, a reading that the OED lists as the earliest reading for this derivative. But this reading is only attested in 11% of the leverage tokens in the BNC. The GENERAL ABSTRACT reading that most tokens show is paraphrased in the OED as a later sense: 'the power of a lever; the mechanical advantage gained by the use of a lever' and 'advantage for accomplishing a purpose; increased power of action'. This confirms Hay's results that high frequency words that are less frequent than their bases often acquire new readings, but their original readings are still present (Hay 2003: 117). In this case, GENERAL ABSTRACT is the sense extension of an earlier ACTION reading, but it is now more frequent than the original reading. Such a GENERAL ABSTRACT reading might principally become productive, as it is potentially transparent. But this will probably not happen in practice, because there is no obvious semantic similarity across a number of derivatives. Derivatives like leverage 'the power of a lever', advantage 'benefit', or mileage 'potential' have too little in common to formulate a common sense that accounts for the semantics of all of them. GENERAL ABSTRACT is thus quite unlike LOCATION , which is another highly frequent reading of -age derivatives. The type and token frequencies of these two readings are similar, but LOCATION derivatives are more transparent. This transparency is probably the reason for the coinage of a few neologisms with a LOCATION reading. New formations with a GENERAL ABSTRACT reading, however, cannot be found in the BNC. This shows again how important transparency is for productivity: If a group of derivatives is transparent, new coinages with a similar reading may occur even if other indicators, like frequency, suggest this reading is unproductive. 149 CHARGE A CHARGE reading occurs only 48 times in the data, so the token frequency is one of the lowest. The type frequency, however, is quite high: 19 different derivatives have such a reading. More than a third of these are hapax legomena, which is a high share of such low-frequency types. The most frequent type is cheminage, which can be linked to the base chimin n. 'a way or road', giving rise to eleven tokens. This distribution - a high number of lowfrequency types and no types with high frequencies - suggests a high productivity of this reading. But most of these formations are old coinages that are used only rarely and in specialised registers, academic texts referring to medieval taxes being the most common. The high number of lowfrequency forms does not signify a currently productive word formation process in this case. Still, some charges are mentioned with current reference, e.g. pilotage 'payment for the services of a pilot', or corkage 'charge for uncorking and serving bottles of wine that have not been bought in the place they are consumed'. These usages show that this reading is not completely historical, but it cannot be considered a productive reading of -age; a finding which is in line with the result from the dictionary-based investigation of PDE neologisms. RIGHT The reading RIGHT is only slightly more frequent than CHARGE . The 56 tokens with this reading are due to only four types, and one type, patronage, is responsible for 52 of them. The other three derivatives commonage, herbage, and pasturage, contribute one or two tokens each. The texts in which these three words are attested all refer to historical periods, and none of these words seems to be used outside of this particular register. The only word in current use is patronage, but this derivative refers to an ACTION (77% of tokens) far more often than to a RIGHT (21% of tokens). This reading is unproductive. TENURE The least frequent reading in the ABSTRACT category, TENURE , occurs only seven times. The two formations burgage and socage are responsible for these tokens. Both contexts are clearly about the medieval period, so that it is save to say that this reading is unproductive. AMOUNT AMOUNT is the second most frequent reading of -age derivatives in the BNC. Although the type frequency of 49 is also the highest within the ABSTRACT category, the type-token ratio is quite low compared to readings in other categories, because AMOUNT contains a few words with extremely high token frequencies. The most common type is package (4,891 tokens), closely followed by average (3,461 tokens) and percentage (2,756 tokens). These derivatives illustrate the different meanings of words subsumed under this read- 150 ing: package often refers to a collection of software programs developed for computers, so it is a collection of abstract entities. The other two derivatives refer to an amount or number, which is the other possible interpretation of words in this group. AMOUNT contains six hapaxes, of which one, ampage 'amount of amperes', is not attested in the OED. Four of the remaining derivatives are 20 th century coinages, and one is already attested in the 19 th century. Most of these can therefore be considered relatively new formations. Also, many of the words with an AMOUNT reading, especially the low-frequency formations, are monosemous. In spite of this, 60% of all derivatives in this group have additional readings, very often ACTION , but also COLLECTIVE and OBJECT . These overlaps are illustrated on the semantic maps in figures 22 and 23. The words that have overlaps with ACTION usually have higher token frequencies in that group than for AMOUNT , e.g. shrinkage 'the act of shrinking' (85% of tokens), 'the amount of such contraction or loss in bulk, volume, or measurement' (15% of tokens). In many of these cases, the OED also lists an ACTION reading earlier than an AMOUNT reading. In fact, an AMOUNT reading is not recorded there at all for a few derivatives that clearly show it in the BNC, like breakage in the following example: “Skull and mandible breakage is also higher in the fledglings' pellets: 22-25 per cent in the pellets from the adults compared with 31-72 per cent for the fledglings” (B2C 806). All of this suggests that AMOUNT can be seen as a sense extension of an ACTION reading in many polysemous derivatives with both interpretations. The AMOUNT reading is then a specialized resultative interpretation of an ACTION reading, which in fact mirrors the relationship between ACTION and COLLECTIVE . Three derivatives, ampage, hectarage, and subpackage, are not attested in the OED, so new formations with this reading are clearly coined. And they are used regularly - hectarage has a token frequency of eleven, one of the highest among the forms not listed in the dictionary. Similarly to LOCATION , the derivatives in this group are very transparent, as most of the derivatives can be paraphrased using as 'amount of' and 'collectivity of abstract entities'. Most of the formations with an AMOUNT reading also have lower token frequencies than their bases, which increases their decomposability and semantic transparency. This can encourage new coinages with similar semantics. In spite of the existence of some highly frequent types in this group, this reading can be considered a productive reading of -age. This result also corresponds with the dictionary investigation, where AMOUNT was one of the most common readings among PDE neologisms. 6.2.3 EVENT Derivatives The only reading in this ontological category that is attested in the BNC is ACTION . But as can be seen in table 20, this reading is very frequently found in the corpus. 151 reading types tokens examples ACTION 126 16,611 leakage, steerage, storage Table 20: EVENT readings of -age derivatives (BNC) ACTION has the highest token frequency of all readings and also by far the highest type frequency: 126 types give rise to 16,611 tokens. This means that half of all -age derivatives found in the corpus have an ACTION reading. Although some of these types exhibit very high token frequencies, e.g. marriage (3,127 tokens) or storage (2,611 tokens), most words have token frequencies of less than ten, and nearly a quarter are hapax legomena. Such a large amount of low-frequency words suggests a productive process. This group also contains seven derivatives that are not attested in the OED, e.g. drystorage and computer-usage. Most of these contain the element usage, and could also be interpreted as compounds, but as a possible first element, e.g. computer use, is attested independently as well, they are included here. Another interesting point is the number of monosemous types. Half of all formations in this group only refer to an ACTION , among them some of the high-frequency words like pilgrimage. The polysemous types usually refer to an ACTION in the majority of cases, and this reading is the origin of sense extensions to a number of readings, e.g. OBJECT and COLLECTIVE . Although some high-frequency derivatives with ACTION readings, e.g. espionage (239 tokens; base: 0 tokens), are more frequent than their bases, a situation that inhibits their decomposability, most of the formations are based on extremely frequent words. Usage (1,150 tokens), which is one of the most frequent types in this group, is considerably less common than its base use (61,233 tokens). The same is true for other derivatives like storage (2,611 tokens) and its base store (4,276 tokens), or drainage (865 tokens) and drain (1,127 tokens). Despite the fact that ACTION contains some established formations that are unlikely to be decomposed, and therefore semantically potentially opaque, many high-frequency derivatives as well as lowfrequency ones are easily decomposable. This contributes to the transparency of this reading. Due to the high type and token frequencies, this reading is clearly central for -age derivatives. It is also a transparent reading, and the amount of lowfrequency types suggests that it is productive as well. This result corresponds with the findings of the dictionary-based investigation, in which ACTION also emerged as a highly productive reading of PDE -age neologisms. 152 6.2.4 STATE Derivatives Two different STATE readings, CONDITION and POSITION , are attested in the corpus. Table 21 shows the token frequencies and examples for the different readings. reading types tokens examples CONDITION 35 6,102 hostage, linkage, marriage POSITION 3 154 peerage, vantage, vicarage Table 21: STATE readings of -age derivatives (BNC) CONDITION Most of the CONDITION tokens are due to the extremely frequent word marriage (5,034 tokens with this reading), and the other 34 types with this reading have far lower token frequencies. The high frequency of marriage is thus clearly an outlier. Five formations in this group are hapax legomena, which equals a share of 15%. Two of these, victimage and interlinkage, are 20 th century coinages, another two, parage and fosterage, are earlier formations, and unmarriage is not attested in the OED. The token frequency of this reading is very low, especially compared to other clearly productive interpretations like ACTION or AMOUNT . This is due to many types that only give rise to a very small number of tokens with CONDITION readings. A good example for this is cleavage, which is attested 349 times in the BNC, but only refers to 'the condition of being cleft' in 6% of all instances. The high amount of polysemous types should be kept in mind as well. Over two thirds of the formations with this reading are polysemous, and most of these have overlaps with ACTION . This close connection is illustrated on the semantic maps in figures 22 and 23. In many of these cases, the CON- DITION reading can be interpreted as a sense extension of ACTION . Consider, for example, shrinkage, which can refer to 'the action of shrinking', or 'the condition of being shrunk'. 90% of shrinkage tokens in the BNC refer to the ACTION , and 2% to the CONDITION . This example is typical for the relationship between these two readings, as ACTION is usually the more common reading of polysemous derivatives. CONDITION is probably not a productive reading, but a possible sense extension of -age formations that already have an ACTION reading. POSITION Only three types, peerage, vantage, and vicarage, give rise to the 154 tokens in the POSITION group. Both token and type frequencies of this reading are extremely low, so that this reading is clearly unproductive. 153 6.2.5 Semantic Maps of -age Derivatives Two semantic maps of the -age derivatives found in the corpus can be found in figures 22 and 23 below. The first map represents the token frequency of the individual readings as well as the number of polysemous derivatives, the second the type frequency and the number of polysemous derivatives. Box sizes represent a reading's share of the total token or type count: the most frequent reading ACTION , for example, gives rise to 16,611 tokens, which equals a share of slightly under a quarter of all tokens. Box sizes start at 0.5 cm x 0.25 cm for shares of 0.1 - 4.9% of all tokens or types, then increase by 0.5 cm in length in steps of 5% (refer to section 4.2 for more details on the construction of semantic maps). The thickness of the connecting lines represents the number of derivatives with a particular polysemy. A dashed line means there are one or two derivatives that have both readings, a continuous line represents three to five polysemous words, a continuous line of 0.1 cm thickness is drawn if six to ten words exhibit a particular polysemy, and so on. The token frequency map clearly shows that ACTION and AMOUNT are the most frequent readings, closely followed by GENERAL ABSTRACT and LOCA- TION . COLLECTIVE , OBJECT , and CONDITION have much lower, but similar, frequencies, and the other reading groups contain only a small number of tokens. Especially interesting about this map is the high amount of overlaps between some of the readings, in particular those between ACTION and OB- JECT , ACTION and COLLECTIVE , ACTION and CONDITION , and ACTION and AMOUNT . The prevalence of polysemous formations with an ACTION reading shows that this reading is central for -age derivatives. Although all readings apart from TENURE are connected to a number of other interpretations through polysemous formations, the regularity of these four overlaps is striking. As was already pointed out in the discussion above, ACTION is probably the origin of a sense extension in these cases. The generally high amount of connections emphasises the impression of a single polysemous category, which confirms the result of the dictionary-based investigation. 154 Figure 22: Semantic map of all readings of -age derivatives (BNC) based on token frequency 155 The semantic map based on the type frequency of the various readings differs from the map based on token frequency in certain regards. The connecting lines are, of course, identical, as they represent absolute numbers, but the share of type frequency greatly differs from the relative token frequency for some readings. The most obvious difference is the size of the reading AC- TION . In figure 22, ACTION has the same size as AMOUNT , and is only slightly larger than GENERAL ABSTRACT or LOCATION . This shows that the token frequency of the ACTION reading is high, but not significantly higher than that of other readings. In figure 23, however, ACTION is by far the largest reading, as it is expressed by the highest number of types. OBJECT and COLLECTIVE are also larger now, as their type frequencies are relatively high as well. With regard to ACTION and COLLECTIVE this should be expected, because productive processes are characterised by a high number of types that give rise to a relatively small number of tokens. Productive readings should therefore be relatively small on the token frequency based map and large on the type frequency based map. Less productive readings like GENERAL ABSTRACT or CONDITION , on the other hand, show the reverse relation: They give rise to a large number of tokens, and are therefore large on the map in figure 22, but are only expressed by a small number of types on the map in figure 23. The comparison of the two semantic maps can therefore help to reveal the productivity of the various readings. 156 Figure 23: Semantic map of all readings of -age derivatives (BNC) based on type frequency 157 6.2.6 Comparison to Dictionary-based Analysis A comparison between the semantic maps drawn on the basis of the different datasets reveals some interesting similarities and differences. For convenience, the maps of ME and PDE neologisms recorded in the OED are reprinted as figures 24 and 25 below. Only the map of PDE neologisms and that based on type frequency in the corpus can be compared directly to each other, as they both represent the type frequency of -age formations in the 20 th century. The dictionary-based analysis has shown that ACTION is the most productive reading of PDE neologisms, and this result is supported by the corpus analysis. ACTION is already a productive reading in ME, and it is not surprising that a reading that is frequently attested in neologisms in both ME and PDE is also found to be productive in the BNC. The other core readings from the dictionary analysis of PDE neologisms, AMOUNT and COLLECTIVE , are also productive in the corpus. Especially COLLECTIVE gives rise to a large number of types with low token frequencies, and is thus very productive in the BNC. Among the dictionary neologisms, this reading is not as frequent as AMOUNT , however. AMOUNT is only found with a high frequency among the PDE, but not the ME neologisms in the OED, which suggests that it has only become productive recently. Corpora that contain more recent data than the BNC might therefore suggest a higher productivity for the reading AMOUNT . In spite of this, the differences between the two data sources dictionary and corpus are only small, and the results obtained from them generally point into the same direction. The corpus contains some readings, like TENURE or RIGHT , that are not recorded in PDE neologisms in the OED, but are attested in ME neologisms. These two readings are mostly found in texts with historical focus, as is the reading CHARGE , and are hardly found in current usage. Their share of types and tokens in the corpus is thus extremely low and they are unproductive readings. All of this corresponds to the dictionary-based results, as hardly any neologisms with this reading are found there in PDE. The patterns that emerge from the analysis of dictionary and corpus data are very similar. The readings that have the largest shares of PDE neologisms as attested in the OED are the most productive in the BNC, and those types that are rare or unattested in the dictionary neologisms are also found to be unproductive and rare in the corpus. The corpus-based investigation thus largely confirms the results of the dictionary study. The second feature of the semantic map, the connecting lines, offers some interesting insights into the structure of this morphological category. The dictionary-based map of PDE neologisms suggests a much clearer, less interconnected, category than the ME map. Recent neologisms clearly have fewer, but also more regular polysemies than ME neologisms. The corpus-based maps show a similar amount of connecting lines as the ME map, however. The -age derivatives used in the corpus are very often polysemous, which 158 results in a closely linked semantic map. This is the result of the high amount of older, established derivatives found in the corpus, as such formations often acquire additional interpretations in the course of time. But the corpus-based maps are not only closely linked, they also contain many polysemies that are very common. The thick connecting lines linked to AC- TION in particular show that some additional readings are regular occurrences. This does not contradict the dictionary-based impression of a more clearly structured network of neologisms in PDE, but it shows that the polysemy of this suffixational category is not lost over time. While recent neologisms concentrate on certain productive readings, established formations often have a number of different interpretations, many of which are still used. But the corpus and dictionary-based maps show a significant difference regarding these connections as well: The corpus-based maps contain some extremely thick lines, which represent a high number of derivatives with a particular semantic overlap. Especially the lines linked to ACTION are thicker than others, which means that polysemous derivatives with ACTION as one of their readings are very numerous. OBJECT , COLLECTIVE , AMOUNT and CONDI- TION could be established as sense extensions of ACTION . A similar tendency can also be found in the dictionary data, but this close relationship is far more pronounced in the corpus. This, and the fact that ACTION is the most frequent and a highly productive reading, shows that ACTION is at the heart of this morphological category. 159 Figure 24: Semantic map of all readings of ME -age neologisms (OED) 160 Figure 25: Semantic map of all readings of PDE -age neologisms (OED) 161 Both corpus and dictionary data point towards ACTION as the most productive reading of -age derivatives in current usage, so this reading is clearly a core sense. Unsurprisingly for a morphological category that gives rise to action nouns, some typical sense extensions are attested here. The OBJECT group contains many words that denote the concrete results of actions, e.g. spillage. CONDITION is another typical result of an action, consider, for example, marriage, which can refer to the action of getting married, and also to the state of being married, or cleavage, which refers to an action in “A thorough knowledge of the cleavage rate and expected fluctuations in the mouse strain being used provides a useful guide to the timing of exposure to the mitotic arrestant” (EV6 1465), and to a condition in “It was this seemingly irreconcilable cleavage that effectively meant that the Council of Europe was in no position to advance by and in itself the concept of European union to any great length” (CLR 173). Such sense extensions are to be expected of any word formation pattern giving rise to action nouns. Similar readings can, for example, be found in nouns derived by -ing: cooking can refer to an action, but also to the result of that action, i.e. a meal, and anchoring is paraphrased in the OED as 'the action or condition of lying at anchor'. The other prominent readings of -age derivatives are somewhat less common in other word formation processes. Although COLLECTIVE and AMOUNT readings are found as possible interpretations of other suffixes as well, these readings constitute a significant share of -age formations. The productivity of these two very similar readings makes them characteristic readings for this particular morphological category. It is thus not only the most productive reading ACTION that defines this word formation pattern. Due to their transparency and productivity, AMOUNT and COLLECTIVE should also be considered core senses of -age. To summarise, suffixation with -age predominantly gives rise to action nouns, and some typical sense extensions of such nominals such as OBJECT and CONDITION , but the high frequency and transparency of other readings, in particular AMOUNT and COLLECTIVE , are important characteristics of this word formation pattern as well. This process is therefore similar to other action noun forming patterns, but its particular semantic structure distinguishes it from other action noun forming processes like, for example, -ing suffixation. 6.3 The Suffix -ery in the BNC Altogether 366 words in the BNC are analysed as -ery derivatives. These follow a similar distribution to the -age derivatives in that many types have a low token frequency - a quarter of them are hapax legomena - and few types have high token frequencies. Such a distribution is predicted by Zipf (1949/ 1972). The most frequent derivative is ministry (4,813 tokens). Alt- 162 hough there are substantially more -ery than -age derivatives in the corpus, the highest frequency -ery formation is much less frequent than the highest frequency -age formation (village: 10,291 tokens). As with -age, a number of -ery formations are not attested in the OED. The amount of these presumably new coinages is slightly higher than the 10% found for -age: 48 -ery derivatives, which equals 13% of all derivatives, are either not attested in the OED at all, or not recorded with the reading they have in the corpus. Unsurprisingly, most of these are low-frequency words, e.g. the hapax legomena mobbery ( ACTION ) in “[...] an area of west Berlin with a reputation for left-wing street mobbery” (BMB 2121) or boozery ( LOCATION ) in “In 'Downtown Beirut', the boozery next door to the infamous 'Village Idiot' on 10th St. [...]” (ED7 3341). The attested -ery derivatives span across all four ontological categories, as is shown in table 22 below. The different readings in all of these categories will be discussed in turn in the following section. ontological category types tokens examples CONCRETE 201 26,923 citizenry, pastry, saddlery ABSTRACT 52 8,922 jewry, pleasantry, poetry EVENT 211 13,334 forgery, harlotry, revelry STATE 62 3,659 canonry, deanery, slavery Table 22: Ontological categories of -ery derivatives (BNC) 6.3.1 CONCRETE Derivatives All the CONCRETE readings that are found in the dictionary are attested in the corpus as well. Table 23 below shows the type and token frequencies of each reading in this ontological category and example derivatives. reading types tokens examples OBJECT 23 2,189 confectionery, forgery, pastry PERSON 6 327 adversary, fairy, justiciary LOCATION 100 7,413 noshery, nursery, old folkery COLLECTIVE 132 16,994 chivalry, jewellery, pottery Table 23: CONCRETE readings of -ery derivatives (BNC) The most frequent reading in this category is COLLECTIVE with 132 derivatives, which equals a third of all -ery derivatives found in the corpus. These 163 derivatives also account for a large amount of all -ery tokens. LOCATION is also quite common, as more than a quarter of all derivatives have this reading. OBJECT and PERSON , on the other hand, are both rare. Each of these readings will be discussed individually below. OBJECT The 23 formations with an OBJECT reading account for 2,189 tokens, e.g. pastry (483 tokens). All of these are attested in the OED, including the hapaxes. At 13%, the share of hapax legomena in this group is much lower than the average of all -ery derivatives, where hapaxes account for a quarter of all formations. Only a third of the words in this group are monosemous, but half of them have overlaps with other CONCRETE readings, and a third may also refer to an ACTION . Other, but less frequent, overlaps are attested with GENERAL ABSTRACT and AMOUNT . The high share of frequent formations and the fact that there are no new-to-the-OED words in this group suggests that these words are established formations and that the productivity of this reading is quite low. This supports the result of the dictionary-based investigation, where OBJECT was also found to be an unproductive interpretation. PERSON Only six -ery derivatives refer to a PERSON , and these give rise to 327 tokens. The most frequent of these is fairy (133 tokens). None of the six words in this group is a hapax, and all of them can be found in the OED. They are also all polysemous and usually denote a person only in the minority of instances. Fairy, for example, denotes a collectivity of fae creatures in 76% of all attested cases, and only 17% of all fairy tokens denote a single fae. The predominance of high-frequency formations with a PERSON reading, their polysemy and the fact that they are all found in the OED suggest that such words are established formations. This reading is clearly unproductive. LOCATION LOCATION is the second largest reading group in the CONCRETE category with 100 types that give rise to 7,413 tokens. Only 14% of these words are hapax legomena, but more than 40% have token frequencies of less than ten, so that the share of low-frequency words here is comparable to that of -ery derivatives in general. As a result, the type-token ratio of this reading at 0.013 is quite high, in spite of the high-frequency formations that can be found in this group, among them nursery (1,651 tokens), brewery (655 tokens), and monastery (614 tokens). Most of the high-frequency words have additional readings, but a number of lower frequency derivatives are monosemous and only denote a location. Examples are beanery 'a cheap restaurant', commandery 'the place where a commander works', or old folkery 'an old people's home'. 164 Almost half of all LOCATION derivatives also have a COLLECTIVE interpretation. Such words usually denote both the location in which certain objects are made or sold, and the collectivity of these objects. Good examples for this overlap are haberdashery 'the goods and wares sold by a haberdasher' and 'the shop or establishment of a haberdasher', greengrocery 'fruit and vegetables sold by a greengrocer' and 'the business or shop of a greengrocer', and chandlery 'a place where candles are kept' and 'candles and other lighting materials'. The slight majority of the derivatives with both LOCATION and COLLECTIVE readings denote a location more often than a collectivity, especially when the COLLECTIVE reading refers to a collectivity of persons or an institution. A good example for this is tannery, which sometimes denotes an institution, e.g. in the following sentence: “Certain city officials with vested interests in the tannery operation had given the tannery written permission to send their chemical waste to the city sewage treatment plant” (ALB 135), but usually denotes a location, e.g. in “[a]nd that was next door to the tannery” (FY5 210). On the other hand, a COLLECTIVE interpretation is more common than a LOCATION reading when a derivative denotes a collectivity of objects and the place in which these are made or sold, as is the case with haberdashery, corsetry, or saddlery. A quarter of the derivatives, most of them high-frequency formations, also have an ACTION reading. Half of these have a locative reading more often than an ACTION reading, and another 15% show both readings with the same token frequency, so that there is no cause to claim a sense extension ACTION → LOCATION or vice versa. Overlaps with other readings exist as well, but these are very infrequent. In general, the pattern formed by the polysemous derivatives in this group is very similar to that found in the ME dictionary neologisms. The ME coinages show a strong overlap with COLLECTIVE and a less significant overlap with ACTION , which is exactly what we find in the corpus data. In spite of the frequency of this overlap both in the dictionary and in the corpus, a clear sense extension from COLLECTIVE → LOCATION or LOCATION → COLLECTIVE cannot be established. In the dictionary, both readings are usually attested at the same time, and it is not clear which is the earlier interpretation. Although COLLECTIVE has a higher type and token frequency than LOCATION in the corpus, the polysemous words with both readings have slightly higher token frequencies in the LOCATION group, so no clear tendency regarding the direction of a possible sense extension can be discerned here. If the COLLEC- TIVE reading is split up further though, a tendency emerges in the corpus data: LOCATION readings are more common for derivatives that also denote a collectivity of people or an institution. For such words, a sense extension LOCATION → COLLECTIVE can be assumed. Exactly this extension has often been described before, e.g. by Lieber (2004: 150), who gives the example “Seattle voted Democratic” (italics in original, MS), where the place name Seattle is used to refer to the collectivity of people who live there. But if a deriva- 165 tive denotes a collectivity of objects, e.g. grocery, the COLLECTIVE reading is more common than the LOCATION reading, and a sense extension COLLECTIVE → LOCATION may be assumed. LOCATION is clearly an important reading for -ery derivatives, as this interpretation has a high type frequency. The slightly below-average share of hapaxes limits the productivity of this reading somewhat, but LOCATION contains many low-frequency formations, which is shown by its high typetoken ratio. Most of the high-frequency types are not transparent, however. Although pottery, nursery, and bakery are less frequent than their bases and can thus easily be analysed as derivatives of -ery, the coinages surgery, poultry, fairy, and laundry, among others, are substantially more frequent than their bases. These words are less likely to be understood as derivatives of -ery, and thus limit the productivity of this reading. In spite of this opacity, LOCATION is not an unproductive reading, an assessment that is also supported by the fact that some of the words in this group, e.g. fish-and-chippery or boozery, are not attested in the OED, which suggests that they are new coinages. But this reading is also defined by a number of established and opaque derivatives with high token frequencies, which indicates a limited productivity of this pattern. COLLECTIVE The largest reading group in the CONCRETE category is COLLECTIVE , which contains 132 derivatives that give rise to 16,994 tokens. The type-token ratio of this reading is quite low at 0.0077, which is due to a high number of very frequent types in this group. The most frequent derivative ministry (4,106 tokens with a COLLECTIVE reading), but also others like machinery (1,890 tokens) and jewellery (1,220 tokens) can be found here. Only 13% of the COL- LECTIVE words are hapaxes, and one third of the formations have a token frequency of less than ten. The majority of the derivatives with this reading therefore seem to be of high or mid-range frequency. Most of the words with this reading are polysemous, and the monosemous formations are exclusively low-frequency words. Almost half the words in this group have an additional ACTION reading. Such words refer to an action or practice and to a collectivity of objects or people, e.g. archery 'the practice or art of shooting with bow and arrow' and 'an archer's weapons', or saddlery 'the occupation or craft of a saddler' and 'articles made or sold by a saddler considered collectively'. About half of the words that show this overlap have an ACTION reading more often than a COLLECTIVE reading, and almost exactly the same number of words have a COLLECTIVE reading more often than an ACTION reading. Because of this balance, the directionality of a putative sense extension cannot be established on the basis of this data. A strong link between COLLECTIVE and ACTION was already found in the dictionary investigation, and the dictionary data supports a sense extension ACTION → COLLECTIVE at least for ME. 166 Another significant overlap exists between COLLECTIVE and LOCATION . These words have already been discussed above, and the analysis will not be repeated here. COLLECTIVE also has links with other readings, e.g. OBJECT , GENERAL ABSTRACT , or CONDITION , but these are much weaker than the connections between COLLECTIVE and ACTION , and COLLECTIVE and LOCATION . The large number of high and mid-frequency words and the small amount of hapaxes and other low-frequency items in this group suggests an unproductive reading. However, a number of words with a COLLECTIVE interpretation cannot be found in the OED, among them socketry 'sockets collectively' or widgetry 'widgets collectively'. All of these words are of low frequency, and are likely new coinages. Another factor that should be considered is that many of the more frequent types in this group are transparent formations. Ministry, machinery, forestry, and pottery, among others, have lower frequencies than their bases, and can therefore easily be analysed by speakers. The existence of new coinages shows that this reading is somewhat productive. As many of the high-frequency formations are transparent as well, the high token frequency of COLLECTIVE does not serve as an argument against the productivity of this reading. 6.3.2 ABSTRACT Derivatives Three different readings are found in this ontological category: GENERAL ABSTRACT , AMOUNT , and CHARGE . TENURE , which can be found in OED neologisms, is not attested in the BNC. Table 24 shows the different readings with their type and token frequencies and examples. reading types tokens examples GENERAL ABSTRACT 39 7,158 chemistry, mystery, poetry AMOUNT 15 1,762 anecdotery, imagery, syllabary CHARGE 1 2 chapelry Table 24: ABSTRACT readings of -ery derivatives (BNC) A CHARGE reading is only attested in a single derivative with two tokens, so this reading is clearly a marginal occurrence. It can be considered unproductive, because there is only one derivative with this reading and this formations is an old, established coinage, as can be seen by its attestation date in the OED (1870). Due to the lack of forms and its obvious unproductiveness, this reading will not be discussed in the following. AMOUNT is more frequent, but GENERAL ABSTRACT is the most common ABSTRACT reading, regarding both token and type frequency. These two readings are going to be analysed individually below. 167 GENERAL ABSTRACT The largest reading group in this ontological category, GENERAL ABSTRACT , accounts for 10% of all -ery derivatives found in the corpus. The token frequency of these few types is, however, comparatively high, which is due to the fact that the derivatives often account for a large percentage of the highest frequency -ery derivatives. For example, 71% of all tokens of poetry, which add up to 1,896 tokens, and 99% of all chemistry tokens (1,999 tokens) have a GENERAL ABSTRACT interpretation. Low-frequency formations are indeed rather rare in this group: Less than 10% of the words in this group are hapaxes. This leads to a low type-token ratio (0.0054), and suggests an unproductive reading. Another argument against the productivity of this reading is the opacity of most of the high-frequency formations like chemistry or mystery. These are more frequent than their bases, which inhibits their decomposability. Additionally, potentially transparent formations like forgery or weaponry primarily have other readings than GENERAL ABSTRACT , namely ACTION and COLLECTIVE in these two cases. But new coinages with a GENERAL ABSTRACT reading are possible, if infrequent, as the new-to-the-OED hapax selfish genery shows. This reading can therefore not be completely unproductive. A quarter of the derivatives are monosemous, most of these are lowfrequency formations like biogeochemistry (one token). The most significant semantic overlap exists with ACTION , as more than half of all items in this group have an additional ACTION reading. Out of the 21 derivatives with this polysemy, nine have substantially more tokens in the ACTION group, and nine have a GENERAL ABSTRACT reading more often. The remaining three are of comparable frequency in both groups. Other semantic overlaps exist with two CONCRETE readings: COLLECTIVE and OBJECT . The six derivatives with additional OBJECT readings refer to an object more often than to an abstract entity. 75% of all tapestry tokens, for example, refer to a physical object, but only 18% refer to an abstract concept. Most of these occur in the formulation “tapestry of life” in the BNC. The words that show this polysemy have many additional readings, however, and a directional link between OBJECT and GENERAL ABSTRACT can therefore not be established. The relation between COLLECTIVE and GENERAL ABSTRACT is similar. Most of the eight derivatives with this polysemy denote a collectivity more often than an abstract concept, but these words have too many other readings to establish a sense extension between COLLECTIVE and GENERAL ABSTRACT . GENERAL ABSTRACT is attested in 10% of -ery derivatives, and these give rise to a relatively large number of tokens. This reading is thus clearly important for the semantic structure of this morphological category, but it is not particularly productive. 168 AMOUNT The second most frequent reading in this ontological category is AMOUNT . 15 derivatives refer to a collectivity of abstract concepts and give rise to 1,762 tokens. The most frequent of these formations are poetry (721 tokens), which refers to poetic works collectively, and machinery (455 tokens), which refers to mechanisms of administration or government collectively. This group also contains three hapax legomena, two of which are not attested in the OED. These are business stuffery 'business stuff collectively' and anecdotery 'anecdotes collectively'. Only three of the 15 derivatives in this group are monosemous, and these are either hapax legomena or dislegomena. Almost half the words in this group have additional COLLECTIVE or ACTION readings. In the case of the AMOUNT - COLLECTIVE polysemy, the COLLECTIVE interpretation is more common than the AMOUNT reading. A good example is machinery, which denotes a collectivity of machines in 79% of all cases, and mechanisms of administration in 19% of all cases. Most of these formations have so many other interpretations that a direct link, and thus a sense extension, between COLLECTIVE and AMOUNT cannot be established. The derivatives with an AMOUNT - ACTION polysemy are mostly of similar frequency. Only popery refers to an ACTION significantly more often than to an AMOUNT , while imagery more commonly refers to an AMOUNT than an ACTION . Most of the formations with this polysemy only represent very few tokens of each type: for example, only 3% of all microcircuitry tokens refer to an AMOUNT , and 2% refer to an ACTION . Derivatives like this predominantly have other readings - microcircuitry, for example, mostly has a COLLECTIVE interpretation. The overlap between ACTION and AMOUNT thus seems to be an epiphenomenon that arises due to the predominant readings of derivatives and their additional senses. AMOUNT is one of the least common readings in the corpus, so it cannot be considered overly productive, but the high share of low-frequency formations in this group and the unattested formations show that this reading is not completely unproductive. 6.3.3 EVENT Derivatives More than half of all -ery types found in the corpus have an ACTION reading, and are thus part of this ontological category. This reading is the only EVENT reading, and table 25 shows the type and token frequency of this interpretation. reading types tokens examples ACTION 211 13,334 flattery, jackassery, robbery Table 25: EVENT readings of -ery derivatives (BNC) 169 ACTION has an extremely high type frequency, but its token frequency is not the highest of all groups. This leads to a very high type-token ration, which indicates a productive reading. A quarter of the derivatives in this group are hapax legomena, and many of them are not attested in the OED, e.g. debunkery in “the recent BAFTA award for best 1991 arts programme went to Channel 4's Without Walls for Robert McKee's constructive debunkery in J'Accuse -- Citizen Kane” (AHK 1896), or siegery in “They could not storm the castle, and there was no time for siegery” (CD8 837). These presumably new coinages suggest a productive reading. But this group also contains some high-frequency words like surgery (2,123 tokens), forestry (905 tokens), or robbery (753 tokens), which are established and frequently used words. A large part of the derivatives in this group, namely 45%, are monosemous. This includes not only many hapax and dislegomena, but also a number of high-frequency formations like robbery (753 tokens), cookery (473 tokens), or adultery (274 tokens). The extremely high number of monosemous derivatives including both low and high-frequency formations shows that this reading is used in both newly coined -ery derivatives and in established formations. ACTION also exhibits a complex network of polysemy relations with most other reading groups. The most frequent overlap exists with COLLECTIVE - 60 derivatives have both readings. These words denote an action or practice, e.g. laundry in “[a]ll Conservatives have done is put out services to tender such as cleaning, laundry and portering” (JJD 315), or a collectivity of objects or persons that are connected to this action, e.g. laundry in “unless you have a separate utility room, that does not leave much space nearby for [...] setting down the just cleaned laundry” (HGW 1420). A similarly strong link between the readings COLLECTIVE and ACTION was already found in the dictionary investigation, where a sense extension ACTION → COLLECTIVE could be established for ME neologisms. A slightly less common overlap exists between ACTION and CONDITION , as 44 types express both readings. It is often impossible to decide whether the CONDITION or the ACTION reading is referred to in a particular expression, so many of the tokens were added to both readings. A good example is handcuffery: “I spy some chafing upon your wrists, suggestive of handcuffery” (HTU 333). Here, handcuffery may either refer to 'the action of handcuffing' or to 'the state of being handcuffed'. This vagueness is a major reason for the pervasiveness of this particular polysemy. In the higher frequency words most of the tokens refer to one of these two readings, which means that most types concentrate on one of these two readings, but a few instances can also be understood as the other interpretation. Idolatry, for example, can be understood as 'the action of worshipping idols' in all instances, but may also be read as 'immoderate attachment to a person or thing', and thus a CONDITION , in the following example: “Where the word of God was adversary and 170 against his authority, pomp, covetousness, idolatry and superstitious doctrine [...]” (CLM 619). Due to this variation it is not possible to establish a sense extension from one reading to the other. About 10% of all ACTION derivatives also denote a location. These derivatives sometimes refer to an action and to the place at which this action is carried out, e.g. laundry in “When I went first went to work at a laundry” (FYD 162) and in the previously given example “All Conservatives have done is put out services to tender such as cleaning, laundry and portering” (JJD 315). This is, however, not the case for all derivatives that show a LOCA- TION - ACTION polysemy. Many formations also denote a location at which objects are kept, e.g. pottery, and only have an ACTION reading in a small number of tokens. Derivatives with a LOCATION reading should thus not be considered a sense extension of ACTION . Apart from these two polysemies, ACTION is connected to all reading groups except for PERSON and CHARGE . But as these are only rarely encountered interpretations of -ery derivatives, the lack of polysemous formations may well be due to a lack of data. Although there are no sense extensions from ACTION to other readings that can be established on the basis of the corpus data, it is clear that ACTION is a reading that is at the very heart of this morphological category. It is extremely frequent and very productive, and it is also connected to almost all other readings. 6.3.4 STATE Derivatives Two different STATE readings are attested in the corpus: CONDITION and POSI- TION . Their type and token frequencies and examples for each reading are shown in table 26. reading types tokens examples CONDITION 54 3,457 cuckoldry, debauchery, harlotry POSITION 10 202 deanery, gentry, sergeantry Table 26: STATE readings of -ery derivatives (BNC) CONDITION CONDITION is the larger group of the two, its 54 types account for 14% of all -ery derivatives found in the BNC. These types account for the comparatively low number of 3,457 tokens, which leads to quite a high type-token ratio. The most frequent derivative with a CONDITION reading is misery 'a condition of external unhappiness' (1,201 tokens). But apart from such highfrequency words, this group contains a high number of low-frequency coinages. More than a quarter of the derivatives with a CONDITION reading are hapaxes, and seven of these are not recorded in the OED. Examples of these 171 presumably new coinages are handcuffery in “I spy some chafing upon your wrists, suggestive of handcuffery” (HTU 333), which may also refer to an action, or songsmithery in “the band already have a healthy roster of tunes which walk the fine line between Sly Stone funk and 'Abbey Road' songsmithery” (CAD 544). Such a high proportion of low-frequency derivatives and the existence of a number of newly coined formations suggests a productive interpretation. There are, however, some factors that limit the productivity of this reading. There are hardly any monosemous derivatives, which makes it doubtful whether these derivatives were coined to express a CONDITION reading or whether this interpretation is merely a sense extension or an additional sense of derivatives which primarily refer to something else. Especially AC- TION commonly co-occurs with CONDITION , but as both readings are similarly frequent in these polysemous formations, a sense extension cannot be established. Another polysemy exists with COLLECTIVE . This overlap affects nearly one fifth of all words with a CONDITION reading. This is, of course, much less substantial than the CONDITION - ACTION polysemy, but it is still noteworthy. The relatively small number and unsystematic nature of the words with this polysemy does not suggest a sense extension, however. Almost all of these words show a COLLECTIVE reading only in a small number of instances, and bar one, they all refer to an ACTION as well. This overlap between CONDITION and COLLECTIVE thus seems to be the result of a more systematic and frequent overlap between ACTION and COLLECTIVE on the one hand and ACTION and CONDITION on the other. In spite of the relatively high token frequency and the high share of hapaxes and other low-frequency formations in this group, CONDITION is probably not an overly productive reading. Most of the types, including the hapax legomena, also have an ACTION reading. This reading is very productive, and the CONDITION reading of many words may be due to the inherent vagueness of many derivatives as to what kind of SITUATION they describe - an eventive or a stative one. POSITION A mere ten words refer to a position or office, which makes the POSITION reading one of the rarest ones encountered in -ery derivatives in the BNC. The low token frequency of this reading is not due to a large number of lowfrequency formations, but rather due to the fact that most derivatives only refer to a position or office in one or two cases. Only 1% of all tokens of midwifery, for example, refer to the position of a midwife. Also, only the two hapaxes, ealdormanry and sergeantry, are monosemous, but both of these are old coinages that can be found in the OED. POSITION can therefore be understood as a rare additional reading of derivatives that primarily refer to something else, such as a collectivity of people, e.g. deanery, or a location, e.g. archdeaconry. This reading is not productive. 172 6.3.5 Semantic Maps of -ery Derivatives The semantic maps below represent the token frequencies (figure 26) and type frequencies (figure 27) of the individual readings relative to the complete number of tokens, and the number of polysemous derivatives. The relative type or token frequency of a reading is reflected by its box size: COL- LECTIVE , for example, gives rise to 16,994 tokens, which equals a share of 34% of all -ery tokens. This is the largest reading with regard to token frequency, so it is the largest box on the map as well. The polysemous derivatives are represented by the connecting lines between readings. For example, 60 derivatives have both an ACTION and a COLLECTIVE reading, making this the most common polysemy in this morphological category. The connecting line between these two boxes is therefore the thickest at 1.1 cm. For more details on how the semantic maps are created please refer to chapter 4. The semantic map based on token frequency shows the two largest boxes for the two most frequent readings, COLLECTIVE and ACTION . All other readings are of a much lower frequency, which can be seen by the much smaller box sizes of LOCATION , GENERAL ABSTRACT , and CONDITION . The remaining readings PERSON , OBJECT , AMOUNT , CHARGE , and POSITION are very small on the map, because some of them are extremely rare and only give rise to a handful of tokens. In contrast to the dictionary investigation of 20 th century neologisms, the box sizes on the map in figure 26 are not necessarily indicative of the productivity of a reading. ACTION , for example, can be considered the most productive reading of -ery, but this reading is smaller than COLLEC- TIVE on the semantic map. This is why a semantic map based on the type frequency of the BNC readings was also created. The comparison of the two corpus-based maps can reveal which readings are particularly productive and which are less productive. Very productive readings give rise to a large number of types and a relatively low number of tokens, while less productive readings are expressed by a small number of types with relatively high token frequencies. A comparison of both maps indeed reveals that ACTION gives rise to the largest number of types - almost two thirds of all -ery derivatives in the corpus have this reading - while the token frequency of this reading is comparatively low. ACTION is thus the most productive reading in the BNC. COL- LECTIVE is less productive, as it gives rise to a high number of types, but also a very high number of tokens. The types with a COLLECTIVE reading therefore have higher token frequencies than those with an ACTION reading. A third reading, LOCATION , is also revealed to be a productive interpretation for the same reason. The analysis of the derivatives with a locative readings has, however, shown that this interpretation is characterised by a number of opaque high-frequency formations, which somewhat limits its productivity. The semantic maps are thus a useful indicator for the productivity of different readings, but they should not be used without an analysis of the derivatives. 173 A striking feature of both maps is the extreme thickness of its connecting lines. ACTION and COLLECTIVE are connected by 60 polysemous types, and ACTION and CONDITION and COLLECTIVE and LOCATION are each connected by more than 40 polysemous derivatives. All three of these overlaps affect a significant share of the types in each reading group. The connecting lines between the smaller readings are obviously thinner, but these can also be highly regular. More than a third of all object-denoting types, for example, may also have a COLLECTIVE reading. The general impression of a highlyinterconnected network arises here as well as in the previously introduced semantic maps. All readings are connected to other interpretations usually by a number of polysemous derivatives. 174 Figure 26: Semantic map of all readings of -ery derivatives (BNC) based on token frequency 175 Figure 27: Semantic map of all readings of -ery derivatives (BNC) based on type frequency 176 6.3.6 Comparison to Dictionary-based Analysis A comparison of the semantic maps based on the BNC (figures 26 and 27) to the semantic maps based on the dictionary data (reprinted as figures 28 and 29 below) reveals some differences, but also many similarities. A direct comparison is only possible between the map of PDE neologisms (figure 29) and the corpus map based on type frequency (figure 27), as both of these refer to the type frequencies of readings attested in roughly the same time period. In both of these maps ACTION is by far the largest reading, and both the dictionary and the corpus analysis have found this reading to be highly productive. ACTION should therefore certainly be considered a core sense of -ery derivatives. With regards to the other groups there are some differences, however. LOCATION is found to be a productive reading of -ery derivatives in the corpus, which substantiates the results from the dictionary analysis, as LOCATION is also productive in PDE neologisms. But COLLECTIVE , which is very productive in the BNC, seems less productive in the 20 th century dictionary neologisms. According to the dictionary analysis, LOCATION is more productive than COLLECTIVE , but based on the corpus data, this relationship is reversed. This is only a minor difference, however, as both readings are productive in both data sources. This suggests that COLLECTIVE and LOCATION should also be considered core readings of -ery derivatives in PDE. It also shows that the analysis of corpus data is an important addition to the dictionary-based discussion, as it adds valuable information to the previously attained results. The connecting lines show more differences between the individual maps. The corpus-based map displays some extremely thick lines that represent very high numbers of polysemous derivatives. None of the dictionarybased maps have such strong connections, but the readings that are most closely connected in the dictionary neologisms are the same ones that show very thick connecting lines in the corpus: In ME, COLLECTIVE and LOCATION , and ACTION and CONDITION have the closest links, and PDE has a close connection of ACTION and CONDITION . These connections are very strong in the corpus as well, but here we also find links between ACTION and COLLECTIVE and ACTION and LOCATION . This shows that certain polysemies are much more regular in the words in current usage than in the dictionary neologisms of both periods. Regarding the thinner connecting lines, which represent a smaller number of polysemous derivatives, especially the Middle English map exhibits a multitude of links between the different readings. These thinner lines create a tightly knit network. The 20 th century neologisms in figure 29 are also connected to each other, but the picture that emerges from this map is that of a more concentrated pattern with fewer readings and fewer semantic overlaps than in ME. The corpus-based map differs from its synchronic dictionary counterpart in certain aspects. It also contains slightly fewer readings than the ME map, but still includes more than the 20 th century dictionary-based 177 map. The connecting lines are also more similar to the ME map, as most readings are connected to multiple others, often by thin lines that only represent one or two polysemous formations. It is not surprising that the corpus-based map is more similar to the ME map in this regard. The 20 th century neologisms are very recent coinages that have not been in use for long, and have had little time to develop regular polysemies. The time period that was investigated for Middle English is much longer than that for Present Day English, which makes additional readings more likely to appear before the end of Middle English. And the corpus contains many words that are old and established coinages, which are expected to be highly polysemous. The same pattern was found for -age derivatives. The corpus-based maps show a high number of polysemous derivatives, and they also show some highly regular semantic overlaps. These regularities are not as obvious in the dictionary-based maps for both suffixes. Although there are some differences between the corpus-based and the dictionary-based maps, the similarities between them are far more pronounced. ACTION is a frequent reading in all of them, and the analysis of the ACTION derivatives has shown that this reading is highly productive in both sources. LOCATION is also found to be productive in both dictionary neologisms and the corpus, but COLLECTIVE is considerably more productive in the corpus than in the dictionary neologisms. All three readings are productive in both data sources, however, and all three should thus be considered core readings of -ery derivatives in PDE. The highly interconnected network that is created by polysemous derivatives found in the BNC is similar to the web of connections on the map of Middle English neologisms. This shows that the derivatives that are in current usage are highly polysemous and create a morphological category with closely linked semantics. The more regular pattern of polysemous derivatives in the corpus shows that the overlaps found to be important in the dictionary data are exploited even more in actual usage. 178 Figure 28: Semantic map of all readings of ME -ery neologisms (OED) 179 Figure 29: Semantic map of all readings of PDE -ery neologisms (OED) 180 6.4 Summary: The Semantic Structure of -age and -ery Derivatives in the BNC The analysis of -age and -ery derivatives in the BNC has shown that there are many similarities between these two morphological categories. This is not too surprising, as previous studies have already stressed the semantic similarity of these two suffixes (e.g. Lieber 2004). The semantic maps reveal that the readings expressed by these formations have an almost identical range, the only exceptions being TENURE and RIGHT , which are shown by -age, but not by -ery derivatives. The individual readings also often show comparable token frequencies relative to the total number of derivatives in each category. ACTION , for example, is very frequent in both -age and -ery derivatives and can also be considered a core reading of both groups of derivatives. Another similarity exists with regard to the amount of connecting lines. Both -age and -ery derivatives are highly connected by an intricate network of polysemous formations. In spite of these similarities, the two word formation processes cannot be considered identical. The semantic maps show that while the range of readings may be very similar, each morphological category concentrates on different readings. Derivatives of -age have ACTION , AMOUNT , and COLLECTIVE as their core interpretations, while -ery derivatives focus on ACTION , COLLEC- TIVE , and LOCATION . The amount and nature of the connecting lines between the various readings also differ substantially between -age and -ery formations. Some of the connections on the semantic maps of -ery (figures 26 and 27) are extremely thick, indicating a large number and high regularity in particular of derivatives with an ACTION - COLLECTIVE , COLLECTIVE - LOCA- TION , ACTION - CONDITION , and ACTION - LOCATION polysemy. The less frequent readings are mostly connected by thinner lines. The semantic maps of -age derivatives (figures 22 and 23) are also highly interconnected, but the links themselves are far thinner, and therefore represent less numerous polysemous derivatives than the ones shown on the map for -ery formations. This may partly be due to the lower absolute number of -age derivatives, but the connecting lines are significantly thinner here, so that we can assume a noticeable difference in the polysemy structure of -age and -ery derivatives. It is, however, not only the overall thickness of the connecting lines that differs between the two corpus-based maps. Derivatives of -ery show particularly strong overlaps between readings that are not as important for -age formations, and vice versa. And there are also structural and semantic differences between the readings themselves. Formations with an ACTION reading illustrate this nicely: -ery derivatives with this interpretation, e.g. milksoppery 'beaviour of a milksop' or foxery 'behaviour of a fox', often refer to a behaviour or the actions that are characteristic of a person or thing. Derivatives of -age that have an ACTION reading usually refer to more generic actions that do not show a behavioural or characteristic element of the person 181 or thing denoted by the base, e.g. in pillage 'the action or an act of plundering' or anchorage 'the action or process of anchoring'. The derivatives of -age and -ery as found in the BNC have many similarities, but the two morphological categories do not show an identical semantic structure. They differ in the range of readings they express, in the exact nature of these readings, their core senses, and in the connections between readings made by polysemous derivatives. The two suffixes can therefore not be seen as semantically identical. 183 7 Conclusion: The Semantic Structure of -age and -ery Derivatives The empirical part of this study investigated the semantic structure of -age and -ery derivatives in two different data sources, the Oxford English Dictionary online and the British National Corpus. A diachronic investigation comparing Middle English to Present Day English neologisms was carried out as part of the dictionary-based account. This conclusion revisits the main findings from both parts of the empirical investigation to describe the semantic structure of -age and -ery derivatives and reviews the differences and similarities between them. It also assesses the usefulness of the adapted semantic map approach developed as part of this work. 7.1 Comparison of -age and -ery Derivatives The combined results of dictionary and corpus investigations suggest that both -age and -ery are predominantly action noun forming processes. Many of the other readings found among the derivatives, e.g. OBJECT or CONDITION , are typical sense extensions of action nouns described elsewhere in the literature (e.g. Bauer et al. 2013, Melloni 2011). But both morphological categories are highly polysemous and ACTION is not their only characteristic reading. Additional core senses of -age derivatives in PDE are AMOUNT and COLLECTIVE . The Middle English pattern differs from this in that AMOUNT and COLLECTIVE do not have the prominence in ME that they acquire later. Instead, CHARGE is an especially productive reading next to ACTION at that point - a fact that illustrates the significant amount of semantic change in this category since ME. Derivatives of -ery show COLLECTIVE and LOCATION as core senses in addition to ACTION in both investigated periods. Both categories have more than one core sense, which shows that a semantic structure other than the often assumed radial structure (e.g. Geeraerts 1997, Tyler & Evans 2001) is possible, and that semantic models should be able to represent this. Even in addition to the shared core reading ACTION , the derivatives of -age and -ery have many similarities. The most obvious similarity revealed by the various semantic maps is the similar range in readings that is shown by both affixes. There are slight differences here between the different time periods and data sets considered, but most of the readings are identical - at least on the surface. Another similarity is the interconnectedness of the semantic maps. Both morphological categories are characterised by a network of polysemous derivatives, which leads to the conclusion that both categories are indeed polysemous rather than homonymous. This polysemy does 184 not develop over time. All Middle English maps already show a high level of interconnectedness of the different readings - a state that is similar to that shown on the corpus-based maps that represent current usage. Both morphological categories are therefore highly polysemous from the earliest time that their derivatives are used in English. As the range of readings shown by Middle English -age derivatives is very similar to that expressed by medieval French -age derivatives (cf. Uth 2011), the level of polysemy may be a function of the borrowing process. Various derivatives of -age and -ery with different readings were borrowed from French into English in ME, resulting in polysemous categories from the start. The development of native affixes may well be different, and a comparison of the results obtained in this study with a similar investigation of native affixes in English would thus be very interesting. The similarity of readings and the high amount of polysemy in both morphological categories makes the maps look very similar, but there are some important differences between the suffixes nevertheless. Apart from the variation in the frequency of some readings, which is connected to the dominance of the core readings on each map, these differences exist with regard to the exact nature of the readings and the polysemous structure of the morphological categories. Both aspects can be illustrated with the LOCA- TION and ACTION readings frequently shown by both groups of derivatives. While -age derivatives mostly refer to places in which actions take place, e.g. parachutage 'a drop of supplies' ( ACTION ) and 'a drop site' ( LOCATION ), -ery derivatives tend to denote locations in which objects are kept, e.g. bootery 'a shop where boots and shoes are sold'. The same -ery derivatives then often refer to a collectivity of the objects kept in a certain place as well, which leads to a very strong semantic overlap between LOCATION and COLLECTIVE readings. The connection between ACTION and LOCATION is, however, far less pronounced for -ery than for -age derivatives, which naturally follows from the kind of location denoted by each set of derivatives. The ACTION readings themselves also differ significantly: -age derivatives mostly refer to rather general actions and processes, e.g. pre-shrinkage 'the process of preshrinking', while -ery derivatives often refer to behaviours and actions that are characteristic for a person or thing, e.g. idiotry 'extremely stupid behaviour'. This shows that although the derivatives have a similar range of readings the readings themselves can be structured in very different ways. The differences between the individual readings and the semantic overlaps can also be seen in the formal make-up of the various derivatives. For example, almost all Middle English -ery derivatives are based on person nouns, while the -age neologisms of the same period frequently take verbal bases as well as person nouns. The fact that -ery coinages often refer to characteristic behaviours or actions is probably connected to the large amount of person noun bases in this group, although it is not clear whether the bases 185 are selected because of the semantic preference of existing derivatives or whether the behaviour reading is a result of the bases. So in spite of the many similarities between the derivatives of -age and -ery, these two suffixes cannot be seen as synonymous. The semantic maps have helped to reveal differences in the semantic structure of each group of formations that make these morphological categories quite distinct from one another. The polysemy patterns are notably different; for this reason it cannot be said that two processes expressing similar readings necessarily also show similar sense extensions. The sense extensions and, more generally, the whole semantic structure seem to be unique to each category and there is no evidence for universal tendencies apart from some wellknown and recurrent sense extensions typical for action nouns, e.g. ACTION → OBJECT . Even these familiar sense extensions are not shown to the same extent by -age and -ery derivatives. While -age derivatives show a sense extension ACTION → LOCATION , for example, -ery derivatives do not exhibit such a strong link between these two readings, which is yet more evidence for localised polysemy patterns. Each of the two morphological categories investigated here is therefore unique in its semantic structure in spite of the many similarities between the derivatives. Another important result of this investigation is the observation of semantic change in the derivatives of these suffixes. The -age formations in particular have changed significantly from ME to PDE. While CHARGE is very frequent among ME neologisms, it is virtually non-existent in PDE neologisms and also hardly occurs in the corpus. CHARGE can be considered a core reading of -age derivatives in ME, and the fact that it is barely attested in PDE at all is striking. At the same time, the AMOUNT reading of -age derivatives has become dramatically more frequent and can be considered a core reading in PDE. If only the neologisms attested in each period are considered, the change in the range of readings is even more obvious. Both -age and -ery derivatives have fewer readings in new PDE words than in ME neologisms, and the readings that are lost are similar. Both groups do not refer to TENURE , PERSON , and RIGHT in PDE, and CHARGE has almost completely disappeared as well. This similarity is probably not accidental. These readings are quite specific, which is somewhat unusual in English derivatives, and with the exception of PERSON , they refer to feudal concepts in ME. With the disappearance of feudal society words that refer to deeply feudal concepts, e.g. taxes on services and actions, have become unnecessary. Taxes and charges have not disappeared, of course, and it would have been possible to use -age derivatives to refer to modern taxes, but this possibility is not exploited. Although the reasons for this cannot be addressed within the scope of the present investigation, it seems likely that the taxation reforms that were part of the Glorious Revolution in the 17 th century have led to a break both with previous methods of taxation and the associated terminolo- 186 gy (see also Hughes 1988 for a discussion of the connection between social and linguistic change). A significant change has also occurred in the range of readings expressed by -ery neologisms: While POSITION is quite frequent in ME, it is not attested in PDE neologisms. This semantic change is probably connected to a change in the formal make-up of derivatives. Middle English -ery derivatives are often based on nouns denoting persons - a situation that leads to many semantic overlaps between the readings COLLECTIVE , LOCATION , and POSITION , as all of these mostly have person nouns as bases. Derivatives with a POSI- TION reading refer to an office or position in society occupied by a person, while especially COLLECTIVE seems to select objects connected with the work or occupation of a person and the derivatives with this reading then denote a collectivity of these objects. The strong connection between COLLECTIVE and LOCATION is due to the fact that LOCATION often denotes the place where a collectivity of objects is kept. Many derivatives in these two groups are thus not directly connected to a person, but rather to objects associated with a person. PDE -ery derivatives are far less often based on person nouns than new ME coinages - object-denoting bases are much more common in PDE. This change may well be a result of the semantic structure of the core reading COLLECTIVE , which denotes collectivities of objects, as the derivation of this pattern may have influenced the semantic structure of subsequent coinages. POSITION , a reading that relies entirely on nouns denoting persons as bases, vanishes in neologisms at the same time as person nouns become less common. Again, the reasons for this change lie outside of the scope of this investigation, but it can be assumed that these simultaneous developments are not accidental. This investigation has shown that the semantic structure of derivatives can change significantly over time. Not only does the range of the readings of -age and -ery derivatives transform quite substantially, the formal makeup of derivatives can also be subject to change. However, this is not necessarily the case. PDE -age neologisms still have the same kinds of base words as their ME counterparts, only the readings they express are concentrated on a subset of those shown by ME neologisms. PDE -ery derivatives, on the other hand, are more likely to be based on nouns denoting objects rather than persons; a situation that is quite different from ME. This structural change happened at the same time as changes in the range of readings expressed by -ery neologisms occurred. It would be interesting to see if the derivatives of other affixes are also subject to change, and whether that change is similar to that observed for -age and -ery coinages. Previous research on a number of derivational suffixes (e.g. Uth 2011 for French -age and -ment) suggests that this degree of development is not the norm. Future research in this area could investigate whether semantic and formal change to the degree observed for -age and -ery 187 is limited to these two suffixes, which seems unlikely, or whether it can be found elsewhere as well. 7.2 Evaluation of the Adapted Semantic Map Model The semantic map method developed as part of this work has proven to be very useful in assessing the semantic structure of -age and -ery derivatives. This method is a bottom-up approach. It is able to incorporate all the readings encountered in the data and can therefore be considered an accurate representation of the readings of -age and -ery derivatives. The readings shown on the semantic maps are abstractions over the semantic paraphrases provided in the OED. The derivatives that are classified as belonging to the same reading group have similar semantics, but they are by no means identical. A more detailed account with more fine-grained and homogeneous readings would also be possible with this highly flexible method. The level of detail shown by the semantic maps can thus be tailored to the questions that drive a particular investigation. The boxes representing each reading are sensitive to frequency: more frequent readings are larger than less frequent ones. Because the box size shows the relative frequency of each reading depending on its share with regard to the overall frequency of the morphological category, the semantic maps of two different categories can be compared directly. This makes differences or similarities between different categories or diachronic changes in the same category immediately visible. The frequency information included on the maps also contributes to a more detailed picture of each morphological category because dominant readings are easily differentiated from readings that occur only rarely and are maybe less central to the semantics of a particular process. The connecting lines representing polysemous derivatives have helped to reveal the semantic structure of each morphological category and have shown that both categories investigated here are polysemous. They have also exposed subtle differences in the structure of -age and -ery derivatives, which distinguish these two categories and prove that they are not synonymous in spite of an extensive overlap regarding the range of readings they express. Perhaps most importantly, this approach makes no a priori assumptions regarding the structure of the readings expressed by derivatives and is able to represent different kinds of structures. This investigation has found that there is no clear single core sense for either of the categories, but that multiple readings are highly productive and strongly connected to other readings. The semantic map method can be used to investigate other morphological categories in the future to find out whether their structure is similar or whether the number of core senses and the relationship of the readings is, 188 like other attributes, unique to each morphological category. The flexibility and openness of this model makes it a particularly suitable tool to investigate the semantic structure of morphological categories. Due to its relatively accurate representation of the range of readings and their connections, this model can be used to provide the basis for theories of semantic change and the development of polysemy in morphologically complex words. In order to generalise and abstract from the semantic structure found for individual morphological categories, more such categories, both in English and other languages, should be investigated. The semantic map model could also show whether affixes of the same origin, e.g. English -ery, French -erie, and German -erei, have a similar range of readings and undergo similar diachronic developments. But this approach also has some limitations. One of them is the fact that readings with the same labels can be generally similar but sometimes differ in specific respects. This can be seen in the differences of the ACTION readings expressed by -age and -ery derivatives: -age often refers to more general actions and -ery to behaviours, but this is not reflected on the semantic maps shown here. This variation could be shown by using more fine-grained readings, however. A more serious problem exists with regard to the frequency information displayed on the maps. The boxes are scaled to represent the share of type or token frequency of each reading with regards to the total type or token frequency of a particular word formation process. As shown in section 4.2, the size of these boxes only increases in steps of 5%. A reading that contains 5.1% of all types would therefore be represented by a box with the same size as that of a reading that contains 9.8% of all types, although the second reading group is almost twice as large as the first. It is, of course, possible to decrease the range of percentages in each stage, but the problem that the box sizes do not exactly mirror the share of each reading remains. This issue could only be resolved by a more advanced version of the model introduced here. The technologically quite crude execution of the semantic maps has the advantage that this method can be used easily and with little training, but this is at the expense of a more accurate representation of the frequency distribution of individual readings. A more advanced model could also tackle the issue that the arrangement of readings on the maps is not meaningful and mostly arbitrary at the moment - apart from the idea of grouping readings in the same ontological category together. Distance between readings could be used as a feature in a two-dimensional model, but this possibility is not exploited in the present version of the semantic map model. The analysis has also shown that semantic maps should not be used as the sole indicator of a reading's productivity. A careful consideration of all derivatives can reveal facts that are not represented on the maps, e.g. the existence of many opaque high-frequency formations with a locative reading 189 in the BNC that limit the productivity of the LOCATION reading of -ery derivatives. In spite of these problematic issues, the adapted semantic map model proves that a detailed data-driven account of the semantics of a morphological category can reveal interesting new insights into topics like semantic change and polysemy. It is a flexible method that can be adapted to individual needs and can be used easily and with little training. 191 8 References Primary Sources The Oxford English Dictionary. Online Version. Oxford, OUP. Available online at http: / / www.oed.com Davies, M. (2004-). BYU-BNC. (Based on the British National Corpus from Oxford University Press). Available online at http: / / corpus.byu.edu/ bnc/ The British National Corpus, version 3 (BNC XML Edition). 2007. Distributed by Oxford University Computing Services on behalf of the BNC Consortium. URL: http: / / www.natcorp.ox.ac.uk/ Secondary Sources About. (11 Jul 2013). Retrieved from http: / / public.oed.com/ about/ Aitchison, J. (2012). Words in the mind. An introduction to the mental lexicon (4. ed.). Malden, Mass.: Blackwell. Ali, A. N. & Ingleby, M. (2010). Gradience in morphological decomposability: Evidence from the perception of audiovisually incongruent speech. JLP, 1, 263-282. Anderson, L. B. (1982). The “perfect” as a universal and as a language-specific category. In P. J. Hopper (Ed.), Tense-aspect: Between semantics & pragmatics. Amsterdam: John Benjamins, 227-264. Arndt-Lappe, S. (2014). Analogy in suffix rivalry: the case of English -ity and -ness. English Language and Linguistics, 18, 497-548. Baayen, R. H. (1993). On frequency, transparency and productivity. In G. E. Booij & J. van Marle (Eds.), Yearbook of Morphology 1992. Dordrecht: Kluwer, 181-208. Baayen, R. H. (2014). Experimental and psycholinguistic approaches. In R. Lieber & P. Štekauer (Eds.), The Oxford handbook of derivational morphology. Oxford: Oxford UP, 95-117. Baayen, R. H. & Lieber, R. (1991). Productivity and English derivation: A corpusbased study. Linguistics, 29, 801-843. Baayen, R. H., Milin, P., Filipovic Durdejevic, D. Hendrix, P. & Marelli, M. (2011). An amorphous model for morphological processing in visual comprehension based on naïve discriminative learning. Psychological Review, 118 (3), 438-481. Baayen, R. H. & Renouf, A. (1996). Chronicling the times: Productive lexical innovations in an English newspaper. Language, 72, 69-96. Baeskow, H. (2012). -ness and -ity: Phonological exponents of n or meaningful nominalizers of different adjectival domains? Journal of English Linguistics, 40(I), 6-40. doi: 10.1177/ 0075424211405156 Bauer, L. (1983). English word formation. Cambridge: Cambridge UP. Bauer, L. (2001). Morphological productivity. Cambridge: Cambridge UP. Bauer, L., Lieber, R. & Plag, I. (2013). The Oxford reference guide to English morphology. Oxford: Oxford UP. Beard, R. (1995). Lexeme-morpheme base morphology. Albany: State University of New York Press. 192 Beard, R. & Volpe, M. (2005). Lexeme-morpheme base morphology. In P. Štekauer & R. Lieber (Eds.), Handbook of word-formation. Dordrecht: Springer, 189-205. Berlin, B. & Kay, P. (1969). Basic color terms. Their universality and evolution. Berkeley: University of California Press. Blank, A. (2003). Polysemy in the lexicon and in discourse. In B. Nerlich, Z. Todd, V. Herman & D. Clarke (Eds.), Polysemy. Flexible Patterns of Meaning in Mind and Language. Berlin: Mouton de Gruyter, 267-294. doi: 10.1515/ 9783110895698.267 Bloomfield, L. (1969 [1933]). Language. London: Allen & Unwin. Booij, G. (1986). Form and meaning in morphology: the case of Dutch 'agent nouns'. Linguistics, 24, 503-517. Booij, G. (2007). Polysemy and construction morphology. In F. Moerdijk, A. van Santen, R. Tempelaars (Eds.), Leven met woorden. Leiden: Brill, 355-364. Bozic, M., Tyler, L., Su, L., Wingfield, C. & Marslen-Wilson, W. D. (2013). Neurobiological systems for lexcial representation and analysis in English. Journal of Cognitive Neuroscience, 25 (10), 1678-1691. Brugman, C. & Lakoff, G. (2006). Cognitive topology and lexical networks. In D. Geeraerts (Ed.), Cognitive linguistcs: Basic readings. Berlin: Mouton de Gruyter. 109- 139. Burani, C. & Thornton, A. M. (2003). The interplay of root, suffix, and whole-word frequency in processing derived words. In R. H. Baayen & R. Schreuder (Eds.), Morphological structure in language processing. Berlin: Mouton de Gruyter, 157- 207. Bybee, J. (1995). Regular morphology and the lexicon. Language and Cognitive Processes, 10 (5), 425-455. Caramazza, A., Laudanna, A. & Romani, C. (1988). Lexical access and inflectional morphology. Cognition, 28, 297-332. Carstairs-McCarthy, A. (2005). Basic terminology. In P. Štekauer & R. Lieber (Eds.), Handbook of word-formation. Dordrecht: Springer, 5-23. Casati, R. & Varzi, A. (2006). Events. In E. N. Zalta (Ed.), The Stanford encyclopedia of philosophy (Spring 2010 ed.). Retrieved from http: / / plato.stanford.edu / archives/ spr2010/ entries/ events/ Chomsky, N. (1970). Remarks on nominalization. In R. A. Jacobs & P. S. Rosenbaum (Eds.), Readings in English transformational grammar. Waltham, Mass.: Ginn, 184- 221. Comrie, B. (1976). Aspect. An introduction to the study of verbal aspect and related problems. Cambridge: Cambridge UP. Copestake, A. & Briscoe, T. (1995). Semi-productive polysemy and sense extension. Journal of Semantics, 12, 15-67. Costello, F. J. & Keane, M. T. (2005). Compositionality and the pragmatics of conceptual combination. In M. Werning, E. Machery & G. Schurz (Eds.), The Compositionality of Meaning and Content. Volume II. Applications to lingusitics, psychology and neuroscience. Frankfurt a.M.: Ontos, 203-216. Crystal, D. (2008). A dictioanry of linguistics and phonetics (6. ed.). Malden, Mass.: Blackwell. Cysouw, M. (2007). Building semantic maps: The case of person marking. In M. Miestamo & B. Wälchli (Eds.), New challenges in typology. Berlin: Mouton de Gruyter, 225-248. Cysouw, M., Haspelmath, M. & Malchucov, A. L. (Eds.) (2010). Semantic maps: Methods and applications. [Special issue]. Linguistic Discovery, 8 (1). Retrieved 193 from http: / / journals.dartmouth.edu/ cgi-bin/ WebObjects/ Journals.woa/ xmlpage/ 1/ archive Dalton-Puffer, C. (1996). The French influence on Middle English morphology. A corpusbased study of derivation. Berlin: Mouton de Gruyter. Dixon, R. M. W. (1991). A new approach to English grammar, on semantic principles. Oxford: Clarendon Press. Dunbar, G. (2005): The goldilocks scenario: Is noun-noun compounding compositional? In E. Machery, M. Werning & G. Schurz (Eds.), The compositionality of meaning and content. Volume II. Applications to lingusitics, psychology and neuroscience. Frankfurt a. M.: Ontos, 217-228. Fleischman, S. (1977). Cultural and linguistic factors in word formation. An integrated approach to the development of the suffix -age. Berkeley, CA: University of California Press. Frauenfelder, U. H. & Schreuder, R. (1992). Constraining psycholinguistic models of morphological processing and representation: The role of productivity. In G. E. Booij & J. van Marle (Eds.), Yearbook of Morphology 1991. Dordrecht: Kluwer, 165- 183. Geeraerts, D. (1997). Diachronic prototype semantics. A contribution to historical lexicology. Oxford: Oxford UP. Grimshaw, J. (1994). Argument structure (4. print.). Cambridge, Mass.: MIT Press. Günther, H. (2005). Mentale Repräsentation morphologischer Strukturen. In G. E. Booij (Ed.), Morphologie. Ein internationales Handbuch zur Flexion und Wortbildung. Berlin: de Gruyter, 1766-1777. Haselow, A. (2011). Typological changes in the lexicon. Analytic tendencies in English noun formation. Berlin: de Gruyter Mouton. Haspelmath, M. (1997). Indefinite Pronouns. Oxford: Clarendon. Haspelmath, M. (2003). The geometry of grammatical meaning: semantic maps and cross-linguistic comparison. In M. Tomasello (Ed.), The new psychology of language (Vol. 2). Mahwah, NJ: Erlbaum, 211-242. Haspelmath, M. (2004). On directionality in language change with particular reference to grammaticalisation. In O. Fischer, M. Norde & H. Perridon (Eds.), Up and down the cline: The nature of grammaticalization. Amsterdam: Benjamins, 17-44. Haspelmath, M. & Sims, A. D. (2010). Understanding morphology (2. ed). London: Hodder Education. Hay, J. (2003). Causes and consequences of word structure. New York: Routledge. Hay, J. & Baayen, R. H. (2005). Shifting paradigms: gradient structure in morphology. TRENDS in Cognitive Sciences, 9 (7), 342-348. Hughes, G. (1988). Words in time. A social history of the English vocabulary. Oxford: Blackwell. Jackendoff, R. (1991). Semantic structures (2. print). Cambridge, MA: The MIT Press. Katamba, F. & Stonham, J. T. (2006). Morphology (2. ed.). Basingstoke: Palgrave Macmillan. Klos, V. (2011). Komposition und Kompositionalität. Möglichkeiten und Grenzen der semantischen Dekodierung von Substantivkomposita. Berlin: de Gruyter. Labov, W. (1973). The boundaries of words and their meanings. In C.-J. N. Bailey & R. W. Shuy (Eds.), New ways of analyzing variation in English. Washington, DC: Georgetown UP, 340-373. Lakoff, G. (1987). Women, fire, and dangerous things: What categories reveal about the mind. Chicago: University of Chicago Press. 194 Langacker, R. W. (1987). Foundations of cognitive grammar (Vol. 1). Stanford: Stanford UP. Lehmann, C. (2007). Motivation in language. Attempt at a systematization. In P. Gallmann, C. Lehmann & R. Lühr (Eds.), Sprachliche Motivation. Zur Interdependenz von Inhalt und Ausdruck. Tübingen: Narr, 105-140. Lehrer, A. (2000): Are affixes signs? The semantic relationships of English derivational affixes. In W. U. Dressler, O. E. Pfeiffer, M. A. Pöchtrager & J. R. Rennison (Eds.), Morphological analysis in comparison. Amsterdam: John Benjamins, 143-154. Lehrer, A. (2003). Polysemy in derivational affixes. In B. Nerlich, Z. Todd, V. Herman & D. Clarke (Eds.), Polysemy. Flexible patterns of meaning in mind and language. Berlin: Mouton de Gruyter, 217-232. doi: 10.1515/ 9783110895698.217 Lieber, R. (2004): Morphology and lexical semantics. Cambridge: Cambridge UP. Lieber, R. (2012). Semantics of derivational morphology. In C. Maienborn, K. von Heusinger & P. Portner (Eds.), Semantics (Vol. 3). Berlin: de Gruyter, 2098-2119. doi: 10.1515/ 9783110253382.2098 Löbner, S. (2002). Understanding semantics. London: Hodder. Lyons, J. (1977). Semantics. Cambridge: Cambridge UP. Marchand, H. (1969). The categories and types of present-day English word-formation. A synchronic-diachronic approach. (2., completely rev. and en. ed.). München: Beck. Marle, J. v. (1992). The relationship between morphological productivity and frequency: A comment on Baayen's performance-oriented conception of morphological productivity. In G. Booij and J. v. Marle (Eds.), Yearbook of Morphology 1991. Dordrecht: Kluwer, 151-163. McQueen, J. M. & Cutler, A. (2001). Spoken word access processes: An introduction. Language and Cognitive Processes, 16(5-6), 469-490. doi: 10.1080/ 01690960143000209 Melloni, C. (2011). Event and result nominals. A morpho-semantic approach. Bern: Peter Lang. Mithun, M. (1997). Lexical affixes and morphological typology. In J. Bybee, J. Haiman & S. A. Thompson (Eds.), Essays on language function and language type. Amsterdam: John Benjamins, 357-372. Murphy, M. L. (2010). Lexical meaning. Cambridge: Cambridge UP. Pelletier, F. J. (1994). The Principle of Semantic Compositionality. Topoi, 13, 11-24. Plag, I. (1999). Morphological productivity. Structural constraints in English derivation. Berlin: Mouton de Gruyter. Plag, I. (2003). Word formation in English. Cambridge: Cambridge UP. Plag, I., Braun, M., Lappe, S. & Schramm, M. (2007). Introduction to English Linguistics. Berlin: mouton de Gruyter. Pustejovsky, J. (1995). The generative lexicon. Cambridge, MA: MIT Press. Raffelsiefen, R. (2010). Idiosyncrasy, regularity, and synonymy in derivational morphology: evidence for default word interpretation strategies. In S. Olsen (Ed.), New impulses in word formation. Hamburg: Buske, 174-232. Reid, A. A. & Marslen-Wilson, W. D. (2003). Lexical representation of morphologically complex words: Evidence from Polish. In R. H. Baayen & R. Schreuder (Eds.), Morphological structure in language processing. Berlin: Mouton de Gruyter, 287- 336. Riddle, E. (1985). A historical perspective on the productivity of the suffixes -ness and -ity. In J. Fisiak (Ed.), Historical semantics. Historical word-formation. Berlin: Walter de Gruyter, 435-462. 195 Rueckl, J. G. & Aicher, K. A. (2008). Are CORNER and BROTHER morphologically complex? Not in the long term. Language and Cognitive Processes, 23 (7-8), 972- 1001. doi: 10.1080/ 01690960802211027 Sandra, D. (1994). The morphology of the mental lexicon: Internal word structure viewed from a psycholinguistic perspective. Language and Cognitive Processes, 9(3), 227- 269. Schröder, A. (2011). On the productivity of verbal prefixation in English. Synchronic and diachronic perspectives. Tübingen: Narr. Schulte, M. (2014). Accounting for affix polysemy with semantic maps - a diachronic study of -age suffixation in English. In S. Augendre, G. Couasnon-Torlois, D. Lebon, C. Michard, G. Boyé, F. Montermini (Eds.), Proceedings of the Décembrettes 8 th International conference on morphology 6-7 December 2012, 300-318. Retrieved from http: / / w3.erss.univ-tlse2.fr/ textes/ publications/ CarnetsGrammaire/ carnGram22.pdf Smolka, E., Preller, K. H. & Eulitz, C. (2014). 'Verstehen' ('understand') primes 'stehen' ('stand'): Morphological structure overrides semantic compositionality in the lexical representation of German complex verbs. Journal of Memory and Language, 72, 16- 36. Taylor, J. (2003). Cognitive models of polysemy. In B. Nerlich, Z. Todd, V. Herman & D. Clarke (Eds.), Polysemy. Flexible patterns of meaning in mind and language. Berlin: Mouton de Gruyter, 31-48. doi: 10.1515/ 9783110895698.31 Timelines. (11 Jul 2013). Retrieved from http: / / www.oed.com/ timelines Trips, C. (2009). Lexical semantics and diachronic morphology. The development of -hood, dom, and -ship in the history of English. Tübingen: Niemeyer. Tyler, A. & V. Evans (2001). Reconsidering prepositional polysemy networks: The case of over. Language, 77 (4), 724-765. Uth, M. (2011). Französische Ereignisnominalisierungen. Abstrakte Bedeutung und regelhafte Wortbildung. Berlin: de Gruyter. Wälchli, B. (2010). Similarity semantics and building probabilistic semantic maps from parallel texts. In M. Cysouw, M. Haspelmath & A. L. Malchucov (Eds.), Semantic maps: Methods and applications. [Special issue]. Linguistic Discovery, 8 (1), 331-371. Retrieved from http: / / journals.dartmouth.edu/ cgibin/ WebObjects/ Journals.woa/ 1/ xmlpage/ 1/ article/ 356 Wälchli, B. & Cysouw, M. (2012). Lexical typology through similarity semantics: Toward a semantic map of motion verbs. Linguistics, 50 (3), 671-710. Wierzbicka, A. (1996). Semantics. Primes and universals. Oxford: Oxford UP. Wittgenstein, L. (1968). Philosophische Untersuchungen. Philosophical investigations (3. ed.). Oxford: Blackwell. Zipf, G. K. (1972 [1949]). Human behavior and the principle of least effort. New York: Hafner. 197 9 Appendix Appendix A: ME -age Neologisms (OED) Derivative Attestation Date Possible Base(s) and Attestation Date of Suitable Sense Origin of Base Reading(s) of Derivative advantage c1300 avaunt v2 1393 ROM CONDITION, LOCATION, AMOUNT, GEN. ABSTRACT alliage c1450 ally v. c1325, ally n. c1380 ROM GEN. ABSTRACT altarage ? c1430 altar n. OE ROM, GER RIGHT, OBJECT, CHARGE arbitrage 1480 arbitre v. a1513 ROM ACTION arrearage c1315 arrear adv. 1330, arrear n. 1340, arrear v. 1399 ROM CONDITION, AMOUNT arrivage c1384 arrive v. c1275, arrive n. c1386 ROM ACTION average n1 1489 aver n. ? a1513 ROM ACTION average n2 1451 aver n. 1330 ROM CHARGE baggage c1430 bag n. ? c1225 ROM, GER COLLECTIVE barbicanage c1415 barbican n. a1300 ROM CHARGE barnage c1400 bairn n. OE GER CONDITION baronage a1300 baron n. a1200 ROM COLLECTIVE, LOCATION bondage 1330 bond n2 c1275 bond adj. 1330 GER CONDITION, POSITION, AC- TION, TENURE borrowage c1440 borrow n. OE, borrow v. OE GER GEN. ABSTRACT boscage c1400 bosk n. 1297 ROM, GER COLLECTIVE; LOCATION, CHARGE brokage 1377 broker n. 1377 ROM ACTION 198 brokerage 1466 broker n. 1377 ROM ACTION burgage 1362 borough n. OE GER LOCATION butlerage 1491 butler n. 1297 ROM CHARGE carriage c1386 carry v. 1330 ROM ACTION, CHARGE, COLL- ECTIVE cartage 1428 cart n. ? c1200, cart v. 1393 GER ACTION, CHARGE chevage 1461- 83 chief n. 1297 ROM CHARGE coinage c1380 coin v. c1330, coin n. 1362 ROM ACTION, RIGHT, COLLECTIVE, GEN. ABSTRACT companage c1325 companion 1297, company c1250 ROM OBJECT concubinage a1425 concubine n. 1297 ROM CONDITION, ACTION cordage 1490 cord n. a1305, cord v. c1430 ROM COLLECTIVE corsage 1481 corse n. c1420 ROM CONDITION costage a1327 cost n2 1297, cost v. c1320 ROM CHARGE cottage c1386 cot n1 OE GER LOCATION coupage a1483 coup v2 a1300 ROM ACTION cousinage a1340 cousin n. c1290 ROM CONDITION, COLLECTIVE cranage 1390 crane n. 1487 GER CHARGE, AC- TION curtilage c1330 curtiler n. a1300 ROM LOCATION, AC- TION disadvantage c1380 disadvance v. 1374 ROM GEN. ABSTRACT disusage 1475 disuse v. 1487 ROM ACTION dotage c1386 dote n1 a1250, dote v1 a1225 ROM, GER ACTION, CON- DITION eremitage c1400 eremite n. c1200 ROM LOCATION falsage a1400 false v. a1225, false n. OE, false adj. OE ROM, GER GEN. ABSTRACT fardellage 1489 fardel n. a1300 ROM OBJECT ferriage c1440 ferry v. OE, ferry n. c1425 GER ACTION, CHARGE forceage c1470 force n1 a1303, force v1 a1330 ROM ACTION 199 fraughtage 1442 fraught v. c1400, fraught n. a1400 GER ACTION, CHARGE furnage c1468 furner n. a1483 ROM CHARGE gainage 1390 gain n. ? c1200, gain v1 ? c1200 ROM, GER COLLECTIVE, GEN. ABSTRACT gavelage c1450 gavel n1 OE, gavel v1 OE ROM, GER CHARGE groundage c1450 ground n. OE, ground v. c1275 GER CHARGE, AC- TION guidage c1450 guide n. 1362, guide v. c1374 ROM CHARGE guyage c1425 guy v1 1362, guy n. a1375 ROM CHARGE harbergage, herbergage c1386 harbour v. c1150 ROM, GER ACTION, LOCA- TION heirage 1478 heir n. c1275, heir v. c1330 ROM ACTION herbage 1390 herb n. c1290 ROM COLLECTIVE, RIGHT herbryage 1488 harboury n. a1300, harbry v. 14.. GER ACTION, LOCA- TION hermitage c1290 hermit n. c1275 ROM LOCATION hidage a1195 hide n2 OE GER CHARGE hostage n1 c1290 host n2 c1290 ROM CONDITION, PERSON, GEN. ABSTRACT hostage n2 c1440 host n2 c1290, host v2. ? c1450 ROM LOCATION labourage a1460 labour n. c1300, labour v. c1390 ROM ACTION landage 1470- 85 land n1 OE, land v. a1300 GER ACTION lastage a1387 last n2 OE, last v2 OE GER CHARGE, COLL- ECTIVE leakage 1491 leak v. c1420, leak n. 1487 GER ACTION lighterage 1481- 90 lighter n. 1487 GER CHARGE, AC- TION lineage .13.. line n. OE ROM, GER PERSON, COLL- ECTIVE, CONDI- TION 200 lodemanage c1405 lodeman n. OE GER ACTION lovage n2 1489 love v2 OE GER GEN. ABSTRACT maritage c1478 marite n. a1398 ROM CONDITION marriage c1300 marry v. a1325 ROM ACTION, CON- DITION, GEN. ABSTRACT, CHARGE, PER- SON, RIGHT massage c1450 mass v1 OE, mass n1 OE ROM, GER ACTION measurage 1460-1 measure n. a1225, measure v. 1340 ROM CHARGE, AC- TION ménage c1325 meinie n. c1300 ROM PERSON, LOCA- TION mockage 1485 mock n1 c1425, mock v. ? a1439 ROM ACTION murage 1424 mure n. OE, mure v. ? a1425 ROM, GER CHARGE, RIGHT, ACTION, OBJECT orage 1477 aura n. a1398 ROM OBJECT out-passage a1398 outpass v. a1398 ROM ACTION parage c1250 peer n. c1300, peer adj. a1325 ROM POSITION, TEN- URE, GEN. ABSTRACT parcage 1449 park n. 1222, park v. 1531 ROM ACTION parentage 1490 parent n. ? a1425 ROM POSITION, AC- TION parsonage c1400 parson n. c1275 ROM, GER CHARGE, LO- CATION partage c1450 part n. OE, part v. ? c1225 ROM, GER ACTION, GEN. ABSTRACT passage c1300 pass v. c1225 ROM ACTION, CHARGE, LO- CATION, RIGHT, GEN. ABSTRACT, OB- JECT patronage 1395 patron n. c1300 ROM, GER RIGHT pavage c1376 pave v. c1325 ROM LOCATION, CHARGE, RIGHT 201 peerage 1454 peer n. c1300, peer v.1 c1400 ROM PERSON pelerinage c1300 pelerin n. c1325 ROM ACTION, LOCA- TION personage c1460 person n. ? c1225 ROM PERSON, OB- JECT, CONDITI- ON pesage c1450 peise v. a1382, peise n. a1382 ROM CHARGE, COLL- ECTIVE pickage 1405 pike n1 eOE, pick v1 1272-3 ROM, GER CHARGE, RIGHT pilgrimage c1275 pilgrim n. OE ROM, GER ACTION, LOCA- TION pillage a1393 pill v1 ? c1225 ROM, GER ACTION, COLL- ECTIVE plankage 1424 plank n. 1294-5, plank v. 1432 ROM, GER CHARGE plumage c1395 plume n. OE ROM, GER COLLECTIVE pontage a1325 pont n. 1279 ROM CHARGE portage c1400 port n1 OE ROM ACTION, CHARGE, COLL- ECTIVE, AMOUNT pottage ? c1225 pot n. OE ROM, GER OBJECT poundage 1422 pound n. OE ROM, GER CHARGE presserage c1450 press v. 1330, pressour n. 1348 ROM OBJECT primage 1476 prime v1 1513 ? CHARGE putage 1480 pute n. c1384 ROM ACTION quarterage 1389 quarter n. c1300, quarter v. a1387 ROM CHARGE, GEN. ABSTRACT quayage 1440 quay n. 1399 ROM, GER CHARGE reclusage 1480 recluse n. and adj. ? c1225, recluse v. a1400 ROM LOCATION recolage a1400 rigole v. a1393 ROM ACTION 202 repassage 1429 repass v. c1460 ROM ACTION, RIGHT, LOCATION, GEN. ABSTRACT rivage c1330 rive n. 1296, rive v2 c1300 ROM ACTION, LOCA- TION rummage 1486 rime v2 eOE, room v1 OE ROM ACTION scavage 1444 show v. OE, show n. a1300 GER CHARGE schoolage 1496 school n. OE, school v1 c1456 ROM, GER CHARGE scourage 1470 scour v. 1297, scour n. a1300 ? ACTION scrimmage, scrummage 1488 skrim v. c1450 ROM ACTION seigniorage 1444 seignior n. 1393 ROM CHARGE senage c1380 sene n3 1380 ROM CHARGE servage c1290 serve v.1 c1175 ROM CONDITION, ACTION, OB- JECT, CHARGE skevinage 1449 skevin n. 1389 ROM LOCATION socage a1325 soc n1 1228 ROM, GER TENURE, LOCA- TION sorage ? a1400 sore adj c1450 ROM CONDITION spousage .13.. spouse v. c1290, spouse n. c1200 ROM CONDITION, ACTION stallage a1387 stall n1 OE, stall v1 1415 ROM, GER CHARGE steerage c1450 steer v1 a1122, steer n2 OE GER ACTION stoppage c1450 stop v. OE, stop n2 1483-4 ROM, GER GEN. ABSTRACT stowage 1390 stow n1 OE, stow v1 1362 GER ACTION, CHARGE summage c1450 soum n1 1488, sum n2 c1450 ROM CHARGE superplusage 1436 superplus n. A1450 ROM AMOUNT surplusage c1407 surplus n. C1374 ROM AMOUNT swannage 1398 swan n. OE GER CHARGE tallage n1 c1290 tail v2 c1330, tail n2 1340 ROM CHARGE 203 tarriage 1488 tarry v. c1320, tarry n. 1451 ? ACTION taxage 1483 tax v. c1290 tax n1 a1327 ROM ACTION testimonage 1483 testimony v. c1330 ROM GEN. ABSTRACT thanage .14.. thane n1 OE GER TENURE, POSI- TION, LOCATI- ON thrillage c1400 thrall n1 OE GER CONDITION, ACTION tillage 1488-9 till v. OE GER CONDITION tonnage 1422 tun n1 OE, tun v. c1430 ROM, GER CHARGE towage a1327 tow v. OE, tow n3 1407 ROM, GER ACTION trewage, truage c1275 trewe n. c1330 ROM CHARGE umbrage 1426 umber n1 c1380 ROM OBJECT usage c1325 use n. ? c1225, use v. a1250 ROM ACTION, AMOUNT vantage c1380 vaunce v. 1303 ROM POSITION, AMOUNT, GEN. ABSTRACT vassalage 1303 vassal n. a1400 ROM GEN. ABSTRACT, AC- TION vicarage 1425 vicar n. a1300 ROM CHARGE, COLL- ECTIVE victorage c1480 victor n1 a1340, victor n2 1390 ROM GEN. ABSTRACT villeinage a1325 villein n. a1325, villain n. 1303 ROM TENURE, LOCA- TION vintage c1450 vinter n. 1297 ROM COLLECTIVE wharfage 1469- 71 wharf n. OE GER ACTION 204 Appendix B: PDE -age Neologisms (OED) Derivative Attestation Date Possible Base(s) and Attestation Date of Suitable Sense Reading(s) of Derivative air mileage 1913 air mile n. 1919 AMOUNT, GEN. AB- STRACT beamage 1902 beam n1 1420 GEN. ABSTRACT bricolage 1960 bricoleur n. 1965 ACTION, OBJECT, GEN. ABSTRACT, COLLECTIVE briquetage 1902 briquette n. 1883 COLLECTIVE cerclage 1920 circle n. OE, circle v. c1374 ACTION coverage 1912 cover v1 a1275, cover n1 c1300 ACTION, LOCATION, AMOUNT, GEN. AB- STRACT creepage 1903 creep v. OE, creep n. 1486 ACTION dressage 1936 dress v. c1540 ACTION ecotage 1971 ecoteur n. 1972 ACTION empennage 1909 impen v2 c1614 COLLECTIVE flamage 1983 flame v. 1981, flame n. 1983 ACTION frettage n2 1938 fret v1 OE, fret n2 1545 GEN. ABSTRACT frottage 1933 frot v. ? c1225 GEN. ABSTRACT, AC- TION fuselage 1909 fusil n1 1486 OBJECT gallonage 1909 gallon n. c1300 AMOUNT gauffrage 1904 goffer/ gauffer v. 1824, goffer n. 1865 ACTION hangarage 1932 hangar n. 1902 ACTION headage 1957 head n1 OE AMOUNT hourage 1924 hour n. c1250 AMOUNT, GEN. AB- 205 STRACT in-leakage 1905 inleak n. 1909 ACTION intertillage 1912 intertilled adj. 1912 ACTION marcottage 1926 marcot n. 1926, marcot v. 1926 ACTION matrilineage 1949 matriline n. 1957 COLLECTIVE, GEN. ABSTRACT megatonnage 1954 megaton n. 1952 AMOUNT, GEN. AB- STRACT megavoltage 1955 megavolt n. 1868 AMOUNT meritage 1989 merit n. 1340 OBJECT metreage 1974 metre n2 1797 AMOUNT milliamperage 1909 milliampere n. 1885 AMOUNT minutage 1984 minute n1 a1393 AMOUNT montage 1930 mount v. 1622 ACTION, OBJECT, COLLECTIVE mud pilotage 1932 mud pilot n. 1856 ACTION narratage 1933 narrate v. 1656 ACTION ohmage 1909 ohm n2 1861 AMOUNT overage n2 1931 over adv. OE, over adj. OE, over v. OE, over n3 1905 AMOUNT parachutage 1945 parachute n. 1784, parachute v. 1812 ACTION, LOCATION patrilineage 1949 patriline n. 1957 COLLECTIVE, GEN. ABSTRACT peonage 1900 peon n1 1609 ACTION petrolage 1904 petrol n. 1540 ACTION photoreportage 1939 photo reporting n. 1935 ACTION, OBJECT Pinotage 1964 Pinot n. 1854 OBJECT placage n2 1932 place n1 a1387, place v. 1442 CONDITION, ACTION plombage 1933 plombe n. 1904 ACTION 206 plottage 1910 plot n. OE, plot v1 a1586 LOCATION, ACTION, GEN. ABSTRACT plussage 1927 plus n. a1721 AMOUNT pointage 1925 point n1 1701 AMOUNT portalage 1903 portal n1 c1400 ACTION pre-package 1946 pre-pack v. 1926, prepack n. 1951 OBJECT, ACTION preshrinkage 1924 preshrink v. 1907 ACTION problemage 1928 problem n. a1382 CONDITION radiosondage 1939 radiosonde n. 1932 ACTION rapportage 1903 rapport n. 1454 ACTION remuage 1908 remue v. a1325, remue n. a1450 ACTION reporterage a1936 reporter n. 1400 ACTION, GEN. AB- STRACT retainage 1902 retain v. c1415, retain n. 1455-6 AMOUNT, ACTION riffage 1991 riff v1 1935, riff n5 1934 ACTION sabotage 1910 sabot n. 1607 ACTION, GEN. AB- STRACT scrappage 1949 scrap n1 1387, scrap v3 1891 ACTION screenage 1929 screen n1 1393, screen v. c1485 OBJECT, COLLECTIVE, ACTION, GEN. AB- STRACT septage 1977 septic adj. 1605, septic n. 1608 OBJECT sex linkage 1912 sex-linked adj. 1905 GEN. ABSTRACT signage 1949 sign n. c1300, sign v1 a1398 COLLECTIVE, GEN. ABSTRACT sondage 1930 sonde n. 1901 LOCATION spillage 1934 spill v. a1340, spill n4 a1849 ACTION, OBJECT spindlage 1908 spindle n. OE AMOUNT spindleage 1921 spindle n. OE AMOUNT 207 stillage 1940 still n1 1562, still v2 a1300 OBJECT strewage 1902 strew v. OE, strew n. 1578 COLLECTIVE teacherage 1916 teacher n. 1382 LOCATION trippage 1941 trip n1 1412-20 trip v. C1380 ACTION, AMOUNT tuneage 1985 tune n. a1387, tune v2 a1527 GEN. ABSTRACT turbinage 1909 turbine n. 1838 ACTION twiggage 1923 twig n1 OE COLLECTIVE victimage 1954 victim n. 1781 CONDITION, ACTION virage 1963 veer n. 1611, veer v2 1633 vire v. c1485 LOCATION, ACTION voidage 1946 void n. a1618 COLLECTIVE, GEN. ABSTRACT warehouseage 1915 warehouse n. 1349, warehouse v.1799 CHARGE wattage 1903 watt n. 1882 AMOUNT, GEN. AB- STRACT weightage 1906 weight n1 OE, weight v. 1747 GEN. ABSTRACT, AMOUNT 208 Appendix C: ME -ery Neologisms (OED) Derivative Attestation Date Possible Base(s) and Attestation Date of Suitable Sense Origin of Base Reading(s) of Derivative achatry c1450 achate n2 c1405 ROM COLLECTIVE, LOCATION adultery 1357 adulter n. c1384 ROM ACTION, CON- DITION adversary 1340 adverse v. a1393, adverse adj. a1393, adversity ? c1225 ROM PERSON, OBJECT advowry c1460 avow v1 c1220 ROM RIGHT, CONDI- TION, PERSON, CHARGE, POSI- TION aldermanry 1384 alderman n. 1275 GER POSITION, COL- LECTIVE almonry 1440 almoner n1 c1330, almoner n2 1340 ROM LOCATION, COLLECTIVE, OBJECT ambassadry c1386 ambassador n. c1374 ROM POSITION ancestry 1330 ancestor n. 1297 ROM CONDITION, COLLECTIVE arbalestry a1423 arbalest n. a1100, arbalester n. 1330 ROM ACTION archery a1400 archer n. 1297 ROM ACTION, CON- DITION, COL- LECTIVE armory 1489 arm n2 1340, arm v1 1250 ROM ACTION armoury 1330 arm n2 1340, arm v1 1250 ROM COLLECTIVE artillery c1405 artiller n. c1453 ROM OBJECT, COL- LECTIVE, AC- TION avauntry 1330 avaunt v1 1303, avaunter n. c1374 ROM CONDITION, ACTION 209 avowry 1330 avow v1 c1220, avowe n. 1297 ROM ACTION, PER- SON, RIGHT babery ? c1450 babe n. a1393 ? OBJECT baboonery c1400 baboon n. C1400 ROM OBJECT bachelry 1297 bachelor n. 1297 ROM CONDITION, COLLECTIVE bailiery 1425 bailie n. 1297 ROM LOCATION baptistery 1460 baptist n. c1200, baptiste n. 1460 ROM LOCATION baronry c1449 baron n. C1200 ROM LOCATION barratry 1427 barrat n. ? c1225 ROM ACTION baudery c1386 baude adj. C1400 ROM, GER CONDITION bawdry c1374 Bawd n1 1362 ? ACTION, CON- DITION beggary 1377 beggar n. a1250, beg v. ? c1225 ? CONDITION bordelry c1440 bordel n. c1305, bordeler n. c1375 ROM LOCATION bribery c1386 briber n. 1377, bribe n. c1386, bribe v. c1386 ROM ACTION broidery 1382 broid v. c1405 ROM ACTION, OBJECT buggery 1330 bugger n. 1340 ROM GEN. ABSTRACT butchery c1340 butcher n. A1300 ROM LOCATION, AC- TION, COLLEC- TIVE butlery 1297 butler n. 1297 ROM LOCATION buttery 1389 butt n2 1423 ROM LOCATION canonry 1482 canon n2 c1275 ROM, GER POSITION, CHARGE carpentry 1377 carpenter n. c1325 ROM ACTION catery 1455 cater n1 c1400 ROM COLLECTIVE chamberlainry 1484 chamberlain n. ? c1225 ROM, GER POSITION chancellery c1300 chancellor n. OE ROM, GER POSITION chantry 14.. chant v. c1405, chanter n1 a1387 ROM ACTION, CHARGE, COL- 210 LECTIVE, LOCA- TION chapmanry 1483 chapman n. OE GER ACTION checkery 1420 check v2 c1440, check n2 c1450 ROM OBJECT chinchery c1386 chincher n. c1386, chinch n2 c1386, chinch adj. a1300, chinch v. c1440 ROM CONDITION chirurgery 1398 chirurgeon n. 1297 ROM ACTION chivalry 1297 chevalier n. 1292 ROM COLLECTIVE, POSITION, CON- DITION, ACTION constablery c1400 constable n. a1240 ROM POSITION cookery 1393 cook n. OE, cook v1 c1380 ROM, GER ACTION croiserie c1290 cross n. OE, croise v. ? c1225 ROM, GER ACTION cutlery c1449 cutler n. C1430 ROM COLLECTIVE dairy c1290 dey n. OE GER LOCATION, COLLECTIVE deaconry 1483 deacon n. OE ROM, GER POSITION deanery a1440 dean n. c1330 ROM POSITION, COL- LECTIVE, LOCA- TION desidery c1450 decide v1 c1380 ROM GEN. ABSTRACT devilry c1380 devil n. OE ROM, GER PERSON, AC- TION dowry c1330 dow v2 1297 ROM COLLECTIVE, GEN. ABSTRACT drapery a1300 draper n. 1362 ROM COLLECTIVE, ACTION, LOCA- TION duchery 1387 duke n. 1129 ROM LOCATION, AC- TION embracery 1450 embrace v3 1475, embracer n. 1495 ROM ACTION 211 embroidery 1393 embrowd v. c1380, embroiderer n. 1413 ROM ACTION enchantery 1297 enchanter n. 1297, enchant v. c1374 ROM ACTION Englishry 1439 English adj. OE, English n. OE GER COLLECTIVE ewery c1460 ewer n1 1361 ROM LOCATION fairy c1330 fay n. 1393 ROM LOCATION, COLLECTIVE, GEN. ABSTRACT, PERSON faitery 1377 fait v1 c1330, faitour n. a1340 ROM ACTION falsary 1435 false n. OE, false adj. OE, false v. ? c1225 ROM, GER PERSON fardry c1430 fard v a1450 GER ACTION, GEN. ABSTRACT flattery c1320 flatter n1 1340, flatter v1 ? c1225 ? ROM, ? GER ACTION forcenery 1480 forcene v. 1490 ROM ACTION foxery c1400 fox n. OE GER ACTION, CON- DITION frary a1400 friar n. c1290 ROM COLLECTIVE freemasonry 1435 freemason n. 1376 ACTION gainery 1424 Gain n1473 „profit“ gainage 1390 GER, ROM ACTION, CON- DITION, COL- LECTIVE, GEN. ABSTRACT gentlery a1275 gentle adj. ? c1225 ROM POSITION, CON- DITION, ACTION gentry c1325 gent adj. a1225 ROM POSITION, CON- DITION, ACTION glovery 1483 glover n. 1464, glove n. OE GER LOCATION gluttery a1340 glut n1 c1394, glut v1 c1315 ROM ACTION gluttonry c1175 glutton n. ? c1225 ROM GEN. ABSTRACT grocery 1436 grocer n. 1427 ROM COLLECTIVE 212 guilery 1303 guile v. ? c1225, guile n. ? c1225, guiler n. 1303 ROM ACTION gunnery 1497 gun n. 1339, gunner n. 1344 ? COLLECTIVE haberdashery 1419 haberdasher n. 1311-12 ? COLLECTIVE harbergery 1303 harbinger n. C1175 ROM, GER ACTION, LOCA- TION harlotry c1325 harlot n. ? c1225 ROM ACTION, OBJECT hazardry 1297 hazarder n. a1300, hazard n. c1300 ROM ACTION hostelry c1386 hosteler n. c1300, hostel n1 a1325, hostel v. c1330 ROM LOCATION hostry 1377 host n2 c1290 ROM LOCATION housewifery 1440 housewife n. C1225 GER ACTION, CON- DITION huckery 1377 hucker n. 14.., huck v. 14.. ? GER ACTION huckstery 1362 huckster n. a1300 ? GER ACTION, LOCA- TION, COLLEC- TIVE husbandry c1290 husband n. OE GER ACTION, COL- LECTIVE, GEN. ABSTRACT idiotry 1488 idiot n. c1384 ROM, GER ACTION, CON- DITION idolatry a1325 idolater n. c1380 ROM ACTION imagery c1350 image n. ? c1225, image v. c1390 ROM OBJECT, ACTION intrusery c1470 intruse v. ? a1500 ROM ACTION Irishry c1475 Irish adj OE, Irish n. c1275 GER, ROM COLLECTIVE janglery c1374 jangler n. 1303, jangle v. a1300, jangle n.1340-70 ROM ACTION japery c1386 japer 1362, jape v. 1362, jape n. C1377 ROM ACTION 213 jewellery c1400 jewel n. c1290, jeweller n. a1382 ROM COLLECTIVE Jewry ? c1225 Jew n. c1175 ROM LOCATION, GEN. ABSTRACT, COL- LECTIVE jugglery a1400 juggler n. a1100, juggle v. 1377 ROM ACTION justry c1425 just adj. 1382, just n1 c1384 ROM LOCATION koffry 1488 coff v. c1425, cofe n. 1471 GER ACTION lavendry 1377 lavender n1 a1300 ROM, GER PERSON lechery c1230 lecher n1 c1175 ROM, GER ACTION lepry c1475 leper n2 a1398 ROM GEN. ABSTRACT lollardry 1414 lollard n. 1390 GER AMOUNT lormery 1419 lorimer n. C1230 ROM COLLECTIVE, LOCATION loselry 1480 losel n. 1362 GER ACTION losengery 1303 losenger n. 1303 ROM ACTION losery c1460 lose v1 OE GER ACTION Mahometry 1481 Mahomet n. C1275 ROM GEN. ABSTRACT mammetry c1330 Mammet n. C1225 ROM ACTION, COL- LECTIVE, OB- JECT, GEN. AB- STRACT, AMOUNT mangery a1400 maunge v. c1400 ROM ACTION martyry c1390 martyr n. OE ROM, GER ACTION, CON- DITION masonry a1425 mason n1 c1275, mason v. c1450 ROM, GER OBJECT, ACTION, CONDITION mastery c1225 master n1 OE, master v. c1225 ROM, GER ACTION, CON- DITION, GEN. ABSTRACT mercery c1300 mercer n. C1230 ROM COLLECTIVE, ACTION, LOCA- TION 214 merchandry a1450 merchant n. c1225, merchant v. c1400 ROM ACTION, COL- LECTIVE meselry a1387 mesel n. c1300, mesel adj. c1300 ROM GEN. ABSTRACT messagery c1430 message n. c1300 ROM ACTION michery a1393 mitcher n. c1230 ROM ACTION midwifery ? c1475 midwife n. c1300 GER ACTION, GEN. ABSTRACT ministry a1225 minister n. c1300 ROM ACTION, POSI- TION mockery ? a1439 mock n1 c1425, mock v. ? a1439, mocker n1 ? c1450 ROM ACTION monastery a1425 monastic adj. c1449 ROM LOCATION, COLLECTIVE muliery c1400 mulier n2 c1400 ROM COLLECTIVE mummery 1465-6 mummer n. a1456, mum v. a1456 ROM, GER ACTION musardry c1450 musard n. c1330 ROM ACTION, CON- DITION musery c1450 muser n1 a1382, muse v. 1340 ROM ACTION mystery n1 c1350 mystic adj. a1382, mystic n. c1350 ROM GEN. ABSTRACT, ACTION, OBJECT, PERSON napery c1400 naperer n. 1225, nape n2 c1400 ROM COLLECTIVE, POSITION nigonry c1430 nigon n. a1400 ? GER CONDITION notary 1340 notar n1 1399, note n2 OE ROM, GER PERSON novelry a1393 novel adj. 1405 ROM CONDITION, OBJECT novicery a1425 novice n. 1340 ROM LOCATION, CONDITION nunnery c1300 nun n1 OE GER, ROM LOCATION, COLLECTIVE, CONDITION nunry c1325 nun n1 OE GER, ROM LOCATION, COLLECTIVE, CONDITION 215 nursery c1330 nurse n1 a1325, nurse v. c1330 ROM LOCATION, AC- TION officialry 1489 official n1 c1330, official adj. a1400 ROM POSITION, PER- SON, COLLEC- TIVE outlawry a1400 outlaw n. OE, outlaw v. OE GER ACTION, CON- DITION paintry c1454 painter n1 1240, paint v1 c1275, paint n1 1290-1 ROM ACTION, OBJECT palmery c1300 palm n2 c1300 ROM OBJECT pantry a1325 panter n2 c1325 ROM LOCATION, COLLECTIVE, GEN. ABSTRACT pastry 1442 paste n. 1288-9, paste v. ? a1425 ROM OBJECT paynimry c1384 paynim n. C1275 ROM CONDITION, ACTION, COL- LECTIVE pelfry ? a1475 pelf v. a1400, pelf n. a1425 ROM COLLECTIVE, OBJECT peltry a1450- 1500 pelt n1 1303, pelter n1 1318 ROM, GER COLLECTIVE, LOCATION pickery 1460 pick v1 ? c1300, picker n1 1350 GER, ROM ACTION pilfery 1489 pelf n. a1425, pelf v. a1400 ROM OBJECT, COL- LECTIVE pillery 1433 pill v1 ? c1225, piller n. c1385 ROM, GER ACTION pleadery c1450 plead v. c1275, pleader n1 ? a1300 ROM CONDITION plumbery c1450 plumber n. 1385- 6 ROM LOCATION, AC- TION, COLLEC- TIVE poetry a1387 poet n. a1382 ROM AMOUNT, GEN. ABSTRACT, AC- TION pomary c1390 pome n1 ? 1435 ROM LOCATION pompery c1460 pomp n1 c1330, pomp v2 c1450 ROM ACTION 216 porkery 1439 pork n. c1300 ROM COLLECTIVE portmanry 1346-7 portman OE GER POSITION pottery 1480 pot n1 OE, potter n1 a1225 GER, ROM LOCATION poultry 1345-6 poulter n. a1400 ROM LOCATION, COLLECTIVE, OBJECT, POSI- TION, ACTION prebendry 1489 prebend n. 1422 ROM POSITION, CHARGE presbytery 1466 presbyter n. OE ROM, GER LOCATION priory c1300 prior n. OE ROM, GER COLLECTIVE, POSITION provendry a1425 provend n. C1300 ROM, GER COLLECTIVE, GEN. ABSTRACT, RIGHT, POSI- TION, CHARGE provostry 1420 provost n. OE ROM, GER POSITION, LO- CATION, AC- TION, CHARGE, GEN. ABSTRACT pullery 1488 pull n2 a1500, pull v. OE ROM, GER COLLECTIVE, OBJECT, LOCA- TION putery c1390 pute n. c1384 ROM ACTION quaintry 1484 quaint adj. c1300, quaint v2 1484 ROM CONDITION ragery a1393 rage v. c1250, rage n. c1330, rager n. 1440 ROM CONDITION, ACTION rascaldry ? 1457 rascal n. a1382 ROM COLLECTIVE reavery c1325 reave v1 OE, reaver n. OE GER ACTION registery 1483 register n2 a1443 ROM OBJECT, GEN. ABSTRACT, LO- CATION regratery c1400 regrater n. c1400, regrate v1 1444 ROM ACTION revelry c1410 revel n1 a1375, revel v1 c1390 ROM ACTION 217 revestry 1412 revest v1 c1300 ROM LOCATION ribaldry 1389 ribald n. a1250 ROM, GER ACTION, CON- DITION ringildry 1483 ringild n. 1483 ROM POSITION, LO- CATION riotry ? a1400 riot n. ? c1225, riot v. a1393 ROM ACTION, CON- DITION robbery a1225 robber n. c1175, rob v. c1225 ROM ACTION, COL- LECTIVE ropery 1329 roper n. a1387, rope n1 OE, rope v2 a1400 GER LOCATION sacristanry 1483 sacristan n. c1480 ROM LOCATION saddlery c1449 saddler n. 1287, saddle n1 OE, saddle v. OE GER ACTION Sarsenry c1440 Saracen n. OE, saracen adj. 1400 ROM, GER COLLECTIVE saucery c1440 sauce n. 1362, sauce v. c1440 ROM COLLECTIVE, LOCATION saumbury 1393 saumbu n. C1330 GER OBJECT Scotry c1475 Scot n1 OE GER, ROM COLLECTIVE scullery c1440 squiller n. 1303 ROM COLLECTIVE, LOCATION seigniory c1290 seignior n. c1330 ROM GEN. ABSTRACT, RIGHT, LOCA- TION, COLLEC- TIVE sergeantry c1400 sergeant n. c1200, sergeant v. c1430 ROM TENURE, POSI- TION servagery c1400 servage n. c1290 ROM GEN. ABSTRACT sextry c1390 sexton n. a1325 ROM LOCATION skinnery a1475 skin n. OE, skinner n. 1255, skin v. c1475 GER COLLECTIVE, LOCATION skulkery ? a1400 skulk v. ? c1225, skulk n. c1320, skulker n. 1387 GER ACTION sophistry 1340 sophister n. c1380 ROM ACTION 218 sorcery a1300 sorcer n. c1400 ROM ACTION specery a1300 spece n. a1300 ROM COLLECTIVE spicery 1297 spice n. ? c1225, spice v. 1377 ROM COLLECTIVE, LOCATION spurriery c1449 spurrier n. 1389 GER ACTION squiry c1327 squire n. c1290, squire v. c1386 ROM COLLECTIVE stewartry 1473-3 steward n. OE GER POSITION, LO- CATION Sub-chantry 1492 sub-chanter n. 1484 ROM CHARGE, POSI- TION, LOCATION subdeanery 1476 subdean n. C1390 ROM POSITION, LO- CATION surfeitry c1425 surfeit n. a1387, surfeit v. c1400 ROM ACTION surgeonry 14.. surgeon n. c1330 ROM ACTION surgery 1398 surger n. a1400- 50 ROM ACTION tannery c1460 tanner n1 OE, tan v1 OE GER, ROM ACTION tapestry 1434 tapester n. 1472-3 ROM OBJECT tapissery 1426 tapisser n. c1405 ROM OBJECT tenantry 1385 tenant n. c1330 ROM CONDITION, LOCATION, GEN. ABSTRACT tormentry a1350 torment v. c1290, torment n. c1290, tormentor n. c1290 ROM COLLECTIVE, ACTION, CON- DITION tracery 1464 trace n1 a1300, trace v1 1374-5 ROM LOCATION traitory 1303 traitor n. ? c1225 ROM ACTION treachery ? c1225 treacher n. c1290, treche v. c1230 ROM ACTION tregetry c1380 treget n. a1400 ROM ACTION triflery a1400 trifler n. a1382, trifle v1 c1305 ROM ACTION truantry 1426 truant n. c1290, truant v. c1400 ROM ACTION, CON- DITION 219 trumpery c1485 trumper n. a1450, trump v2 1487 ROM ACTION, AMOUNT tutory c1400 tutor n. 1377 ROM POSITION tyrantry 1340 tyrant n. c1290 ROM ACTION, POSI- TION vassalry a1470 vassal n. a1400 ROM COLLECTIVE vauntery a1492 Vaunter n. 1484, vaunt v. 14.. ROM ACTION vestry 1388 vest v. c1425 ROM LOCATION vinery c1420 vine n. a1300, viner n2 1390 ROM LOCATION vintry 1297 vinter n. 1297 ROM LOCATION, COLLECTIVE voutry a1382 vouter n. c1386 ROM ACTION wardenry c1420 warden n. ? c1225 ROM LOCATION, PO- SITION waxchandlery 1398 wax candle n. OE COLLECTIVE Welshry ? a1400 Welsh adj. OE, Welsh n. OE GER COLLECTIVE yeomanry 1386 yeoman n. 1345-8 GER COLLECTIVE, CONDITION 220 Appendix D: PDE -ery Neologisms (OED) Derivative Attestation Date Possible Base(s) and Attestation Date of Suitable Sense Reading(s) of Derivative air gunnery 1926 air gunner n. 1916, air gun n. 1685 CONDITION, ACTION, COL- LECTIVE astrochemistry 1901 astrochemist n. 1849 GEN. ABSTRACT Babbitry 1920 Babbit n2 1921 ACTION, CON- DITION banditry 1922 bandit n. 1594, bandit v. 1611 ACTION bardolatry 1901 bardolater n. 1903 ACTION beat poetry 1959 beat poet n. 1955 AMOUNT biogeochemistry 1935 biogeochemist n. 1952 GEN. ABSTRACT, CONDITION blossomry 1901 blossom n. OE, blossom v. OE COLLECTIVE bootery 1920 boot n3 c1325, boot v3 1468 LOCATION brigandry 1909 brigand n. ? a1400, brigand v. 1886 ACTION, COL- LECTIVE cartoonery 1902 cartoon n. a1684, cartoon v. 1884 ACTION, GEN. ABSTRACT circuitry 1946 circuit n. 1746, circuit v. 1895 COLLECTIVE, GEN. ABSTRACT cokery 1923 coke n1 1669, coke v1 1804 OBJECT componentry 1959 component n. 1563, component adj. 1664 COLLECTIVE computery 1960 compute n. 1531, compute v. 1579, computer n. 1613 COLLECTIVE, ACTION 221 comstockery 1905 comstocker n. 1905 ACTION condensery 1909 condense v. 1477, condense adj. 1610, condenser n. 1868 LOCATION corsetry 1904 corset n. 1795 COLLECTIVE crookery 1927 crook n. 1879, crook v1 a1340 CONDITION, ACTION, GEN. ABSTRACT cryosurgery 1962 cryosurgical adj. 1962 ACTION do-goodery 1959 do-good n. 1654, dogooder n. 1901 ACTION eatery 1901 eat n. OE, eat v. OE LOCATION gannetry 1913 gannet n. OE LOCATION geekery 1947 geek n. 1876, geek v. 1946 AMOUNT, AC- TION, CONDI- TION god-wottery 1939 God wot phrase c1300 ACTION, CON- DITION hot doggery 1923 hot dog n. 1884, hot dog n. 1894, hot dog v. 1959 LOCATION, AC- TION lappery 1937 lap n3 1867, lap v2 1847 ACTION madcappery 1905 madcap n. 1589 ACTION martial artistry 1990 martial artist n. 1970 CONDITION mascotry 1900 mascot n. 1881 CONDITION, ACTION micorcircuitry 1959 microcircuit n. 1959 COLLECTIVE, GEN. ABSTRACT micormachinery 1986 micromachine n. 1988 COLLECTIVE microbrewery 1982 microbrewer n. 1983 LOCATION microsurgery 1927 microsurgeon n. 1959 ACTION milksoppery 1925 milk sop n. C1390 ACTION, CON- DITION 222 mopery 1907 mope v. 1568, mope n. 1693, moper n. 1721 ACTION Mormonry 1930 Mormon n. 1833, Mormon adj. 1833 COLLECTIVE, AMOUNT, GEN. ABSTRACT mulletry 1902 mullet n1 1393 LOCATION Mumbojumbery 1923 mumbo-jumbo n. 1738 ACTION mumpery 1913 mump v2 1685, mumper n. a1652 ACTION mystagoguery 1927 mystagogue n. C1540 ACTION neoslavery 1958 neoslave n. 1970 GEN. ABSTRACT neurochemistry 1945 neurochemist n. 1957 GEN. ABSTRACT neurosurgery 1904 neurosurgeon n. 1925 ACTION nitery 1934 nite n2 1928 LOCATION nitwittery 1931 nitwit n. 1914, nitwit adj. 1928 CONDITION noshery 1952 nosh n. 1873, nosh v. 1892, nosher n. 1917 LOCATION, OB- JECT nothingbuttery 1961 nothing-but adj. 1937, nothing-but n. 1953 GEN. ABSTRACT outbackery 1961 outback adj. 1893, outback n. 1904 ACTION, CON- DITION pargetry 1908 parget n. c1400, parget v. a1398 OBJECT parish pumpery 1962 parish pump n. 1840, parish pump adj. 1923, parish pumper n. 1963 CONDITION, COLLECTIVE pen-andinkery c1909 pen and ink n. c1500, pen and ink adj. 1672, pen and ink v. 1801 ACTION 223 perchery 1985 perch n1 c1385, perch v1 a1425 COLLECTIVE Peter-Pannery 1960 Peter Pan n. 1908 ACTION, CON- DITION platypussary 1938 platypus n. 1799 LOCATION plushery 1951 plush adj. 1890, plush n1 1590 LOCATION press agentry 1907 press agent n. 1814, press agent v. 1901 ACTION, CON- DITION pressure cookery 1918 pressure cook v. 1922 ACTION Proustery 1928 Proustian n. 1919, Proustian adj. 1925 CONDITION pseudery 1972 pseud n. 1954, pseud adj. 1962 CONDITION, ACTION psychosurgery 1936 psychosurgeon n. 1945 ACTION pudibundery 1910 pudibund adj. 1542 ACTION, CON- DITION puffinry 1954 puffin n1 1337 LOCATION, COLLECTIVE punditry 1926 pundit n. 1661, pundit v. 1940 ACTION, CON- DITION purchasery 1927 purchase v. c1300, purchase n. c1325, purchaser n. c1384 ACTION purple patchery 1933 purple patch n. ? 1704 ACTION, CON- DITION quippery 1933 quip n. 1532, quip v. 1542, quipper n. 1589 ACTION, AMOUNT radiochemistry 1904 radiochemist n. 1917 GEN. ABSTRACT radiosurgery 1929 radiosurgical adj. 1928 ACTION Reichschancellery 1932 Reichschancellor n. 1759 LOCATION reptillery 1976 reptile n. a1393 LOCATION robotry 1924 robot n2 1922 ACTION, CON- 224 DITION Scottishry 1958 Scottish adj. OE, Scottish n. OE CONDITION smart-aleckry 1918 smart alec n. 1865,smar alec adj. 1877 ACTION, CON- DITION smuggery 1928 smug adj. 1551, smug n2 1891 CONDITION, ACTION snackery 1936 snack n2 1757, snack v. 1807 LOCATION snoopery 1935 snoop v. 1832, snooper n. 1889, snoop n. 1891 ACTION stokery 1901 stoker n. 1660, stoke v2 1735 LOCATION summitry 1958 summit n. 1950, summit v.2 1972 ACTION targetry 1977 target n. a1400-50, target v. 1611 ACTION tartanry 1976 tartan n1 a1500, tartan v. 1881 COLLECTIVE, CONDITION teenagery 1960 teenager n. 1941, teenage adj. 1921, teenage n2 1934 CONDITION twiggery 1909 twig n1 OE COLLECTIVE wagery 1917 wage n. 1338, wage v. c1330 GEN. ABSTRACT, COLLECTIVE waivery 1903 waive v1 1297, waive n. 1544 ACTION weedery 1908 weed n2 OE COLLECTIVE wormery 1952 worm n. OE LOCATION, OB- JECT 225 Appendix E: -age Derivatives (BNC) Derivative Adjusted Token Frequency Readings (Percentage of Attested Tokens) abusage 1 ACTION (100) acreage 110 LOCATION (6), AMOUNT (94) advantage 7121 GEN. ABSTRACT (100) ampage 1 AMOUNT (100) anchorage 107 OBJECT (5), LOCATION (57), COLLEC- TIVE (3), GEN. ABSTRACT (1), ACTION (32), CONDITION (10) anecdotage 6 AMOUNT (50), CONDITION (50) appendage 48 OBJECT (33), PERSON (27), LOCATION (14.5), COLLECTIVE (10), GEN. AB- STRACT (14.5) arbitrage 258 ACTION (100) assemblage 209 COLLECTIVE (96), AMOUNT (4) average 3461 AMOUNT (100) badinage 13 ACTION (100) baggage 451 PERSON (4), COLLECTIVE (81), AMOUNT (15) bailliage 1 LOCATION (100) bandage 213 OBJECT (100) baronage 18 COLLECTIVE (100) baronetage 4 COLLECTIVE (100) barrage 499 OBJECT (38), ACTION (61.5) barrelage 7 AMOUNT (100) beerage 1 COLLECTIVE (100) beguinage 4 LOCATION (100) beverage 96 OBJECT (100) blockage 146 OBJECT (17), ACTION (6), CONDITION (83) bondage 137 ACTION (29), CONDITION (71) bossage 1 ACTION (100) breakage 250 OBJECT (1), COLLECTIVE (4), AMOUNT (10), ACTION (50), CONDITION (61) 226 bricolage 3 COLLECTIVE (33), AMOUNT (66) brigandage 9 ACTION (100) briquetage 2 COLLECTIVE (100) brokage 1 ACTION (100) brokerage 70 COLLECTIVE (7), AMOUNT (7), ACTION (86), CONDITION (3) burgage 2 TENURE (100) carriage 1803 OBJECT (84), ACTION (15), CONDITION (1) cartage 5 CHARGE (20), ACTION (80) cellarage 1 LOCATION (100) chaperonage 5 ACTION (100) cheminage 11 CHARGE (100) chummage 1 ACTION (100) clearage 1 ACTION (100) cleavage 349 OBJECT (12), ACTION (82), CONDITION (6) clientage 2 COLLECTIVE (50), CONDITION (50) coinage 296 OBJECT (3), COLLECTIVE (89), GEN. ABSTRACT (4), ACTION (3) colportage 2 ACTION (100) commonage 1 RIGHT (100) compangnonnage 2 COLLECTIVE (100) computerusage 1 ACTION (100) concubinage 4 ACTION (100) consulage 1 CHARGE (100) cooperage 6 LOCATION (50), ACTION (50) cordage 6 COLLECTIVE (100) corkage 2 CHARGE (100) corsage 16 OBJECT (100) costage 1 CHARGE (100) cottage 2432 LOCATION (100) cousinage 3 COLLECTIVE (66), CONDITION (33) coverage 2608 AMOUNT (64.5), ACTION (35.5) cranage 2 ACTION (100) cribbage 5 ACTION (100) crimpage 1 ACTION (100) curbage 3 LOCATION (100) 227 curettage 17 ACTION (100) curtilage 22 LOCATION (100) decolletage 8 OBJECT (100) delinkage 3 CONDITION (100) demurrage 2 CHARGE (100) disadvantage 1119 GEN. ABSTRACT (100) dosage 122 COLLECTIVE (1), AMOUNT (81), AC- TION (17), CONDITION (1) dotage 15 CONDITION (100) drainage 1115 OBJECT (11), COLLECTIVE (11), ACTION (77.5), dressage 84 ACTION (100) driveage 10 LOCATION (100) dry-storage 1 ACTION (100) ecotage 1 ACTION (100) embassage 8 COLLECTIVE (12.5), GEN. ABSTRACT (50), ACTION (37.5) empennage 3 COLLECTIVE (100) entourage 205 COLLECTIVE (100) equipage 6 COLLECTIVE (100) espionage 239 ACTION (100) ex-hostage 3 PERSON (100) finnage 19 COLLECTIVE (100) flowage 1 ACTION (100) foggage 1 ACTION (100) foliage 706 COLLECTIVE (100) footage 209 OBJECT (94), AMOUNT (7) fosterage 1 CONDITION (100) fraughage 1 COLLECTIVE (100) frontage 145 OBJECT (36.5), LOCATION (59), AMOUNT (5) frottage 4 OBJECT (25), ACTION (75) fruitage 1 COLLECTIVE (100) fuselage 244 OBJECT (100) gallonage 13 AMOUNT (100) gaufrage 2 OBJECT (50), ACTION (50) grindage 2 OBJECT (50), ACTION (50) hand-baggage 1 COLLECTIVE (100) hangarage 3 ACTION (100) 228 harbourage 1 LOCATION (100) haulage 246 COLLECTIVE (1), ACTION (99) headage 14 AMOUNT (100) hectarage 11 AMOUNT (100) herbage 40 COLLECTIVE (87.5), CHARGE (10), RIGHT (2.5) heritage 1182 COLLECTIVE (62), GEN. ABSTRACT (11.5), AMOUNT (25), ACTION (2) hermitage 14 LOCATION (100) hidage 2 CHARGE (100) hostage 566 PERSON (77), CONDITION (23) ill-usage 1 ACTION (100) interlinkage 1 CONDITION (100) intermarriage 39 ACTION (95), CONDITION (5) land-usage 1 ACTION (100) leafage 8 COLLECTIVE (100) leakage 234 OBJECT (5), AMOUNT (8), ACTION (87) leverage 239 COLLECTIVE (1), GEN. ABSTRACT (89), ACTION (10) lineage 330 COLLECTIVE (92), GEN. ABSTRACT (7), AMOUNT (7) linkage 337 OBJECT (15), ACTION (18), CONDITION (67) lockage 8 AMOUNT (12.5), ACTION (87.5) low-wattage 1 AMOUNT (100) luggage 585 COLLECTIVE (100) maquillage 1 OBJECT (100) marriage 8314 GEN. ABSTRACT (2), ACTION (38), CONDITION (60.5) massage 433 ACTION (100) matrilineage 3 COLLECTIVE (100) megatonnage 1 AMOUNT (100) menage 12 LOCATION (42), COLLECTIVE (17), CONDITION (42) metreage 3 AMOUNT (100) mileage 401 GEN. ABSTRACT (19), AMOUNT (80.5) mintage 1 ACTION (100) mirage 71 OBJECT (54), GEN. ABSTRACT (46) miscarriage 159 ACTION (100) 229 montage 46 OBJECT (35), COLLECTIVE (15), ACTION (50) on-carriage 1 ACTION (100) orphanage 177 LOCATION (100) outage 15 AMOUNT (7), CONDITION (93) overdosage 8 ACTION (100) package 5606 OBJECT (7), COLLECTIVE (4), GEN. AB- STRACT (2), AMOUNT (87) parage 1 CONDITION (100) parentage 99 COLLECTIVE (83), GEN. ABSTRACT (15), ACTION (1), CONDITION (1) parsonage 46 LOCATION (98), COLLECTIVE (2) partage 1 ACTION (100) passage 3852 OBJECT (2), LOCATION (12), GEN. AB- STRACT (40), ACTION (45.5) pasturage 16 OBJECT (19), LOCATION (31), RIGHT (12.5), ACTION (37.5) patronage 848 RIGHT (6), ACTION (94) peerage 187 OBJECT (15), COLLECTIVE (35), POSI- TION (71) pelage 1 COLLECTIVE (100) peonage 4 GEN. ABSTRACT (50), ACTION (25), CONDITION (25) percentage 2756 AMOUNT (100) persiflage 4 ACTION (100) personage 51 PERSON (100) pesage 3 CHARGE (100) photoreportage 2 OBJECT (50), ACTION (50) pilferage 14 ACTION (100) pilgrimage 448 ACTION (100) pillage 27 ACTION (100) pilotage 18 CHARGE (11), ACTION (89) plumage 310 COLLECTIVE (100) pointage 1 AMOUNT (100) portage 18 LOCATION (67), ACTION (44) porterage 5 LOCATION (20), ACTION (80) postage 578 OBJECT (1), AMOUNT (59), ACTION (40) pottage 17 OBJECT (88), AMOUNT (12) poundage 60 AMOUNT (33), ACTION (66) 230 primage 1 CHARGE (100) pupillage 47 CONDITION (100) quackage 1 ACTION (100) quadrillage 1 COLLECTIVE (100) ramblage 1 ACTION (100) rampage 89 ACTION (100) rapportage 1 ACTION (100) re-assemblage 2 ACTION (100) recoinage 8 ACTION (100) remarriage 120 ACTION (98), CONDITION (6) remuage 8 ACTION (100) repackage 1 ACTION (100) reportage 41 OBJECT (22), GEN. ABSTRACT (7), AC- TION (71) retroussage 2 ACTION (100) roughage 41 COLLECTIVE (100) rummage 12 COLLECTIVE (8), ACTION (92) sabotage 167 ACTION (100) salvage 136 COLLECTIVE (26), ACTION (80), scottage 1 CHARGE (100) scrappage 1 ACTION (100) scrimmage 2 ACTION (100) scrummage 56 ACTION (100) seepage 41 OBJECT (19.5), ACTION (95) seigniorage 2 CHARGE (100) sewage 833 OBJECT (93), COLLECTIVE (8), ACTION (8) sewerage 256 OBJECT (52), LOCATION (1), COLLEC- TIVE (34), ACTION (86) shortage 1454 GEN. ABSTRACT (100) shrinkage 117 COLLECTIVE (2), AMOUNT (10), AC- TION (90), CONDITION (2) signage 10 COLLECTIVE (100) silage 104 OBJECT (99), ACTION (1) sinkage 3 AMOUNT (33), ACTION (66) siphonage 10 ACTION (100) slippage 64 AMOUNT (34), ACTION (66) soakage 3 AMOUNT (33), ACTION (66) 231 socage 5 TENURE (100) spillage 101 OBJECT (26), ACTION (75) spingleage 1 AMOUNT (100) splintage 1 ACTION (100) spoilage 12 OBJECT (25), ACTION (75) stackage 1 ACTION (100) stallage 1 CHARGE (100) standage 1 CHARGE (100) steam-haulage 1 ACTION (100) steerage 26 LOCATION (85), ACTION (15) stillage 1 OBJECT (100) stockbrokerage 2 COLLECTIVE (100) stoppage 126 ACTION (98), CONDITION (2) storage 3046 OBJECT (1), LOCATION (6), AMOUNT (2), ACTION (85.5), CONDITION (6) stowage 30 LOCATION (50), ACTION (50) stumpage 1 CHARGE (100) sub-lineage 1 COLLECTIVE (100) subpackage 2 AMOUNT (100) sufferage 1 GEN. ABSTRACT (100) sullage 1 OBJECT (100) surplusage 2 AMOUNT (100) tallage 5 CHARGE (100) tankage 3 COLLECTIVE (66), ACTION (33) tentage 5 COLLECTIVE (100) tillage 24 LOCATION (8), ACTION (79), CONDI- TION (12.5) tonnage 88 AMOUNT (93), CHARGE (7) towage 12 CHARGE (8), ACTION (92) trellisage 1 OBJECT (100) triage 15 ACTION (100) tutelage 67 ACTION (97), CONDITION (3) umbrage 24 CONDITION (100) umpirage 2 ACTION (100) un-marriage 1 CONDITION (100) under-usage 1 ACTION (100) underdrainage 2 ACTION (100) 232 usage 1162 AMOUNT (1), ACTION (100) vantage 219 LOCATION (79), GEN. ABSTRACT (1), POSITION (19) vassalage 12 ACTION (8), CONDITION (92) vicarage 184 LOCATION (97), COLLECTIVE (2), POSI- TION (1) victimage 1 CONDITION (100) village 10291 LOCATION (97), COLLECTIVE (3) villeinage 3 CONDITION (100) vintage 517 OBJECT (15.5), GEN. ABSTRACT (78.5), ACTION (6) voltage 931 GEN. ABSTRACT (14), AMOUNT (86) waftage 1 ACTION (100) warpage 2 OBJECT (50), ACTION (50) wastage 282 COLLECTIVE (26), AMOUNT (17), AC- TION (93) water-storage 1 ACTION (100) water-usage 1 ACTION (100) waterfrontage 1 AMOUNT (100) wattage 41 AMOUNT (100) wharfage 3 LOCATION (66), ACTION (33) wordage 4 AMOUNT (100) wreckage 423 OBJECT (29), COLLECTIVE (64), AMOUNT (6), ACTION (1) yardage 27 AMOUNT (100) 233 Appendix F: -ery Derivatives (BNC) Derivative Adjusted Token Frequency Readings (Percentage of Attested Tokens) adultery 274 ACTION (100) adversary 159 OBJECT (3), PERSON (88), COLLECTIVE (7.5), GEN. ABSTRACT (1) advisery 4 ACTION (100) affrontery 2 ACTION (100) almonry 3 LOCATION (100) ancestry 315 COLLECTIVE (60), GEN. ABSTRACT (39) ancientry 1 COLLECTIVE (100) anecdotery 1 AMOUNT (100) anglo-jewry 2 COLLECTIVE (100) archdeaconry 15 LOCATION (67), POSITION (33) archery 105 COLLECTIVE (4), ACTION (96) armoury 235 LOCATION (24.5), COLLECTIVE (22.5), ACTION (53) artillery 699 COLLECTIVE (100) artisanry 1 ACTION (100) artistry 94 COLLECTIVE (2), ACTION (6), CONDI- TION (91) bailliary 4 LOCATION (100) bakery 270 LOCATION (99), COLLECTIVE (1) balladry 5 COLLECTIVE (80), ACTION (20) banditry 23 ACTION (100) baptistery 37 OBJECT (3), LOCATION (97) bardolatry 4 ACTION (100) basketry 3 COLLECTIVE (67), ACTION (33) battery 1254 OBJECT (71), COLLECTIVE (15), AMOUNT (8), ACTION (6) bawdry 6 ACTION (100) beanery 2 LOCATION (100) 234 beat-poetry 1 ACTION (100) beggary 1 CONDITION (100) begrudgery 1 ACTION (100) bigotry 96 ACTION (20), CONDITION (80) bindery 15 LOCATION (80), COLLECTIVE (20) biochemistry 243 GEN. ABSTRACT (100) biogeochemistry 1 GEN. ABSTRACT (100) blazonry 1 OBJECT (100) bluery 1 ACTION (100) bodgery 1 OBJECT (100) boffinry 3 COLLECTIVE (67), ACTION (33) bookbindery 1 ACTION (100) boozery 1 LOCATION (100) bowery 1 LOCATION (100) bravery 343 ACTION (26.5), CONDITION (73.5) brewery 750 LOCATION (87), COLLECTIVE (11), AC- TION (2) bribery 222 ACTION (100) bric-bracery 1 COLLECTIVE (100) brigandry 3 ACTION (100) budgetry 2 ACTION (100) buffoonery 5 ACTION (100) buggery 80 LOCATION (4), GEN. ABSTRACT (4), ACTION (92.5) butchery 47 LOCATION (15), ACTION (85) butlery 1 LOCATION (100) buttery 25 LOCATION (100) cabinetry 4 COLLECTIVE (75), ACTION25) cajolery 7 ACTION (100) campery 1 ACTION (100) cannery 5 LOCATION (80), COLLECTIVE (20) canonry 18 GEN. ABSTRACT (6), POSITION (94) carpentry 71 COLLECTIVE (10), ACTION (90) 235 carvery 25 OBJECT (52), LOCATION (48) casuistry 15 ACTION (93), CONDITION (7) cattery 10 LOCATION (100) cavalry 519 COLLECTIVE (100) chancellery 43 LOCATION (25.5), COLLECTIVE 72), POSITION (2) chandlery 8 LOCATION (62.5), COLLECTIVE (37.5) chantry 52 OBJECT (40), COLLECTIVE (8), GEN. ABSTRACT (23), ACTION (29) chapelry 8 COLLECTIVE (75), ACTION25) chemistry 2019 GEN. ABSTRACT (99), ACTION (1) chestnutery 1 CONDITION (100) chicanery 20 ACTION (100) chirurgery 1 ACTION (100) chivalry 146 COLLECTIVE (13), ACTION (3), CONDI- TION (83) choppery 1 ACTION (100) circuitry 131 COLLECTIVE (95), AMOUNT (3), AC- TION (2) citizenry 40 COLLECTIVE (100) colliery 389 LOCATION (96), LOCATION (1), AC- TION (3) commandery 2 LOCATION (100) componentry 2 COLLECTIVE (100) con-servatory 2 LOCATION (100) confectionery 106 OBJECT (1), COLLECTIVE (99) controversery 1 ACTION (100) cookery 473 ACTION (100) coparcenary 2 CONDITION (100) coquetry 7 ACTION (100) corpsebumlickery 1 ACTION (100) corsetry 3 LOCATION (33), COLLECTIVE (67) creamery 16 LOCATION (69), COLLECTIVE (31) crockery 135 COLLECTIVE (100) crookery 5 ACTION (100) 236 cuckoldry 2 ACTION (50), CONDITION (50) cuppery 1 COLLECTIVE (100) cutlery 303 COLLECTIVE (100) cytochemistry 2 GEN. ABSTRACT (50), ACTION (50) dairy 871 OBJECT (8.5), LOCATION (40.5), COL- LECTIVE (51) daredevilry 2 ACTION (100) day-nursery 2 LOCATION (100) deanery 68 LOCATION (19), COLLECTIVE (72), PO- SITION (9) debauchery 60 ACTION (83), CONDITION (17) debunkery 1 ACTION (100) demagoguery 4 ACTION (100) dentistry 60 COLLECTIVE (5), ACTION (95) devilry 13 COLLECTIVE (8), ACTION (69), CONDI- TION (23) distillery 78 LOCATION (97.5), COLLECTIVE (4) donnery 2 ACTION (50), CONDITION (50) dowry 143 COLLECTIVE (100) drapery 67 LOCATION (7), COLLECTIVE (82), AMOUNT (1.5), ACTION (9) dream-imagery 2 AMOUNT (50), ACTION (50) drollery 2 ACTION (100) dromedary 8 OBJECT (100) drudgery 93 ACTION (90), CONDITION (10) duckery 1 LOCATION (100) duncery 2 ACTION (50), CONDITION (50) dybbukry 3 COLLECTIVE (33), ACTION (67) ealdormanry 1 POSITION (100) eatery 2 LOCATION (100) effrontery 29 ACTION (24), CONDITION (76) ego-tossery 1 ACTION (100) electrochemistry 19 GEN. ABSTRACT (63), ACTION (37) electronickery 1 ACTION (100) 237 embroidery 317 OBJECT (52), LOCATION (1), ACTION (47) englishry 1 CONDITION (100) enslavery 1 ACTION (100) eunuchry 2 ACTION (50), CONDITION (50) fairy 1014 PERSON (13), LOCATION (28.5), LOCA- TION (58.5) fakery 3 ACTION (100) falconry 45 ACTION (100) farriery 35 ACTION (100) fernery 1 LOCATION (100) finery 81 COLLECTIVE (99), CONDITION (1) fish-andchippery 1 LOCATION (100) fishery 193 LOCATION (41.5), COLLECTIVE (14), ACTION (44.5) flackery 4 COLLECTIVE (25), ACTION (75) flattery 146 ACTION (100) floristry 22 COLLECTIVE (14), ACTION (86) folkery 2 LOCATION (100) foodery 1 LOCATION (100) foolery 2 ACTION (100) foppery 1 COLLECTIVE (100) forestry 1148 COLLECTIVE (21), ACTION (79) forgery 196 OBJECT (37), GEN. ABSTRACT (1), AC- TION (62) foundry 163 LOCATION (83), LOCATION (9), AC- TION (8) freemasonry 69 COLLECTIVE (10), GEN. ABSTRACT (68), ACTION (19), CONDITION (3) friary 24 LOCATION (92), COLLECTIVE (8) frippery 5 COLLECTIVE (80), AMOUNT (20) fuckery 1 ACTION (100) gadgetry 46 COLLECTIVE (96), ACTION (4) gallantry 132 COLLECTIVE (1), ACTION (58), CONDI- TION (41) gannetry 3 LOCATION (100) 238 garmentry 1 COLLECTIVE (100) genery 1 GEN. ABSTRACT (100) gentry 675 COLLECTIVE (96), POSITION (4) geochemistry 66 GEN. ABSTRACT (94), ACTION (6) geometry 469 GEN. ABSTRACT (97), ACTION (3) gimmickry 14 ACTION (50), CONDITION (50) goldsmithery 1 ACTION (100) granary 35 LOCATION (94), COLLECTIVE (6) greenery 168 LOCATION (20), COLLECTIVE (75), AC- TION (5) greengrocery 9 LOCATION (11), COLLECTIVE (78), AC- TION (11) greyfriary 1 LOCATION (100) grocery 185 LOCATION (2), COLLECTIVE (98) gunnery 47 COLLECTIVE (36), ACTION (64) haberdashery 21 LOCATION (5), COLLECTIVE (95) hackery 1 ACTION (100) handcuffery 2 ACTION (50), CONDITION (50) harlotry 4 ACTION (50), CONDITION (50) hatchery 17 LOCATION (94), ACTION (6) heraldry 73 OBJECT (8), COLLECTIVE (55), ACTION (37) heronry 8 OBJECT (12.5), LOCATION (75), COL- LECTIVE (12.5) hoc-ery 1 ACTION (100) hosiery 83 LOCATION (2.5), COLLECTIVE (96), ACTION (1) hostelry 66 LOCATION (100) housewifery 28 ACTION (68), CONDITION (28.5), POSI- TION (3.5) husbandry 184 COLLECTIVE (1), ACTION (99) idolatry 59 ACTION (95), CONDITION (5) imagery 642 COLLECTIVE (39), AMOUNT (54), AC- TION (7) immunochemistry 1 GEN. ABSTRACT (100) infirmary 509 LOCATION (99), COLLECTIVE (1) 239 ironmongery 27 COLLECTIVE (89), ACTION (11) jackassery 1 ACTION (100) japanesery 1 CONDITION (100) japery 2 ACTION (100) jewellery 1220 COLLECTIVE (100) jewry 83 LOCATION (1), COLLECTIVE (66), GEN. ABSTRACT (32.5) jobbery 6 ACTION (100) joinery 63 COLLECTIVE (62), ACTION (38) jugglery 2 ACTION (100) justiciary 6 PERSON (17), GEN. ABSTRACT (83) knavery 4 ACTION (100) knight-errantry 2 ACTION (100) lamasery 2 LOCATION (100) laundry 583 LOCATION (28), COLLECTIVE (51), AC- TION (21) lechery 19 ACTION (37), CONDITION (63) left-wingery 1 ACTION (100) lottery 346 ACTION (100) love-poetry 2 AMOUNT (100) luthiery 1 ACTION (100) macabrery 2 ACTION (50), CONDITION (50) machinery 2345 COLLECTIVE (81), AMOUNT (19) manufactory 28 LOCATION (89), COLLECTIVE (7), AC- TION (3.5) masonry 328 COLLECTIVE (97), ACTION (3) mastery 445 ACTION (12), CONDITION (88) mercery 4 LOCATION (75), COLLECTIVE (25) merchantry 1 COLLECTIVE (100) microcircuitry 4 COLLECTIVE (50), GEN. ABSTRACT (25), ACTION (25) microsurgery 17 ACTION (100) midwifery 106 COLLECTIVE (7.5), GEN. ABSTRACT (1), ACTION (90.5), POSITION (1) millinery 15 COLLECTIVE (73), ACTION (27) 240 mimicry 115 GEN. ABSTRACT (4), ACTION (96) ministry 4813 COLLECTIVE (85), ACTION (7), CONDI- TION (5), POSITION (3) misery 1213 PERSON (1), CONDITION (99) missilry 2 COLLECTIVE (100) mobbery 1 ACTION (100) mockery 316 ACTION (100) monastery 614 LOCATION (100) mugwumpery 1 ACTION (100) mummery 5 ACTION (100) musketry 28 COLLECTIVE (29), ACTION (71) mystery 2086 GEN. ABSTRACT (95), ACTION (5) nailery 1 LOCATION (100) napery 2 COLLECTIVE (100) neurochemistry 2 GEN. ABSTRACT (100) neurosurgery 23 ACTION (100) nicknackery 1 LOCATION (100) nobbery 1 COLLECTIVE (100) noshery 1 LOCATION (100) notary 40 PERSON (100) nunnery 64 LOCATION (87.5), COLLECTIVE (12.5) nursery 2016 LOCATION (82), ACTION (18) orangery 24 LOCATION (100) oratory 122 LOCATION (18), COLLECTIVE (5), AC- TION (77) outlawry 9 ACTION (56), CONDITION (44) pageantry 41 COLLECTIVE (17), ACTION (80.5), CONDITION (2.5) palmistry 7 ACTION (100) pantomimicry 1 ACTION (100) pantry 147 OBJECT (1), LOCATION (99) papistry 2 COLLECTIVE (50), ACTION (50) parquetry 6 OBJECT (100) pastry 483 OBJECT (100) 241 peacockery 2 ACTION (100) peasantry 512 COLLECTIVE (100) pedantry 31 ACTION (29), CONDITION (71) penitentiary 22 PERSON (5), LOCATION (95) perchery 2 COLLECTIVE (100) perfumery 40 LOCATION (12.5), COLLECTIVE (27.5), ACTION (60) pheasantry 3 LOCATION (100) photochemistry 15 GEN. ABSTRACT (47), ACTION (53) photogrammetry 4 ACTION (100) phrontistery 1 LOCATION (100) piggery 10 LOCATION (80), ACTION (20) pleasantry 11 GEN. ABSTRACT (82), ACTION (9), CONDITION (9) plumery 1 COLLECTIVE (100) podsnappery 2 ACTION (50), CONDITION (50) poetry 2617 GEN. ABSTRACT (72.5), AMOUNT (27.5) popery 63 GEN. ABSTRACT (1.5), AMOUNT (1.5), ACTION (92), CONDITION (5) pottery 884 LOCATION (5), COLLECTIVE (82), AC- TION (13) poultry 369 LOCATION (1), COLLECTIVE (98), AC- TION (1) presbytery 123 LOCATION (28), COLLECTIVE (71), AC- TION (1) priggery 2 ACTION (50), CONDITION (50) priory 355 LOCATION (92), COLLECTIVE (8) prose-poetry 2 GEN. ABSTRACT (50), AMOUNT (50) prudery 26 ACTION (19), CONDITION (81) psychiatry 230 GEN. ABSTRACT (83), ACTION (17) psychometry 12 GEN. ABSTRACT (25), ACTION (75) psychosurgery 2 ACTION (100) punditry 3 ACTION (67), CONDITION (33) puppetry 17 COLLECTIVE (6), ACTION (94) quackery 10 ACTION (100) 242 quixotry 2 ACTION (50), CONDITION (50) rabbitry 2 COLLECTIVE (100) radiochemistry GEN. ABSTRACT (100) raillery 7 ACTION (100) re-upholstery 1 ACTION (100) refinery 168 LOCATION (96), ACTION (4) revelry 32 ACTION (100) revery 4 CONDITION (100) ribaldry 10 ACTION (100) right-wingery 4 ACTION (50), CONDITION (50) rivalry 532 ACTION (17.5), CONDITION (82.5) robbery 753 ACTION (100) rockery 35 LOCATION (97), COLLECTIVE (3) rocketry 8 COLLECTIVE (37.5), ACTION (62.5) roguery 5 ACTION (80), CONDITION (20) rookery 4 LOCATION (100) ropery 2 LOCATION (100) rosary 83 OBJECT (57), GEN. ABSTRACT (1), AC- TION (42) routery 1 ACTION (100) rudery 1 ACTION (100) saddlery 36 LOCATION (8), COLLECTIVE (64), AC- TION (28) savagery 179 COLLECTIVE (1), ACTION (28), CONDI- TION (70) scallywaggery 1 ACTION (100) scapulary 1 OBJECT (100) scavengery 1 ACTION (100) scenery 739 COLLECTIVE (100) scullery 172 LOCATION (100) seigniory 5 LOCATION (80), CONDITION (20) self-mockery 28 ACTION (100) sergeantry 1 POSITION (100) servery 4 LOCATION (100) 243 sextry 1 LOCATION (100) show-bizzery 2 COLLECTIVE (50), ACTION (50) shrubbery 54 LOCATION (52), COLLECTIVE (48) siegery 1 ACTION (100) skin-flintery 2 ACTION (50), CONDITION (50) skulduggery 27 ACTION (100) slavery 573 GEN. ABSTRACT (1), ACTION (66), CONDITION (34) smokery 3 LOCATION (100) snobbery 119 ACTION (4), CONDITION (96) socketry 1 COLLECTIVE (100) soldiery 35 COLLECTIVE (88.5), ACTION (11.5) songsmithery 1 CONDITION (100) sophistry 12 ACTION (100) sorcery 101 ACTION (100) spicery 1 COLLECTIVE (100) spinnery 6 LOCATION (100) spivvery 1 ACTION (100) stationery 313 COLLECTIVE (100) stereochemistry 7 GEN. ABSTRACT (100) stewartry 5 LOCATION (100) stitchery 2 COLLECTIVE (100) stonemasonry 3 ACTION (100) stuffery 2 AMOUNT (50), ACTION (50) superministry 5 COLLECTIVE (100) surgery 2673 LOCATION (19), COLLECTIVE (2), AC- TION (79) swannery 8 LOCATION (100) syllabary 2 AMOUNT (100) tanistry 2 GEN. ABSTRACT (100) tannery 37 LOCATION (76), COLLECTIVE (24) tapestry 247 OBJECT (73), GEN. ABSTRACT (18), AC- TION (9) targetry 1 ACTION (100) 244 tartanry 1 COLLECTIVE (100) tastery 1 LOCATION (100) tenantry 9 COLLECTIVE (100) thermochemistry 3 GEN. ABSTRACT (100) thievery 2 ACTION (100) thuggery 29 ACTION (100) tilery 1 LOCATION (100) tinkery 1 COLLECTIVE (100) toiletry 8 COLLECTIVE (100) tomfoolery 11 ACTION (100) tracery 72 OBJECT (99), GEN. ABSTRACT (1) treachery 196 ACTION (100) trickery 71 ACTION (100) trinketry 2 COLLECTIVE (100) trumpery 2 COLLECTIVE (100) turnery 4 LOCATION (25), COLLECTIVE (50), AC- TION (25) twiggery 1 COLLECTIVE (100) ungallantry 1 ACTION (100) unpleasantry 1 ACTION (100) upholstery 244 OBJECT (53), COLLECTIVE (30), ACTION (17) vestry 126 LOCATION (69), COLLECTIVE (28), AC- TION (4) vintry 3 LOCATION (100) waafery 3 LOCATION (100) waggery 1 ACTION (100) watergatery 1 ACTION (100) weaponry 186 COLLECTIVE (99), GEN. ABSTRACT (1) weirdery 2 ACTION (50), CONDITION (50) whammery 1 ACTION (100) whiggery 13 ACTION (100) whippery 1 ACTION (100) whitefriary 3 LOCATION (100) 245 whizzery 1 ACTION (100) widgetry 1 COLLECTIVE (100) winery 28 LOCATION (82), COLLECTIVE (18) witchery 2 ACTION (100) wizardry 72 COLLECTIVE (7), ACTION (92), CONDI- TION (1) wormery 9 OBJECT (100) yarnery 1 ACTION (100) yeomanry 47 COLLECTIVE (100) zealotry 3 ACTION (67), CONDITION (33) Language in Performance Edited by Rainer Schulze Bisher sind erschienen: Frühere Bände finden Sie unter: http: / / www.narr-shop.de/ reihen/ l/ languagein-performance-lip.html 17 Roswitha Fischer Lexical Change in Present-Day English A Corpus-Based Study of the Motivation, Institutionalization, and Productivity of Creative Neologisms 1998, X, 209 pages €[D] 39,- ISBN 978-3-8233-4940-2 18 Sylvia Kalina Strategische Prozesse beim Dolmetschen Theoretische Grundlagen, empirische Fallstudien, didaktische Konsequenzen 1998, 304 pages €[D] 44,- ISBN 978-3-8233-4941-9 19 Willis Edmondson Twelve Lectures on Second Language Acquisition Foreign Language Teaching and Learning Perspectives 1998, VIII, 287 pages €[D] 48,- ISBN 978-3-8233-4942-6 20 Andrea Sand Linguistic Variation in Jamaica A Corpus-Based Study of Radio and Newspaper Usage 1999, 197 pages €[D] 39,- ISBN 978-3-8233-4943-3 21 Eija Ventola (ed.) Discourse and Community Doing Functional Linguistics 2000, 397 pages €[D] 54,- ISBN 978-3-8233-4944-0 22 Christopher J. Gledhill Collocations in Science Writing 2000, X, 268 pages €[D] 48,- ISBN 978-3-8233-4945-7 23 Ilka MIndt Intonation im Lancaster/ IBM Spoken English Corpus Falls and fall-rises , Sprecherwechsel, paratones , declination 2001, VIII, 305 pages €[D] 43,- ISBN 978-3-8233-4947-1 24 Beate Hampe Superlative Verbs A corpus-based study of semantic redundancy in English verb-particle constructions 2002, 274 pages €[D] 48,- ISBN 978-3-8233-4948-8 25 Norbert Schlüter Present Perfect Eine korpuslinguistische Analyse des englischen Perfekts mit Vermittlungsvorschlägen für den Sprachunterricht 2002, XIV, 374 pages €[D] 48,- ISBN 978-3-8233-4949-5 26 Paul Skandera Drawing a map of Africa Idiom in Kenyan English 2003, XVI, 238 pages €[D] 54,- ISBN 978-3-8233-5820-6 27 Anne Schröder Status, Functions, and Prospects of Pidgin English An Empirical Approach to Language Dynamics in Cameroon 2003, 284 pages €[D] 58,- ISBN 978-3-8233-5821-3 28 Dagmar Barth-Weingarten Concession in spoken English On the realisation of a discourse-pragmatic relation 2003, XVIII, 328 pages €[D] 78,- 978-3-8233-5822-0 29 Ulrike Altendorf Estuary English Levelling at the Interface of RP and South-Eastern British English 2003, XIV, 187 pages €[D] 48,- ISBN 978-3-8233-6022-3 30 Brigitta Mittmann Mehrwort-Cluster in der englischen Alltagskonversation Unterschiede zwischen britischem und amerikanischem gesprochenen Englisch als Indikatoren für den präfabrizierten Charakter der Sprache 2004, 407 pages €[D] 68,- ISBN 978-3-8233-6089-6 31 Vivian Raithel The Perception of Intonation Contours and Focus by Aphasic and Healthy Individuals 2005, VIII, 180 pages €[D] 39,- ISBN 978-3-8233-6167-1 32 Rolf Kreyer Inversion in Modern Written English Syntactic Complexitiy, Information Status and the Creative Writer 2006, XII, 254 pages €[D] 58,- ISBN 978-3-8233-6227-2 33 Sandra Mollin Euro-English Assessing Variety Status 2006, XII, 230 pages €[D] 58,- ISBN 978-3-8233-6250-0 34 Uwe Vosberg Die Große Komplementverschiebung Außersemantische Einflüsse auf die Entwicklung satzwertiger Ergänzungen im Neuenglischen 2006, 302 pages €[D] 64,- ISBN 978-3-8233-6265-4 35 Michaela Albl-Mikasa Notationssprache und Notizentext Ein kognitiv-linguistisches Modell für das Konsekutivdolmetschen 2007, 452 pages €[D] 78,- ISBN 978-3-8233-6310-1 36 Torsten Müller Football, Language and Linguistics Time-critical Utterances in Unplanned Spoken Language, Their Structures and Their Relation to Non-linguistic Situations and Events 2007, 390 pages €[D] 78,- ISBN 978-3-8233-6356-9 37 Christina Sanchez Consociation and Dissociation An Empirical Study of Word-Family Integration in English and German 2008, 294 pages €[D] 64,- ISBN 978-3-8233-6384-2 38 Eva Lavric, Gerhard Pisek, Andrew Skinner, Wolfgang Stadler (eds.) The Linguistics of Football 2008, IV, 418 pages €[D] 78,- ISBN 978-3-8233-6398-9 39 Tamsin Sanderson Corpus · Culture · Discourse 2008, 351 pages €[D] 68,- ISBN 978-3-8233-6426-9 40 Edith Szlezák Franco-Americans in Massachusetts “No French no mo’ ‘round here” 2010, 325 pages €[D] 64,- ISBN 978-3-8233-6449-8 41 Silke Höche Cognate Object Constructions in English A Cognitive-Linguistic Account 2009, XII, 311 pages €[D] 58,- ISBN 978-3-8233-6489-4 42 Thomas Wagner Interlanguage Morphology Irregular Verbs in the Mental Lexicon of German-English Interlanguage Speakers 2010, 193 pages €[D] 58,- ISBN 978-3-8233-6547-1 43 Saskia Kersten The Mental Lexicon and Vocabulary Learning Implications for the foreign language classroom 2010, XIV, 197 pages €[D] 54,- ISBN 978-3-8233-6586-0 44 Anne Schröder On the Productivity of Verbal Prefixation in English Synchronic and Diachronic Perspectives 2011, 375 pages €[D] 68,- ISBN 978-3-8233-6587-7 46 Sandra Handl The conventionality of figurative language A usage-based study 2011, 371 pages €[D] 68,- ISBN 978-3-8233-6624-9 47 Sebastian Patt Punctuation as a Means of Medium- Dependent Presentation Structure in English Exploring the Guide Functions of Punctuation 2013, 307 pages €[D] 68,- ISBN 978-3-8233-6753-6 48 Anne-Kristin Cordes The role of frequency in children’s learning of morphological constructions 2013, XIV, 266 pages €[D] 58,- ISBN 978-3-8233-6840-3 This book presents a synchronic and diachronic investigation of two derivational English affixes. The suffixes -age and -ery are analysed on the basis of dictionary and corpus data and an adapted semantic map method is introduced as a new way of accounting for the semantic structure of derivatives. This study shows that the semantic structure of morphological categories can change significantly over time, and that semantic maps can represent this change in a straightforward manner. The semantic maps visualise the relations and interdependencies of the readings expressed by derivatives, which leads to a new understanding of the semantic complexity of these categories.