Layering of Size and Type Noun Constructions in English 9783110252927, 9783110252910

Citation preview

Layering of Size and Type Noun Constructions in English

Topics in English Linguistics 74

Editors

Elizabeth Closs Traugott Bernd Kortmann

De Gruyter Mouton

Layering of Size and Type Noun Constructions in English

by

Lieselotte Brems

De Gruyter Mouton

ISBN 978-3-11-025291-0 e-ISBN 978-3-11-025292-7 ISSN 1434-3452 Library of Congress Cataloging-in-Publication Data Brems, Lieselotte. Layering of size and type noun constructions in English / by Lieselotte Brems. p. cm. ⫺ (Topics in English linguistics; 74) Includes bibliographical references and index. ISBN 978-3-11-025291-0 (alk. paper) 1. Grammar, Comparative and general ⫺ Noun. 2. Grammar, Comparative and general ⫺ Case. I. Title. P271.B74 2011 4251.54⫺dc23 2011037051

Bibliographic information published by the Deutsche Nationalbibliothek The Deutsche Nationalbibliothek lists this publication in the Deutsche Nationalbibliografie; detailed bibliographic data are available in the Internet at http://dnb.d-nb.de. ” 2011 Walter de Gruyter GmbH & Co. KG, 10785 Berlin/Boston Cover image: Brian Stablyk/Photographer’s Choice RF/Getty Images Printing: Hubert & Co. GmbH & Co. KG, Göttingen 앝 Printed on acid-free paper 앪 Printed in Germany www.degruyter.com

Acknowledgements This book could not have been written without the input and encouragements of several people. First of all I want to thank my supervisor Kristin Davidse for taking me on as a PhD student, for generously sharing her time to discuss ideas and carefully commenting on texts, always alert and to the point. It is probably not done to use the topic of the book to quantify my gratitude, but I'll do it anyway: heaps and heaps of thanks, Kristin, for everything you have done for me over the years. I also thank my co-supervisor Hubert Cuyckens for introducing me to grammaticalization in his MA course and for his comments on the grammaticalization chapter in this book. I am grateful to a number of other people for careful comments on the manuscript and help with the publication process, i.e. Lieven Vandelanotte, my series editor Elizabeth Traugott, Birgit Sievert, Angelika Hermann and Jurgen Benteyn. I also warmly thank my colleagues at the Department of Linguistics for interesting chats on linguistic and non-linguistic topics. For having faith in me and encouraging me I thank my parents, my sister, and my boyfriend Dries, my personal MacGyver. Finally, I want to dedicate this book to the memory of my grandmother, Leonie Verbeeck.

Table of contents

Acknowledgements Table of contents

v vii

Part I: Preliminaries

1

Chapter 1: Description of the topic and state of the art

2

1.1 Description of the topic: Sort of a lot of size and type nouns

2

1.1.1 Size noun constructions

8

1.1.2 Type noun constructions

16

1.2 State of the art: reference grammars and theoretically-oriented literature on size and type nouns

19

1.2.1 Discussions of SNs in the literature

19

1.2.2 Discussions of TNs in the literature

27

1.2.3 Conclusions: a joint treatment of SN- and TN-constructions

37

Chapter 2: A cognitive-functional constructional framework for the English NP

41

2.1 Requirements of a framework for SN- and TN-constructions

41

2.1.1 Halliday’s systemic-functional approach

42

2.1.2 Langacker (1991): A radical functional account underpinned by dependency

47

2.1.3 McGregor’s semiotic grammar: Syntactic combinatorics in the NP

56

2.1.4 Conclusion: An eclectic functional model of the NP

59

2.2 A constructional framework of the English NP

60

2.2.1 Elaborating the functional model: SN- and TN-patterns as partially filled constructions

60

2.2.2 Background: From idioms to Construction Grammar

62

viiiTable of contents 2.2.3 CxG as a family of constructional approaches: General tenets

63

2.2.4 A construction grammar approach to SN- and TN-patterns

68

2.3 Conclusion: A dynamic model of the NP for SN- and TN-constructions

78

Chapter 3: Grammaticalization, delexicalization and subjectification in SN- and TN-constructions

79

3.1 Grammaticalization, synchronic variation and emergent grammar

79

3.2 Grammaticalization: Changing perspectives

80

3.2.1 A morphology-based approach to grammaticalization

82

3.2.2 Criticism of Lehmann’s parameters and semantico-pragmatic approaches to grammaticalization

83

3.2.3 A construction-based approach to grammaticalization

86

3.2.4 Criticism of grammaticalization research

91

3.2.5 Discussion and conclusion

92

3.3 Grammaticalization of SN- and TN-constructions: From source to target construction

93

3.4 Factors at work in the grammaticalization of SN- and TN-constructions

94

3.4.1 Reanalysis and analogy

96

3.4.2 Semantic changes, layering and persistence, (inter)subjectification

103

3.4.3 Decategorialization: A positive interpretation

111

3.4.4 Paradigmaticization: A dynamic interpretation

113

3.4.5 Syntactic extension

113

3.4.6 Coalescence and phonetic erosion

115

3.4.7 Frequency

116

3.4.8 Cyclical processes of renewal

117

3.4.9 Lexicalization and grammaticalization

118

Table of contents

ix

3.5 Conclusion: Grammaticalization as functional and formal approximation of a target construction

119

Part II: Synchronic and diachronic corpus studies

123

Chapter 4: SN-constructions

125

4.1 Introduction to SNs and issues overarching SNs and SSNs

125

4.2 Description of the synchronic data set, selection of the SNs and method of analysis

126

4.3 Head and quantifier constructions: heap(s) versus pile(s) and lot(s)

132

4.3.1 Heap(s) versus pile(s)

132

4.3.2 Lot of and lots of: Head, modifier and ambivalent uses

157

4.4 Towards valuing (quantifier) uses

160

4.4.1 Load and loads of

160

4.4.2 Bunch and bunches of

176

4.5 Premodification patterns and decategorialization: A case for collocationally constrained constructions

191

4.6 Degrees of grammaticality

201

4.7 Diachronic corpus study of heap(s) and lot(s)

204

4.7.1 Introduction

204

4.7.2 Data sets and methodology used

206

4.7.3 Heap and heaps

207

4.7.4 Lot and lots

210

4.7.5 Discussion of findings

213

4.7.6 Conclusions

215

4.8 Comparative study: English heap(s) and bunch versus Dutch hoop, hopen and stel, stelletje, zooi and zooitje

216

4.8.1 Head and (valuing) quantifier constructions of English and Dutch SNs

216

xTable of contents 4.8.2 Introduction to the data and methodological issues

217

4.8.3 Quantitative and qualitative data analysis

218

4.8.4 Discussion and conclusion

228

4.9 Synchronic layering of SN-constructions

229

Chapter 5: Small SN-constructions

234

5.1 Introduction: Similarities and differences between regular SNs and SSNs as sets of meso-constructions

234

5.2 Description of the corpus data and methodological issues

238

5.3 Hoffmann (2004): Frequency and analogy in low-frequency complex prepositions

240

5.4 Polarity sensitive contexts in SSN-constructions

243

5.5 Individual discussions of SSN-constructions

245

5.5.1 Bit of

248

5.5.2 Positive polarity SSNs

252

5.5.3 Negative polarity SSNs

254

5.6 Discussion and conclusions

263

Chapter 6: TN-constructions

270

6.1 Introduction

270

6.2 Description of the synchronic data sets

273

6.3 Classification of NP-internal TN-uses and extended uses

274

6.3.1 Head use construction

275

6.3.2 Quantifier construction

281

6.3.3 Modifier uses

284

6.3.4 Postdeterminer use

292

6.3.5 Qualifying use

307

6.3.6 Discourse marker use

317

Table of contents

xi

6.3.7 Marker of onomatopoeia and quoted speech and thought

319

6.3.8 Ambivalent contexts

320

6.3.9 Conclusion: Synchronic layering of TN-constructions and comparison with S(S)N-constructions

326

6.4 Register analysis of the functions of TN-constructions

331

6.5 Diachronic case study of sort(s), kind(s) and type(s)

335

6.5.1 Introduction

335

6.5.2 Data sets and methodology

336

6.5.3 Nominal constructions with type nouns

337

6.5.4 Diachronic paths of change and chronology

345

6.5.5 Summary

353

6.6 Conclusion

355

Chapter 7: Descriptive-theoretical consequences and prospects for further research

357

Corpora

362

Notes

364

References

377

Index

404

Part I Preliminaries

Chapter 1 Description of the topic and state of the art 1.1 Description of the topic: Sort of a lot of size and type nouns This study aims at a systematic corpus-based description of two, originally, binominal structures and the development of their extended uses. On the one hand, I look at structures that incorporate a size noun expression, i.e. a nominal expression that describes size or shape, implying a measure, such as a bunch of, heaps of, a bit of, a jot of, etc. On the other hand, I look at structures with a type noun expression, i.e. a nominal expression to do with (sub)categorization, i.e. sort of, kind of and type of. These two types of structures will be referred to as size noun constructions and type noun constructions respectively (henceforth also SNconstructions and TN-constructions). The specifics of each construction type as well as potential functional overlap between them will be accounted for within a constructional framework in the vein of usage-based Goldbergian Construction Grammar (Goldberg 2006).1 Both SN-constructions and TN-constructions are originally binominal noun phrases, in that they consist of two nouns connected by the preposition of. These two nouns each allow for premodification by adjectives and/or determiners, i.e. identifiers or quantifiers. Both initial construction types can hence be schematized as (D) (M) N1 of (D) (M) N2, where N1 is instantiated by a size noun (henceforth also SN)2, or a type noun3 (henceforth also TN) respectively. (D) refers to optional determiners and (M) to optional (adjectival) modification of both nouns. In certain extended uses of SN- and TN-constructions the N2slot opens up to non-nominal syntactic categories, such as adjective, adverb and verb. In addition, there are some uses that are more or less external to the structure of the NP and hence equally stretch beyond the boundaries of the schema proposed here. Binominal and extended uses of both types of construction display what seem to be polysemous uses in synchronic corpus data. The main aim of this study is to classify these various uses in terms of synchronic layering, i.e. to see them as the co-existing synchronic sediments of diachronic processes of change (cf. Hopper 1991). As many of these uses have not been described adequately yet, this will require identifying clear “form-meaning pairings ([subsuming] lexical collocation, syntactic structure and semantics-pragmatics)” (Traugott 2010b: 45), based on the study of extensive sets of synchronic corpus data. On the basis of carefully chosen diachronic case studies, it will be argued that the synchronic classification attested by SNconstructions and TN-constructions can be interpreted in terms of paths of grammaticalization and (inter)subjectification.

1.2 State of the art

3

Let us take a look at a number of instantiations of both construction types. In the original binominal syntagms N2 is typically a plural count noun or uncount noun in the case of SN-constructions, whereas it can be singular count, plural count or uncount in the case of TN-constructions.4 (1.1)-(1.4) are examples of NP-internal SN-constructions; (1.5)-(1.8) exemplify TNconstructions, with (1.8) being an NP-external use (the SNs and TNs are in bold):5 (1.1)

Such is the fable of “The Fox and the Grapes”, in which a fox, unable to reach a bunch of grapes that hangs too high, decides that they were sour anyway (implied moral: It is easy to spurn what we cannot attain. (CW-USbooks)6

(1.2)

He threw himself on the floor, he kicked a couple of dishwashers and washing machines and knocked a load of stuff off the top of the washers. (CW-UKspoken)

(1.3)

He said: “A bunch of drunken, brain-dead louts seem determined to disgrace our team.” (CW-Sunnow)

(1.4)

Isn't there a bit of a double standard there? (CW-NPR)

(1.5)

My grandfather had a curly moustache and offered two sorts of kisses: an ordinary flat kiss or a tickling kiss reserved for when we had been extra good. (CW-OZnews)

(1.6)

The problem was that the Bush administration said the funding should come from the cities and states. The cities and states said they didn’t have this kind of money. (CW-NPR)

(1.7)

This is why im crazy about u. in a non-‘worship and serve me forever’ typa way ;-) (www.7thrimofhell.blogspot.com/1990/ blog-post)

(1.8)

I’ve sort of become part of the mountain bike world in a way without actually having a mountain bike; it’s quite strange. (CWUKspoken)

The SN- and TN-constructions focused on in this study are part of a larger set of structures that fit the proposed binominal schema.7 Some of these constructions do not seem to be explicitly related to the very specific constructional semantics of SN- or TN-constructions, e.g. of-apposition patterns such as the city of Rome or genitive constructions, such as a friend of

4 Chapter 1: Description of the topic and state of the art Warhol’s. However, this more extensive binominal set also includes (time) prepositional expressions, such as on the edge of, on the brink of and on the verge of, which display patterns of polysemy and seem to have engaged in processes of change similar in certain respects to those observed for SNand TN-constructions (cf. Vanden Eynde 2003). Nevertheless, this set of prepositional expressions have their own particular semantics and will not be discussed in this study, unless when relevant in comparison with the evolutions observed for SN- and TN-constructions, or as corroborative evidence for some of the claims made for SNs and TNs. This study zeroes in on SN- and TN-constructions for a number of reasons, both theoretical and descriptive in nature. The main reason for looking at TN- and especially SN-constructions is of a descriptive nature. Examples (1.1) to (1.8) already to some extent bear out what constitutes the main descriptive-analytical challenge of both constructions, and why they are so often considered as intractable material in the literature. The obvious question in any binominal syntagm is: which of the two nouns is the head and what is the status of the other noun? The central question with regard to both SNs and TNs, within the specific binominal structure they instantiate, is hence also one of a structural and semantic nature. The two main analytical options can be simplified at this point as head versus non-head or modifier status. Corpus and Internet examples (1.1) to (1.8) serve to illustrate that this decision is not a straightforward one in terms of either/or in each instance separately, let alone in all instances of SN- or TN-constructions as a whole. Instead, both SNs and TNs, each in their own way, show a very intricate and multidimensional interplay between grammatical and lexical status, which in addition to clear-cut black or white cases also gives rise to a substantially large grey area, itself a microcosm of different shades. Important factors determining structural and semantic status are collocational patterns, based on the habitual co-occurrence of SN- or TN-nodes and collocates (cf. Robins 1971 and Crystal 1991), as well as semantic prosody, i.e. the context-dependent preference of nodes for positive, neutral or negative collocates, e.g. the verb cause as a node typically has a negative semantic prosody as it usually teams up with negative collocates, such as accident or damage (cited in Stubbs 1995: 25; see also Robins 1971, Crystal 1991, Bublitz 1996 and Louw 1993). Other lexical-syntagmatic factors are premodification patterns of the SN/TN and N2, e.g. a foot-high pile of manure (CW-Today), a whole load of new clothes (CW-Sunnow), Not the remotest jot of nostalgia (CW-Times), a rare type of infection (CWUKephem) and (1.7) a non-‘worship and serve me forever’ typa way. This will be discussed in more detail in Sections 1.1.1 and 1.1.2. In addition and partially overlapping, the interaction between the functional zones of categorization, modification and determination in the prenominal string of the English NP is important, where categorization has to do with the specifica-

1.2 State of the art

5

tion of a type as such, modification adds further classifying and/or qualitative attributes to this type specification, and determination is concerned with signalling identifiability status of the referent of the NP (cf. Bache 2000; see Section 2.1.2). Let us return to examples (1.1) to (1.8) and quickly run through them to get a better grasp of these analytical issues. In a bunch of grapes (1.1) and two sorts of kisses (1.5) the SN and TN seem to be the dominant elements within the binominal NP, both on semantic and formal grounds. In (1.1) the SN has the lexically very specific and collocationally restricted meaning of a constellation of entities growing or fastened together at one end. The noun in the of-phrase then specifies what the bunch consists of, in this case grapes, and both nouns are coextensive. Ultimately, the NP is about a bunch, which in this context refers to a spatially continuous composite entity (cf. Halliday’s definition of the Thing within the nominal group: see Section 2.1.1). In example (1.5) the TN expresses its literal meaning of (sub)categorization with regard to the noun in the of-phrase, in this case two subtypes of the superordinate type ‘kisses’. The specific subtypes are then described after the colon. In A bunch of drunken, brain-dead louts (1.3) and I’ve sort of become part of the mountain bike world (1.8) the meanings of the SN and TN display a clear semantic shift from those expressed in examples (1.1) and (1.5). In (1.3) bunch of refers to a quantity of people, without reference to the specific and collocationally restricted cluster meaning, and seems to express added evaluative meaning. In (1.8) sort of moves beyond the binominal structure and does not denote strict subcategorization; rather, it seems to do the opposite, i.e. hedge the denotation of the verbal predicate become part of. Examples (1.3) and (1.8) hence exemplify non-head or modifier status. A formal corollary of this non-head status in (1.3) is that, as opposed to (1.1), the SN does not seem to control verb agreement, since the verb is plural. Examples (1.2) as well as (1.4), (1.6) and (1.7) then are more difficult to classify, in that they each in their way lie at different points between the analytical poles of head versus non-head status. In a load of stuff (1.2) the meaning of the SN hovers between referring to an actual load and just designating quantity as such. In a bit of a double standard (1.4) bit of a expresses more than just a small quantity and further activates implicatures of intersubjective hedging meaning, mitigating the force of the proposition. In a non-‘worship and serve me forever’ typa way (1.7) the TN fits into the binominal structure in a different way than in two sorts of kisses (1.5). There is greater cohesion between the TN, the nonce attribute in front of it as well as of, which appears to have merged with the TN orthographically as well. Typa seems to have suffix-like status, which ties in with a specific stress pattern of non-salience for the TN-expression. In this kind of money (1.6) the TN together with the demonstrative in front of it form a (complex)

6 Chapter 1: Description of the topic and state of the art anaphoric determiner which provides complex identification information by referring to some generalized standard amount of money that, in this case, cities and states may have (cf. De Smedt, Brems and Davidse 2007). All of these types of meaning will be discussed in greater detail in Sections 1.1.1 and 1.1.2 and especially in the corpus studies in Part 2. At this point, examples (1.1) to (1.8) serve to indicate that the formal and semantic status of SNs and TNs within binominal and other structures is not clear-cut. The structural ambivalence of SNs and TNs also has repercussions at clause level, most notably on the question of subject-verb concord whenever the SN/TN-nominal occurs in subject position, as already noted for (1.1) and (1.3). In the extended uses where the N2-slot is sometimes taken up by non-nominal categories or when the SN/TN has an NP-external function, structural status is also an important and knotty question. Some instances of SN- and TN-constructions ultimately seem to resist disambiguation for several (good) reasons and prove to remain irreducible ‘blends’ (Bolinger 1961) of head and modifier status. In addition to similarity in descriptive concerns, there is another reason for looking at SN-constructions as well as TN-constructions, which pertains to a particularity in the literature on both constructions. As the literature survey in Section 1.2 will show, SN- and TN-constructions are traditionally discussed together; in general reference grammars, such as the Quirk grammars, as well as in (topic-specific) theoretically-oriented ones, such as Halliday and Matthiessen (2004). However, the literature does not put forward well-argued reasons for this joint treatment, in terms of fundamental synchronic and/or diachronic correspondences in their constructional semantics for instance. Most authors lump SN- and TN-constructions together almost exclusively on the basis of formal similarities in the surface structures of both constructions (as indicated in the binominal schema described earlier). In both constructions the SN or TN usually receives a default head analysis. Both constructions have, however, also been noted to display incongruous concord patterns, where the verb and the apparent head noun disagree in number, as in (1.3) for instance. These incongruities are treated as unsystematic “idiomatic anomal[ies]” (Quirk, Greenbaum, Leech and Svartvik 1985: 765) or contextual idiosyncrasies restricted to informal contexts. Hardly ever are they considered as potentially systematic symptoms of SNs and TNs’ constructional semantics possibly in flux, and if so, no further comments or generalizations are made. The incongruity is mostly described as notional or proximity concord violating strict grammatical concord. The possibility that head function can shift from the SN/TN to another element of structure in at least some cases is usually not considered. Yet, this would explain why the SN/TN does not behave like a head, both semantically and formally, in

1.2 State of the art

7

some of the incongruous instances and would make ad hoc and a posteriori justifications in terms of notional or proximity concord unnecessary. Sometimes cautious remarks are made about quantifier-like uses of SNexpressions such as load(s) of, but the only further explanation of this quantifier function is that it is restricted to informal or colloquial registers and disapproved of in other contexts, e.g. (willy-nilly prescriptive)descriptive Quirk et al. (1985: 264 footnote) and explicitly corpus-based Biber, Johansson and Leech (1999: 254-255). With regard to TNs too systematic remarks are restricted to observing rather well-established regional differences in distribution between TNs, e.g. that kind of is preferred in American English, while sort of is typical of British English (Biber et al. 1999: 256). No unified classifications of different SN/TN-uses are made other than distinguishing between various SNs in terms of their source semantics, e.g. container SNs versus shape SNs (cf. Biber et al. 1999 and Dodge and Wright 2002). In sum, the joint discussion of SN- and TNconstructions is not accounted for by explicit reference to fundamental parallels between the similarities in structural realization on the one hand, and similarities in the semantics of both constructions on the other. Both SN- and TN-constructions are mainly presented as intractable material ‘riddled with idiosyncrasies’ (Jackendoff 1981: 103) that seems to slip through the system of grammar in a supposedly unpredictable way. The similarities and anomalies are furthermore rather straightforwardly posited in most grammars, without in-depth descriptive back-up by means of corpus analysis. The literature on SNs and TNs will be reviewed in greater depth in Section 1.2. The main objective of this study is to account for the individual descriptive specifics of SN-constructions and TN-constructions separately. In addition, both construction types will also be related by assessing similarities and dissimilarities between them from an essentially semiotic constructionist perspective, i.e. by looking at them in terms of form-meaning pairings, thereby filling a gap in the literature, which only saw similarity in the pure shape of the constructions. The question of similarities will be mainly one of potential functional overlap between or engagement in similar changes of the two constructions. The key question overarching SN- and TNconstructions hence is: in which way or ways can and do they instantiate the determination – modification - categorization schema referred to earlier, with determination subsuming both identification and quantification (cf. Langacker 1991)? The various constructional uses attested by SN- and TN-syntagms are essentially looked at in terms of synchronically coexisting meanings. In addition, it is argued that both SN-constructions and TN-constructions have and are engaged in (ongoing) processes of delexicalization and grammaticalization, with grammaticalization understood as

8 Chapter 1: Description of the topic and state of the art the change whereby lexical items and constructions come in certain linguistic contexts to serve grammatical functions, and once grammaticalized, continue to develop new grammatical functions. (Hopper and Traugott 2003: 18)

Hopper and Traugott’s definition is one of the standard definitions of grammaticalization, but certainly not the only one. The concept of grammaticalization will be discussed in much more detail in Chapter 3, where certain vexed questions in the debate on what should be included in the definition are reviewed, and a motivation is provided for the definition that will be used in this study. The polysemy and structural ambivalence of the constructions attested in corpus and Internet data, such as examples (1.1) to (1.8), is then argued to be the synchronic result of diachronic processes of pragmatic inferencing interacting with grammaticalization. Grammaticalization has become a busy field of linguistic study these days and studies have prototypically dealt with verbal structures. The present study wants to make a contribution to a relatively under-studied environment in grammaticalization, i.e. the noun phrase (some exceptions are Adamson 2000, Aijmer 2002, Breban 2010, Brems 2001, 2003a and 2004a, Denison 2002, Paradis 2000 and Traugott 2008a and b). Chapter 2 presents a cognitive-functional model of the English NP that accommodates the synchronic polysemy of SN- and TN-syntagms, and adds a specific grammaticalization framework to fully account for the dynamicity of these patterns within the NP and beyond, incorporating basic tenets from usage-based construction grammar (Goldberg 2006) and collocation studies (e.g. Hunston and Francis 2000). Based on extensive sets of synchronic corpus data and diachronic case studies, Part 2 then discusses the synchronically layered uses of several sets of SN- and TN-expressions in great detail. In addition, these descriptive observations evoke theoretically-focused research questions about the nature of grammaticalization as such, and its relation to key notions like subjectification, collocational patterns, constructional schematicity, reanalysis, analogy and frequency. First, Sections 1.1.1 and 1.1.2 of this introductory chapter will delineate SN- and TN-constructions separately with attention to the descriptive issues specific to each construction, as well as the potential theoretical repercussions of these on a theory of grammaticalization. Section 1.1.1 will introduce a further distinction within the set of SN-constructions between regular SNs and small SNs, which each bring in issues particular to them, as well as share the fundamental reanalysis from head to non-head status. 1.1.1 Size noun constructions

1.2 State of the art

9

The main focus of this study is a systematic description of the layered polysemous uses of SN-constructions as manifested by extensive sets of synchronic corpus data. Compared to TN-constructions, SN-constructions have received very little attention in the literature and in corpus research, and attempts at functionally motivated classifications within a unified constructional account are virtually non-existent. In a second step it will be argued that the semantic and structural ambivalence, already succinctly exemplified by (1.1) to (1.4), can be interpreted as the synchronic result of ongoing processes of delexicalization and grammaticalization, which are motivated by the language user’s need for expressive means of quantification and evaluation. The grammaticalization framework is dynamic enough to account for this layered polysemy as a semi-stable system of SN-uses, instead of the unpredictable chaos it is often made out to be (see status quaestionis in Section 1.2). It allows me to describe the various SN-uses as the combined (interim) results of diachronic processes of change. SNs were described earlier as nouns with size implications. The SNconstruction is a productive pattern for non-canonical coding of quantification that is furthermore sensitive to fashion and renewal (Hopper and Traugott 2003). For instance, in the sixties things typically came in bags or bagsful, as in bagsful of fun (personal communication M. A. K. Halliday and Eirian Davies), which is now a dated expression that has been superseded by new topical expressions such as loads of and heaps of fun. The list of SNs to choose from for this study is endless, as new (nonce-)expressions continue to crop up, potentially associated with regionally restricted varieties of English (see Trousdale 2010b on such SNs as hella, gob and wodge). Therefore several restrictions were made. In this study two subsets of SNs are included, distinguished on the basis of the kind of size they imply or denote. On the one hand, there are the regular SNs, which designate a relatively large quantity, and on the other there are the semantically more specific small SNs (henceforth also SSNs), which all imply or incorporate small to very small quantities in their lexical make-up. Semantically more schematic quantifiers such as many/much and little/few can equally be scaled into such subsets.

1.1.1.1 Regular SNs For the case studies on regular SNs corpus extractions were made on the following singular and plural expressions occurring in and beyond binominal constructions: bunch, bunches, heap, heaps, load, loads, pile and piles. The synchronically coexisting uses of this subset revolve around a difference between uses with head versus modifier status. Examples (1.9) and

10 Chapter 1: Description of the topic and state of the art (1.10) exemplify SN-constructions with pile in which the SN functions as the dominant element of the binominal noun phrase: (1.9)

A pile of bodies mounted up, like a barricade of flesh, around the sweating, cursing, straining Soviet strongpoint. (CW-UKbooks)

(1.10)

A jilted girlfriend got revenge on the boyfriend who dumped her by dumping a foot-high pile of manure in his bed. (CW-Today)

In (1.9) the verb phrase mounted up makes clear that pile is used in its lexical meaning of ‘a heap of things laid or gathered upon one another’ (The Concise Oxford Dictionary of Current English) and calls up a specific constellation that takes up space and hence has referential meaning. Similarly, in (1.10) premodifier foot-high makes it clear that pile refers to a particular constellation consisting of what is specified by the of-phrase. In (1.1) in delete an additional syntactic argument for head status was the concord between the SN and the verb. I will refer to these uses as lexical head constructions. In examples (1.11) to (1.14), on the other hand, the SNs display different semantic and syntactic behaviour: (1.11)

Modesty was eventually restored, but not before a heap of people in Pine Rivers had become privy to intimate knowledge what Santa wears beneath those baggy pants. (CW-OZnews)

(1.12)

All this idea of ‘there's nothing for people to do’ it's a load of codswallop of course. (CW-UKspoken)

(1.13)

What can beat calling the entire politically correct movement a bunch of second-rate drinkers and third-rate Marxists? (CWOZnews)

(1.14)

The ‘surrogate mum’ to princes William and Harry shared heaps of fun with them at a fair yesterday while father Charles was otherwise engaged. (CW-Today)

The N2s in a heap of people (1.11) and a bunch of second-rate drinkers (1.13), just like in a pile of bodies (1.9), refer to humans, but in (1.11) and (1.13) reference is not to a literal heap or bunch of them, but a large quantity of people as such. The lexical constellation meaning of the SNs heap and bunch is nearly completely backgrounded. The N2 collocates in a load of codswallop (1.12) and heaps of fun (1.14) refer to abstract and uncount concepts that cannot be stacked. In (1.14) heaps of fun hyperbolically refers

1.2 State of the art

11

to ‘a lot of fun’. In addition to, or maybe instead of, quantifier meaning (1.12) and (1.13) seem to incorporate increased subjective, evaluative meaning. Rather than quantifying the N2 referent, the SN-expressions serve to further evaluate them (negatively). For this reason Chapter 4 will distinguish between two main grammaticalized uses, a quantifier use on the one hand and a more evaluative quantifier use on the other. The latter will be called valuing quantifier, which reflects its evaluative function as well as refers to the emotive value of N2 in these uses. The developmental path of valuing quantifiers involves grammaticalization-cum-attitudinal subjectification, with subjectification here referring to the shift from propositional meaning to the encoding of speaker-related meanings, in the sense of attitudes or beliefs, with regard to the propositional content (e.g. Traugott 2010b). Besides examples that seem to have either head or modifier status there is also an extensive set of data in which the lexical constellation semantics and the grammatical quantifier meaning are both referred to in various ways, intentionally or not, as in (1.15)

I would take up a pile of commonplace books like Lord David Cecil’s Library Looking Glass, John Julius Norwich’s Christmas Crackers, Rupert Hart-Davis’s A Beggar in Purple, etc. (CWUKmags)

(1.16)

We found loads of tapes parcelled up for delivery. Many were recent releases. (CW-Sunnow)

(1.17)

Those who must deal with them on a regular basis and wish to get the better of them must employ much patience, deviousness and duplicity as well as lorry loads of insincere flattery. (CW-Today)

(1.18)

Well, possibly not, but if you were expecting to find spade loads of hubris heaped high on DPW's first post-EMI outing, you'd be sorely mistaken. (CW-UKmags)

(1.19)

The British have forged a fine tradition of gardening and cannot afford to sit on their well-clipped laurels. Striding past the compost heap of nostalgia, comes Christopher Lloyd. (CW-Times)

(1.20)

Rather than doing it word for word Yeah he's wanting to do bunches of words or something at the same time. (CWUKspoken)

12 Chapter 1: Description of the topic and state of the art Examples (1.15) to (1.20) show the quintessential feature of SNexpressions, i.e. the often intricate interplay between their lexical and grammatical status, but in an intensified way. In a pile of commonplace books (1.15) and loads of tapes (1.16) pile and loads respectively can be interpreted as referring either to the lexical denotation of the SN, or as having quantifier meaning based on the size implications of these SNs. In (1.17) to (1.19) the premodifiers of the SN loads, lorry and spade, rekindle the lexical semantics of the SN, but at the same time a quantifier meaning is conveyed in what are highly expressive contexts. In the compost heap of nostalgia (1.19), again there seems to be both reference to a lexical heap that can be walked around and a quantity of N2, i.e. nostalgia. In (1.20) bunches of words refers to ‘several words’, but also evokes a cluster meaning which derives from the lexical meaning of bunch, and which for instance incorporates temporal contiguity (at the same time). Examples (1.15) to (1.20) hence all incorporate an ambivalence, even though they do not all have the same status. The classification of SN-uses proposed in Chapter 4 will account for such ambivalent uses in terms of vague and ambiguous uses, and grammaticalized quantifier uses with lexically persistent SNs. I will set up generalizations about their constructional semantics, as well as frame them within an ongoing process of grammaticalization and delexicalization. I will also look at differences in terms of degrees of grammaticalization, and will discusses prosodic and lexical factors in this as well as differences between the singular and plural forms of SNs. As a further sign of the intricate interplay between lexical and grammatical meaning layers, it will be noted that the lexical source semantics of most SNs always remain (latently) present in their grammaticalized quantifier uses and can be alluded to or exploited to fit the speaker’s need for more expressivity, as in lorry loads of insincere flattery (1.17) (cf. Hopper 1991). Meanings expressed by SNs outside of the binominal construction are also taken into account, either as extended, NP-external uses arising in a diachronic process, or in order to check whether these constructional semantics feed into other SN-uses. Regular SNs display what one might label adverbial or degree modifier uses, in which of is typically left out (cf. Quirk et al. 1985, Langacker 2009 and to appear, and Traugott 2010b): regular SNs can be used to modify verbs, e.g. (1.21) to (1.23), (mainly comparative) adjectives, as in (1.24) and (1.25), and adverbs (1.26): (1.21)

Zoe, whose favourite tipple is Jack Daniels, insists she doesn't drink loads. (CW-Sunnow)

(1.22)

“Perhaps you could suggest you're too old,” he offered sullenly. “Thanks a bunch.” (CW-UKbooks)

1.2 State of the art

13

(1.23)

“Having them down here during the season will help me heaps,” said Vander-Kuyp, who was fifth in the 110m hurdles. (CWOZnews)

(1.24)

I miss the old Valley Pool. The new one is heaps better, of course, but I pity the children who have to attend school carnivals there. (CW-OZnews)

(1.25)

With Crass they failed to notice it totally and we're like f ing loads softer than that. (CW-UKmags)

(1.26)

And there'll be plenty more spectacular action with 6-a-side matches and mini soccer for the younger visitors, video displays of some of the world's greatest games, interviews with stars past and present and loads, loads more. (CW-UKmags)

The adverbial uses are very similar in meaning to the hyperbolic quantifier uses in binominal constructions, only the syntactic environment is different. In addition to case studies on English SN-expressions, a study using data on Dutch SN-expressions such as hopen (‘heaps’), stelletje and zootje (‘bunch’) will be included as corroboration of the two developmental paths set out for the English expressions (see Section 4.8 in Chapter 4). This comparative study confirms the existence of these two main paths, even though it also points up cross-linguistic differences in their specific development.

1.1.1.2 Small size nouns In this subset of SNs distinct factors are at work in addition to collocational shifts. The notions of polarity sensitivity (Israel 2004), frequency and analogy (Hoffmann 2004) as well as pragmatic values of downtoning and intensifying come into play which are not that prominent in the other subset of SNs. As opposed to the regular SNs, the SSNs looked at in this study, except for bit, are all infrequent and rare lexical items, e.g. jot, skerrick, scintilla. Frequency has alternatively been seen as a prerequisite, symptom or consequence of grammaticalization and the possibility of infrequent items grammaticalizing hence touches on the basic tenets of grammaticalization as such, and calls for alternative explanations. Following Hoffmann (2004), a bit of will be considered but also partially rejected as possible analogical model for the grammaticalization of the other, infrequent SSNs in Chapter 5 (cf. Hoffmann 2004). The concepts of frequency and analogy as possible mechanisms of change add very specific issues to the gram-

14 Chapter 1: Description of the topic and state of the art maticalization processes of SSNs in comparison with the regular SNs and TNs. Examples (1.27) to (1.42) exemplify all of the SSNs looked at in this study within a binominal context and beyond. Incidentally, the grammaticalized uses of SSNs, except bit and jot, normally stay within this binominal context. Examples (1.27), (1.29), (1.31) and (1.33) illustrate the original, fully lexical meaning of the SSNs while the other examples represent non-head uses. In the examples with bit, in addition to the head use, at least four kinds of non-head uses need to be distinguished, which can be related to increased subjectification, intersubjective values, and various discourse schemata. The head use example for bit (1.33) derives from a diachronic corpus, i.e. The York-Toronto-Helsinki Parsed Corpus of Old English Prose (henceforth YCOE)8 (1.27)

If you’re lucky you’ll catch a whiff of cat’s pee! (CW-UKephem)

(1.28)

There is an unpleasant whiff of scapegoat-hunting in the latest assault on the banks to come from Deputy Premier Tom Burns. (CW-OZnews)

(1.29)

“Why is it,” he said in a voice soaked in Irish charm, “that I want so much to tear every shred of clothing off your body?” “What's the matter, you don't like the dress?” (CW-UKbooks)

(1.30)

“I accept absolutely Mr Costello's statement to me that there's not a shred of truth in these allegations,” said Senator Evans. (CWOZnews)

(1.31)

A flicker of light from the flames licked her lips and stained them crimson. (CW-UKbooks)

(1.32)

It's all I look forward to; at the end of the day it gives me a flicker of comfort. (CW-UKbooks)

(1.33)

[…] wid attorcoppan bite […] (‘As a treatment against the bite of a spider’) (YCOE)

(1.34)

He took a bit of apple.

(1.35)

“I love the colour brown,” says Nyree “which is why I chose a brown carpet and curtains. Add a little bit of pink and you have a very cosy room on winter nights.” (CW-UKmags)

1.2 State of the art

15

(1.36)

“I’m going to test you now Gwyneth so you we’ve got an examination coming up now bit of a an oral examination.” (CWUKspoken)

(1.37)

Life hasn't got any better under the new manager. The first thing he said to me was “I hear you are a bit of a troublemaker.” (CWUKmags)

(1.38)

“A word, Matt. I'd like you to do a run-through with Lotte. She's a bit iffy – ” the director rotated his right hand palm down – “know what I mean?” (CW-UKbooks)

(1.39)

The pace of change introduced by technological progress is such that if our attention lapses for even one generation, a smidgen of time on any geological scale, transformations in the land that are truly dangerous could catch us unawares. (CW-USbooks)

(1.40)

Mr Bullock says there is not one skerrick of evidence to involve this man. (CW-OZnews)

(1.41)

They may yet force through the proposals but they won't make a jot of difference to the way the game's mishandled here. (CWSunnow)

(1.42)

This latest twist gives credibility to our argument all along that there is not a scintilla of evidence against Michelle of (a) having used a banned substance or (b) physically manipulating the sample. (CW-Sunnow)

Bit of seems to have gone through a typical process of grammaticalization with the quantifier meaning gradually emerging in particular restricted contexts that activate the size implications of the SN, still visible in synchronic examples such as a bit of apple (1.34). In later stages it can function as a quantifier without the support of these specific contexts, e.g. a little bit of pink (1.35), and eventually it extends to other syntactic environments, as in a bit iffy (1.38). The quantifier uses of infrequent SSNs seem to arise in specific polarity sensitive contexts and are tied to specific collocational patterns and discourse schemata, which, contrary to bit of, they continue to be restricted to. In not one skerrick of evidence (1.40) for instance skerrick as a quantifier typically features in negative polarity contexts in which it has emphatic meaning, such as the reactive discourse characteristic of court cases (cf. Brems 2010). Such restrictions render Hoffmann’s proposal for

16 Chapter 1: Description of the topic and state of the art analogy with a frequent structural variant somewhat problematic: more complex analogical mechanisms seem to be at work. These specific questions with regard to the grammaticalization of SSNs feed into the general debate on issues such as grammaticalization by analogy and the levels of constructional schematicity involved in grammaticalization (see Fischer 2007, Trousdale 2010a, and Chapter 3). The corpus study on SSNs presented in Chapter 5 will argue for an account that integrates these notions in a careful way.

1.1.2 Type noun constructions TN-constructions are mainly discussed here in comparison with SNconstructions. As opposed to the latter, the versatile behaviour of TNs already attracted the attention of the great descriptive grammarians of the first half of the 20th century such as Kruisinga (1925) and in recent years their relevance to the study of grammaticalization and discourse patterns has made them into something of a hot topic in English studies. However, there is no agreement in the literature about a systematic, formally motivated classification of their different uses, especially NP-internal functions, which also pose the greatest descriptive problems. Chapter 6 will work towards such a systematic grammatical description, again based on extensive sets of synchronic and diachronic data of three common TNs, i.e. sort, kind and type. This allows me to investigate whether and where SNconstructions and TN-constructions show functional overlap in terms of what their constructional semantics can express. Just as with SN-uses the basic distinction within the TN-uses centres on head versus modifier status of the TN. Example (1.43) illustrates a head noun construction with the TN type. (1.43)

The late-17th century console has an elaborate top inlaid with agate and pietra paesina, a rare type of marble which in its colours and veining looks like a paesaggio or landscape. (CW-UKmags)

Such head noun uses typically construe relations of (sub)categorization and express generic meaning. In (1.43) for instance pietra paesina is classified as a rare subtype of the superordinate category ‘marble’. In modifier uses, on the other hand, the TN is semantically demoted from this taxonomic meaning and takes on various other semantics depending on the type of modifier use. Some examples are (1.44) to (1.53):

1.2 State of the art

17

(1.44)

A little cheaper Ole Works Inn (495 4837) seems like a laid-back sort of a place and sits handily on Cane Garden Bay across the road from Quito's Gazebo. (CW-Times)

(1.45)

This is why im crazy about u. in a non-‘worship and serve me forever’ typa way ;-). (http://www.7thrimofhell.blogspot.com/ 1990/01/blog-post)

(1.46)

I could never afford a boat, nor an oceanside house (Marianne is editor of the Nantucket Inquirer and Mirror) because we don't have that kind of money. (CW-Times)

(1.47)

“When we came under pressure the lads literally threw themselves in front of the ball,” Wilkins said. “If we continue with that sort of attitude we can stay up.” (CW-Times)

(1.48)

Jonathan Cope was the kind of guy who feels slightly ill at ease at parties but manages to look terrific nonetheless. (CW-Times)

(1.49)

The principle of earth sheltering is the same, whether applied to homes above or below the ground. An 18in covering of earth acts as a kind of blanket. (CW-Times)

(1.50)

Then later in the night we took a walk in our underwear around the campus. That was sorta weird. (http://www.yaledailynews.com/ article.asp)

(1.51)

My father was a skilled worker and er he was part of er sort of er a gang which was sort of recruited to be taken overseas for construction of railway. (COLT)9

(1.52)

How do they behave. Well sort of like. Oh it's hard to explain isn't it. It's like. She thinks she's the best but she's not. (COLT)

(1.53)

He kinda went, ‘Yeah, I think so!’ (www.rockconfidential.com/ Testament)

In a laid-back sort of a place (1.44) and a non-‘worship and serve me forever’ typa way (1.45) sort of and typa are phonetically reduced, reflected in writing in the case of typa, and do not indicate a true subtype. Rather, together with the adjective preceding it, the TN-expression is concerned with ascribing an attribute to a concrete hotel in (1.44), or with describing a very specific instance of loving in (1.45). In (1.46) to (1.48) the TN is preceded

18 Chapter 1: Description of the topic and state of the art by a demonstrative or definite determiner, and together with it expresses a phoric relation as well as gives information about identifiability status. In (1.47) that sort of attitude anaphorically refers back to the previous sentence and generalizes over it to set up a contextual type that includes the concrete instance referred to in the previous discourse and potentially extends over more instances not explicitly referred to in the discourse. The same goes for the kind of guy in (1.48); the only difference is that the information for decoding the referent’s identity, which the definite determiner signals to be retrievable, is located in the following discourse, i.e. cataphorically, in the relative clause. In that kind of money (1.46) identification of the NP’s referent is not established by means of phoric reference to qualities, but based on quantitative notions, which have to be retrieved anaphorically in this example, i.e. the amount of money assumed necessary to buy a boat. In (1.49) again not a real subtype of, in this case, a blanket is described, but the principle of earth sheltering is described metaphorically in terms of a non-prototypical member of the category ‘blanket’. The type specification ‘blanket’ is qualified, i.e. the appropriateness of the categorial label is questioned or mitigated (cf. Aijmer 2002, Denison 2002). Examples sorta weird (1.50) and sort of recruited (1.51) are adverbial uses of the TN-expression, similar to those of SNs in examples (1.21) to (1.26) and (1.38). The semantics here are also of the qualifying kind, as in (1.49), but apply to another syntactic context. In (1.52) the TN-expression functions as a discourse marker, similar to like, which is often used with it. It then functions as a hesitation marker, or can, among other things, function as a mere filler. The TN-expression in (1.53) functions as a quotative marker introducing quoted speech. In conclusion, it can be noted that both (S)SNs and TNs take on various functions, centering around head versus modifier status, as attested by synchronic and diachronic corpus data. General reference grammars, such as Quirk et al. (1985) and Biber et al. (1999), however, cannot account for this variation, either synchronically or diachronically. In addition, theoreticallyoriented work, such as transformationalist accounts in the vein of Akmajian and Lehrer (1976) lack the dynamicity necessary to tackle construction types that are essentially on the move within and beyond the NP, with nouns proceeding to take on functions not typically associated with them. Even Halliday’s (1994) metafunctional proposal cannot fully accommodate the different uses of SNs and as such is too static to be descriptively accurate. It is only the cognitive framework proposed by Langacker (1991) and grammaticalization studies by Aijmer (2002), Denison (2002) and Traugott (2010b) that offer a point of departure for a systematic classification of the various uses of SNs and TNs, which allows me to compare both in terms of

1.2 State of the art

19

potential zones of functional overlap with regard to the constructional semantics they can express. Section 1.2 will present a detailed overview of the literature on the subjects of SN-expressions and TN-expressions.

1.2 State of the art: reference grammars and theoretically-oriented literature on size and type nouns This section offers a survey of the relevant literature for SNs and TNs and the constructions in which they function. As noted earlier, SN-expressions and TN-expressions are often discussed together in the literature, but this body of text rarely points out semantic or constructional similarities between the two syntagms. Systematic functional descriptions of each individual construction are likewise hard to find. In the case of SN-constructions the literature most clearly lacks a unified functional-constructional approach and is largely restricted to separate observations scattered over different sections in reference works. In more traditional grammars, such as Quirk et al. (1972) and (1985) for instance, they are not treated as a topic and construction in their own right, but discussed in sections dealing with their deviant subject-finite concord or other agreement patterns. Systematic analyses of the quantifying potential of SNs are largely lacking in existing grammars and are mostly restricted to observing that they are incongruent means of quantification. Because of the joint treatment of SNs and TNs found in more traditional grammars, the literature often overlaps for both constructions. Nevertheless, I will survey the literature for both constructions in two separate sections, as TNs have received much attention in recent years, especially in the field of grammaticalization. In addition there are a small number of specialist studies on both subjects.

1.2.1 Discussions of SNs in the literature The literature reviewed here subsumes traditional word class grammars, such as Quirk et al. (1972) and (1985), as well as some corpus-based grammars, such as Biber et al.’s (1999) Longman Grammar of Spoken and Written English, and several older descriptive grammars, such as Kruisinga (1925) and (1932), Poutsma (1928) and Jespersen (1932) and (1970). Finally, more specialized studies, such as Akmajian and Lehrer (1976), Lehrer (1986) and Vos (1999) will be briefly discussed, all of which are situated in the transformational-generative framework. Cognitive and functional accounts of SN- and TN-expressions will be discussed in Chapter 2.

20 Chapter 1: Description of the topic and state of the art The literature survey will be organized thematically rather than purely chronologically. Studies will be grouped together on the basis of shared general approach and methodology, e.g. corpus-based or introspective, or theoretical framework, e.g. functional or formalist. In each case the proposed structural and semantic analyses of SN-constructions will be discussed, as well as any mention of quantifying function. This status quaestionis will make clear that the literature so far makes for a rather fragmented point of departure, lacking explanation of the multi-layered polysemy of functions attested by SN- and TN-constructions in corpus data (see examples (1.1) to (1.4) and (1.9) to (1.42) above) and mostly treats them as intractable material. Such gaps and inaccuracies in the literature reveal the need for a systematic and comprehensive description of SNconstructions as well as a dynamic interpretation of NP-structure. These will be presented in Chapters 2 and 3. Traditional pedagogic grammars and general reference grammars tend to be class-based or bottom-up, i.e. they work their way up in grammatical structure, starting from the description of word classes. Traditional grammars identify what they alternatively call “quantifying nouns” (Biber et al. 1999: 252) or “measure partitive nouns” (Quirk et al. 1985: 251) as belonging to the class of nouns, but nouns with a special or incongruent function, sometimes referred to as individualization. SNs are hence not related to the quantifying function within the NP, but are treated as marginal nouns or at most colloquially restricted alternatives to the regular quantifiers. Bottom-up grammars are generally at a loss with regard to SNs, which seem to slip through the word class approach in a rather unpredictable way. A case in point of this is found in how they treat conflicting principles of subject-verb concord. In these accounts the SN is by default considered to be the head of the binominal NP. Verb concord is nevertheless noted to fluctuate, in the sense that the finite verb does not always display agreement with the grammatical number of the SN, as in A heap of people are aware of it. This is then explained as a conflict between strict grammatical concord and notional concord or proximity concord, e.g. Quirk et al. (1985: 757-765). The element determining concord is still considered to be the head, i.e. the SN, but different, i.e. semantic, aspects of this head determine the number of the verb rather than grammatical number, e.g. the sense of unity and plurality in heap. Note that the possibility of another element in the NP being the head is not really considered. Only in proximity concord is another element than the presumed head allowed to control the number of the verb, because of its proximity to the verb, but not in any systematic way. The account in terms of conflicting types of concord is very much an a posteriori account that explains with hindsight why verb concord is the way it is. It does not help language users to predict verb concord by offering criteria on which type of concord wins out. It essentially fails to see

1.2 State of the art

21

fluctuating verb concord as a symptom of langue change, rather than a grammatical mistake or figure of speech, and therefore does not accurately assess the impact of SNs on NP and clause structure. Both the Quirk grammars and the older descriptive ones mostly limit their observations to the more restricted set of idiomatic measure noun expressions, such as a loaf of bread, a pack of wolves and an acre of land. Only Kruisinga (1925/1932) has some interesting and useful observations on the specific NP structure of SN-constructions, relating it to the conceptual semantics of of (see Section 1.2.1.1). In other accounts, such as Biber et al. (1999), SNs are merely classified into smaller subsets depending on the source semantics of the SN, e.g. “container”, “shape” and “standardized measure terms” (Biber et al 1999: 252-254). All of this makes for rather fragmentary accounts that do not address the perceived incongruity in terms of the construction’s inherently dynamic semantics. Variable subject-finite concord is not linked up with synchronic variation and/or a diachronic shift in the status of SNs. Of course one has to keep in mind that all of the reference works reviewed in this section aim at providing a comprehensive description of the grammar of English and this aim for comprehensiveness naturally leads to less depth in favour of breadth of coverage.

1.2.1.1 Jespersen (1932/1970), Kruisinga (1925/1932) and Poutsma (1928) The older descriptive grammars by Kruisinga, Poutsma and Jespersen are wonderfully descriptive in their observations of all kinds of patterns and curiosities in the English language, but they are not strongly geared towards functional or explanatory generalizations. Jespersen and Kruisinga both describe SNs as a means of quantifying “mass-words” or “uncountables”, called “individualization” (Jespersen 1932: 117-125 and Kruisinga 1925: 33). Jespersen discusses SNs in a section on types of nouns, whereas Kruisinga first brings them up in a section on numeratives. Jespersen’s and Kruisinga’s remarks are restricted to the set of collocationally restricted measure nouns which individuate specific N2s, e.g. a loaf of bread (instead of a bread). Kruisinga (1925: 33) in addition notes that some SNs “have retained more of their independent meaning” than others, as illustrated by his examples: A little pair of scissors versus a flight of steps. He also notes that these kinds of expressions occur in front of plural nouns, as well as uncount nouns. Jespersen explains incongruous verb concord in SN-constructions as a conflict between grammatical concord and attraction (i.e. proximity concord) (Jespersen 1932: 179). What distinguishes Kruisinga from Jespersen and even some of the more modern linguists is that he observes a specific structural versatility in

22 Chapter 1: Description of the topic and state of the art of when it features in N of N-patterns, such as TN and SN-constructions. This versatility is moreover related to a diachronic dimension (Kruisinga 1925: 396). In such patterns of can either make the noun preceding it or the following noun into an adjunct, i.e. either noun can be subordinated (ibid.: 391-395). He connects this with the semantic subordination of the noun in question and with the specific behaviour of the article and adjectives in such binominal noun phrases. The structural versatility of of in such patterns also translates into versatile concord patterns. Kruisinga, however, unlike Jespersen, does not resort to conflicting concord principles, but explicitly states that the noun following of sometimes determines concord (ibid.: 306). This is accounted for by the statement that “what is formally the adjunct is in thought the dominant member” as in There was heaps of time (example from Waugh’s Loom of Youth cited in Kruisinga 1925: 306). However, no systematic indications are provided for when which element determines concord. Hence in a way one can look at this account as a variation on notional concord conflicting with strict grammatical concord.

1.2.1.2 Quirk, Greenbaum, Leech and Svartvik (1972/1985) and Biber, Johansson and Leech (1999) Both Quirk et al. (1972) and (1985) are general reference grammars in the true sense of the word. Despite some influence from Hallidayan systemicfunctional grammar, the Quirk grammars do not start from the demarcation of functional zones in the NP, such as determination, modification and categorization, in order to work their way down to the word classes that can express these functions. Instead they primarily use the basic constituents of the NP to structure the grammatical and other information they provide. The Quirk grammars hence mainly define and describe word classes. SNs and TNs first appear in the chapter on nouns and determiners. SNs, and numerals in general, are characterized as incongruent members of both the noun category and the quantifier category (Quirk et al. 1985: 73). This incongruent function is called “partition”, i.e. a means to achieve “the expression of quantity and thus countability” of uncountables such as mass nouns (cf. “individuation”). SNs and TNs are furthermore treated together as “partition expressions”, with SNs expressing “partition in respect of quantity” and TNs in terms of quality. For this function SNs and TNs appear in “partitive constructions” (ibid.: 249) and SNs are further subdivided into general partitive nouns (e.g. a piece of cake), typical partitive nouns (e.g. a loaf of bread/a speck of dust) and measure partitive nouns (e.g. an acre of land), ranging respectively from collocationally unrestricted over collocationally restricted expressions to standardized measures in terms of length, area, volume and weight (ibid.: 249-251). The SNs focused on in

1.2 State of the art

23

the present study are also brought up in a section on open-class quantifiers such as plenty of, a lot of and lots of which are claimed to be used chiefly informally. Heaps of and loads of are said to be “roughly synonymous with lots of”, but restricted to “familiar spoken English” (ibid.: 264 footnote). The head status of the latter two SN-expressions is not questioned. In fact, even for plenty, lot and lots head status is only cautiously questioned on the basis of the fluctuating verb concord they display. In the section on subjectverb concord, then, it again seems that Quirk et al. (1985) assume that these open-class quantifiers constitute the head of the NP, since divergence from concord with these open-class quantifiers is explained in terms of notional concord or proximity concord (i.e. “attraction”) overriding strict grammatical concord. The possibility of the noun in the of-phrase constituting the head in at least some cases is not considered (ibid.: 764-765). Observations about fluctuating concord patterns or head status are moreover strictly synchronic. Largely sharing the descriptive framework of the Quirk grammars is Biber et al. (1999), which explicitly presents itself as the data-based complement to the Quirk grammars. The corpus-based nature of the Longman Grammar of Spoken and Written English translates largely into charting statistical differences in language use across regional varieties, such as American versus British English, or genres and registers, such as academic and conversational English (Biber et al. 1999: preface viii and Introduction p. 4). Biber et al. (1999: 252) also approach the quantifier semantics of “quantifying nouns” as a means of expressing quantities of “uncountable nouns and plural countables” (cf. Kruisinga 1925). They add a classification of quantifying nouns in terms of their lexical source semantics, i.e. “nouns denoting type of container”, e.g. barrel of; “nouns denoting shape”, e.g. heap of and wedge of; “standardized measure terms”, e.g. gallon; “plural numerals”, e.g. dozens; and “nouns denoting large quantities”, e.g. a load of, loads of, a mass of and masses of. Quantifying collective nouns, such as bunch of, clump of, pack of and unit nouns such as bit of, piece of, loaf of and scrap of, discussed before the quantifying nouns as such, are said to be special cases of quantifying nouns used with count nouns and uncount nouns respectively (Biber et al. 1999: 252-254); 10 a dichotomy that is contradicted in at least one direction by the corpus data in Chapter 4, in which a bunch of legalese/practicing/time for instance are attested.11 Most quantifying nouns are said to have specific collocates, such as a loaf of bread, while others are noted to have “a more general or metaphorical use” (Biber et al. 1999: 249). In terms of corpus findings this is reflected in lists of frequent collocations for all types of quantifying nouns. In some cases, e.g. bunch of, this gives rise to twin lists as with bunch of daffodils/grapes/bananas versus bunch of idiots/amateurs/perverts, but this

24 Chapter 1: Description of the topic and state of the art collocational variation is not accounted for in terms of a change in the status of SNs, even though the two lists clearly reflect head use collocates on the one hand and quantifier collocates on the other. Only one quantifying noun is said to act like an emergent “quantifying determiner” (i.e. a quantifier), namely loads of, because it systematically displays the same type of subject-verb concord as true quantifiers. Other than that Biber et al. (1999: 250) only alert the reader to “[n]ote the similarity with forms such as a number of and a couple of which are treated as quantifying determiners”. In addition to these two a set of quantifiers with of are also described as “a novelty” and further includes a (great) deal of, plenty of, a lot of and lots of (ibid.: 277), but not loads of or heaps of or any of the other SN-expressions of the present study. The main arguments attributed by Biber et al. (1999) in favour of noun status of most quantifying nouns are that they still have both a singular and a plural form and allow for adjectival modification, as in great bunches of dried herb. Nevertheless, Biber et al. (1999: 250, 252, 257) are aware of the fact that even core quantifiers still allow for premodification in certain contexts, e.g. a select few, and that a lot of and lots of have retained the singular/plural contrast as well. The quantifying nouns are hence claimed to lack decategorialization, i.e. the loss of properties typically associated with a more lexical category, such as the noun class, in processes of grammaticalization (See Section 3.4.3). Quantifying nouns are associated with fluctuating verb concord and are in this context related to “species nouns”, i.e. TNs, because they “are found in patterns which are superficially like those of quantifying nouns” (Biber et al. 1999: 255; see Section 1.2.2.1). As in the Quirk grammars and other reference works discussed earlier, head noun status of SNs and TNs is considered to be the default. Even though Biber et al. (1999) devote a separate section to the specific problem of identifying the head in quantifying noun phrases, the choice between verb concord with the SN or the noun in the ofphrase is, rather unhelpfully, concluded to depend “upon circumstances” (ibid.: 257). Only general tendencies of verb concord for certain groups of quantifying nouns are pointed out, but these do not explain the fluctuation in terms of an actual shift in the status of the SNs. Corpus-based Biber et al. (1999) do not provide frequency counts of this variation in verb concord, but do include information about the distribution of the various species nouns and some quantifying nouns across different registers and genres (ibid.: 255-256).

1.2 State of the art

25

1.2.1.3 Transformational-generative approaches to SN-expressions This section reviews the main analyses of SN-constructions offered in early transformational approaches. Transformational-generative grammar has spent considerable attention to the issue of SN-structures and the problem of determining the head of such syntagms. These are very theory-specific studies, which point out interesting questions and often syntactically wellargued analyses. Nevertheless, they fall short in describing the dynamicity of SN-structures compared to functional accounts. In early transformational grammar considerable attention was devoted to SN-expressions as part of a larger set of (determiner) N + of + (determiner) N-syntagms, which subsumed, besides SNs, structures such as, A review of a new book about cooking, A photograph of a man with three arms, as well as collective noun expressions, e.g. a herd of and a family of, and sometimes TN-expressions, such as kind of (Akmajian and Lehrer 1976: 109, Lehrer 1986: 395-396 and Jackendoff 1981). SNs are primarily analyzed within partitive constructions, i.e. “with a definite of-phrase”, e.g. many of the men. Structures like A bunch of men are called “pseudo-partitives” (Jackendoff 1981: 106 and 119; compare Vos 1999). Overall SNconstructions are presented as “riddled with idiosyncrasies” that form possible counterexamples or exceptions to “rule operations”, such as prepositional phrase extraposition12 (Jackendoff 1981: 103; cf. Akmajian and Lehrer 1976, Selkirk 1977 and Lehrer 1986). This is mostly related to another observed structural peculiarity, namely the problem of determining the head of an NP, which, incidentally, is also the title of Akmajian and Lehrer’s (1976) article on the subject of SNs. In addition transformational accounts have difficulties with assessing the value of of in these structures. This section will focus on reviewing the historically central analyses offered by Chomsky and Jackendoff, as summarized by Langacker (1991: 87), while also referring to additional articles of the same authors and Selkirk (1977), as well as the more semantically-oriented account of Akmajian and Lehrer (1976) and Lehrer (1986). Whereas Jackendoff (1981) deals with SNs under the heading of “NP specifiers” and “group nouns”, Akmajian and Lehrer (1976) refer to them as “NP-like quantifiers” and Lehrer (1986) as classifiers. Up until Chomsky (1970) generativist accounts of a structure such as several of the men treated the noun following of as the head and everything before this final noun as a complex determiner, more specifically a prearticle and an article: [[several of the]DET[men] N] (cited by Jackendoff 1981: 103). In these accounts of is considered to be a more or less meaningless grammatical formative that does not add to the inherent constructional semantics of the structure, but is deleted or inserted transformationally depending on certain constraints (of is not even explicitly represented in the

26 Chapter 1: Description of the topic and state of the art tree diagram in Jackendoff 1981: 107). In this case of has to be inserted because of the definite article. (see Chomsky 1970, Jackendoff 1981: 107). This analysis was untenable, also for Langacker (1991), if only for overlooking the fact that the men also is a constituent in several of the men and because of the disregard for of. This analysis was also rejected within transformational grammar, because it could not explain why extraposition of the of-phrase is possible in Of the leftover turkey a lot has been eaten, because it was not considered to be a constituent. The pre-article analysis was rejected in favour of the proposal made in Jackendoff (1968), who did recognize that the of-phrase functions as a kind of complement to the first noun, which is the head, but again treats of as a meaningless particle. Several men for instance is equated with several of the men by of-deletion. Cognitive Grammar, and Langacker (1991: 35) specifically, has always strongly objected to the postulation of meaninglessness for such elements of structure, since it violates one of its basic semiotic tenets that each grammatical structure has conceptual import, however abstract. Langacker (1991: 35 and 88 and further) hence argues that of does have specific semantic content and profiles an “intrinsic relationship between its trajector and landmark” (cf. Kruisinga 1925). Nevertheless, Jackendoff (1981) keeps to the complement analysis as the most plausible one for “group nouns”,13 a category that subsumes a rather varied set of expressions with more grammatical and more lexical items, e.g. a group of, a lot of, a bunch of, a number of, but also narrowlydefined measure nouns such as a gallon of, a pound of (Jackendoff 1981: 107-108 and 119). In conclusion, transformational accounts have proposed both head and non-head analyses for SN-constructions but they are all synchronic and of the either/or-type, in the sense that in each proposal only one analysis is offered for all SN-constructions (which, incidentally, is the opposite of what Halliday proposes: see Section 2.1.1). The two central analyses are not linked to each other in terms of diachronic developments (but see Akmajian and Lehrer 1976 for some exceptions). However, proposed constituency tests, such as the preposing of the of-phrase and preposition stranding, as well as testing selectional restrictions, etc. can be useful in the case studies, presented in Part 2 (see also Aarts 1998 with reference to Hudson 1987). The structural and semantic dynamicity of SN-constructions hence seriously challenged early transformational generative grammar, which worked with a notion of NP structure that is detached from a functional organization of the NP and from the functions performed by that NP within a clause.

1.2 State of the art

27

1.2.2 Discussions of TNs in the literature The versatile behaviour of TNs already attracted the attention of the great descriptive grammarians of the first half of the 20th century such as Kruisinga and in the mid-to-late nineties they were commented on in readers or in the margins of chapters in historical grammars (e.g. Tabor 1993, Kay 1997, Denison 1998). As opposed to SNs, TN-expressions such as sort of, kind of and type of have also received considerable attention over the last few years in grammaticalization studies and have even become somewhat of a hot topic (e.g. Aijmer 2002, Denison 2002, Margerie 2010). The main focus of this recent body of studies is on qualifying uses, extended discourse marker uses of TNs and other subjectified meanings which differ most clearly from the head use. However, despite the number of articles and papers devoted to them, there is no agreement in the literature about a systematic, formally and functionally motivated classification of all the different uses of TNs, including NP-internal ones besides the head use. In this section I will consider three basic grammatical classifications of TN-patterns which contribute essential insights: Kruisinga (1932), Denison (2002) and Aijmer (2002). These classifications complement each other in interesting ways, because they invoke partly different types of formal evidence and because each author points out patterns not necessarily considered by the others. As we will see, the NP-internal functions of TNs pose the greatest descriptive problems and have been covered least well. By contrast, the uses of TNs involving clear meaning shifts, for instance to qualifying uses (Denison 2002) and discourse particles, have already received much more attention and have been covered better. It is NP-internal functions of TNs that form the main focus of the present study, because these have been covered least well and also because it is the point at which TNs compare most to SNs. Part 2 will present a comprehensive corpusbased description with NP-internal functions as its main, though not exclusive, focus.

1.2.2.1 Discussions of TNs in general reference grammars We have seen in the accounts of SN-constructions that TNs are often brought up in the same context as a variant instantiation of the N of Nscheme. Most general reference works as well as theoretically-oriented grammars do not go beyond noting that TNs can occur in the same structural patterns as SNs and that they display similar fluctuation in their verb concord patterns (e.g. Jespersen 1932: 117-125, Akmajian and Lehrer 1976: 109, Lehrer 1986: 395-396, Jackendoff 1981, Biber et al. 1999: 255258, Halliday and Matthiessen 2004: 333-334). TNs are alternatively re-

28 Chapter 1: Description of the topic and state of the art ferred to as “species nouns” or “varietal classifiers”, or “SKT-nouns” in Biber et al 1999, Allan 1977, Keizer 2001 respectively. If any semantic distinction between SN- and TN-constructions is made, it mainly amounts to reference to quantity versus reference to type or quality respectively. This distinction is either dealt with in terms of two distinct types of “partition” (Quirk et al. 1985: 249 and 764-765) or in a more covert way (e.g. Jespersen 1932: 117-125, Biber et al. 1999: 255-258). In the case of TNs not only verb concord comes into play, but also concord in number between the TN and the noun in the of-phrase, and between determiners in front of the TN and the TN itself (see Denison 2002/2005 in Section 1.2.2.2). Moreover, these types of concord interact. Quirk et al. (1985) point out what they call incongruous examples such as These/Those sort/kind/type of parties are dangerous and describe them as “an idiomatic anomaly”. They provide rephrasal strategies, such as Those kinds of parties are dangerous, “to avoid the anomaly”. The anomalous verb concord in the original structure is explained as “a more acceptable” instance of notional concord overriding grammatical concord (Quirk et al. 1985: 765-766). Just like SNs then, TNs seem to be elements of structure that deviate, in unpredictable ways, from some norm of default behaviour, which calls for a posteriori or ad hoc explanations. The corpus-based Longman Grammar of Spoken and Written English (1999) is more accurate in its description of the incongruous concord patterns displayed by TN-constructions, because Biber et al. (1999) connect it with ambiguous head status of the TN. Nevertheless, in keeping with their descriptive manifesto perhaps, they refrain from making theoretical generalizations over such observations in terms of possible grammatical reanalysis of TNs. They use corpus data mainly to confirm regional and register preferences of the various TN-expressions (Biber et al. 1999: 256), as they also did for SN-expressions, for instance the fact that sort of is used more in British English while kind of is more typical of American English. With regard to concord patterns, Biber et al. (1999: 255) note that normally there is agreement in number between the TN and the noun in the ofphrase, as well as between the determiner preceding the TN and the TN itself. However, there are also instances were a singular TN teams up with a plural noun in the of-phrase and vice versa. More important is the problem of incongruous concord between a plural determiner preceding a singular TN, where the determiner agrees in number with the noun in the of-phrase. This typically also entails incongruous verb concord with the noun in the of-phrase, as in These kind of decisions are normally made by the teacher alone (Biber et al. 1999: 255 and 258). In a subset of these cases the TN is what grammaticalization theory would call clearly delexicalizing and possibly grammaticalizing, e.g. I hate these sort of things, even though Biber et al. do not use these terms. They warily describe the phenomenon as a sub-

1.2 State of the art

29

ordinate use of the species noun “in much the same way as a determiner”. On other occasions TNs are vaguely described as having “a close relationship” with determiners (Biber et al. 1999: 256 and 258) (see Kruisinga in Section 1.2.1.1). Based on corpus findings, Biber et al. (1999: 256) further note two kinds of extended functions for singular sort of and kind of “in conversation and fiction”. Besides “serving the purposes of exact definition” “in academic prose”, these TN-expressions can be used to “introduce greater vagueness”, as in It’s a very difficult sort of situation; A silly sort of drink really; There’s a kind of mystery here, wasn’t there?. In addition to this use, they can also be used “more generally as vagueness markers, or hedges”, as in I kind of danced into work and Yes, yes, it’s sort of all a bit naked isn’t it?. In the latter two cases the species nouns are not regarded as nouns anymore, but hedges, though without any explanation in terms of grammaticalization or an explicit reanalysis claim (ibid.: 256-257). The corpus studies in Part 2 will make clear that this set of hedging uses as well as the set of singular TNs with plural determiner and plural noun in the of-phrase in fact subsume more types of grammaticalized TN-functions than observed in Biber et al. Can There’s a kind of mystery here and I kind of danced into work be said to be semantically different, apart from the element over which they have scope? Is It’s a very difficult sort of situation truly a vagueness marker, is it a head use of the TN, or is it yet another TN-use? Biber et al.’s distinctions do not provide clear answers to these questions and lack delicacy.

1.2.2.2 Denison (2002/2005) on SKT-patterns Denison (2002/2005) offers a descriptive account of TN-constructions that does attempt to arrive at a more unified and grammar-based classification of the various kinds of TN-uses attested in actual corpus data. Denison’s (2002) conference talks on “SKT patterns in present-day English” are presented “in an informal construction grammar framework” (Denison 2002: 1 and 2005: 1).14 He posits two, or possibly three, basic constructions with sort, kind and type of in present-day English, and some “semiconventionalised variants” (2002: 3) of these basic patterns, with the adverbial construction as the most important one. He also considers historical pathways of change, linking up the basic SKT patterns diachronically and thereby explaining their emergence in English grammar. The three basic constructions, which are distinguished on the basis of clusters of syntactic, semantic, formal and discourse features (Denison 2002: 2), are represented in the following table (N1 refers to the TN and N2 to the noun following of):

30 Chapter 1: Description of the topic and state of the art Table 1.1: Denison’s (2002: 3) classification of sort/kind/type of-constructions Semantic head

Discourse function

N1

N1 number

E.g.

Binominal

N1

Discourse topic or anaphor

sg. or pl.

the sort of material

Qualifying

N2

Hedge

sg.

Postdeterminer/ complex determiner

?

Anaphoric

sort kind type sort kind sort kind type

A sort of holiday these sort of skills

sg.

The first construction is called binominal because the TN functions as a full noun, which is the head of the noun phrase, while the of-phrase with N2 functions as a postmodifier to that head, e.g. (1.54)

Collagen is the sort of material that is found already [...] in the dermis of the skin. (Denison 2002: 2)

According to Denison, N1 and N2 typically agree in number in this construction and premodification of N2 is rare (cf. Biber et al 1999 in Section 2.2.1). “Apparent premodification of N1 is really premodification of the whole construction” (Denison 2002: 2). In the qualifying construction, sort and kind + of form a unit which “qualifies”, i.e. mitigates, the categorial meaning of N2 for ironic or other purposes, as in (1.55)

But it I suppose it’s as a that’s as a sort of holiday, kind of doing you know doing nothing but sitting around (Denison 2002: 2)

According to Denison, only kind and sort (in their singular form) are used in this construction, while type does not appear in it. In the qualifying construction N2 is considered to be the head of the noun phrase. The postdeterminer or complex determiner construction has an uncertain status in Denison’s current analysis, both in terms of its independence from the binominal and qualifying construction, and the status of the TN in it. It is distinguished mainly on the basis of number incongruence between a singular TN on the one hand, and a plural anaphoric determiner and plural N2 on the other, as in

1.2 State of the art

(1.56)

31

I mean I don’t associate you with uh you know one of these sort of skills like like driving. (Denison 2002: 3)

The construction is noted to be “common only in speech” (Denison 2002: 2). Denison leaves it open whether it is the TN or N2 that constitutes the head of the noun phrase in postdeterminer constructions. More fundamentally, he raises the question of whether this really is a distinct third pattern, or merely a reanalysis of the binominal construction with singular N1+of+plural N2. It is also said to bear resemblances to both of the previously described basic SKT-constructions (Denison 2002: 2 and 11). The additional “semi-conventionalised” patterns identified by Denison are viewed as intrinsically variants of the basic constructions. Two of these are particularly productive. Firstly, there are several variants of the qualifying structure, in which there is no N2 and in which the sort/kind of string modifies adjectives or verbs and sometimes adverbs or prepositions instead of nouns, as in (1.57)

I sort of saw his point. (Denison 2002: 3)

Denison calls this the adverbial construction. The adverbial use of sort of has also given rise to a semantically bleached discourse marker use, as in (1.58)

As I remember it used to be sort of like fairly common for a Tuesday [...]. (Denison 2002: 4)

The meaning of this discourse marker use is similar to that of colloquial like in examples such as (1.59)

I should have lLLke just whipped up this amazing meal. (cited in Denison 2002: 4)

Example (1.58) shows that both discourse markers often occur together. Denison is again not sure whether this bleached construction is a separate construction or not (Denison 2002: 4). In addition to the adverbial construction, there is a second main “semiconventionalised pattern” only found with type (of), which Denison (2002: 4) refers to as the semi-suffix use, as illustrated in (1.60). In these uses the TN comes to function more or less as a qualifying suffix that attaches to a previous description: (1.60)

what you’re saying is we need multiple type of. I mean ideally we need a multiple type building [...] sorry a building with multiple type rooms. (cited in Denison 2002: 4)

32 Chapter 1: Description of the topic and state of the art Other variants are considered as “partial lexicalizations” (Denison 2005: 2), which include sort/kind/type of thing, as in (1.61)

If he had visitors he used to send her out of the room type of thLLng. (cited in Denison 2002: 3)

(1.62)

It’s been a an about town sort of thLLng. (cited in Denison 2005: 2)

In these examples the sort/kind/type of thing phrase typically takes up clause-final position and functions as “adverbial hedge” according to Denison (2002: 3 and 2005: 2). Those strings are considered to be lexicalized instances of the binominal construction (Denison 2002: 6) Denison (2002/2005) also points out expressions such as the phrase all sorts of, which has quantifier meaning similar to ‘much, many’, as in (1.63)

And the fact was causing all sorts of gossip and unrest. (Denison 2005: 2)

The semantic shift in this pattern is one from ‘many kinds’ to ‘many’ as such (Denison 2005: 2). None of these semi-conventionalized structures receives an explicit analysis. One can assume that they retain the analysis of the construction of which they are a variant. It can be concluded that Denison distinguishes two basic sets of TNconstructions. On the one hand, there is the binominal construction in which the TN is used in its original sense of ‘subclass’. On the other hand, there is the qualifying construction in which sort/kind of have a hedging function; this pattern has led to the adverbial construction and the bleached discourse use. In the historical part of his paper Denison proposes several pathways of change, relating the basic construction types (as well as the adverbial construction) diachronically. This historical account also looks at other TNs, such as manner, kin, variety, class, brand, species and category, which can or could all appear in Determiner N1 of N2-patterns (except kin) and several of these show the number incongruence typical of Denison’s postdeterminer category (Denison 2002: 7). Germanic words such as kind and kin already acquired a general ‘class’ meaning in Old English; sort was borrowed from French in the Middle English period and has general class meaning from the late 14th century onwards, while type does not acquire its general class meaning until the late 19th century (Denison 2005: 3-4 and Oxford English Dictionary; henceforth OED). When the SKT-nouns in their general class meaning team up with an of-phrase, SKT-constructions as such emerge. Denison then puts forward a number of hypothetical pathways of change from one type of SKT-construction to another. The idea is that older stages

1.2 State of the art

33

persist alongside stages with newer constructional meanings and constructions may have multiple parents. The first pathway moves from the binominal construction to the qualifying construction by means of a generalization or defocusing of the meaning of the SKT-noun, through conversational implicature, to the point where it indicates a peripheral or possible member of a class. Even though people might feel this to be a recent use of SKT-constructions, it in fact already appeared in the 17th century. Since type only became part of the lexical set of class meanings in the 19th century, its inability to feature in qualifying constructions might be explained by it arriving too late to take part in the change. It generally shifted later than the other TN-expressions. Another pathway sees the complex determiner construction as being derived directly from the binominal construction by means of a reanalysis of the binominal pattern D N1sg of N2pl with a determiner that is not marked for number as in: (1.64)

as giving him some kind of hopes that… (1627 [CEECS]) (Denison 2002: 11)

The adverbial construction then is hypothesized to have arisen from the qualifying construction, where the hedging semantics are simply syntactically extended to non-nominal elements in the of-phrase. Another pathway suggests a change from the complex determiner category to the qualifying and adverbial construction, mainly put forward because of the early attestation of the complex determiner construction in historical corpus data, preceding both qualifying and adverbial uses. Because these three constructions furthermore share the fact that of belongs to sort or kind, Denison (2002:12) sees the earlier complex determiner construction as potentially feeding into the later two.15 Questions to do with the diachronic development of TN-constructions will be returned to in Part 2 which subsumes a detailed diachronic corpus analysis of sort(s), kind(s) and type(s).

1.2.2.3 Aijmer (2002) Using different labels and introducing further distinctions, Aijmer (2002) draws a distinction between different TN-uses which is similar to the fundamental distinction made by Denison between the binominal construction and the qualifying construction. Aijmer zeros in on the various qualifying uses of TNs and NP-external discourse marker uses. On the one hand, she distinguishes the basic NP structure consisting of a determiner and a TN, used as head, and a postmodifier: [a [sort/kind/type

34 Chapter 1: Description of the topic and state of the art [of NP]]], as in A robin is a sort of bird. This head use, she notes, encodes the meaning ‘X is a hyponym of Y’ (Aijmer 2002: 176), in which robin is the X-element and bird is the Y-element. On the other hand, the strings sort of and kind of fulfil a great variety of what are called “particle” uses by Aijmer (2002: 176). She (2002: 180) views these particle uses as the result of grammaticalization, pointing to formal evidence such as the bonding of of to the TN and its phonological reduction as reflected in spellings like sorta/kinda. Their different meanings come about through processes of pragmatic enrichment (cf. Traugott 1988, 1995a), in which the literal type meaning has shifted to a hedging meaning and further, through subjectification, to an affective discourse particle. In the particle uses, the scope of sort of/kind of has also extended to other syntactic environments than the NP, such as adjective phrases, verb phrases, and even whole sentences. Aijmer (2002) mainly restricts herself to observations on sort of and kind of, since these are the only ones that have developed systematic particle uses. Within the particle use Aijmer makes a general distinction between evidential and affective function. As an evidential marker sort of functions as “an adjuster word, as an indicator that the following word or construction functions on a different level of talk (the meta-level sort of), as an indicator of number approximation, signalling a lexical gap or lexical imprecision or as a self-repair signal” (Aijmer 2002: 192). The meta-level sort of subsumes a quoting use, introducing reported speech or onomatopoeic phrases (1.65), as well as a use marking a “special idiom” in front of “technical, rare, foreign, formal, vulgar, idiomatic” words, or ad hoc invented phrases. In all cases the meta-level sort of signals that the following word or phrase has a special status or does not belong to the regular vocabulary of the speaker, as in (1.66): (1.65)

[…] was still - pretty wealthy – I’ve never seen a sort of bottle after bottle, sort of pop pop popping all the time- -- and everybody got awfully drunk I remember. (Aijmer 2002: 186)

(1.66)

Sorry I didn’t mean to I didn’t mean to sound sort of prissy. (Aijmer 2002: 195)

The second major particle use is the affective, or interpersonal, use, in which sort of mainly serves to hedge the illocutionary force of utterances in order to avoid disagreement with the hearer and instead claim common ground (Aijmer 2002: 191). Further submeanings include detensifying or downtoning; compromising; signalling intimacy between the speakers, or hedging of strong opinions, as strategies of positive and negative politeness

1.2 State of the art

35

respectively (Aijmer 2002: 202-207).16 An example of the affective use is (1.67) cited in Aijmer (2002: 200): (1.67)

And I haven’t contacted her and she hasn’t contacted me. I haven’t really felt I wanted to cos it was a little sort of rather unpleasant in the end and it’s a shame really, cos we you know before […]

Formal indications for non-head status of particle uses, though by no means foolproof, are sorta and kinda spellings, graphologically reflecting the bonded status of the strings. Stress patterns similarly differ in head uses and particle uses, with primary stress on the TN (or its determiner or premodifier) or on N2 for head uses, while TN-expressions in particle uses are typically unstressed or only medium-stressed (Aijmer 2002: 177). The presence of hesitation pauses between the TN and the noun in the of-phrase is taken to be another indication of particle status. Finally, the co-occurrence of TNs with other discourse particles is also seen as further corroboration of particle status of the TN itself, as in (1.68): (1.68)

I was sort of you know heaving sighs and tutting. (cited in Aijmer 2002: 204)

In conclusion, the main contribution of Aijmer’s study lies in her finegrained description of the various particle uses of sort of, observed in her data, which she accounts for within a coherent framework of grammaticalization, semantic shift and pragmatic enrichment. She does not, however, discuss the non-head NP-internal functions of TNs that span the grey area in between head and particle uses. Denison (2002/2005) already pointed out a postdeterminer use, and Kruisinga (1932), who will be discussed in the next section, observes yet other NP-internal uses and adduces formal reflections of these patterns as well.

1.2.2.4 Kruisinga (1925/1932) Contemporary descriptions of TNs seem to assume head status for all the semantically more neutral TN-uses in the NP, as in Denison (2002/2005) and Aijmer (2002). It is only for uses which have undergone a very clear semantic shift such as Denison’s (2002) qualifying construction that a concomitant grammatical shift from head to modifier is posited. As we saw, Denison (2002/2005) hesitates to view the postdeterminer use as a distinct construction for instance. By contrast, Kruisinga (1932), in his early treatment of TN-uses, did not only discuss their nominal head use as well as their ability to function “as

36 Chapter 1: Description of the topic and state of the art adjuncts to verbal forms […], and to adjectives” (1932: 399), but he also identified non-qualifying modifier uses within NPs. Whereas in head uses the TN functions as an “independent element of the group” (1932: 395), it is “entirely subordinated in meaning” (1932: 395) in modifier uses such as (1.69)

What sort of a man is he to see? (The Strange case of Dr Jekyll and Mr Hyde, by Robert L. Stevenson 1886, quoted in Kruisinga 1932: 178)

(1.70)

What sort of weather are we going to have? – It doesn’t look very promising at present, but you never know! (Spoken English, by Collinson, quoted in Kruisinga 1932: 178)

In these examples, Kruisinga (1932: 178) notes, what sort of is used to inquire about the quality of persons or things. In other words, what sort of asks the hearer to attribute a quality to the entity designated by the noun following sort of, such as ‘not very promising’ in the case of weather in (1.70). Parallel with his description of semantically subordinated uses of TNs, Kruisinga also discusses a number of grammatical corollaries of their modifier status. The shift from head to modifier status of TNs is made possible by the special nature of the preposition of, which “can sometimes make a preceding noun (instead of the following noun) into an adjunct” (Kruisinga 1932: 391). Kruisinga points out these two possible structural analyses for nominal expressions with both "species" nouns, i.e. TNs, and "measure" nouns such as dozens of. In general, Kruisinga (1932: 397) points out that “the subordination of sort, kind, manner” also influences their stress. As a modifier, the TN is typically only medium-stressed (Kruisinga 1932: 397). Another formal characteristic of the modifier status of TNs is that adjectives in front of the TNs have scope over N2 (i.e. the noun in the of-phrase) (Kruisinga 1932: 397), as in (1.71)

But Kezia bit a big piece out of her bread and dripping, and then stood the piece up on her plate. With the bit out it made a dear little sort of a gate. (Bliss, by Katherine Mansfield 1920, quoted in Kruisinga 1932:397)

Kruisinga (1932: 398) also takes number incongruence between the TN and the determiner preceding it is as a general sign of the modifier status of the TN, e.g. (1.72)

It is a charming talent: all manner of arts and graces proceed from it. (The Times, 29 August 1913, quoted in Kruisinga 1932: 398)

1.2 State of the art

37

Finally, Kruisinga (1932: 396) regards the presence of an “indefinite article before the prepositional noun” as an indication of head status of the second noun, as in what sort of a man (1.69), a dear little sort of a gate (1.71) and (1.73): (1.73)

He is a good sort of a fellow after all. (The Last Chronicle of Barset, by Anthony Trollope, 1867, quoted in Kruisinga 1932: 396)

In conclusion, there is general agreement on the fact that the qualifying use of TNs involves a shift to modifier status. However, only Kruisinga (1925/1932) unhesitatingly posits a head to modifier shift for some nonqualifying uses within the NP, such as an attributive use of TNs. Denison (2002/2005) adds interesting observations about postdeterminer, quantifier, adverbial and semi-suffix uses as possible NP-internal TN-uses, without, however, providing structural analyses of these.

1.2.3 Conclusions: a joint treatment of SN- and TN-constructions The joint treatment of SNs and TNs is deeply entrenched in the literature, yet it is largely unaccounted for in terms of fundamental similarities in their respective constructional semantics. This raises the question of whether such a combined account is desirable, let alone possible, in the present study. The main focus of this study is not a search for commonalities between SN- and TN-constructions per se, but the development of a theoreticaldescriptive approach which does justice to the constructional features and reflexes of diachronic processes such as grammaticalization in each construction. This framework should naturally bring out a number of similarities between both constructions, which then have to be accounted for. A straightforward one is the similarity in structures that serve as input for the delexicalization and grammaticalization processes, i.e. the binominal schema introduced in Section 1.1. SN- and TN-constructions both display a fundamental shift from head to modifier status that lies at the core of these processes, reflected as structural and semantic ambivalence in synchronic corpus data. I want to argue that the synchronically coexisting uses of SNs and TNs constitute semi-stable layered (Hopper 1991) systems of head, modifier and ambivalent uses. Different subfunctions have to be distinguished for the various modifier functions, both for SN- and TN-constructions. It was already clear from the overview in Sections 1.2.1 and 1.2.2 that SNs have led to fewer layers of modifier uses than TN-expressions. For (S)SN-

38 Chapter 1: Description of the topic and state of the art constructions the grammaticalized modifier functions seem to be situated primarily in the determination zone of the NP, more precisely quantification. For TN-constructions, in addition to discourse marker uses and quotatives which are external to the NP, the following NP-internal uses have been pointed out: postdeterminer, quantifier and qualifying uses (Denison 2002/2005), as well as an attributive use (Kruisinga 1932). The first two uses relate to both types of determination in the NP, i.e. identification and quantification, whereas qualifying uses and the attributive use tie in more with the modification zone. However, the NP-internal modifier uses have not yet been analyzed systematically in terms of semantic and structural features, and also pose the greatest descriptive problems (cf. De Smedt, Brems and Davidse 2007 and Brems & Davidse 2010a). They are not only important uses observed in synchronic data, but they also constitute important steps in the diachronic development of TN-constructions. The uses of TNs that involve clear meaning shifts have been covered considerably better, e.g. qualifying and discourse marker uses in Denison (2002/2005) and particle uses in Aijmer (2002). These uses mostly involve syntactic extension to other word classes for the N2-slot, e.g. He was sort of weird (CW-UKmags) in adverbial constructions (Denison 2002/2005). Following De Smedt, Brems and Davidse (2007), I will introduce further distinctions within the set of qualifying uses in terms of the element of structure over which the TN has scope, i.e. nominal qualifiers, adverbial qualifiers and sentential qualifiers. With each extension to new syntactic environments, TNs increasingly rid themselves from the structural constraints of the NP, AP or VP. As discourse markers, TN-strings are not tied to grammatical class boundaries anymore, lack clear notions of scopal domain and can be scattered throughout the discourse. Hesitation markers and fillers are alternative functions of discourse markers. In addition to the source-construction and the fundamental grammaticalization from head to modifier status that they share, the question arises whether SN- and TN-constructions display additional deeper-lying analogies and whether they bring to the fore aspects of grammaticalization and other processes of change that are not yet included in present theorizing. It is a fact that until recently the NP as a locus of change and phenomena characteristic of it have not received much attention, for instance in terms of its structural and collocational individuality. Do SN-expressions and TNexpressions allow for similar pathways of change in their grammaticalized and subjectified uses, i.e. do they show zones of constructional overlap in terms of the types of meaning they can construe and the way in which they construe it? This is suggested by sets of corpus data of SNs, SSNs and TN. For instance, both the SSN-expression bit of and the TN-expressions sort of and kind of can construe hedging meaning, as in examples (1.74) and (1.75) respectively. They can also express similar qualifying or approximator

1.2 State of the art

39

uses, in the sense of referring to non-prototypical members of a category, e.g. (1.76) and (1.77): (1.74)

I thought Leeds would keep a clean sheet more often than not, but they have had a bit of a dip in fortune. (CW-Times)

(1.75)

I haven’t contacted her and she hasn’t contacted me. I haven’t really felt I wanted to cos it was a little sort of rather unpleasant in the end. (COLT, quoted in Aijmer 2002: 200)

(1.76)

I am a bit of a gypsy. For the last four years, I haven't lived in one place longer than three months." (CW-OZnews)

(1.77)

I hear he’s sort of a mystic. I hear he's sort of crazy. I hear he's unpredictable. (CW-USbooks)

SN-expressions such as a bunch of and a load of and TN-expressions can feature in similar emphatic constructions, such as what + (a) +TN/SN +of + (a) + N2, e.g. (1.78) and (1.79) (1.78)

What a load of crap you write! Your article was Naff, Numb, and Bland! You can't write. (CW-UKmags)

(1.79) What kinda jerk would I be if I simply turned my back on her? (www.bigblack.blogspot.com/2005_01_01) In what a load of crap (1.78) the SN-expression seems no longer concerned with specifying the size of crap, but rather further negatively evaluates it. Similarly, in what kinda jerk (1.79), kinda does not refer to an actual subtype of the superordinate type ‘jerks’, but further emphasizes the negative concept. Hence, at least hedging, qualification and intensification of the type specification can be expressed by both SN- and TN-constructions. This means that they are constructional semasiological variants of these particular onomasiological meanings. In this respect the SSN-expression bit of (a) seems to form a hinge between SN- and TN-constructions, displaying many of the uses referred to. Chapter 2 will set out the framework needed to arrive at a comprehensive description of all the layered (S)SN and TN-constructions. This subsumes setting up an adequately dynamic model for the internal organization of the English NP, as well as a specific interpretation of the grammaticalization framework and the factors at work in it. As the literature survey in Chapter 1 has shown, many frameworks cannot cater for the dynamic prop-

40 Chapter 1: Description of the topic and state of the art erties of the NP as a locus of change. In Chapter 2 a dynamic model of the NP will be set up that can interact with processes of change such as grammaticalization, delexicalization and (inter)subjectification. This framework will be functional-cognitive in nature and enhanced with a strong constructional pillar. Constructional models of grammar, such as Construction Grammar (e.g. Goldberg 2006), allow me to describe the semantic and structural properties of SN- and TN-patterns as true form-meaning correlates. The concern for the multidimensional description of the specifics of constructions as well as for linking up construction types in larger networks is essential for the theoretical-descriptive purposes of the present study. The constructional approach opted for will be an eclectic one which meshes with the diachronics of grammaticalization theory, and collocational notions, and displays a usage-based character in view of the corpus studies. Chapter 2 will present the functional-cognitive aspects of the framework as well as its construction-based properties. Chapter 3, then, will integrate this with grammaticalization, (inter)subjectification and delexicalization. In a way Chapters 2 and 3 can be seen as an extensive glossary of the technical terms used in the corpus studies in Part 2.

Chapter 2 A cognitive-functional constructional framework for the English NP 2.1 Requirements of a framework for SN- and TN-constructions This section will set up a framework for the English NP that allows for a dynamic description of SN- and TN-constructions and has general validity for other NPs. Such a model has to meet a list of requirements in order to be sufficiently flexible, both synchronically and diachronically, and truly account for the NP as a locus of variation and change. Most importantly, it has to be able to accommodate non-canonical codings of such functions as quantification and categorization. It also has to allow for emergent codings of functions that come about through diachronic shifts in lexical and grammatical status, and more general mechanisms of change. The model of the NP furthermore has to account for syntactic phenomena as well as incorporate semantics and pragmatics in a broad sense, which includes lexical syntagmatic relations such as collocational patterns and semantic prosody, encyclopaedic knowledge, polarity sensitivity, expressivity motivations, prosodic features, sound symbolism, etc. The latter three phenomena are often considered to be extra-linguistic, but can have descriptive and explanatory power within a sufficiently dynamic and functional framework. If well argued, they can clarify restrictions in the use or developments of certain SNs/TNs in comparison to others, when other parameters and conditions are equal. Ultimately, a model of the NP has to treat structures as essential formmeaning pairings, i.e. constructions, which might display interrelations with other constructions. The more traditional class-based approach to the NP, such as Quirk et al.’s (1972) and (1985) descriptions, are hence not suitable. Such grammars describe the NP in terms of the typical word classes that feature in it, which are then rather rigidly correlated with their functions, such as article and determiner, adjective and attribute, noun and head. In such an outlook, non-canonical realizations of nominal functions and elements on the move within the NP, such as SNs and TNs, are referred to as exceptions or anomalies as we have seen in Chapter 1. My eclectic cognitive-functional framework of the NP will be based mainly on Langacker’s (1991) account of the NP in terms of four basic functions, underpinned by a dependency structure, which allows elements to shift from head to modifier status. Additional input comes from observations from various functional grammars, such as Halliday (1985a/1994), Bache (2000) and McGregor (1997). I will voice some criticism against

42 Chapter 2: A cognitive-functional constructional framework certain specific observations of Halliday and Langacker on SN- and TNconstructions, but will also argue that other parts of their approaches to the organization of the English NP are crucial for the kind of cognitivefunctional NP-model that I will be using in the present study.

2.1.1 Halliday’s systemic-functional approach In An Introduction to Functional grammar (Halliday 1985a, 1994 and Halliday and Matthiessen 2004), Halliday proposes a systemic-functional approach to grammar in which he distinguishes three functional components of meaning or metafunctions, i.e. the ideational, interpersonal and textual. These metafunctions “are realized throughout the grammar of a language” and each in its own way structurally organizes a particular type of meaning in the grammar of a language (Halliday 1985a: 158). The ideational function is concerned with meaning as representation of our experience of the world, more commonly referred to as propositional content. The interpersonal metafunction structures meaning as exchange between speaker and hearer and is concerned with, for instance, organizing speech function and the expression of attitude. The textual metafunction, then, deals with meaning as message and is concerned with relating it to the surrounding discourse and the context of situation, for instance by means of theme structure, given/new ordering etc. At the level of the, mainly nominal and verbal, group17 these are not represented in the form of separate structures for each metafunction (as Halliday claims is the case in the grammar of the clause); rather, each metafunction makes a partial contribution within one and the same structure. With regard to the topic of SN-constructions I will restrict the discussion to the analysis of the nominal group in terms of the ideational component. When analyzing group structure, Halliday further splits this ideational level up into two subcomponents of meaning, namely the experiential and the logical layer, both of which have to be simultaneously taken into account in order to grasp the full meaning expressed by the nominal group. This twofold analysis is central to Halliday’s observations of what he calls “measure nominals” (Halliday 1985a: 173) and will at the same time be the main point of criticism with regard to his approach in the present study. The experiential structure of the nominal group is “multivariate”, i.e. it constitutes a constellation of distinct functional slots, such as Deictic, Numerative, Epithet and Classifier, which each in their own way characterize the Thing of the nominal group. The Thing in its turn is “the semantic core of the nominal group” and specifies “(i) a class of things, and (ii) some category of membership within this class”. (Halliday 1985a: 159-172). Figure 2.1 visualizes the experiential structure in its maximal instantiation

2.1 Requirements of a framework for SN- and TN-constructions 43

with two epithets, i.e. qualitative adjectives, and a qualifier following the Thing, as analyzed by Halliday (1985a: 159): Deictic Those

Numerative two

Epithet 1

Epithet 2

splendid

old

Classifier electric

Thing

Qualifier

trains

with pantographs

Figure 2.1. The experiential structure of the nominal group

The functional slots visualized in Figure 2.1 include the Deictic, Numerative, Epithet, and Classifier. Classifiers indicate a subclass of the class indicated by the Thing and are typically non-gradable, e.g. electric in electric trains, whereas the Epithet indicates a typically gradable quality of the designated subset, either by referring to an objective property, e.g. old trains, or subjectively by expressing the speaker’s attitude towards the subset, e.g. splendid trains (Halliday 1994: 184-185). The Numerative ascribes a “numerical feature” to the subset in terms of quantity or order (Halliday 1994: 183). The Deictic, then, serves to indicate “whether or not some specific subset of the Thing is intended”, i.e. it signals its identifiability, or it serves to identify the subset in terms of deictic features such as proximity and possession by means of demonstratives, possessive determiners and genitives respectively. In addition, this primary Deictic may be accompanied by a secondary post-Deictic, not represented in Figure 2.1, which aids identification of the subset by giving information about “its fame or familiarity, its status in the text, or its similarity/dissimilarity to some other designated subset” (Halliday 1994: 183). It typically immediately follows the primary Deictic. Examples are same and usual in the same two trains and his usual silly self (cited in Halliday 1994: 183). This post-Deictic is generally more lexical than the primary determiner. The postnominal Qualifier, then, also further characterizes the Thing; its function remains more vaguely defined, but seems to be rather broad (Halliday 1994: 187-188). Basically, the analysis of the experiential structure of the nominal group is one in terms of constituency, i.e. part-whole relations (cf. McGregor 1997: 58-59). Halliday's fine-grained semantic characterization of the functional slots in the NP will be crucial in the description of SN- and TN-uses in the corpus studies and in arguing for distinct constructional uses of TNs and SNs. It is important to distinguish between the various premodifiers within the (recursive) modifier structure of the English NP, since they enable or accompany much of the diversity of head/modifier uses of SNs and TNs (see Chapter 6 on TNs). I will largely follow Halliday’s semantic characterizations of premodifiers, distinguishing between Deictic, post-Deictic, Numerative, Epithet and Classifier. Instead of Epithets, Classifiers, Numera-

44 Chapter 2: A cognitive-functional constructional framework tives and (Post-)Deictics, I will refer to qualitative adjectives, classifying adjectives or nouns, quantifiers and (post)determiners respectively. In keeping with Langacker (1991), the present study furthermore argues that determiners subsume both identifiers and quantifiers. As we will see in Chapter 6, the semantic differences between classifying and qualitative adjectives for instance, their formal reflexes and how they are bracketed within the NP are important to differentiate between the head use of TNs, e.g. Pneumocystis carinii pneumonia is a very rare type of infection and the attributive modifier use, e.g. He’s an ordinary sort of bloke. The logical structure of the nominal group is one in terms of dependency. Halliday calls it a “univariate” structure because he views the recursive head-modifier relation displayed throughout the nominal group as one and the same generalized logical-semantic relation. Whereas the multivariate structure of the experiential layer focuses on the different functions expressed by different word classes, the logical structure focuses on “the recurrence of the same function”, namely the head-modifier function (Halliday 1985a: 170-172). The logical structure is visualized in Figure 2.2 (cf. Halliday 1985a: 170): Those

ȗ

two

İ

splendid

į

old

Ȗ

electric

trains

Modifier

Head

ȕ

Į

Figure 2.2: The logical structure of the nominal group (Head modifier)

How these two principles work in the nominal group is explained in the following quote: Let us consider the same example, but this time starting with the most general term, trains. Moving to the left, we get: (which trains? —) electric trains; (which electric trains? —) old electric trains; (which old electric trains? —) splendid old electric trains; and so on. Calling trains the Head, we can represent this [LB: as in Figure 2.2], using the letters of the Greek alphabet. The basis of the subcategorization of course shifts as we move to the left: ‘what type of…?’ ‘what quality of…?’, ‘how many …?’ and so on — this is the principle underlying the experiential structure [LB: cf. Figure 2.1]. Here however we are not concerned with the differences but with the similarities: with the general relationship that runs throughout the pre-Head modification of the nominal group, whatever the experiential function of the individual elements. (Halliday 1985a: 170)

2.1 Requirements of a framework for SN- and TN-constructions 45

Figure 2.3 is another example of the logical structure of the nominal group, in this case with both premodifiers and postmodifier, which are seen as two distinct chains of modification, each dependent on the Head (cf. Halliday 1985a: 172). Modifiers can themselves be further modified by submodifiers. Ȗ

Premodifier ȕ

Head Į

Postmodifier ȕ

Those

electric

trains

with pantographs

Figure 2.3: The logical structure of the nominal group (Head-Premodifier/Postmodifier)

In Figure 2.3 trains is the Head, which is premodified by Those and electric and postmodified by with pantographs. Both modifiers express the same function of dependency with regard to the Head, which is the logical core of the nominal group. In nominal groups that express a measure of something, i.e. “measure expressions” or “ ‘measure’ nominals”, “which include collectives, e.g. a pack of cards; partitives, e.g. a slice of bread; and quantitatives, e.g. a yard of cloth” (Halliday 1985a: 173)18 the logical Head and experiential Thing do not coincide, which they normally do. With regard to the status of and relation between the two levels of analysis within measure nominals Halliday (1985a: 173) observes that the measure word (e.g. pack, slice or yard) functions as the Head in the logical structure, with the of-phrase functioning as its Postmodifier. The Thing, however, is not the measure word but the thing being measured: here cards, bread, cloth. The measure expression functions as a complex Numerative. In the experiential structure, therefore, it is the Numerative that is embedded. (Halliday 1985a: 173)

Figure 2.4 represents the twofold analysis of measure nominals, which leads to a discrepancy between the dominant elements on the two levels of structure (cf. Halliday 1985a: 174; Halliday and Matthiessen 2004: 332): A

pack Numerative

Premodifier

Head

of

cards Thing

Postmodifier

Experiential structure Logical structure

Figure 2.4: Internal structure of a nominal group with measure expression

46 Chapter 2: A cognitive-functional constructional framework Halliday (1985a: 172-173) furthermore argues that [i]t is not that one [analysis] is right and the other wrong; but that in order to get an adequate account of the nominal group and a concept of what is meant by a ‘group’ as the grammatical resource for representing things, we need to interpret it from both these points of view at once.

Halliday’s twofold analysis in a way acknowledges a sort of shift in the structural status of SNs and hence seems to allow some flexibility in their structural interpretation. However, Halliday does not consider both levels as displaying a diachronic shift from an original logical analysis to an experiential one, i.e. from former head status to (quantifying) modifier status. Instead his analysis merely posits a synchronic ambivalence in each instance of an SN-construction. What is unhelpful about the proposed analysis is that Numerative and Head status are divided over two simultaneous levels of analysis, thus suggesting that in each use the SN is always both, and that both the SN and N2 are equally dominant elements of structure. This analysis is in some ways reminiscent of the conflicting concord explanation offered in general reference grammars. A more recent example of a dual account is the multi-modular analysis posited in Francis and Yuasa (2008). The main problem with Halliday’s twofold analysis of the NP is that his two layers of analysis, one dependency and the other constituency-based, have to be reinterpreted as two synchronically and diachronically distinct realizations of one dependency-based structure.19 Instead of claiming that each SN-expression is always head as well as modifier in any SN-structure, as Halliday does, head and modifier analyses have to be seen as two distinct realizations of head-modifier structure. The description of SNconstructions put forward in this study will involve two distinct synchronic and diachronic analyses, with the second one being treated as a reanalysis of the first, as exemplified in figure 2.5, and illustrated by (2.1) and (2.2): A lot Heaps Head

of land of clinker Postmodifier

A lot of Heaps of Quantifier

people/nonsense people/fun/etc. Head

Figure 2.5: Shift from head noun status to modifier status

(2.1)

Pulham, scion of the Portland cement family, experimented and perfected in the 1840s the art of using liquid cement poured over heaps of clinker to make rock formations. (CW-UKmags)

2.1 Requirements of a framework for SN- and TN-constructions 47

(2.2)

What’s interesting is how many of the sexual researchers and observers were driven by self-interest? Heaps of them at least. (CWOZnews)

2.1.2 Langacker (1991): A radical functional account underpinned by dependency Langacker (1991) presents a radical functional(-cognitive) approach to the NP that abandons traditional constituency and accounts for the internal organization of the NP in terms of configurations of functions, which are underpinned by dependency relations. This is mainly inspired by the fact that the English NP can take so many different shapes that it cannot be reduced to one structural template capturing all the elements that can feature in it, as Halliday’s constituency analysis for instance suggests. Langacker proposes an account in terms of four semantic functions, namely specification of a type, instantiation of that type, quantification and grounding. These semantic functions are argued to be present in each NP, even when they do not have an overt structural realization (Langacker 1991: 5154). In canonical codings of the NP each function is represented by a distinct structure, but in non-canonical codings, such as proper names, they are fused into one single element (but see Van Langendonck 2007). The proper name Iraq for instance expresses several functions at once in a single word. It has a type specification, i.e. ‘nation’; it expresses quantity, i.e. singular; and grounding, i.e. definiteness. The fact that Langacker’s account allows for non-canonical coding already indicates a first level of dynamicity necessary to account for elements of structure on the move, as in grammaticalization. In canonically coded NPs, such as those three black cats, “what is traditionally known as the head noun” forms the ultimate core or “innermost functional layer” of the NP, specifying a type as such, e.g. generic cats. The addition of modifiers to this type specification, e.g. black cats, turns this “basic type specification” more specific (Langacker 1991: 146). To this, quantification by three and instantiation of the type are added as further layers, with grounding of the instantiated type by those as a final layer. Grounding pertains to relating the designatum of the NP to the speech-event and speech-participants, i.e. the ground, and can roughly be equated with what is more commonly referred to as the functional system of determination. The composition of the NP via the consecutive application of these four basic functions is Langacker’s interpretation of “grammatical constituency” (Langacker 1991: 143). The structural relationship between the symbolizations of these layers

48 Chapter 2: A cognitive-functional constructional framework “relies almost exclusively on modifier relations”, i.e. dependency structures (Langacker 1991: 146). In his discussion of the function of quantification, Langacker successively deals with the two main types of devices for indicating quantity in NPs, namely the category of number (mainly singular versus plural in English) and a system of quantifiers which enable a more specific indication of quantity. The discussion of number is relevant here in that it sets out the conceptual differences, as well as overlap, between the profiles of singular count nouns, plural count nouns and mass nouns. Singular nouns designate a discrete entity; whereas plural count nouns and uncount nouns are grouped together as mass nouns. The difference between the latter two is that plural count nouns such as books designate a “replicate mass”, i.e. a mass that consists of a replication of a number of discrete entities, all of the same type. Uncount nouns such as gravel, on the other hand, designate a “non-replicate mass”, i.e. their mass is profiled as continuous and internally homogeneous (Langacker 1991: 75-81).20 These noun profiles are important for the description of the conceptual semantics of quantifiers later on, in that quantifiers are often restricted to certain categories of nouns for their functioning, and also incorporate the notion of a mass for their own conceptualization (Langacker 1991: 84-85). The section on the system of number is also important because it concludes with a pivotal claim in Langacker’s analysis of instantiation, which in its turn feeds his particular interpretation of quantification. This fundamental claim states that every NP profiles only one instance of the instantiated type. Hence, also in the case of plural noun NPs incorporating a quantifier, such as the seven pebbles, only one instance of the plural noun category pebbles is instantiated and not seven instances of the category pebble. The function of a quantifier is therefore not to specify the number of instances (which is always just one) but rather to indicate the size [or cardinality] of the [single] profiled instance. (Langacker 1991: 81; insertions are alternative formulations by Langacker, same page)

In the case of plural nouns the size is always the number of component entities together making up the single profiled instance, e.g. seven (cf. Langacker 1991: 54, 81). The particular operationalization of size in the semantic function of quantification has been one of the main motivations for using the term size noun constructions in this study, in favour of earlier names such as measure nouns (see note 2). Within the function of quantification, Langacker (1991) makes a fundamental semantic distinction between absolute and relative quantifiers, which depends on the way in which they supply their quantitative informa-

2.1 Requirements of a framework for SN- and TN-constructions 49

tion, and on whether they have an (additional) grounding function or not. This conceptual distinction is supported by systematic formal and distributional recognition criteria; however, it is not yet a familiar distinction in descriptive grammars of English (e.g. Quirk et al. 1985, Bache 2000). Since it will be useful when we compare canonical codings of quantification and non-canonical SN-expressions, I will discuss it here. Put very simply, absolute quantifiers only have a quantifying function and serve to express the intrinsic magnitude of the designated mass directly, whereas relative (or proportional) quantifiers always specify the quantity of the designated mass indirectly by comparing its size to that of a reference mass and in this way also ground the instance: A relative quantifier is so called because it specifies a quantity in relation to a reference mass; in the default-case interpretation, this reference mass consists of the maximal instantiation of the pertinent category (i.e. its full extension in all conceivable worlds). [in other cases it is] a contextually determined set (Langacker 1991: 110). […]. [Relative quantifiers] are best regarded as grounding predications that fulfill this function through a quantitative assessment made relative to a reference mass. On the other hand, an absolute quantifier specifies the size of the profiled instance without referring to the maximal extension of the relevant category, [i.e.] independently of any estimate as to how this total stacks up in relation to the set of all [instances of the pertinent category] (Langacker 1991: 82-83 and 110;; insertions are mine).

Examples (2.3) and (2.4) illustrate absolute and relative quantification respectively. The quantifiers are in bold: (2.3)

Three mice were found dead in the shed.

(2.4)

Most mice in the shed are still alive.

In (2.3) three merely counts the number of mice without referring to the total amount of mice present in the specific context of a mice-infested shed. By contrast, in order to know what the cardinality is of most mice in (2.4), one first has to know the cardinality of the reference mass. Most supplies its quantitative information indirectly, relative to the entire number of mice in the present context, namely as a large subpart or proportion of this reference mass, very similar to the way a fraction such as ¾ expresses its quantitative information. 21 First, I will look in more detail at the way Langacker describes the various internal conceptualizations of what he considers the core set of purely quantifying quantifiers, i.e. absolute quantifiers. The conceptual semantics of these quantifiers incorporate the following key notions: the conception

50 Chapter 2: A cognitive-functional constructional framework of an appropriate scale of magnitude, the conception of a mass, a landmark and a trajector. I will focus on the interaction between the first two features. The notion of a scale of magnitude is present in the conceptual semantics of each absolute quantifier, though implemented differently depending on the type of noun they quantify. This scale can be conceptualized as a ruler or a measuring rod against which a particular mass is measured off. We have seen earlier that each type of noun profiles a specific kind of instantiation, e.g. replicate mass in the case of plural count nouns and nonreplicate mass in the case of uncount nouns. A quantifier likewise incorporates the notion of an instantiation of which the size is assessed with respect to a scale of magnitude. In the case of uncount noun quantifiers, such as much and (a) little, this scale is continuous, whereas it is discrete in the case of plural count noun quantifiers. The latter can be either specific expressions, i.e. cardinal numbers, or more schematic ones, such as the indefinite quantifiers many, several, numerous and (a) few. In the case of numbers a quantity is pegged “to a particular step along the counting scale”, i.e. its value is a specific point on the counting scale. Schematic quantifiers merely place their value “within a vaguely-delimited range” instead of translating into a specific number on the scale. In addition they make reference to a norm or baseline value for their conceptualization (Langacker 1991: 84). In the case of many and much the value is conceptualized as lying above that implicit norm, whereas the values of few and little fall well under the implicit norm. In some cases the actual quantity profiled by two different quantifiers can be the same, but by using one quantifier instead of the other, this quantity is conceptualized very differently. A well-known example of this is the alternation between few/little on the one hand and a few/a little on the other hand. A few and a little take zero as their baseline and designate a value which is slightly higher than this. Few and little, however, designate a small quantity with the implication that this quantity is well below expectations, i.e. it is ‘too few/little’ (cf. Langacker 1991: 84). Figure 2.6 visualizes the conceptualization of the quantified NP in Three tables ought to be sufficient, incorporating the scale of magnitude, a mass being sized up (indicated as tr for trajector) and the value of the mass on the scale (indicated as lm for landmark) (Langacker 1991: 84-85):

2.1 Requirements of a framework for SN- and TN-constructions 51

tr

lm

 Figure 2.6: Conceptual quantification provided by a cardinal number (cf. Langacker 1991: 85)

In Figure 2.6 the oval in bold represents the single instance of the plural noun category tables, which consists of three component entities. These component entities jointly constitute the mass of which the magnitude is assessed with regard to the counting scale. The mass is called the trajector or relational figure in Langacker’s terminology. In the case of plural count nouns such as tables the size of the replicate mass is determined by the number of component entities (cf. Langacker 1991: 80-81). The landmark indicates the range to which the value of the quantifier is confined; in this case the landmark is number 3 on the counting scale. Figure 2.7 represents the conceptual mechanism of relative quantification in terms of a comparison of the actually predicated mass P and a reference mass RT, e.g. Most of the people agreed with his claim. ti refers to an instance of type T.  T    

ti

ti

P

ti

RT

Figure 2.7: Relative quantification (Davidse 2004: 4, based on Langacker 1991: 76, 108)

RT then refers to the contextually determined set of ‘all people involved’, of which the actually predicated set is a large proportion, i.e. most people. It is

52 Chapter 2: A cognitive-functional constructional framework this operation of comparison that serves as a basis for the grounding of the instances typically expressed by relative quantifiers. As mentioned earlier, the term size noun constructions was chosen to signal agreement with Langacker’s specific approach to quantification as assessment of size. This agreement is with the conceptualization of quantification in terms of assessment of size relative to a scale of magnitude. The non-equation of component entities with instances in the case of replicate masses, however, seems unnecessary for the present purposes of this study. Langacker himself admits that specification of the number of instances and specification of the size of the single profiled instance are extensionally equivalent (Langacker 1991: 81). He also, rather inconsistently and confusingly, rephrases his take on quantification as “quantification of instances” (Langacker 1991: 89). Langacker sets out his view of SN-constructions towards the end of the discussion of absolute quantifiers, when he turns to more complex expressions, particularly those in which a quantifier is followed by a prepositional phrase with of: three of the tables; many of our friends; much of the difficulty; etc. (Langacker 1991: 87)

As we have seen in Section 1.2.1.3, such (partitive) structures have received ample attention in transformational approaches and have led to diverse and conflicting constituency analyses. Langacker starts out with his analysis of three of the tables, which he compares to the parallel, but distinct construction three tables. In early transformational grammar these two constructions are sometimes equated through derivational operations and transformations. Three in three tables functions as a simple quantifier and is likened to a special sort of adjective that modifies the head noun tables (Langacker 1991: 85-86 and 88). Three of the tables on the other hand displays a pronominal-like three which is said to function as the nominal head, and of is a “true, meaningful preposition that profiles an intrinsic relationship between its trajector and landmark”, in which the trajector is characterized as an inherent subpart of the landmark elaborated by the prepositional object (Langacker 1991: 23 and 88). Of is still very often seen merely as an empty particle, devoid of specific conceptual semantics. As we saw in Section 1.2.1.1, Kruisinga (1925) is an early exception to this view and offers an analysis compatible with Langacker’s cognitive approach. These two analyses are then linked up with a whole family of expressions, such as a bunch of carrots, a bucket of water and a lot of sharks. The nouns featuring in this constructional family constitute a highly varied and open-ended class: mess, pile, heap, loaf, sprig, head, stack, group, array, pinch, tad, bit, barrel, crate, jar, tub, vat, keg, box, ton, pound, yard, foot,

2.1 Requirements of a framework for SN- and TN-constructions 53

flock, herd, pride, school, bevy, gallon, pint, liter, cup, spoonful, pocketful, mouthful, closetful, truckload, shipload, etc. (Langacker 1991: 88). All of these nouns can take up the three-slot in three of the tables. The list includes the SNs that will be the main focus of the case studies in Part 2: bunch, heap, load and pile, in addition to members of the more narrowly defined class of measure nouns, e.g. yard and gallon and other nouns that will not be the focus of this study. Since a “conception of its typical size, which is part of its encyclopaedic characterization” is common to each noun in the list, we will gloss over the specific semantic differences here and label them all size nouns. Interestingly, Langacker not only notes that the above set of nouns do not always straightforwardly constitute the head in each construction they are used in. He also, though cautiously, implies the head and modifier analyses to be distinct diachronic stages. He treats SN-constructions as emergent means of quantification. This is a very important difference with Halliday’s account, which saw the two proposed analyses of SNs as always simultaneously present in each SN-construction, thus disregarding degrees of grammaticalization. The first analysis that Langacker proposes for size nouns can to a large extent be equated with my interpretation of head noun status of SNs. In this use SNs designate a physical, spatially-continuous entity that either serves as the container for some portion of a mass (bucket, cup, barrel, crate, jar, tub, vat, keg, box) or else is constituted of some such portion (bunch, pile, heap, loaf, sprig, head, stack, flock, herd). Thus a bucket of water may in fact be a physical bucket that is filled with water, and a bunch of carrots may actually be a lump-like object formed by tying a number of carrots together into a bunch. (Langacker 1991: 88)

In addition to this literal lexical sense, most of the SNs have developed an extended use in which the size specification, which was backgrounded in the head use, comes to the fore. In this extended use “size becomes the most salient specification. […] A noun interpreted in this fashion can be regarded as a quantifier.” (Langacker 1991: 88): For instance, a bathtub may contain a bucket of water without there being any bucket in it—it is only implied that the water would fill a bucket were it placed in one. Similarly, a bunch of carrots may refer to a certain number of carrots irrespective of their physical distribution. The notion of a discrete physical object has faded, leaving behind the conception of a schematically characterized mass (the mass that, in the original sense, either fills or constitutes the object) whose projection on the scale of magnitude then provides its primary semantic content. (Langacker 1991: 88-89)

54 Chapter 2: A cognitive-functional constructional framework Langacker argues that in their extended sense SNs are “directly analogous to the pronominal sense of three in three of the tables”, where three was analyzed as the head noun (Langacker 1991: 89). This is strange since he observed earlier on that the extended use “can be regarded as a quantifier” (Langacker 1991: 88). Even though his discussion seems to imply that the semantic change lies at the basis of a potential reanalysis of SNs as quantifiers, Langacker at this point remains cautious about claiming that the noun in the of-phrase has already been reanalyzed as the new head in SNconstructions. Only then would the SN together with of have the same function and structural status as three in three tables: A further step in this evolutionary sequence would be for the second noun to be reanalyzed as the head, leaving the remainder as a complex quantifier: [[a lot of] [sharks]]. I leave open the question of whether this reanalysis has actually occurred, for a lot of or any other quantifying expression (cf. Selkirk 1977). But suppose it has, so that a lot of tables displays the same grammatical organization as three tables (apart from the morphological complexity of the quantifier). In this case a lot of must be adjectival at the composite structure level, i.e. roughly comparable to the adjectival sense of three, regardless of whether this value arises compositionally from the meanings of its parts. The integration of a lot of with the following noun will also then be similar to that of three. (Langacker 1991: 89)

Langacker (1991: 89) makes some further observations in a footnote that could similarly be seen as hinting at structural reanalysis, even though he does not present it as such. For instance, he notes that verb agreement is determined by the noun in the of-phrase in the case of extended uses of size nouns: the shift of meaning is also predictably accompanied by a change in verb agreement (e.g. from is to are in A bunch of carrots——in the sink).

Hence, Langacker does connect fluctuating verb agreement with the shift of semantic head in SN-constructions, as opposed to all of the word class approaches as well as Halliday (1985a). However, again he refrains from interpreting this as a syntactic criterion for structural reanalysis. More recently Langacker (2009) and (to appear) discuss a lot of and other, related SN-expressions explicitly within the framework of constructional “grammaticization”. At the outset of Langacker (to appear: 1) a lot of is claimed to have “become a basic element of the English quantifier system”, but again and rather conflictingly, he (ibid.: 2) argues that this phrase, as well as a bunch of for instance, have not yet been reanalyzed as quantifiers, essentially because they lack monomorphemic status. His main arguments against reanalysis are that these phrases can still be analyzed into

2.1 Requirements of a framework for SN- and TN-constructions 55

determiner, noun and of, which each component “mak[ing] definite contributions to the meaning of the whole”. The possibility of premodification of lot, as in a whole lot and plural formation into lots for instance are cited as further arguments for the nounness of lot. Tests for constituency boundaries point towards a break between lot and of according to Langacker (ibid.: 2) and he analyzes a lot of geese with a lot as quantifier and of geese as a prepositional phrase, without an actual head noun, which is allowed in his cognitive grammar. These arguments will be discussed in more detail when the basic notions of grammaticalization will be discussed in Chapter 3. In sum, Langacker is primarily interesting because he explicitly sees SNs as an emergent means of quantification and links the head noun analysis and the quantifier analysis of SNs in terms of distinct diachronic stages. Similarly interesting is the fact that he works with the concepts of absolute and relative quantification, even though he does not apply this distinction to SNs here. He does, rather confusingly, refrain from claiming that phrases such as a lot of or a bunch of have been reanalyzed as quantifiers, though they are said to have become basic members of the quantifier system. The present study will use Langacker’s basic functions at work in the NP, with categorization referring to the description of a general type and quantification viewed in terms of measuring the size of the profiled instance. The notions of absolute and relative quantification can serve to describe the semantics of the various SN-expressions. The function of type specification will be the starting point in the description of TN-expressions. Langacker’s radically functional dependency analysis will be complemented with the semantic analysis of the NP’s functional slots, as offered by Halliday (1985a/1994). This will also help to refine some of Langacker’s functions. Langacker, for instance, argues that in canonically coded NPs the type specification is symbolized by the head noun and its modifiers, i.e. adjectives and all other elements that make it more precise. Quantification and grounding can be realized by quantifiers and determiners. The specific semantic import of these modifiers and determiners, however, is rather understated in his analysis. Langacker mainly keeps to describing the general functions that they symbolize. As such, the difference between qualitative and classifying adjectives, for instance, is not made. Both are said to make the basic type specification of the head noun more specific, but it seems untenable to argue that they do so in the same way (cf. Davidse 1997). Following Bolinger’s (1967) “referent versus referencemodification” distinction, Breban (2010) further illustrates that in NPs with both types of adjectives, such as “a serious social problem”, only the classifying adjective social adds to the refinement of the type specification, whereas serious pertains to an instance of that specific type. Since the distinction between qualitative and classifying adjectives is blurred in Lan-

56 Chapter 2: A cognitive-functional constructional framework gacker (1991), different extended modifier uses of SNs and TNs incorporating these meanings cannot be described properly. Bache’s (2000) approach to the English NP in terms of the functional zones of categorization, modification (subsuming specification, description and classification), and determination links up with Halliday’s analysis as well as with Langacker’s semantic functions, though there are several differences. Bache’s model is more flexible than Halliday’s in that he sees “no strict separation between” the different zones and instead argues that there is a continuum of values from determination to categorization: from the left determination fades into modification via specification and from the right categorization fades into modification via classification. (Bache 2000: 239)

However, in terms of the structural realizations of the separate functions, he (2000: 160), somewhat contradictorily, imposes restrictions similar to Halliday’s approach, in that Bache connects categorization exclusively with the head noun, determination with articles, pronouns and genitives, and modification with adjectives. This also signals a difference with Langacker’s type specification which, in addition to the head noun, also includes all modifiers that make it more precise. We saw in Langacker’s account that the modifiers in the NP, except grounding predications perhaps, are understood to relate to the head noun in the same way. However, the precise nature of these dependency relations is not further described. McGregor’s (1997) Semiotic Grammar owes a lot to Halliday’s metafunctional account, but adds an explicit and fine-grained theory about the types of syntagmatic relations that hold between the various elements in the NP (and at other levels of linguistic organization). These include, in addition to constituency and dependency, notions specific to his account, namely conjugational and linking relations. Langacker’s dependency analysis enriched with distinctions between premodifiers such as classifying elements, qualitative adjectives and postdeterminers can then be further refined by taking on board McGregor’s theory of combinatorics within the NP.

2.1.3 McGregor’s semiotic grammar: Syntactic combinatorics in the NP McGregor’s Semiotic Grammar (1997) takes up Halliday’s functional slots of Deictic, Numerative, etc. in the NP, but combines this with a theory of four basic types of syntagmatic relations at work in it: constituency, dependency, conjugational relations and linking relations. Each is connected with one of the four types of grammatical signs he distinguishes for lan-

2.1 Requirements of a framework for SN- and TN-constructions 57

guage, i.e. experiential, logical, interpersonal and textural signs respectively. These signs are based on Halliday’s metafunctions and interpreted as truly semiotic form-meaning correlates. The signs and types of syntagmatic relations they link up with are described mainly at sentence level, but also apply to the level of the NP. I will restrict discussion of McGregor’s Semiotic Grammar to the NP and SN- and TN-syntagms in particular. McGregor’s main claim is that depending on the nature of the element of structure, relations with the head noun and other elements in the NP and the discourse differ. Constituency is defined by indirect part-whole relationships, i.e. how an element of structure relates to the larger whole it is part of. It is connected with experiential signs, but only the Entity Noun (i.e. Halliday’s Thing), or head noun, is considered experiential. The other elements in the NP display other syntagmatic relations, i.e. dependency, conjugational and linking relations (McGregor 1997: 119-121). Dependency is defined by part-part relationships, i.e. relations between elements of structure which are both part of a larger structure. McGregor (1997: 21) argues that constituency has held an unwarranted monopoly position in linguistics in the twentieth century, and argues that his additional types of combinatorics have to be added in order to grasp syntagmatic relations fully. Dependency relationships are more direct than constituency relations in that, as part-part or “sister-sister” relationships (cf. Hudson 1971), they do not necessitate mediation through “a shared mother” (McGregor 1997: 60), nor do they rely on meaning as much as constituency does (McGregor 1997: 55; see also Croft 2001: 186). Two subtypes are distinguished, namely hypotaxis and parataxis (McGregor 1997: 60). Hypotaxis is defined by dissimilarity between two elements, with one element being the head and the other the dependent. Parataxis is defined by equality between the two parts. SN- and TN-syntagms, just like most NPs, basically involve hypotactic relationships, but in order to apply McGregor’s analysis to them, his interpretation has to be explicitly extended to allow shifts between head and dependent status. McGregor (1997: 64) succinctly proposes distributional equivalence as the main criterion for headness: “the head is that part which shows the same distribution as the entire unit”, but in the case of binominal NPs this is rather unhelpful.22 In the specific discussion of hypotactic dependency in the NP, McGregor (1997: 176-180) introduces a distinction between attribution and classification relations. In attribution relations a quality or quantity of the Entity Noun is described. Quality attribution is similar to Halliday’s notion of the Epithet; quantity attribution corresponds to Halliday’s Numerative. Classification relations correspond to Halliday’s Classifiers, i.e. they serve to distinguish a subclass of the larger class of things designated by the Entity nominal. McGregor connects quality attribution with adjectives and

58 Chapter 2: A cognitive-functional constructional framework quantity attribution with number words, indefinite quantifiers as well as what he calls size adjectives, e.g. a little sugar (McGregor 1997: 176). Whereas Halliday sees relations between Classifier, Epithet or Numerative and Thing in terms of constituency, McGregor relates them directly to the head noun in terms of dependency. In addition to more traditional recursive dependency, McGregor (1997: 69) introduces conjugational syntagmatic relations, which hold between interpersonal signs, i.e. signs that show focus on speaker-hearer interaction. They are described in terms of (interpersonal) modifier relations. In conjugational relations the elements are in a whole-whole relationship where “one unit applies to another in its entirety, moulding it, as it were, into a particular shape” (McGregor 1997: 69), i.e. one unit affects the other by “indicating how it is intended to be taken or viewed by the addressee” (McGregor 1997: 210). Conjugational relations can semantically construe three types of modification: illocutionary, attitudinal and rhetorical modification (McGregor 1997: 66). Illocutionary modification “concerns the illocutionary force of an utterance, with [sic] how the speaker intends it to be taken interactively”, e.g. “frankly in Frankly, they couldn’t save themselves if they tried” (ibid.). Attitudinal modification serves to indicate the speaker’s subjective attitude towards his/her statement, e.g. “fortunately in Fortunately, they found their way to the waterhole”. Rhetorical modification, finally, indicates how an utterance fits within “the framework of knowledge, beliefs, expectations, etc. of the interactants in the speech situation”, e.g. counter-expectation by already in “It has already started to rain” (ibid.). Conjugational relations are further subdivided into scoping and framing relations. Framing relations can be illustrated by quoted speech for instance, where “the quoted unit is encompassed within the whole construction, by the quoting unit, which applies to it in its entirety, indicating that it is to be regarded as something attributable (usually) to another speaker” (McGregor 1997: 64). In scoping relations then the encompassing unit is argued to “appl[y] over” the encompassed unit in terms of “leaving its mark” on the entire unit, e.g. luckily in Luckily he did not come (McGregor 1997: 210). McGregor (1997: 69) notes that such “[m]odifying items may be scattered throughout the item over which they have scope”, as in the French negation marker ne… pas. The final type of syntagmatic relation is called linking relations, which “serve linking functions, establishing referential links to the linguistic context” or “connect[ing] one linguistic unit to another, or to some extralinguistic phenomenon” (McGregor 1997: 70). They are “free relations” (McGregor 1997: 58) compared to the others, in that they need not be structurally related to the elements that they link, and because they can link items up with non-linguistic elements. Linking relations can be indexical,

2.1 Requirements of a framework for SN- and TN-constructions 59

e.g. demonstrative pronouns. They may be connective, as in conjunctions and prepositions. They can express marking in labels for constructions or category types. Linking relations can also be covariate, as exemplified by the lexical cohesion between pursue and escape. Finally, they may also be collocational, as in the habitual co-occurrence between syntagmatic and relationship for instance. These various parameters are not mutually exclusive, with one linking item possibly establishing links of more than one type (examples cited in McGregor 1997: 58). Conjugational framing and scoping can clarify the structural relations expressed by certain extended SN- and TN-uses, such as discourse marker and quotative uses. McGregor (1997) himself for instance considers the possibility that determiners might be scoping elements.

2.1.4 Conclusion: An eclectic functional model of the NP The model of the NP used in this study mainly follows Langacker’s description of the basic functions that are expressed within the NP and sees the syntagmatic relationships between its elements in terms of dependency. In addition, McGregor’s conjugational relations of scoping and framing are incorporated to account for the syntagmatic relations construed by more interpersonal modifiers such as discourse makers, quotatives and possibly determining elements. In keeping with Langacker (1991), such a model allows for noncanonical codings and does away with rigid correlations between functions and their structural realizations as proposed by some authors. This is necessary to accommodate the extended uses of SNs and TNs. Bache (2000) sees the zones of determination, modification and categorization in the NP as a continuum. Such accounts suggest that synchronic variation as well as emergent functions can be captured. However, a true diachronic dimension has to be added. This has already been suggested by Langacker (1991) and (2009 and to appear), who presents head/modifier readings as distinct diachronic stages. Langacker’s and McGregor’s dependency analyses still have to be explicitly extended to implement these reanalyses as shifts within dependency relations and possibly to conjugational relations. The functional labels adopted in this study are from Halliday’s (1985a/1994) semantic description of the various prenominal modifier slots, here referred to as classifying nouns/adjectives, qualitative adjectives, postdeterminers, and determiners, which subsume both identifiers and quantifiers (cf. Langacker 1991). Despite the many strong points of the functional-cognitive model of the NP developed so far, its dynamicity has to be enhanced even further by

60 Chapter 2: A cognitive-functional constructional framework bringing in the notion of construction more explicitly, in keeping with recent trends in constructional approaches to grammar.

2.2 A constructional framework of the English NP 2.2.1 Elaborating the functional model: SN- and TN-patterns as partially filled constructions One additional grammatical model has to be incorporated into the dynamic model of the NP that has not been discussed in an explicit way yet, i.e. Construction Grammar23 (henceforth also CxG). There are several reasons for this. For one thing, an empirical study that wants to investigate different uses of SNs and TNs in terms of SN- and TN-constructions requires the incorporation of a constructional component in its theoretical framework. Up till now this constructional status has been posited in a pre-theoretical way, without reference to CxG, mainly to indicate that a string of words and a specific interaction with co(n)text is involved instead of just the SN or TN in itself. However, an explicitly constructional approach to SNs and TNs has much more to offer in terms of systematizing certain important factors in the attested synchronic variation and grammaticalization of both types of syntagm. CxG is a multidimensional grammatical model that will allow me to describe and interpret SN- and TN-patterns as constructions, i.e. semiotic/symbolic pairings of form and meaning, and describe in great detail their component elements and how these construe specific constructional meanings (cf. Croft and Cruse 2004: 241 and Goldberg 2006: 5). By explicitly bringing in the CxG framework, we also avoid the offhand use of the notion of construction in grammaticalization studies, which has recently come in for criticism (see Noël 2007). CxG furthermore shares fundamental tenets with cognitive approaches to grammar, despite some divergences (see Langacker 2005) and hence links up naturally with the functionalcognitive part of the theoretical framework set out in the previous sections. On the other hand, the specific SN and TN case studies may reveal some flaws in the concrete application of a CxG framework within the context of nominal syntax, and when pragmatic factors are central to the construction’s description. The main reason for adding CxG is that, despite Halliday’s delicate semantic characterization of functional slots in the NP and McGregor’s detailed discussion of types of combinatorics within the NP, the dynamic framework described so far remains a rather schematic model for SNs and TNs. It allows me to describe the structural relations between elements in the NP and the functions they express, but in doing so assumes a nominal

2.2 A constructional framework of the English NP 61

structure in which all positions are relatively free and not predetermined in keeping with construction-specific properties.24 It cannot accommodate the great amount of lexical determination in SN- and TN-patterns for instance in the form of collocations and premodification patterns. These cannot be predicted from the general principles and combinatorics it proposes. McGregor’s framework treats collocational relations as a subtype of the "free" linking relations. However, for the description of SN- and TNpatterns, where each use relates so strongly with particular collocational patterns, these need to be integrated much more profoundly with dependency itself. McGregor (1997) and other functional-cognitive models of combinatorics do not explicitly allow such specific lexical restrictions and changes in collocational range and will have to be extended in order to validate them as true lexical-syntagmatic relations (cf. Crystal 1991). The main asset of incorporating CxG into a framework of the NP is that it allows a detailed description of patterns in terms of all kinds of construction-specific constraints and peculiarities. It fleshes out the functionalcognitive model discussed so far, which basically describes relations for “purely skeletal patterns, which only consist of empty positions representing classes of possible elements”. CxG, on the other hand, also allows one to describe “more specific templates in which one or more of the slots is already lexically filled-in” (Cappelle 2005: 49 and 9; see Section 2.2.2), or rather collocationally constrained as in SN- and TN-patterns (see Section 2.2.4). A constructional approach is better equipped to incorporate lexical constraints that influence the entire construction so profoundly and relate to specific layers of SN- and TN-uses. It also links up with a more recent interest in grammaticalization studies. Section 2.2 will work towards a specific implementation of this constructional approach within the functionalcognitive framework and the corpus studies. Section 2.2.2 will introduce CxG as such, sketching its background in the study of idioms, and the insight that such structures are more pervasive in language than normally assumed. Section 2.2.3 will summarize the main strands within CxG, focusing on their shared tenets. A full and systematic implementation of CxG would lead too far for the present study. A CxG approach will be put forward here mainly to further systematize empirical observations about SNand TN-patterns. Section 2.2.4 will summarize how CxG has looked at the NP until now. Extending CxG to nominal syntax as such is a challenge, considering its longstanding penchant for verbal (argument) structures and the clause level in general (e.g. Fillmore’s Case Grammar and study of the let alone construction, Goldberg’s studies of argument structure). In order to overcome this, several proposals of CxG need rethinking. As we will see, the case studies of SN- and TN-constructions also require us to take specific stands on such topics as the level of schematicity of constructions

62 Chapter 2: A cognitive-functional constructional framework and the interpretation of pragmatics. In addition, such Firthian-Sinclairian notions as collocational range, colligation (roughly “the grammatical company a word keeps”, Hoey 1998), and semantic prosody have to be implemented in this constructional approach25 to SN- and TN-patterns, as well as a grammaticalization framework (cf. Traugott 2010b). There is a growing interest in diachronic constructional approaches and serious efforts are being made to integrate or show possible interaction between the framework of CxG and grammaticalization (e.g. Trousdale 2005, Traugott 2006, Noël 2007). Finally, the constructional component has to be integrated with the functional-cognitive model proposed in Section 2.1. 2.2.2 Background: From idioms to Construction Grammar26 CxG’s origins lie in the study of (conventional) schematic idioms27, i.e. idioms in which some slots are lexically open, discussed in Fillmore et al. (1988) for instance. They posed a fundamental challenge to traditional generative grammar which organizes language knowledge into several components, usually a syntactic, phonological and semantic one. Each component consists of all of the highly general rules and constraints governing the structure of a sentence in a language. Each component is self-contained and the three of them are firmly distinguished one from the other. Within syntactic structure a further distinction is made between two levels, i.e. the deep structure and the surface structure. In addition, the lexicon is distinguished, which overlays the three components, in that it supplies the sound structure, syntactic category and meaning of each word. It is also set up as a sort of repository that contains all the perceived idiosyncrasies and arbitrary elements in the language. Such a componential model still needs to provide a way to map these separate components onto each other, in order to end up with a sentence that has a syntactic structure with a particular phonological structure to which a meaning is attached. This is done by means of linking rules, which are similarly highly general rules that apply to all of the sentences of the language. Nevertheless, in transformational generative grammar (henceforth TGG) the main interest is in the general componential rules, not the linking rules. In CxG this will be the other way round (cf. Croft and Cruse 2004: 225-229). TGG does not consider the notion of a construction a useful or necessary one in grammatical analysis and hence does not worry about construction-specific rules or constraints. Instead, TGG tries to reduce as many of such construction-specific properties as possible to the general rules and constraints of the various components. Any idiosyncratic property that

2.2 A constructional framework of the English NP 63

these componential rules cannot account for is assigned to the lexicon (Croft and Cruse 2004: 228). Especially schematic idioms, as partially substantive constructions, are a challenge to this TGG-model, since their syntactic, semantic (and pragmatic) properties cannot be predicted from the general rules of the syntactic, semantic and pragmatic components, or from the rules that link these various components. Fully substantive idioms, i.e. idioms in which all of the elements are fixed, are less of a challenge since their idiosyncratic characteristics can simply be assigned to the lexicon. In other words, substantive idioms, such as It takes one to know one can be seen as multi-word lexical items without affecting the componential model too much (Croft and Cruse 2004: 226). This is not possible with schematic idioms, e.g. the comparative conditional construction The X-er, the Y-er as instantiated by The longer you practice, the better you will become (cited in Croft 2001: 16). Fillmore et al.’s seminal (1988) study of the let alone-conjunction did not look at idioms as atypical or problematic phenomena, and instead argued for constructions and construction-specific rules as the basic elements of grammatical analysis. This was the start for describing expressions in terms of constructions that have syntactic, semantic, pragmatic, etc. properties that are rule and constraint-governed within the context of that construction. CxG-studies also typically look at a specific construction within a larger family of constructions that share something with it. The let aloneconstruction for instance is related to other constructions and constitutes just one of an entire family of paired focus constructions (Croft and Cruse 2004: 240). Since large portions of a speaker’s linguistic knowledge can and have to be analyzed in terms of constructions, construction grammarians wondered whether the notion of a construction could be generalized so as to account not only for partly idiomatic constructions, but for all linguistic facts, both above and below sentence level, from morphemes to larger stretches of texts. From a specific need for constructions as a level of analysis and syntactic representation, constructional theorists moved to the claim that the construction is the basic unit of grammatical analysis as such. This led to the development of the systematic and comprehensive framework of CxG.2.2.3 CxG as a family of constructional approaches: General tenets CxG itself is not one unified theory, but is best described as a family of several (partially overlapping) “constructionist”28 approaches (Goldberg 2006: 3). They share the fact that they all present an approach that is fundamentally semiotic, or symbolic, in nature, in that it stresses that all linguistic expressions are specific conventionalized form-meaning pairings, i.e. constructions, which should be studied as such. The construction is the basic level of linguistic representation for all the expressions of a language.

64 Chapter 2: A cognitive-functional constructional framework Most constructionist models moreover claim psychological reality for CxG, in that constructions are argued to be the basic units of the mental storage of language knowledge. CxG very much stresses that constructions themselves carry meaning which can be described relatively independently from their concrete instantiations, e.g. the passive construction as such has a particular meaning. However, most versions of CxG also make observations about the specific interaction between constructions and lexical items (themselves constructions). Differences between various constructionist approaches mainly lie in just how cognitive, formalized, usage-based or typologically universal they want to be. Some important strands discussed in Croft and Cruse (2004)29 and Goldberg (2006)30 are: (Unification) Construction Grammar as practised by Fillmore and Kay; (cognitive) Goldbergian/Lakovian construction grammar; and Croft’s Radical Construction Grammar (henceforth also RCxG) (Croft 2001 and Croft and Cruse 2004). Langacker’s (1991) Cognitive Grammar is also often considered a constructionist approach and CxG in general has firm roots in cognitive linguistics and especially cognitive semantics, particularly Lakovian/Goldbergian CxG and RCxG (see Lakoff’s 1977 paper on Linguistic Gestalts). Within the list of constructionist approaches affinity is strongest between Goldberg’s cognitive and psychological CxG, Croft’s (2001) RCxG and Langacker’s Cognitive Grammar. Their commonalities also seem the most useful ones for the CxG approach to SN- and TN-constructions presented in Section 2.2.4. It would lead us too far to discuss each theory of CxG in detail, so I will mainly refer to the basic tenets that they share and report on essential points of divergence between the theories when this is relevant for a constructional approach to SN- and TN-patterns. As a way into the continuously growing CxG literature, I use Mirjam Fried’s definition (taken from www.constructiongrammar.org, accessed September 16, 2010): The trademark characteristic of Construction Grammar as originally developed consists in the insight that language is a repertoire of more or less complex patterns – CONSTRUCTIONS – that integrate form and meaning in conventionalized and often non-compositional ways. Form in constructions may refer to any combination of syntactic, morphological, or prosodic patterns and meaning is understood in a broad sense that includes lexical semantics, pragmatics, and discourse structure. A grammar in this view consists of intricate networks of overlapping and complementary patterns that serve as ‘blueprints’ for encoding and decoding linguistic expressions of all types. (Original formatting)

Linguistic units of varying internal complexity, such as morphemes, words, “partially lexically filled and fully general phrasal patterns”, idioms,

2.2 A constructional framework of the English NP 65

clauses and entire texts are equally treated as having constructional properties, and all units “are taken to be of equal value in describing the overall grammar of a language”. No units are said to be more central than others, as is often done in traditional grammars, where for example active declarative clauses with a transitive argument structure are considered to be the most basic clause type (Goldberg 2006: 5 and Fried and Östman 2004: 12). Instead, every linguistic unit is a construction to the same extent; there are only differences in their respective schematicity and internal symbolic complexity. Hence a constructionist approach can be extended to research into morphology, as well as the lexicon. In order to do this, CxG puts forward two clines: one ranging from schematic to substantive constructions, the other from atomic to complex constructions. As mentioned in Section 2.2.2, in schematic constructions some element or elements are lexically open,31 indicated by a category label, whereas in fully substantive constructions all elements are lexically fixed (Croft and Cruse 2004: 248). Constructions range from maximally substantive, as in the idiom It takes one to know one; over partially schematic, as in [give NP the lowdown] and more schematic, as in the let alone construction, e.g. He doesn’t eat fish, let alone red meat. Finally, there are maximally schematic constructions, such as the ditransitive construction, which is described as [SBJ DITRVERB OBJ 1 OBJ 2], and instantiated by He gives her a flower. By combining the internal complexity and schematicity parameters, traditionally distinguished linguistic units of all sorts rank as constructions. A syntactic category label like NP for instance is an atomic and schematic construction. Both clines also show up in taxonomies or inheritance network of constructions, which all versions of CxG posit in some way. These networks relate constructions in terms of schematicity, e.g. [VERB PHRASE] | [VERB OBJ] | [k ick OBJ] | [k ick [the buck et ]]

This taxonomy moves from schematic to substantive kick the bucket (cited in Croft and Cruse 2004: 263). Each level designates an independent and separate construction, but also shows that it is related to other ones in terms of schematicity. A concrete utterance, or “construct” (Fried and Östman 2004: 18), then, is a combination of various types of constructions and their respective constraints, as long as these are not in conflict.

66 Chapter 2: A cognitive-functional constructional framework [A]ny specific utterance’s structure is specified by a number of distinct schematic constructions. Conversely, a schematic construction abstracts away from the unspecified structural aspects of the class of utterances it describes. (Croft and Cruse 2004: 265)

The “multiple parents” of the utterance I didn’t sleep for instance are (at least) the Intransitive construction and the Negative construction (Croft and Cruse 2004: 264). The outlook of such taxonomies is again meant to reflect the psychological reality of how a language user generalizes over concrete utterances, stores these generalizations and productively uses them. The precise outlook of inheritance relations differs between the various strands of CxG, but this need not concern us in the context of SN- and TN-patterns. Constructional taxonomies, and the combination of the schematicity and complexity continua in general, also explain why in CxG the distinction between grammar and the lexicon is a non-issue. Since everything is a construction they are hence just as (construction-)grammatical, with words being substantive atomic constructions (Croft and Cruse 2004: 255). For our purposes we can interpret this as the lexicon and grammar forming a continuum in CxG (cf. Hunston and Francis 2000, Langacker 2005). This is an argument in favour of a CxG approach to grammaticalization, which can only exist by virtue of it (cf. Fried and Östman 2004: 15).32 The ultimate aim of CxG is to describe construction-specific properties in all of their formal and semantic constraints, and relate these to each other. In keeping with this aim, CxG argues for a very broad interpretation of semantics. The semantic model most strongly associated with it is Frame Semantics. Frame semantics was originally put forward by Fillmore (1975) and relates linguistic semantics with encyclopaedic knowledge. It has much in common with Langacker’s notion of profiling (Langacker 1991). The basic idea is that a word evokes an entire frame of semantic knowledge relating to the specific concept it refers to. To give a textbook example discussed by Schank and Abelson (1977), a restaurant is associated with such concepts as ‘customer’, ‘waiter’, ‘ordering’, ‘eating’ and ‘bill’ (cited in Croft and Cruse 2004: 7). Together these form a coherent structure of concepts that are related in such a way that without knowledge of all of them, one does not have complete knowledge of any of them. When a word is pronounced, usually not its entire frame is activated; instead certain specific parts of the frame can be foregrounded, while others are backgrounded. Frames are based on habitual occurrences in human experience. Semantics in CxG can be defined as semantics of understanding and hence everything that aids understanding can in principle be included. Register variation and information about dialect and attitude can also be part of the meaning of a construction. This model of semantics ties in with the preference for a usage-based model in CxG, which stresses the fact that

2.2 A constructional framework of the English NP 67

properties of the actual use of utterances in communication also determine how language users store grammatical units mentally. Frequency of occurrence and meaning in use are particularly important in this (e.g. Goldberg 2006: 45-65, Croft and Cruse 2004: 292). The let alone construction for instance invokes a scalar model or frame for its interpretation in the sense that it typically puts the more specific of its two propositions in the second conjunct, e.g. (2.5)

Yamoussoukro, many of whose citizens are not even Christian, let alone Catholic. (CW-BBC)

The idea is that if those people are not members of the superordinate category ‘Christians’, they can certainly not be Catholic. Words (i.e. lexeme constructions) such as tall evoke a frame that refers to erect human beings in which tall pertains to their vertical extent, indicating a high point along this vertical axis. It can also invoke a frame that refers to buildings and their vertical extent. The semantically similar notion high, on the other hand, does not evoke the human frame in this particular meaning. A frame semantic model can hence capture rather subtle differences between what seem to be semantically very similar words (Fillmore 1977 cited in Croft and Cruse 2004: 9). A frame typically evokes multiple specifications of different types, and this allows a language user to focus on one element of the evoked frame and background others. Polysemy or subsenses of a verb construction can hence be explained as instances in which certain parts of the frame evoked by the verb are backgrounded and others are foregrounded.33 Since the semantic model is so broadly defined, the distinction with pragmatic meaning is not always clear. Pragmatic meaning traditionally refers to parts of meaning that are not coded by the items as such, but arise in addition to them, through interaction with both linguistic and nonlinguistic context. In CxG studies pragmatics often seems to coincide with discourse structure and information structure, or they are at least somewhat conflated, e.g. Lambrecht (1994) and (2004). It is important to set all of these notions apart to some extent, while maintaining a broad interpretation of semantics and acknowledging pragmatic meaning. Finally, CxG also takes syntactic properties of constructions into account. These may subsume relations between the elements in a construction and the types of grammatical categories allowed in specific constructions. As such CxG does not work with either a constituency or dependency analysis to model syntagmatic relations; rather, dependency or other structural relations will be pointed out only if they are characteristic of a particular construction (cf. Fried and Östman 2004: 19, who work within a HPSG model). The different strands within CxG also differ with regard to this

68 Chapter 2: A cognitive-functional constructional framework topic. Fillmorean Construction Grammar and Goldbergian CxG basically describes constructions as “complex combinations of a set of primitive atomic units.”34 (Croft and Cruse 2004: 268). Goldberg is mostly interested in relations between constructions, looked at in terms of members of prototype categories (Croft and Cruse 2004: 272). CxG case studies in the end seem to describe constructions mainly as word class structures, specifying which grammatical categories are allowed, and point out constituency relations and semantic/syntactic roles, e.g. I didn’t sleep combines [Sbj IntrVerb] and [Sbj Aux-n’t Verb] (cited in Croft 2001: 26).35 CxG is a multi-dimensional model, in that it claims to take into account all of the dimensions of language necessary for the description of construction-specific features. In addition to adopting a broad view on semantics, as well as incorporating syntax, morphology, phonology, and discourse, CxG also takes into account information about prosody, assimilation patterns, etc. (e.g. Goldberg 2006: 10, Fried and Östman 2004 and Croft 2001: 16).36 Other theoretical frameworks such as TGG often discard the latter as extralinguistic or at least consider them to be only peripheral to descriptive analysis.

2.2.4 A construction grammar approach to SN- and TN-patterns 2.2.4.1 Construction grammar and the NP “Constructionist approaches excel at being descriptively adequate, since both generalizations and idiosyncratic particulars can be captured” Goldberg (2006: 11). A constructionist approach would therefore contribute to showing the systematic interactions between elements within SN- and TNsyntagms, since its main aim is to describe everything that is particular to a construction in a constraint-based way.37 As we have seen in the previous sections, CxG is typically very broad in the types of constraints it allows to be part of the description of a specific construction. Prosodic characteristics as well as register and dialectal features can be equally defining for a construction as syntactic or purely semantic constraints, i.e. every aspect is incorporated in the description of a construction as long as it pairs up with an aspect of the global constructional semantics. This is very important for SN- and TN-patterns, which incorporate many different aspects. Furthermore, for a corpus-based approach to SN- and TN-constructions a usagebased version of CxG is most suitable. Both Goldberg (2006) and Croft’s Radical CxG are usage-based,38 but RCxG has a very specific typological agenda (see Croft 2001). I therefore mainly follow Goldberg (2006).39 However, before we can elaborate a CxG approach to SN- and TNpatterns, we first have to see what a constructionist approach applied to

2.2 A constructional framework of the English NP 69

nominal syntax looks like. So far, CxG has displayed a predilection for describing the clause level in general and verbal (argument) structure (at various levels of schematicity) specifically, e.g. Fillmore (1968), Goldberg (1989, 1991, 1992 and 1995), Kay and Fillmore (1999) and Croft (2001). Despite CxG’s theoretical claim of bestowing equal value to all levels and units of linguistic analysis, there are few to no precedents of CxG studies that focus on accounting for the internal organization of particular NPconstructions (but see Croft 2006). The descriptive tools at hand are consequently inherently tied up with verbal notions and many of the pragmaticosemantic and syntagmatic issues involved in SN- and TN-patterns have no straightforward parallels with verb valency phenomena. A nominal implementation of CxG will hence probably require explicit reinterpretation of certain tenets and further adjustments. Even if nominal patterns are not often described as constituting constructions, lexical items effectively in NP function are addressed in a more indirect way, in terms of their external functioning, i.e. “how a construction fits in a larger grammatical pattern” (Fried and Östman 2004: 26). One such context they appear in is that of coercion, as described by Michaelis (2004) and (2005) for instance. As mentioned earlier, CxG stresses the fact that constructions themselves carry meaning, relatively independently of the specific lexical items that feature in it. As a result, nominal constituents can be coerced into an interpretation because of the requirements of the construction they appear in, as in (2.6): (2.6)

You have apple on your shirt. (cited in Michaelis 2005: 52)

In (2.6) apple has to be interpreted as a mass noun, which is not its typical use, due to the fact that “implicit type-shifting occurs because a verb’s object function is filled by a bare nominal.” (Michaelis 2005: 50). Even though the coercion effect described here is located at clause-level, this mechanism can also be invoked at NP-level, e.g. to capture the behaviour of count nouns acting as category labels in constructions such as There are two sorts of comedian (CW-Times), where the count noun comedian is in some ways similar to an (abstract) uncount noun in its conceptual reference (see Section 6.3.1). NPs and lexical items are also dealt with in the context of a fundamental question that receives quite a lot of attention in CxG, especially in usagebased versions: how do lexical items interact with more general constructions in determining overall sentential constructional meaning (cf. Goldberg 2006)? Goldberg’s (2006: 12 and 45-65) usage-based CxG makes the explicit claim that a language user’s mental construction grammar not only consists of generalizations, i.e. schematic40 constructions, but also incorporates item-specific knowledge. In a usage-based model, constructions, in

70 Chapter 2: A cognitive-functional constructional framework the sense of generalizations or prototype categories, are learned through input-driven inductive learning which is combined with general cognitive mechanisms (i.e. not specifically linguistic mechanisms). In keeping with psychological concerns, Goldberg (2006: 12) argues that language users in addition to generalizations also store and make use of item-specific knowledge about individual patterns associated with a specific verb for instance. However, in concrete CxG studies the discussion of the interaction between lexemes and constructions is restricted mainly to explaining why a certain lexeme cannot be used in a construction, when all other prerequisites of the construction are satisfied. An example cited by Goldberg (2006: 50) is the idiosyncratic behaviour of English adjectives with regard to predicative versus attributive uses, such as mere and aghast. Prototypically, English adjectives can be used both predicatively and attributively, e.g. beautiful, which can be used to modify a noun, as in A beautiful dress, as well as function as the complement of a copular verb, as in The dress is beautiful. Mere and aghast, however, are restricted to attributive and predicative usage respectively, which makes *That child seems mere or *An aghast man ungrammatical. Goldberg (2006) argues that since speakers of English know these idiosyncratic constraints, one has to claim that this information is stored separately in their minds, in addition to generalizations about prototypical adjectival properties.41 In the case of SN- and TNpatterns, the original lexical semantics of the SNs and TNs are sometimes essential to the general constructional meanings they help construe or to what the constructions can look like. The question of how lexemes interact with constructions is a very important one and deserves more fundamental answers than usually provided for in CxG. As mentioned in Section 2.2.3, CxG typically does not start from a construction-independent model to account for the syntagmatic relations within constructs, even though they do use notions like dependency when this is a noticeable property of a specific construction. In most CxG studies descriptions of constructions seem word class based, in that they mainly point out the constituents and syntactic categories to which a construction is constrained. Hence, within the domain of the NP too, dependency relations are not as such assumed to underlie the structural relations between elements, but are invoked in particular constructions when necessary. It should in principle be possible to implement the functional dependency model presented in Section 2.1 in CxG studies and describe relations mainly in terms of head-modifier relations, and additionally scoping and framing. Only RCxG differs from other constructional approaches in that Croft (2001: 203) explicitly argues that

2.2 A constructional framework of the English NP 71 the representation of the syntactic structure of a construction should not include any syntactic relations between the elements that make it up. In other words, the only syntactic structure in constructions is the part-whole relation between the construction and its elements.

This of course ties in with the radicality of his CxG, which sees everything, including syntactic categories and hence also syntactic relations as radically construction-specific. Only semantic relations and syntactic roles, i.e. partwhole or constituency relations, are posited independently from specific constructions because they have cross-linguistic validity (Croft 2001: 1825). It is essential to combine the CxG approach that I will propose in the present study explicitly with a functional dependency-based analysis of the NP. It is much more insightful, especially for the description of SNs and TNs which, as decategorialized Ns, function in the premodifier structure of the NP.

2.2.4.2 SN- and TN-patterns as collocationally constrained constructions SN- and TN-patterns qualify as constructions; they meet the strict prerequisites for a constructionist approach in that they display unpredictability in function and form. When one only looks at their [(D) (M) SN/TN + of + (D) (M) N2] surface structure, head, modifier or ambivalent status, as well as any (inter)subjective inferences are unpredictable. SN- and TN-patterns, as attested in corpus data, furthermore resemble what Goldberg (2006: 5) calls partially filled constructions, which are constructions in which some positions are variable, while other configurational slots are lexically fixed in order to arrive at a grammatical construct. Goldberg’s (2006: 5) examples of this are partially filled idioms, such as jog someone’s memory and send someone to the cleaners, with someone(‘s) in both cases acting as the variable slot. As we will see, in the case of the various SN- and TN-constructions specific SNs and TNs act as fixed lexical items in the sense of collocational nodes with different collocational sets. Hence, in SN- and TN-patterns partial lexical determination takes the shape of strong node-collocate coselection related to the various constructional uses of SNs and TNs, and dependent on the lexical semantics of these specific nodes themselves. The head use of SNs for instance requires specific collocates that instantiate the constellation or size/shape denoted by the lexically used SN. The quantifier use, however, allows for an enlarged set of collocates and imposes far fewer restrictions on the collocational range than the head use. Example (2.7) exemplifies a head use of bunch in which its fully lexical meaning

72 Chapter 2: A cognitive-functional constructional framework restricts N2 to a limited set of concrete collocates that can appear in this specific constellation, in this case grapes: (2.7)

A comely model is dressed in a toga and draped over a love seat with a bunch of grapes in her hand. (CW-OZnews)

In modifier uses such as (2.8) and (2.9), then, the N2-collocates are differently constrained than in the head noun use, allowing animate and abstract count nouns respectively. (2.8)

Yes, the world that awaits me is a bit scary; there is going to be a whole bunch of idiot journalists waiting to ask me stupid questions-but then that's life. (CW-UKmags)

(2.9)

It'll be something else and they'll start a whole bunch of new workshops. (CW-UKspoken)

Such quantifier uses often have a premodifier like whole which further emphasizes this use, and which cannot occur in head uses in this sense (see Ghesquière 2010 on the development of whole from a descriptive modifier into an emphasizer). We will also see in the corpus studies that specific semantic prosody patterns are very important for SN- and TNconstructions. As noted earlier, semantic prosody pertains to a specific relation between various collocates of the same node. Depending on whether a node in a specific context typically evokes neutrally, negatively or positively coloured collocates, it has a neutral, negative or positive semantic prosody. To use Bublitz’s words: Apparently, words can have a specific halo or profile, which may be positive, pleasant and good, or else negative, unpleasant and bad. Whenever some such word is uttered, it prompts ahead and ‘sets the scene’ (Sinclair 1992: 8) for a particular type of subsequent item. It opens not just any slot but one with a definite semantic set-up. […] With prosody we refer to the fact that a feature extends its domain, stretches over and affects several units. (Bublitz 1996: 11)42

Especially in SN-constructions, semantic prosody can give rise to collocational reclustering in the sense of new collocational patterns emerging within a grammaticalized use of an SN-expression, e.g. Bunch of morons/crap/bullshit. Firthian-Sinclairian research can also help show that semantic prosody can accompany motivations in grammaticalization, in that it appeals to the language user’s need for expressivity. Bublitz (2003) refers to semantic prosody as emotive prosody, which brings this out even more.

2.2 A constructional framework of the English NP 73

These collocational ties are not exactly the same as partially fixed or filledin constructions, but as Chapters 4, 5 and 6 will describe in greater detail, the head and modifier uses of specific SN- and TN-constructions are always related to systematically recognisable collocational patterning, semantic prosody or colligation. Colligation is also a Firthian term elaborated by for instance Hoey (1998) and Bublitz (1996). It is essentially a more grammatical penchant of purely lexical collocations in that colligation refers to “grouping a set of words on the basis of their similarity in entering into syntagmatic grammatical relations”. For instance, agree, choose and decline are in colligation with to-infinitives (Crystal 1991: 61-62). Collocations, on the other hand, refer to habitual co-selection between node and collocates “irrespective of their grammatical classes or relations” (Robins 1971: 225). Colligation too is a type of collocational structure that relates to particular layers of SN and TN meaning, e.g. the systematic association between determiners and TNs. Considering the importance of collocational relations in SN- and TNpatterns, I will argue that SN- and TN-patterns need to be referred to as collocationally constrained constructions. These are then understood as a specific subtype of partially filled constructions, namely different constructional meanings are marked by specific collocational patterning. Can CxG integrate collocational patterning of this sort in its description and tie it up with dependency and diachronic relations? In CxG collocations are discussed in the context of frame semantics and as a type of constructional patterning in itself, but not often as a defining property of a particular non-lexeme construction. Firstly, in a frame semantic model differences in collocational tendencies of two semantically similar words are explained by saying that such preferences are part of the frame invoked by the specific word; or, more precisely, the node and collocate both invoke the same frame, e.g. hound and baying (Fillmore 1985: 230-231 summarized in Croft and Cruse 2004: 12). The frame semantic model is potentially also dynamic enough to interpret a shift of focus within the frame of SNs/TNs, as suggested by Langacker (1991), which then leads to semantic and grammatical reanalysis. In such a view, the restructuring of the frame is taken to involve a change in functional status of the enclosing construction. Still, as it has been formulated in Fillmore (1985) and other CxG studies, the frame semantic approach to collocations is insufficient to account for the strong and peculiar interdependence between collocational patterns and the shift from head to modifier status of SNs and TNs. Secondly, in CxG collocations in themselves are viewed as a pattern that is somewhere midway on the cline of conventionality in semantic composition, which has selectional restrictions as one extreme and idiomatically combining expressions, such as spill the beans, as the other. Selectional restrictions are restrictions on possible combinations of words determined

74 Chapter 2: A cognitive-functional constructional framework by the semantics of the concepts, e.g. mud oozed onto the driveway is all right, but the car oozed onto the driveway is not because of a clash in semantic features in the concepts of ‘car’ and ‘ooze’. Combinations based on selectional restrictions are considered to be semantically compositional in that the meaning of the whole can be predicted from its parts. In noncompositional idiomatically combining expressions “the syntactic parts can be identified with parts of the idiom’s semantic interpretation (e.g. ‘divulge’ and ‘information’ in spill the beans)” (Croft and Cruse 2004: 249). Selectional restrictions do not rely on conventional forms in which the concepts are expressed, e.g. Goo/mud oozed onto the driveway (examples cited in Croft and Cruse 2004: 249), whereas idiomatically combining expressions do rely on conventional forms. Intermediate between selectional restrictions and idiomatically combining expressions are collocations like roasted meat and toasted bread. These are considered to be semantically compositional, in that the meaning of the whole can be predicted from the meanings of the parts (i.e. node and collocate). However, this semantic compositionality relies on convention within a speech community (examples Matthews 1981: 5, cited in Croft and Cruse 2004: 250). Collocations are, in sum, not looked at in terms of features characteristic of a particular construction, but are themselves an object of CxG studies as a peculiar kind of semantically compositional idiom licensed by convention. For our purposes collocational patterning has to be incorporated more firmly within a CxG approach to SN and TNs in order to truly discuss these as collocationally constrained constructions, a subtype of partially filled-in constructions. In addition, they have to be integrated with dependency relations as a true lexical syntagmatic relation and linked up with a diachronic dimension. Firthian-Sinclairian work has extensively studied collocational patterning as essentially lexico-semantic in nature, but not so much as a factor in language change (but see some recent studies such as Kurtböke 2001, Lorenz 2002 and Hilpert 2008). Sinclairian studies and Pattern Grammar can integrate collocational patterning and semantic prosody more easily with a CxG approach as lexico-grammatical notions that relate to constructional semantics in a very fundamental way. In Sinclair (1991) structure refers to a lexical item, i.e. node, and the collocations and other patterns associated with it. More precisely, Sinclair wants to point out that structure not only consists of units such as phrase, word and clause, but that collocation is also essential in realizing structure. In addition to collocation, Sinclair and Renouf (1991: 128) describe collocational frameworks, which “consist of a discontinuous sequence of two words, positioned at one word remove from each other.” Examples include the one involved in SN- and TN-constructions, i.e. a(n) + ? + of. Such

2.2 A constructional framework of the English NP 75

frameworks are highly selective of their collocates and, interestingly, it is noted that an item can be frequent in a certain collocational framework, but not necessarily so in the corpus as such (Hunston and Francis 2000: 25). This can be an explanation for some of the frequency effects encountered in small size noun-constructions (see Chapter 5). Pattern Grammar also stresses the strong association between sense and syntax: “words that occur in a particular framework are not a random selection but belong to particular groupings.” (Hunston and Francis: 25). These observations largely describe what collocationally constrained constructions introduced earlier are and how they are a variant of CxG’s notion of partially filled-in constructions. Another important claim in Hunston and Francis (2000: 3) is that different senses of a word are typically associated with different collocational patterns and words that share a pattern, also share meaning elements. This is relevant for SN- and TN-constructions in that the status of the SN and TN within the construction strongly relates to node-collocate coselection. The Sinclairian tradition can hence help integrate all of these collocational notions with CxG. Francis (1993) and De Beaugrande (1997) are studies that hint at such interaction between both ways of thinking. Francis (1993), within the larger context of a corpus-driven lexical approach to grammar, makes the point that appositive that-clause qualifiers for lexical items like fact can only be legitimated in restricted phraseologies. For instance, in The reason that this has been so successful is … reason cannot be appositively lexicalized by the that-clause; the that-clause is a standard relative clause here. In For the simple reason that it hasn’t worked, however, it can, because of the prepositional phrase with for, the definite article and the modifier simple (Francis 1993: 152-153). In De Beaugrande (1997: 58) the point is made that extensive corpus extractions can point up the fact that a verb like warrant determines limited collocational sets in variable positions. Its subjects typically refer, among others, to actions, resources or knowledge, whereas its direct objects refer to (in)appropriate reactions, messages, knowledge-gathering, problem-solving, etc. These two sets often show correlations, such as ‘knowledge’ and ‘knowledge-gathering’ in There is enough evidence to warrant an investigation (De Beaugrande 1997: 62). Additional constraints on the warrant construction pertain to attitudes, with pejorative values predominating, lexicalized in either the subject, direct object or both (De Beaugrande 1997: 63). Studies like these show that collocational notions can be a valuable tool outside of purely lexicographic studies, i.e. in truly lexico-grammatical studies which recognize lexical determination as integral to syntagmatic combinatorics. This perspective was lacking or not explicitly provided for by the functional dependency model developed in Section 2.1.

76 Chapter 2: A cognitive-functional constructional framework In addition to collocational patterning in a broad sense, several other defining factors of SN and TN constructional meanings can be accommodated. These phenomena overlap to some extent for SN- and TN-constructions, but there are also differences. Within the set of SN-constructions, different aspects have to be incorporated for SSNs and regular SNs. Overlapping properties for both these construction types are: premodification patterns, determiner/noun interaction, (inter)subjective values, vagueness, lexical persistence, assimilation processes, subject-verb concord, anaphoric concord, syllable structure, frame semantics of each (S)SN/TN and dialectal and register characteristics. All of these will be discussed in greater detail in the corpus studies in Part 2 and are listed here for the sake of completeness. However, in order to validate them accurately in the case studies and assess how these more or less variable elements of the constructions interact with the collocationally nodal SNs and TNs, the remainder of this section will present a specific delineation of such concepts as semantics and pragmatics. These notions, as we have seen in Section 2.2.3, are treated rather vaguely in CxG studies. Some of the constructional factors or values listed are nevertheless semantic, whereas others are pragmatic only or based on pragmatic inferencing. It is important to make these basic distinctions, while also remaining sufficiently flexible in the interpretation of both types of meaning. Semantics in CxG incorporates basically everything that aids understanding of the construction, such as encyclopaedic knowledge, register variation and dialectal variation and information about attitudes which can be incorporated in the model of frame semantics. A frame semantic model can capture rather subtle differences between what seem to be semantically very similar words (Fillmore 1977 cited in Croft and Cruse 2004: 9), such as tall and high. Tall evokes a frame that refers to erect human beings referring to their vertical extent, indicating a high point along this vertical axis. It can also invoke a frame that refers to buildings and their vertical extent. The semantically similar notion high, on the other hand, does not evoke the human frame in this particular meaning. A frame typically evokes multiple specifications of different types, and this allows a language user to focus on one element of the evoked frame and to background others. Cases of polysemy or subsenses of the SN- or TN-construction can hence be explained as instances in which certain parts of the frame evoked by the SN/TN are backgrounded and others are foregrounded. This approach is similar to Langacker’s account of SNs, where the focus on size implications to the detriment of lexical constellation meaning in their semantic profiles leads to quantifying uses (see Section 2.1.2). As we will see, such an account could explain differences in the synchronic degree of delexicalization and grammaticalization between pile and

2.2 A constructional framework of the English NP 77

heap (see Section 4.3). In the case of quantifier uses of SNs such as bunch, loads and heaps there seem to be restrictions in terms of register, as already noted in traditional reference works. In the case of TNs, dialectal effects are at work in the sense that British English prefers TN-constructions with sort of/sorta in them, whereas American English prefers kind of/kinda (see Biber et al. 1999). With regard to the latter especially, the question is whether this is a truly semantic or pragmatic meaning value. In CxG pragmatics is sometimes equated with discourse status or otherwise vaguely defined. It can refer to connotations, general cross-linguistic functional forces such as iconicity, Gricean maxims of relevance and economy, processing demands, speech act theory or affective stance (Croft and Cruse 2004: 241 and their Chapter 9). At times some of these factors are subsumed under semantics as well in CxG studies. I will distinguish between pragmatics and semantics in terms of conventionalized versus nonconventionalized meaning, as proposed by Croft and Cruse (2004: 258) who “use the terms ‘meaning’ and ‘semantic’ to refer to any conventionalized function of a construction”. Hopper and Traugott (2003: 76) make a similar distinction: semantics refers to meaning that is relatively stable out of context, whereas pragmatic meaning is typically inferential and defeasible in that it refers to meaning that is not encoded in the structure as such, but is meaning beyond structure, which is typically attributed cognitive universality. For example questions coded as inquiring into ability, such as can you pass the salt? have a common hortative implicature. In the corpus studies encoded versus not encoded will be the main fault line between semantic meaning and pragmatic inference. This will be important for the assessment of subjective and intersubjective values of certain constructions. Semantic prosody is not always or completely conventionalized, and hence seems mainly pragmatic in nature, despite its name. Maybe its name should be rephrased as affective or pragmatic prosody in certain cases (cf. emotive prosody in Bublitz 2003), but the term is probably too established to be jettisoned. In addition, there is the question of directionality of semantic prosody, i.e. is it the emotive colouring of the collocate that radiates to the node or vice versa? How do specific semantic prosodies emerge? (see note 42) Elements of structure that affect pragmatic and syntactic values are labelled pragmatico-syntactic, e.g. the (negative) polarity sensitivity of shred in There was not a shred of evidence against him. Negation pertains to syntactic structure and interacts with such discourse values as reactive contexts in court cases for instance. Collocations will be referred to as lexical syntagmatic relations.

78 Chapter 2: A cognitive-functional constructional framework 2.3 Conclusion: A dynamic model of the NP for SN- and TN-constructions Since the various uses of SN- and TN-constructions revolve around head and modifier status, the CxG model demands the implementation of the functional-cognitive dependency model proposed in Section 2.1 for the successful description of SN- and TN-constructions. In addition it has to be enriched by means of collocational notions. A dynamic framework of the NP in this study needs both the functional and the constructional-collocational component. NP models proposed by Langacker and others cannot provide for the strong co-selection between node and collocates in SN- and TN-patterns, and CxG in itself does not provide for the link between these collocational patterns and dependency. A strong point of CxG is that other factors than collocational ones can be taken into the analysis of SN- and TN-patterns. In addition, comparisons with related constructions at different levels of schematicity can be incorporated, e.g. the quantifying (S)SN-constructions can be compared with the Quantifier construction as such, as instantiated by many, much, etc.; or the SN-construction can be compared with the TN-construction. These comparisons will be discussed in the corpus studies in Part 2. Finally, in addition to the functional-constructional model, an explicit diachronic dimension has to be added to the framework of the NP. Chapter 3 will introduce grammaticalization as a dynamic field of study that can account for the diachronic dimension of structures as well as its synchronic reflexes. Grammaticalization has been defined and studied from various angles. My main aim is to mesh it with the constructional-functional approach developed in the previous sections.

Chapter 3 Grammaticalization,43 delexicalization and subjectification in SN- and TN-constructions 3.1 Grammaticalization, synchronic variation and emergent grammar In addition to describing the synchronically coexisting uses of SNs and TNs as attested by corpus data, it will be argued that these differences in status can be accounted for as the result of (diachronic) processes of grammaticalization. Over the last decades grammaticalization has become one of the most productive frameworks, especially within functional-cognitive linguistics. Any summary of the topic’s history is necessarily incomplete because of its prolific output. Section 3.2 will therefore only discuss the main changes in interest that grammaticalization studies have gone through, both in terms of the topics studied and the aspects focused on. It will eventually zoom in on core issues of grammaticalization as they affect SN- and TN-constructions by discussing the application of grammaticalization and subjectification to the level of the NP, the interaction between grammaticalization, constructions and CxG. In discussing the interaction between grammaticalization and constructions, I will distinguish between pre-theoretical approaches, which are not linked up with construction grammar, and theoretical ones, which are explicitly embedded in construction grammatical approaches. In addition, I will address some more recent criticism on (the theoretical status of) grammaticalization as an independent framework with explanatory power (e.g. Campbell 2001). Such scepticism correctly points out that certain key concepts in grammaticalization studies, such as reanalysis and semantic generalization, are often too vaguely defined, and that grammaticalization claims are sometimes too straightforwardly posited, without sufficient formal foundation. Section 3.3 will try to overcome such criticism by using it constructively to remedy imprecisions in grammaticalization theory. One way of going about this is through interaction with other grammatical models, such as CxG. In doing so, I affirm the value of the theoretical and especially the empirical use of grammaticalization studies. SN- and TN-constructions, as engaging in ongoing processes of grammaticalization, delexicalization and subjectification, especially drive home the point that the influence of collocations has been underrated. In addition, the case studies in Part 2 will demonstrate just how important pragmatic factors can be in grammaticalization. Given the fact that the grammaticalization framework hinges on the interpretation of language as essentially dynamic in nature and always on the move, as language users try to cater

80 Chapter 3: Grammaticalization, delexicalization and subjectification for their (expressive) communicative needs, this claim is not new, even though seldom explicitly argued for in such a concrete way. First, Section 3.2 will present an overview of the history of grammaticalization research by dividing it into various stages. Changing perspectives produce different factors, parameters, mechanisms of grammaticalization and focus on different motivations for it. There is some confusion in the literature with regard to the terminology used to label the various factors involved in grammaticalization and their status in the change, as also noted by Fischer (2007: 99-155) and Croft (2000: 63-78). Partly depending on their theoretical affiliation and interpretation of grammaticalization, authors have referred to factors as principles, parameters, triggers, mechanisms and motivations, with factor normally used as a more neutral cover term. These concepts are sometimes presented as interchangeable terms, when in fact their status differs. Some pertain to truly causal mechanisms, triggering actual change, whereas others are diagnostics that can be used to measure the degree of grammaticalization. Yet other factors pertain to motivations, explaining why causal mechanisms were triggered in the first place. In Section 3.2, I will use the terminology as it is used by the authors that are being reviewed. In Section 3.4, I will return to the topic of terminology and will propose how I interpret the various factors at work in the specific grammaticalization of SN- and TNexpressions.

3.2 Grammaticalization: Changing perspectives Following preparatory stages in which observations by Von Humboldt and Von der Gabelentz eventually led up to Meillet’s (1912) first explicit reference to grammaticalization, the 1990s witnessed a surge in grammaticalization studies. An important set of articles and monographs was published then which set out the bulk of grammaticalization factors and definitions still used today, e.g. Heine, Claudi and Hünnemeyer (1991), Hopper (1991), Hopper and Traugott (1993), Bybee, Perkins and Pagliuca (1994) and Lehmann ([1982] 1995). Glossing over some moot points here, we can say that all of these basically define grammaticalization as the study of a specific kind of language change in which originally lexical material develops grammatical meaning and function or involves changes from already grammatical to more grammatical. Several stages can be recognized in the development of grammaticalization theory as a fully-fledged framework. First, the focus was on morphosyntactic aspects of grammaticalization, reflected in a generally morphemebased approach to grammaticalization. Then, the main interest was in reconstructing semantico-pragmatic and functional change, which tied in with

3.2 Grammaticalization: Changing perspectives 81

a functional approach. Finally, and very recently, a constructional approach to grammaticalization has been taking centre-stage, e.g. Heine (2003), Bisang, Himmelmann and Wiemer (2004), Traugott (2003) and (2008a), Trousdale (2005) and (forthcoming) and Noël (2007) and (2007). The introduction of new interests does not mean that older interests disappear. Rather, these stages currently co-exist and can be said to display layering, in analogy to grammaticalization phenomena themselves (cf. Hopper 1991). Factors referred to in grammaticalization studies ideally should relate to formal as well as semantic reflexes of grammaticalization. Construction-based approaches aim at such holistic accounts (see Goldberg 2006). Grammaticalization studies are also gradually extending the topics of their studies. Just like CxG, grammaticalization research from the outset has had a penchant for studying verbal structures and the clause level in general. Changes from lexical verbs to (semi-)modal verbs, auxiliaries and the development of the (going) to infinitive for instance constitute examples par excellence of grammaticalization, e.g. Traugott (1989) and (1990), Sankoff (1990), Heine (1993), Bybee et al. (1994), Goossens (1999), Kuteva (2000), Krug (2000), Haspelmath (1989), Fischer (1997), Mair (1997), Los (1999) and Fischer (2000). Recently the NP has been receiving somewhat more attention (e.g. Denison 2002, Traugott 2008b and 2010b), but nominal syntax in general and shifts in NP-internal functions specifically are relatively underresearched. Within the NP as constructional environment of grammaticalization processes, several paths of development have been described (cf. Heine 2003: 594-595): the evolution of definite articles from demonstrative pronouns, of indefinite ones from the numeral one, of relative clause markers from demonstratives or interrogative markers, and of complex prepositions from location and body part nouns (e.g. Schwenter and Traugott 1995 and Hoffmann 2004). This is symptomatic of how grammaticalization research has neglected nominal syntax; mostly only grammaticalization processes that shift a nominal expression to constructions outside of the NP have been looked at. As noted in Chapter 1, SNs and TNs have received different sorts of attention in grammaticalization studies. There is a present boom in the literature on sort/kind of, describing their shifts from head nouns to discourse markers,44 e.g. Tabor (1993), Aijmer (2002), Denison (2002) and Margerie (2010). However, as noted in Chapter 1 as well, these studies only look at the big shift from NP-internal head noun use, e.g. (3.1) (3.1)

My grandfather had a curly moustache and offered two sorts of kisses: an ordinary flat kiss or a tickling kiss (CW-OZnews)

82 Chapter 3: Grammaticalization, delexicalization and subjectification to (inter)subjectified NP-external discourse markers with qualifying or filler function, e.g. (3.2) (3.2)

Those advert features that they do in the States, they’re like sort of like half hour adverts (COLT)

In addition, they look at other uses in which TNs have scope over propositions, e.g. markers of onomatopoeia and quoted speech (see Aijmer 2002 in Section 1.2.2.3). Shifts to NP-internal TN-uses that are intermediate between head noun use and NP-external uses receive little attention (but see Denison 2002 discussed in Section 1.2.2.2). SNs have altogether been left out of grammaticalization studies (but see Traugott 2010b), probably because they display shifts that are less obvious than the ones attested for subjectified TNs. It can be concluded that in studies of grammaticalization in the NP so far attention has gone to shifts in which nominal expressions move outside of the level of nominal syntax, whereas NP-internal shifts have rarely been looked at in a systematic way (but see Denison 2002 and Traugott 2010b). Sections 3.2.1 to 3.2.3 will review the various stages in grammaticalization research. Section 3.3 will present a grammaticalization approach to SN- and TN-constructions. The factors involved in their grammaticalization processes will be discussed and their status within the changes will be assessed.

3.2.1 A morphology-based approach to grammaticalization Within the morphology-based approach (or “element-based view”, in Himmelmann 2004: 31) the focus is on describing morphosyntactic change and its parameters, i.e. diagnostics to measure the degree of grammaticalization. This approach typically focuses on individual words in isolation, rather than strings of words, and tries to trace their lexical origins along socalled clines, which ultimately describe a trajectory of decrease in substance and autonomy. Lehmann’s (1985) six correlated parameters are a well-known example of such a reductive approach.45 Attrition, paradigmaticization, obligatorification, condensation, coalescence and fixation all describe degrees of grammaticalization in terms of decrease of substance (“weight”), and internal complexity of grammaticalizing items and loss of their autonomy (“cohesion” and “variability”). The first three parameters pertain to paradigmatic aspects, and the latter three to syntagmatic ones. Attrition is the loss of semantic and phonological substance. Paradigmaticization is the process of increasingly stronger integration of an item within a paradigm, e.g. grammaticalized French primary

3.2 Grammaticalization: Changing perspectives 83

auxiliaries avoir and être are completely integrated within the conjugational paradigm, whereas their Latin predecessors habere and esse/stare were not (example cited in Lehmann 1985: 307). Obligatorification is the loss of paradigmatic variability. For instance, whereas Latin de was often substitutable by ab or ex, grammaticalized French de can no longer be substituted or omitted (cited in Lehmann 1985: 308). Condensation pertains to the shrinking of scope of a grammaticalizing sign, i.e. a grammaticalizing item combines with structurally increasingly less complex constituents, e.g. Latin habere takes a full NP as its complement, even in its auxiliary use, e.g. habeo epistulam scriptam, whereas French avoir as an auxiliary only has scope over the perfect participle with which it combines, as in J’ai écrit la lettre (example cited in Lehmann 1985: 308; cf. Ramat 1982) (but see Section 3.2.2). Coalescence refers to the increase in bondedness, from juxtaposition to cliticization and beyond, which makes “autosemantic signs become synsemantic”, e.g. Latin clara mente grammaticalizes in French to clairement (Lehmann 1985: 308). Finally, fixation refers to the loss of syntagmatic variability when grammaticalized signs typically occupy a fixed syntactic slot. An example is French de and à which have to precede the NP, whereas Latin de and ad could occupy various positions within it (Lehmann 1985: 308). Lehmann (1985: 312 and further) also considers possible functional motivations for grammaticalization and the synchronic variation between more lexical and more grammatical means of expression in languages. He (1985: 315) concludes that this variation links up with language as a creative activity and reflects a speaker’s choice to use either received grammaticalized expressions, or more vivid and less grammaticalized ones.

3.2.2

Criticism of Lehmann’s parameters and semantico-pragmatic approaches to grammaticalization

The morphology-based approach to grammaticalization was supplemented by a more semantico-pragmatic or functional approach, which describes the semantic changes in, and functional-cognitive motivations behind, grammaticalization and sometimes focuses less on the formal reflexes of these processes. The latter were considered to have been overemphasized in previous morphosyntactic approaches. Within this line of research the necessity of obligatorification, paradigmaticization and increased syntagmatic fixing has been questioned by Bisang et al. (2004). In addition, Traugott (1988), (1989), (1995b) and (2003), and Hopper and Traugott (2003 [1993]) have also questioned some of Lehmann’s parameters, especially the simplicity of the semantic model behind them, expressed in the concept of semantic bleaching, as well as the

84 Chapter 3: Grammaticalization, delexicalization and subjectification decrease in scope parameter. At least with discourse markers precisely the opposite tendency is attested, i.e. an increase in scope to the inclusion of larger stretches of discourse (cf. Traugott 1995b).46 Against Lehmann’s semantic bleaching model, Traugott (1982), (1988), (1989), (1995b) and (2003) argue for an increase in pragmatic meaning in the initial stages of grammaticalization, instead of a simple loss of concrete and referential lexical meaning. This is argued to be a more exact representation of the semantic changes involved in grammaticalization, which may also include metonymy and metaphor. Traugott (1982: 257) formulates a cline for the semantic changes involved in grammaticalization as a unidirectional path from ideational or propositional over textual to expressive meaning. Traugott (1988) and (1989) then reformulate this as three semantico-pragmatic tendencies: Tendency I: Meanings based in the external described situation > meanings based in the internal (evaluative/perceptual/cognitive) described situation. Tendency II: Meanings based in the external or internal described situation > meanings based in the textual and metalinguistic situation. Tendency III: Meanings tend to become increasingly based in the speaker’s subjective belief state/attitude towards the proposition. (Traugott 1989: 3435)

In later studies, such as Traugott (1995a) and (2010b), this last tendency specifically is rephrased in terms of subjectification, which is argued to intersect with grammaticalization in various ways (see Section 3.4.2.2).47 Ultimately, then, semantic changes are argued to amount not to a general loss, but to a redistribution of meaning: the loss of referential lexical meaning is compensated for by an increase in grammatical meaning. Pragmatic strengthening or enrichment, as a mechanism of semantic change, leads to context-induced reinterpretation through repeated conversational inferences. Semantic bleaching is associated with later stages of grammaticalization only (e.g. Traugott 1988, 1989 and 2003). Hopper (1991: 21) similarly observes that Lehmann’s parameters apply to fairly advanced stages of grammaticalization, in which it is already very obvious that grammaticalization is taking place. Lehmann’s parameters are not conclusive or even applicable to incipient grammaticalization, considering the great ambivalence and variability typical of such stages. Hopper (1991: 22-33) then introduces several important principles which are argued to detect (incipient as well as later) grammaticalization and which supplement Lehmann’s parameters. These five heuristic principles are layering, divergence, specialization, persistence and decategorialization.48 Layering, persistence and decategorialization are operative in SN- and TN-

3.2 Grammaticalization: Changing perspectives 85

constructions and will be discussed in more detail in Section 3.4. Layering, referred to earlier, is comparable to Lehmann’s concept of synchronic variation and pertains to the fact that older layers of meaning coexist with newer layers with which they possibly interact. Persistence describes the phenomenon that (lexical) features of the original meaning of grammaticalizing material may persist and continue to have an impact on the specific path of grammaticalization, for instance by constraining semantic generalization and distribution over increasingly more contexts. An example is the volitional source semantics of futurity auxiliary will in some uses. Original lexical features may also resurface in grammaticalized uses, e.g. lorry in lorry loads of flattery discussed earlier as (1.17). Decategorialization builds on the notion of “degree of categoriality” argued for in Hopper and Thompson (1984) and pertains to the fact that as a unit grammaticalizes, it loses categorial properties prototypical of the (more) lexical category it was formerly a member of. For instance, an element that was originally a noun loses the potential of premodification by articles, quantifiers or adjectives, or can no longer vary between a singular and plural form. Heine (2003), like Hopper (1991) and Hopper and Traugott (1993/2003), aims to explicitly distinguish principles that can detect early stages of grammaticalization. He sees Lehmann’s parameters as formal reflexes or “concomitants” of these principles (cf. Himmelmann 2004: 33). Heine’s (2003: 578-580) four interrelated principles are desemanticization, extension, erosion and decategorialization. Desemanticization or bleaching, is understood as semantic reduction or loss in meaning content. Extension, or context generalization, refers to using grammaticalizing structures in new contexts. Erosion pertains to phonetic reduction and decategorialization is similar to Hopper’s notion, which it subsumes, but also includes “the loss of independent word status (cliticization, affixation”, which is similar to Lehmann’s concept of coalescence (Heine 2003: 578). Heine (2003: 583) sees desemanticization as the central principle and as such sees grammaticalization as “above all a semantic process”. His desemanticization seems to be in keeping with Lehmann’s notion of semantic attrition, in that it refers to the loss in lexical content, but he remains agnostic as to whether he adheres to the bleaching model or the loss-and-gain model, as advocated by Traugott, both of which he considers (Heine 2003: 592). Like Traugott (2003) and Hopper and Traugott (2003), Heine does not consider his principles of grammaticalization to have any explanatory value in themselves. They do not explain why lexical material starts to grammaticalize in the first place, even though they are centrally semantics-based. Syntactic reanalysis does not explain grammaticalization either, as some formalists claim, but is just a conditioning mechanism at work in it. Within functional-cognitive theories of language the real motivations for grammaticalization have to be sought in language’s main functions, i.e. “to

86 Chapter 3: Grammaticalization, delexicalization and subjectification communicate successfully” (Heine 2003: 578, but also Lehmann 1985: 315). Traugott (2003: 634) sees speaker-hearer negotiation of meaning in discourse as the main motivation for, but also a semantic mechanism of, grammaticalization, based on invited inferencing and semanticizing of conversational implicatures. Haspelmath (1999) makes a similar point when he notes that the interplay between the economy maxim, i.e. “talk parsimoniously”, and the clarity maxim, i.e. “talk clearly”, are insufficient to explain why a language changes (Haspelmath 1999: 1043). Following Keller (1994), Haspelmath’s usage-based view sees language change as the result or side-effect of “the cumulation of countless individual actions of speakers” who mainly follow the maxim of extravagance, i.e. “talk in such a way that you are noticed” or attract attention (Haspelmath 1999: 1043 and 1055). He prefers extravagance to the term expressivity used by Hopper and Traugott and others because expressivity in his opinion is too similar to clarity. In addition, rather than expressing meaning, Haspelmath (1999: 1057) interprets the notion of extravagance in terms of speakers trying to impress hearers by being “imaginative and vivid [; i.e.] they want to be extravagant poets”. In conclusion, in a reaction to morphosyntactic approaches to grammaticalization almost all of Lehmann’s parameters have been criticized and some authors have made proposals for a more semantico-pragmatic or functional approach. The criticism partly derives from the observation that grammaticalization more often than not is ongoing, rather than completed or deterministic in nature, i.e. it does not have to go on to completion, but can stop halfway or even reverse to a certain extent (cf. Fischer 2000). This issue in its turn ties in with the much-debated question of unidirectionality, which will be discussed in Section 3.2.5. Whereas the first stage in grammaticalization research, which focused on morphosyntactic change, typically undervalued semantic characteristics of grammaticalization as well as general functional-cognitive motivations behind it, the semantic approaches may have a tendency to disregard the formal reflexes of grammaticalization. Both deficits can potentially be remedied in a more recent stage of grammaticalization research, namely one that seeks to mesh it with constructionalism and CxG.

3.2.3 A construction-based approach to grammaticalization Attention to constructions in grammaticalization research comes in two kinds, depending on whether it is explicitly linked up with the CxG framework and its tenets or not. If not, the approach can be called "pretheoretical" (Traugott 2003: 625) and if it is embedded in CxG I will call it theoretical. The construction-based approaches proposed for instance by

3.2 Grammaticalization: Changing perspectives 87

Bybee, Perkins, Pagliuca (1994), Heine (2003), Hopper and Traugott (2003), Traugott (2003) and Himmelmann (2004) are pre-theoretical, in that they mainly point out that grammaticalization only occurs in specific constructions and contexts. Pre-theoretical attention to constructions will be discussed in Section 3.2.3.1. Bisang and Wiemer (2004)49, Noël (2007) and Trousdale (2008a and b), (2010a) and (forthcoming) are examples of theoretical approaches that seek to integrate grammaticalization research and CxG. Theoretical approaches will be discussed in Section 3.2.3.2.

3.2.3.1 Pre-theoretical constructional approaches to grammaticalization Pre-theoretical approaches either simply use the term construction to make the point that grammaticalization processes can work on multiple-word strings, in addition to single words (as advocated in morpheme-based definitions, such as Meillet 1912 and Lehmann 1985). Alternatively, and more importantly for the present purposes, the term construction shows up in definitions of grammaticalization that explicitly argue that grammaticalization works on (strings of) items in very specific environments and not on items in isolation. Authors writing within this constructional line of research emphasize that factors in grammaticalization apply to a lexical item within a well-defined construction and context of which the morphosyntactic properties and pragmatic conditions are specifiable. Grammaticalization and its criteria do not apply to all instances of some lexical unit, but only to instances of this unit in constrained contexts with particular constructional properties. For example, relational nouns shift to case markers, provided they are in an adpositional relationship with another nominal, and the source of the future auxiliary gonna is not just the verb go in itself, but the entire be going to-construction, involving progressive aspect and an infinitival clause with purposive meaning. Often construction is just synonymous with context in these definitions, although the term construction in fact refers more specifically to syntagmatic context. Context, on the other hand, usually has a broader sense and incorporates pragmatic-semantic functioning. The two pre-theoretical uses of the term construction are exemplified in Hopper and Traugott (2003: 18), when they refer to grammaticalization as The change whereby lexical items and constructions come in certain linguistic contexts to serve grammatical function and, once grammaticalized, continue to develop new grammatical functions. (Italics added)

88 Chapter 3: Grammaticalization, delexicalization and subjectification In Traugott (2003: 645) the explicit reference to constructions as grammaticalizing elements is taken out, but the second sense, which refers to them as being environments, is made explicit. Grammaticalization is [t]he process whereby lexical material in highly constrained pragmatic and morphosyntactic contexts is assigned grammatical function, and once grammatical, is assigned increasingly grammatical, operator-like function.50

Like Heine (2003) and Himmelmann (2004), Traugott (2003: 627) emphasizes the importance of both the properties of the lexical material and the larger context and construction it appears in: grammaticalization then is centrally concerned with the development of lexemes in context-specific constructions (not merely lexemes and constructions). (Original italics)

Bybee (2003b: 602) similarly concludes that it may be more accurate to say that a construction with particular lexical items in it becomes grammaticized, instead of saying that a lexical item becomes grammaticized.

Langacker (2009: 60), from the more specific point of view of Cognitive Grammar, states that [i]f grammaticization is characterized as the evolution of grammatical elements from lexical sources, it is with the understanding that the locus of change is an encompassing construction.

Similar observations can be found in Bybee (2003a), Heine (2003), Himmelmann (2004) and Traugott (2010b). Himmelmann (2004) provides an explicitly constructional approach to grammaticalization in the context of a comparison with lexicalization (see Section 3.4.9). He (2004: 31-33) sees grammaticalization in terms of three types of context expansion working at different levels, i.e. host-class expansion, syntactic extension and pragmatic extension. The first expansion is construction-internal and involves expansion of the host-class, i.e. the class of elements a grammaticalizing element is in construction with widens. For instance, demonstratives grammaticalizing into articles combine with nouns they could not co-occur with before, such as proper names. Secondly, a grammaticalizing construction may extend beyond the syntactic environment in which it is used by means of syntactic context expansion. Emerging articles, for instance, expand from core argument positions to adpositional expressions and other syntactic environments they could not occur in as demonstratives. Thirdly, the semantic and pragmatic contexts,

3.2 Grammaticalization: Changing perspectives 89

in the sense of usage contexts, can be extended in a process called semantic-pragmatic context expansion. The usage contexts of articles, for instance, are broader than those of demonstratives, including associative anaphoric uses, such as a wedding-the bride, and larger situational uses such as unique reference in the queen and a locative sense in the pub. This last type of expansion is considered to be the most important one and can be interpreted as Himmelmann’s loss-and-gain alternative to semantic bleaching. Changes at the element-level, such as Lehmann’s parameters of erosion, fusion and paradigm formation, are regarded as epiphenomena of these expansions, dependent on the type of language and construction one is dealing with.

3.2.3.2 Grammaticalization and CxG: Theoretical perspectives In addition to pre-theoretical constructional approaches to grammaticalization, recent studies look at the interaction between CxG and grammaticalization. Noël (2007) formulates an important qualification with regard to how grammaticalization studies and CxG in the strict sense might interact. His main question is whether a diachronic application of CxG would coincide with constructional grammaticalization in terms of scope of study. His conclusion is that their scopes differ: diachronic CxG, as a hypothesized separate framework, has a broader scope than a constructional approach to grammaticalization. So far, CxG has been mainly restricted to synchronic descriptions of constructions. Noël (2007) predicts that authors envisaging a diachronic approach to CxG will naturally turn to the question of how constructions, as form-meaning pairings, emerge and not to the question of how a more lexical construction acquires grammatical meaning. As observed earlier, the shift from more lexical to more grammatical meaning in a way is a non-issue in CxG. As Noël (2007) notes, some construction grammarians have hence applied the term grammaticalization to the emergence of grammar as such, which Noël (2007) refers to as "symbolization". The latter pertains to “the initial formation of a construction, i.e. a primary association of a meaning with a particular (morpho)syntactic configuration” (Noël 2007: 19). In order to maintain Traugott’s (2003) narrow definition of grammaticalization and combine it with CxG, the study of symbolization is left to diachronic CxG, as a potentially separate discipline from constructional grammaticalization (Noël 2007: 19).51 What does a CxG-approach to grammaticalization in the narrow sense entail for the parameters and factors discussed in the previous sections, both the formal and semantico-pragmatic ones? Interpreting Goldberg

90 Chapter 3: Grammaticalization, delexicalization and subjectification (2006: 220) positively, lexicon and grammar can be argued to form a continuum of constructions (but see note 32). In the present study, grammaticalization presupposes extant substantive constructions as its input. Nevertheless schematicity of some sort is involved in grammaticalization. On the one hand, I follow Traugott (2006: 2), who argues that grammaticalization can have an impact at different levels of constructional schematicity, in the sense that a specific grammaticalization process can affect sets of similarly behaving constructions. Traugott (2006: 2) distinguishes between the following levels of schematicity. Firstly, there are “macro-constructions”, i.e. “high-level schemas, the highest level relevant for the discussion at hand”. For our purposes this level includes what I will refer to as the Binominal, Quantifier and Categorization constructions, generalizing all instances of binominal, quantifier and categorizing expressions into schematic constructions. Secondly, there is the level of “meso-constructions”, or “sets of similarly-behaving constructions”, e.g. generalized kind/sort of constructions as distinct from the set with a bit/lot of constructions. Thirdly, there are “micro-constructions” or “individual construction-types”, e.g. a generalized a lot of-construction versus the a bit of-construction (examples adjusted from Traugott 2006). Finally, at the lowest level of schematicity are fully substantive “constructs”, which are specific empirically attested tokens of a construction that serve as “the locus of innovation”, e.g. (3.3)

He threw himself on the floor, he kicked a couple of dishwashers and washing machines and knocked a load of stuff off the top of the washers (CW-UKspoken).

On the other hand, I follow Noël (2007) and will argue that grammaticalization processes always start off in concrete (substantive) constructs, as manifested by the corpus data in this study, and cannot work directly on various levels of schematicity.52 In this interpretation, a posteriori generalizations can be made from these concrete processes at various levels of schematicity. Translated into Traugott’s theory of pragmatic enrichment, this means that inferences attested frequently in constructs can conventionalize or semanticize at a micro-constructional level and at increasingly higher levels of schematicity. The micro-constructions looked at in this study for instance each have individual schematic characteristics, e.g. the SN-expression shred of typically appears with uncount abstract nouns, whereas a lot of allows for both concrete and abstract count and uncount nouns. At the meso-constructional level too generalizations can be made, i.e. over sets of SN-expressions as distinct from sets of TN-expressions (or other N of N-structures), for instance in terms of the different developmental paths and the number of layers of meaning they construe. At the macro-

3.2 Grammaticalization: Changing perspectives 91

constructional level, generalizations can be made over SN- and TNconstructions and N of N-structures in general, involving for instance shared factors and processes such as pragmatic inferencing, metonymic shifts, delexicalization, decategorialization, etc. Finally, I hold that the macro-constructional level is also important as a final layer of schematization over specific grammaticalization processes, in that newly grammaticalizing elements in constructs may be recruited for the Quantifier and Categorization construction, analogously to older members of these two categories, the categorial properties and behaviour of which they gradually (or partly) inherit. From this point of view, the macroconstructional level can be interpreted as a prototype category (cf. Goldberg 2006)53 that may serve as a target-model or analog for new members. The macro-constructional level is then affected in the sense that it expands as new members find their way into it. In all this, it should be kept in mind that the original, more lexical meaning of grammaticalizing constructs does not necessarily disappear, as a meaning can be polysemous (cf. Hopper’s 1991 notion of layering).

3.2.4 Criticism of grammaticalization research Recent years saw the inevitable criticism against the grammaticalization framework, for instance in a thematic issue of Language Sciences (Campbell and Janda 2001), which primarily challenged the independent theoretical status of the concept grammaticalization. It is instead argued that this concept is derivative and consists of various grammaticalizationindependent mechanisms, common to language (change) as such. According to Campbell (2001) the notion of grammaticalization is mainly useful as a heuristic, i.e. grammaticalization studies have focused attention on some very interesting phenomena that would otherwise have remained uncovered. In this line of thought, it is claimed that labelling a phenomenon as grammaticalization has no explanatory value whatsoever, since it does not subsume “a distinct set of principles for its explanation” (Newmeyer 1998: 226). Other important correlations between formal and functional reflexes of grammaticalization are similarly deconstructed as being epiphenomena of more fundamental grammaticalization-independent principles. As pointed out earlier, most people working within grammaticalization research acknowledge that the changes involved in grammaticalization processes are not unique to it, but emphasize the great correlation between all of these within this specific process (e.g. Heine 2003) (see note 48). Joseph (2001) and Campbell (2001) have certainly addressed genuine vexed questions in grammaticalization theory and have pointed out inaccuracies or outright mistakes in concrete case studies of grammaticalization.

92 Chapter 3: Grammaticalization, delexicalization and subjectification However, at times they undeservedly criticize grammaticalization theory for not being able to account for the emergence of all grammatical forms, irrespective of lexical origins. This cannot be expected from any linguistic framework, and is not part of grammaticalization theory’s explicit goals, nor is it part of the definition used in this study (but see e.g. Givón 1995, Heine 2003: 595, Bybee 2003a, Bisang et al. 2004,54 Hoffmann 2004, Haspelmath 2004, Trousdale 2005 and 2008b, some of which roughly equate grammaticalization with emergence of grammar or grammatical change as such). It is clear that vague definitions in grammaticalization studies often lie at the origin of such discussions. Many of these points of discussion can hence be resolved by clearer definitions of grammaticalization, its criteria and goals. Contributions by Campbell (2001) and Joseph (2001) have justifiedly criticized the cavalier definitions of several key concepts in grammaticalization such as reanalysis and semantic generalization, as well as the range of its objects itself. In addition, it has been pointed out that grammaticalization claims are often too lightly made, without sufficient formal foundation, even though this is key in grammaticalization claims. Section 3.4 will attempt to remedy this by means of clear interpretations of the parameters and mechanisms involved in the grammaticalization of SN- and TNpatterns.

3.2.5 Discussion and conclusion Over the years several key issues in grammaticalization have become the focus of debate, both within and outside of the grammaticalization framework. Particularly contentious is the notion of unidirectionality, i.e. the idea that grammaticalization moves from less grammatical to more grammatical and not the other way around (see Newmeyer 1998, Fischer 1999 and 2000, Haspelmath 1999, Fitzmaurice 2000 and Norde 2009 on degrammaticalization).55 In addition, the relationship between grammaticalization and lexicalization and possible intersection with subjectification has been given ample thought (e.g. Wischer 2000, Himmelmann 2004, Brinton and Traugott 2005, Traugott 2010b). The role of semantics and pragmatics as motivating factors has been queried and in general the overall boundaries of grammaticalization are up for debate. Growing interaction with corpus linguistics has led to a more usagebased approach to grammaticalization and has enhanced empirical validation of grammaticalization claims (e.g. Aijmer 2002, Brems 2003a, Lindquist and Mair 2004, Breban 2010). An extreme usage-based interpretation is found in Hopper (1987), who sees all grammar as emergent grammar, i.e. grammar is not static, but constantly under construction; there is

3.2 Grammaticalization: Changing perspectives 93

no grammar, only grammaticalization. The significance of collocational notions (e.g. Lorenz 2002, Hilpert 2008, and the importance of very specific contexts for triggering change have been stressed by Traugott (2003) and Hopper and Traugott (2003) amongst others. Nevertheless, Himmelmann (2004: 32) correctly points out that “most work in grammaticization has not yet begun to take the implications of this view seriously”. This brings us to the following definition that integrates all of the issues cited above. Grammaticalization is The process whereby lexical material in highly constrained pragmatic and morphosyntactic contexts is assigned grammatical function, and once grammaticalized, is assigned increasingly grammatical, operator-like function. (Traugott 2003: 645, repeated from Section 3.2.3.1)

This is the definition that will be used in this study. It dissociates, from the start, grammaticalization and development of grammar as such. Secondly, Traugott’s definition adds the important point that grammaticalization takes place in very specific contexts and/or constructions that trigger the grammaticalization in the first place. An alternative formulation in Hopper and Traugott (2003: 18) focuses more on the interaction of these different factors in grammaticalization, which is defined as the research framework for studying the relationships between lexical, constructional, and grammatical material in language, diachronically and synchronically.

This definition is interesting from a CxG perspective on grammaticalization as well, adding an important diachronic dimension to the kinds of constructional interaction typically studied in CxG.

3.3 Grammaticalization of SN- and TN-constructions: From source to target construction How, then, should one go about describing grammaticalization as a change from a more lexical/less grammatical construction to a more grammatical one? If, with Himmelmann (2004: 41), one argues that “the unit to which grammaticalization properly applies are constructions”, then ideally, a complete account of grammaticization must include a description of the construction at various evolutionary stages, from source to end result. (Langacker 2009: 60)

94 Chapter 3: Grammaticalization, delexicalization and subjectification This statement by Langacker combines some of CxG’s and grammaticalization’s basic tenets. A CxG approach to grammaticalization involves accounting for all of the configurations a grammaticalizing construction goes through as it approximates the target construction in its various dimensions. The path leading from source to target seems to be essentially one of decategorialization and recategorialization, or decategorialization in a lossand-gain way (see Section 3.4.3), potentially reflected by grammaticalization parameters such as coalescence, erosion, layering and persistence (cf. Heine 2003: 588, who claims that formal concomitants of grammaticalization, as described by Lehmann 1985, are “effects of decategorialization”). The formal and functional decategorialization is itself motivated by semantic-pragmatic factors to do with language users' expressive needs and metonymic shifts in the concepts of certain SNs. The original lexical meaning of constellations, containers, shapes, etc. is backgrounded and the measuring of size is foregrounded in keeping with the target construction. Despite the fact that the source-to-target metaphor may suggest a clear goal and a linear, unidirectional trajectory for grammaticalization processes, it should be kept in mind that grammaticalization processes are always potential in nature, may pause or reverse and involve very complex paths of change (cf. Fischer 2000). In addition, it should be noted that not all grammaticalization processes have well-defined source constructions in the common understanding of construction, but may be definable only in more general semantic terms. Section 3.4 will discuss the factors at work in the grammaticalization processes of SN- and TN-constructions as well as epiphenomena and processes of change often associated with grammaticalization. This entails deciding on the status of these factors in terms of labelling them as causal mechanisms, parameters and motivations for instance. In deciding on the status of the various factors involved in the grammaticalization of SN- and TN-constructions, I will attempt to contribute to refining the terminological apparatus of grammaticalization research. All of these factors are implemented within a constructional approach to grammar.

3.4 Factors at work in the grammaticalization of SN- and TN-constructions Several of the factors in and parameters of grammaticalization, as well as concomitants or phenomena often associated with it in the literature, apply to SN- and TN-constructions. In keeping with the semiotic fundamentals of both CxG and the cognitive-functional framework set out in Chapter 2, these are formal, semantic and pragmatic in nature and interpreted as inter-

3.4 Factors in grammaticalization

95

acting. They are ultimately motivated by more general functional-cognitive factors, such as expressivity or extravagance in meaning negotiation (Haspelmath 1999). In addition, in keeping with construction-based grammaticalization in the vein of Traugott (2003) and Hopper and Traugott (2003), special attention is given to the syntagmatic relations involved in grammaticalization, especially collocations, as a specific lexicosyntagmatic relation that has been undervalued in grammaticalization research (but see Hilpert 2008). The factors relevant to SN- and TN-constructions overlap to some extent at what we could call the macro-constructional level, i.e. the level of schematicity generalizing over both construction types (as well as other grammaticalizing N of N-structures, such as on the verge/brink of studied in Vanden Eynde 2003). These include reanalysis manifested in bridging contexts and delexicalization. At the meso-constructional level, i.e. the two subsets separately (kind/sort/type of versus bit/bunch/loads/etc. of), or subsets within both construction-types (for instance the SSNs versus the large SNs), developmental paths may differ. For instance, TN-constructions display many more layers of uses than SN-constructions. Not all factors apply to the same extent at the micro-constructional level either, with for instance individual SN-expressions differing in terms of extent of delexicalization, semantic prosody, (inter)subjectification, grammaticalization and syntactic extension. In the following sections, the relevant factors will be discussed separately. Certain of these are argued to be associated with incipient grammaticalization, whereas others apply to later stages of grammaticalization (e.g. Hopper 1991 and Traugott 2003). Since the grammaticalization processes in this study are essentially ongoing, with most grammaticalizing constructions still having their original more lexical use in addition to grammaticalized and ambivalent uses, both types of factors will be considered. This will allow me to compare SN- and TN-constructions, both as sets and individually, in terms of degree of grammaticalization (cf. Lehmann 1985). Several authors have expressed their view on how the various factors relate. Heine (2003: 588-593) for example sees Lehmann’s (1985) structural parameters as effects of the more general principle that he calls decategorialization. Nevertheless, references to factors, motivations, mechanisms and parameters intertwine to such an extent that it is far from clear what really motivates what. Furthermore, with each specific grammaticalization study it becomes more and more clear just how variable critical factors in grammaticalization can be. Given the fact that grammaticalization parameters can differ in nature depending on the part of grammar they work on (cf. Traugott 2003), it is probably impossible to claim a specific hierarchy between the factors discussed in the literature.

96 Chapter 3: Grammaticalization, delexicalization and subjectification In the present study I will argue that parameters refer to diagnostics that allow to measure, with hindsight, the degree of grammaticalization (cf. Fischer 2007). These include Lehmann’s parameters, but also Hopper’s principles of layering and lexical persistence, since these variables can only be assessed when grammaticalization is already well underway.56 The main truly causal mechanism which directly effects structural and semantic changes will be argued to be reanalysis (based on analogy). Reanalysis will be discussed in Section 3.4.1 and will be argued to be a key element in any grammaticalization argument. I will attempt to formulate a construction-based approach to reanalysis, which ties it in more closely with analogy (compare Fischer 2007, De Smet 2009 and Langacker 2009). Following the discussion of other factors in grammaticalization, I will compare grammaticalization with another process of change, namely lexicalization in Section 3.4.9 and will assess possible interference between the two. The discussion in the following sections hence moves from more central factors over parameters and more general factors to epiphenomena and a comparison with another language change.

3.4.1 Reanalysis and analogy Reanalysis and analogy are generally considered to be the most important mechanisms in all types of language change (cf. Harris and Campbell 1995), including grammaticalization. Most of the time they are distinguished strictly, with reanalysis being considered the primary mechanism of syntactic change and analogy the secondary one. Hopper and Traugott (2003: 39) for instance define the two mechanisms as follows: In reanalysis, the grammatical ʊ syntactic and morphological ʊ and semantic properties of forms are modified. These modifications comprise changes in interpretation, such as syntactic bracketing and meaning, but not at first change in form. Reanalysis is the most important mechanism for grammaticalization, as for all change, because it is a prerequisite for the implementation of the change through analogy. Analogy, strictly speaking, modifies surface manifestations and in itself does not effect rule change, although it does effect rule spread either within the linguistic system itself or within the community.

Section 3.4.1.1 will discuss reanalysis. Via a discussion of the layered semi-stable systems of SN- and TN-constructions in Section 3.4.1.2, I will present an interpretation of reanalysis that integrates it more closely with analogy in Section 3.4.1.3.

3.4 Factors in grammaticalization

97

3.4.1.1 Reanalysis There has been some debate in the literature about the precise relation between reanalysis and grammaticalization, or even the necessity of reanalysis for grammaticalization. For Meillet (1912) and Heine and Reh (1984) reanalysis and grammaticalization are coextensive or synonymous, even though Meillet did not use the term reanalysis. Others consider them to be distinct, but rather idiosyncratically argue that grammaticalization does not need reanalysis (e.g. Haspelmath 1998, who regards reanalysis as saltation). The position assumed in this study, as well as in Hopper and Traugott (2003) and Campbell (2001) for instance, is that reanalysis is a criterial property of grammaticalization, in the sense that every case of grammaticalization involves reanalysis; however, not every instance of reanalysis is a case of grammaticalization. Reanalysis can for instance also be involved in compounding or conversion, often generalized over as lexicalization (see Section 3.4.9), or in word order changes, none of which are considered instances of grammaticalization here. In criticism of grammaticalization as a framework, authors working in more formalist approaches, such as Campbell (2001), have pointed out that reanalysis is a moot point in grammaticalization, in that it is often too lightly posited without sufficient formal foundation. In functional approaches, then, reanalysis risks being reduced to semantic reinterpretation, or semantic generalization. Ultimately, the interpretation of reanalysis is dependent on theoretical orientation. Authors working in a more formal framework, such as Campbell (2001), see syntax and mechanisms such as syntactic reanalysis as explanatory, whereas functional approaches consider syntax itself to be in need of explanation (cf. Heine 2003: 593). This is why Campbell (2001: 151) insists that reanalysis is the only mechanism that can provide grammaticalization with some independent explanatory power: reanalysis (also sometimes extension) is the determining mechanism that explains grammaticalization and without appeal to these mechanisms, grammaticalization has no explanatory power of its own.

He (2001: 7) claims that semantic changes such as semantic bleaching are “not an empirical property that can be tested”, since “it is an artefact of the definition of grammaticalization” as “a shift of lexical or grammatical to more grammatical”. Phonological reduction, unidirectionality and other concomitants of grammaticalization are similarly deconstructed as neither necessary nor sufficient for grammaticalization. Syntactic reanalysis is the only criterion left that can lend some independence to grammaticalization claims, as grammaticalization is not a mechanism of change in its own right.

98 Chapter 3: Grammaticalization, delexicalization and subjectification In this context it has to be noted that, despite the importance attributed to it, reanalysis as such seldom appears in definitions of grammaticalization and is rarely defined explicitly in functional accounts, but sometimes just used as a synonym for change in semantics or category status observed in grammaticalization. Heine (2003: 593) even wants to do away with the term, partly because it causes terminological confusion and partly because it has no explanatory power in his functional framework. With Campbell (2001: 30), I argue in favour of explicitly incorporating the term in a definition of grammaticalization, even though, against Campbell (2001), I do not consider it explanatory in itself. Like Hopper and Traugott (2003) and Campbell (2001), then, I view reanalysis as the typically covert reinterpretation of a given construction, both on the semantic and formal level. Reanalysis is covert in that at first no overt changes are apparent in the surface realization of the construction (cf. Langacker 1977). Reanalysis is not just a renaming of categories, but involves changes in “constituency, hierarchal structure, semantic-syntactic category labels, grammatical relations, and cohesion (type of boundary)” in the structural organization of a construction (Harris and Campbell 1995: 61 and cited in Hopper and Traugott 2003: 51, 63). Reanalysis essentially depends on the possibility of more than one analysis of a particular structure becoming available. This is similar to Hopper and Traugott’s (2003: 52) claim that reanalysis presupposes pragmatic enrichment brought on by meaning negotiation between speaker and hearer. Frequent conversational and context-dependent implicatures gradually become semanticized into the conventional meaning of the grammaticalizing expression, independent of the specific context that triggered off the pragmatic inferencing. This constructional approach ties in with the observation by Bisang and Wiemer (2004: 9) that [r]eanalysis does not happen out of the blue, it happens with regard to a concrete syntagmatic environment within which a certain linguistic item can be interpreted as belonging to a position differing from its erstwhile positional properties.

For SN- and TN-expressions this syntagmatic environment is the binominal construction described in Chapter 1 as [(D) (M) N1 + of + (D) (M) N2]. Within these syntagms the structural combinatorics, grammatical relations, category labels and boundary type are modified covertly. The original structure of SN- and TN-constructions, [SN/TN]+[of+N2], is rebracketed as [SN/TN+of]+[N2], which entails an inversion of the hierarchic dependency relations, i.e. the head/modifier-relations. The original head noun becomes part of the modifying string, SN/TN+of, in the quantifying and extended categorization uses respectively. Category shifts go from noun to, among

3.4 Factors in grammaticalization

99

others, quantifier, determiner, adverbial, etc. uses. The string nature of SN/TN+of marks the shift in morphological boundary, which, in cases such as kinda, sorta, typa, alotta and loadsa, is also reflected graphologically, especially in more informal registers.

3.4.1.2 Layered semi-stable systems of head, modifier and ambivalent uses On the basis of the corpus studies to be discussed in Part 2, I argue that the ongoing grammaticalization processes and reanalyses at work in SN- and TN-constructions have at present resulted in layered semi-stable systems of SN- and TN-uses (cf. Brems 2003a). These semi-stable systems include head and specific modifier uses and incorporate differences in degree of grammaticalization between the individual SNs and TNs. In addition, especially SN-constructions also include cases which are ambivalent between head and modifier use. These often involve metaphorization and/or lexical persistence (Hopper 1991) and subsume two further categories, namely ambiguous versus vague cases, which are two different things (e.g. Crystal 1991, Lessau 1994 and Geeraerts 1993). In this study ambiguity refers to structures that allow two structurally different parsings, which rely on two distinct ways of contextualization. The idea is that typically either interpretation (e.g. head or quantifier) cancels the other one out in a more specific context, but that in certain contexts contextual clues support both readings, e.g. (3.4) (3.4)

We still have to move loads of furniture and stuff.

in which loads (of) can refer to an actual (car/van-)load, or ‘an amount of’. The former interpretation is encouraged by the presence of the concrete uncount noun furniture and the verb move, whereas the latter interpretation is encouraged by the abstract uncount noun stuff. Following Evans and Wilkins (2000), ambiguous instances of SN/TN-patterns are argued to point towards bridging contexts, a phase in semantic change in which, in addition to an item’s established meaning A, a new meaning B comes into existence because a regularly occurring context supports an inference-driven contextual enrichment of A to B (Evans & Wilkins 2000: 550).

Bridging contexts “essentially involve the co-existence of two possible parsings, in the sense that either reading makes sense and is supported by the context” (Willemse 2007: 26).

100 Chapter 3: Grammaticalization, delexicalization and subjectification Vagueness, by contrast, is used here in the sense of Geeraerts (1993: 228) and Willemse (2007: 26), who notes that it involves two or more semantic features simultaneously playing a role in the interpretation of a structure: grasping the meaning of such a structure involves incorporating two or more different semantic features into one global interpretation. (Original italics)

A vague unit of language cannot and in fact need not be reduced to one of the meaning layers it incorporates, because vague cases thrive on this tension. In the specific case of SNs for instance, vague readings are irreducible blends that appeal to both the lexical and grammatical meaning of SNs, but do not allow disambiguation, e.g. (3.5)

The British have forged a fine tradition of gardening and cannot afford to sit on their well-clipped laurels. Striding past the compost heap of nostalgia, comes Christopher Lloyd. (CW-Times)

Example (3.5) revolves around an extended gardening-metaphor, in which heap has literal reference, but at the same time expressively measures the years of tradition in gardening and nostalgia. The abstract nature of the latter N2 fuels the overall vagueness, which is very consciously construed in this context. As we will see, ambiguity contexts often reflect ongoing semantic and structural shifts, whereas vagueness can be a more definitive stage. Actual reanalysis comes about through pragmatic inferencing, or abduction (Hopper and Traugott 2003), which is associated with bridging contexts. The motivation behind such pragmatic enrichment or context-induced reinterpretation ultimately resides in the general functions of language. In the case of SN-expressions for instance the main reason for recruiting them as new quantifying expressions is the need of (hyperbolic) expressivity (cf. Haspelmath’s 1999 notion of extravagance): Meaning changes and the cognitive strategies that motivate them are central in the early stages of grammaticalization and are crucially linked to expressivity. (Hopper and Traugott 2003: 76)

Since reanalysis is a covert process, it is by definition very hard to find formal reflexes. However, the constructional approach presented in Chapter 2 can reveal more evidence, as it can take changing syntagmatic relations such as collocations into account. Section 3.4.1.3 will interpret reanalysis within a CxG approach to grammaticalization, which involves a closer integration with analogy.

3.4 Factors in grammaticalization

101

3.4.1.3 Analogy and reanalysis: A closer integration Just like reanalysis, analogy has been subject to terminological confusion, but it is generally distinguished from reanalysis as a mechanism of syntactic change. Hopper and Traugott (2003: 68) and most standard treatments of the subject see them as interdependent: analogy, as a secondary mechanism of change, overtly “modifies surface manifestations” by spreading the reanalyzed interpretation over contexts in which the former analysis is no longer possible, and properties of the former environment that allowed for a new interpretation are no longer obligatory (Hopper and Traugott 2003: 39). However, analogy “does not involve immediate modification of underlying structure” (Campbell 2001: 24). Since reanalysis is typically covert, and actual proof of it is notoriously hard to find, it is analogical extension that mostly provides the prime evidence that a change has taken place. Analogy is further defined as “the attraction of extant forms to already existing constructions” (Campbell 2001: 63-64), and alternatively labelled “rule generalization”, “generalization”, also in Traugott (2006: 3), and “extension” (Harris and Campbell 1995). An example is the spread of will as a future auxiliary to contexts with all kinds of subjects, whereas it was first restricted to contexts with an animate subject (cf. Campbell 2001). Reanalysis, analogy and grammaticalization are in this standard view causally and chronologically linked: first reanalysis takes place, followed by analogy in the sense of extension (cf. Himmelmann 2004). Reanalysis focuses on syntagmatic combinatorics and has been associated with abruptness and the introduction of new material in language created ex nihilio (Hopper and Traugott 2003: 46, Haspelmath 1998 and De Smet 2007). Analogy, on the other hand, is said to focus on the paradigmatic axis and on similarities between formerly established categories and category members, and new readings of structure that compare to them. In sum, despite their obvious interdependence, reanalysis and analogy are normally distinguished and attributed different tasks, effects and locations in the synchronic organization of grammar. Thus, whereas reanalysis is attributed truly innovative power, analogy gradually spreads or generalizes this new analysis to increasingly more contexts. Several observations can be made with regard to this traditional distinction between reanalysis and analogy. For one thing, the present study, as well as most case studies of grammaticalization in fact, looks at cases of reanalysis that do not lead to the establishment of radically new categories or structures that were not yet part of the grammar before.57 In many cases of reanalysis there is a clear existing model, especially on a rather schematic level, for instance in the emergence of new futurity markers (e.g. Mair 1997, Bybee 1994), new semi-modals (e.g. Krug 2000), new complex prepositions (e.g. Hoffmann 2004), new postdeterminers (Breban 2010). In

102 Chapter 3: Grammaticalization, delexicalization and subjectification the case of SN- and TN-constructions the SN- and TN-expressions are reanalyzed, among other things, as quantifiers and postdeterminers respectively, categories that have already been established in the English grammar, especially the quantifier category. On a more fine-grained level, grammaticalized form-meaning combinations can be partially new, as with subjectively coloured uses of bunch in examples like a bunch of crap. In light of my CxG-approach to grammaticalization, it is important to note that analogy is well established in CxG as a means of synchronically interpreting and producing structures in terms of their parallels with other, more general, constructions (Goldberg 2006). From this perspective, reanalysis is defined as the semantic and syntactic reinterpretation of a structure based on similarities with an already existing category, the grammatical meaning and behaviour of which it gradually adopts through analogy (cf. Fischer 2007: 122 and further). By connecting this with CxG and the levels of schematicity set up by Traugott (2006), I can distinguish a lexical or less grammatical source-construction in the grammaticalization process at the level of constructs, and a (more) grammatical target-construction, which is situated at a more schematic level. Langacker (2009: 74) makes a similar point when discussing the development of the quantifier use of a lot of. The quantifier-construction is independently present in the grammar of English and “offers a model for an alternative analysis” of the SN-expression, “whether it facilitates the changes or merely provides a way to assimilate the resulting structure in the grammatical system.” A lot of and a bunch of function analogously to much and many in those respects (Langacker 2009: 75). In keeping with Langacker, it can be noted that the model for analogy does not have to display the same syntagmatic make-up as the items following it, even though this is sometimes presupposed in grammaticalization studies. Traugott (2010b) discusses a bit of as model for a shred of, and Hoffmann (2004) uses complex prepositions as model in his study on preposition noun prepositionstructures. In a functional CxG-model, however, it is important to distinguish models at different levels, e.g. a bit of as well as some and any in the case of SSNs (see Chapter 5). The general applicability of such a reanalysis/analogy concept is not clear. It does seem that at least an important set of grammaticalization processes involves a concept of reanalysis that is more closely associated with analogy than commonly assumed. I argue that analogy is certainly more than mere extension or generalization to other contexts, as suggested by Harris and Campbell (1995) and Himmelmann (2004). It involves extension, but also the re-construction of change, most noticeable in the extension and reclustering of collocational sets in the case of SN-expressions for instance (cf. Fischer 2007: 123). Furthermore, it involves the analogical effect of functional, conceptual features. Hopper and Traugott (2003: 235

3.4 Factors in grammaticalization

103

footnote) suggest such an intimate relation between reanalysis and analogy, when stating that “[i]n some sense every new analogy is a minor reanalysis.” In this study, analogy is not just a vehicle that transports a reanalysis to other contexts, but itself shapes this reanalysis. The factors and processes that will be looked at in Sections 3.4.2 to 3.4.9 also to a large extent interact with each other. I take (metonymic and metaphorical) pragmatic inferences and the cognitive (un)conscious strategies of expressivity that motivate these to be the ultimate triggers of the grammaticalization process, which are paralleled by decategorialization. Metonymic and metaphorical inferencing tie in with the syntactic mechanism of reanalysis, incorporating a notion of analogy. Other semantic, phonetic and morphological changes, involving erosion or attenuation of substance, are hypothesized to take place at later stages, as a consequence of conventionalization and routinization, as a force competing with expressivity (cf. Hopper and Traugott 2003: 98, Haspelmath 1999).

3.4.2 Semantic changes, layering and persistence, (inter)subjectification 3.4.2.1 Delexicalization As noted by authors critical of grammaticalization studies, such as Campbell (2001), semantic processes involved in grammaticalization are often too vaguely defined, or the only proof for a grammaticalization claim.58 The concept of reanalysis is also at times restricted to semantic reanalysis without bringing in formal reflexes (cf. Heine 2003, who sees grammaticalization as essentially a semantic process). Grammaticalization claims then risk being equated with simple semantic reanalysis or semantic change as such. I have argued instead for a fundamentally semiotic interpretation of reanalysis, with both semantic and formal reflexes. Moreover, I operationalize the semantic process of delexicalization mainly as a broadening of collocational range. This approach to delexicalization can be seen as a variant of Hopper and Traugott’s (2003) loss-and-gain model. For instance, with SNconstructions, the more abstract size implications of SN-expressions invite pragmatic inferences in the very early stages, paralleled by a backgrounding of the lexical semantics, which denote, for instance, a type of configuration or container. This results in an overall redistribution of meaning, rather than overall loss (cf. redistributional decategorialization, see Section 3.4.3). Pragmatic inferences draw on one or a selected set of semantic features of the grammaticalizing SN and TN, to the detriment of others (i.e. loss), but the inferred meaning components themselves gain in importance and can even develop their own further inferences, which can motivate additional

104 Chapter 3: Grammaticalization, delexicalization and subjectification developments, such as the more subjectively coloured uses of bunch of in a bunch of morons and of bit of a in a bit of a jerk (cf. Sweetser 1988) (see Section 3.4.2.2 on subjectification). This redistribution of meaning also entails overall semantic generalization in the SN (and TN), which allows for the collocational extension of N2, and makes the delexicalization concept used here empirically testable. Delexicalization hence subsumes two interacting steps: semantic generalization in the SN/TN and extension of collocational range (cf. Himmelmann’s host-class expansion). In addition to extension of collocational range, SN-constructions may also display collocational reclustering. Some SNs may develop rather restricted and specific collocational sets, in the sense of significantly favoured collocates, in their quantifier use, e.g. shred/scrap/(skerrick) of evidence, or, in the case of subjectified bunch of and bit of a, attract collocates with a specific semantic prosody, namely predominantly negative ones. Collocational patterning, i.e. extension and reclustering of collocational range in terms of semantic prosody, is a lexical syntagmatic relation that has received little explicit attention in grammaticalization studies as an overt marker of reanalyzed grammaticalized status59 (but see Lorenz 2002 and Hilpert 2008). Still, such Sinclairian notions are an important contribution to a usage-based constructional approach to grammaticalization. Since reanalysis is a covert process, it is hard to find formal reflexes; a constructional approach taking into account changing (lexical) syntagmatic relations such as collocations enhances the visibility of the reanalyzed status of SNs and makes it empirically attestable. Semantic prosody, as “emotive prosody” (cf. Bublitz 2003), for a large part determines the meaning of valuing quantifying uses of SNs such as a bunch of crap and a load of codswallop. Both reclustering and semantic prosody can be seen as going against grammaticalization and delexicalization, which are typically associated with ever increasing generalization and distributional extension (cf. Himmelmann 2004), whereas semantic prosody is considered by some to be an indication of incomplete grammaticalization and delexicalization, blocking it from running its full course (cf. De Smet & Ghesquière 2010). However, semantic prosody patterns appear to be part of the quirky developmental path grammaticalization can display, often because of the individuality of the original lexical semantics of SNs, for instance in the form of lexical persistence. In view of its pragmatic power of subjective expression and the explanatory focus such factors receive in this study, it will be argued that it can be seen as a positive facilitating factor as well, that promotes emergent grammaticalizing readings. Traugott (2010b) remarks that grammaticalizing elements sometimes congeal in their facilitating, but constrained contexts for grammaticalization, similar to idiomatization. Lexical persistence or conceptual image persistence refers to the idea that semantic features of the original lexical semantics persist in a gram-

3.4 Factors in grammaticalization

105

maticalizing construction instead of attenuating fully. Such persisting features can continue to influence grammaticalization developments, in terms of constraints or the direction of developmental paths. Lexical persistence can block delexicalization and hence subsequent grammaticalization in some SNs, e.g. flicker (see Chapter 5). As just noted, it may also show up in the semantic prosody that the various SNs develop. Lexical persistence can also impose restrictions on the felicitous use of quantifier SNs, for instance with regard to boundedness and spatial proximity of the N2 referents. For instance, particularly British English speakers will feel that an example such as A bunch of rivers, is acceptable only if the rivers are spatially contiguous and bounded within a certain area, as in a delta (p.c. Paul Thibault). Langacker (to appear: 12) describes the semantic profile of bunch as designating “a set of entities all of the same sort, spatially contiguous and physically attached to constitute a functional unit.” Nevertheless, against Thibault, he considers the spatial proximity constraint to be absent in the quantifier sense and gives the following examples: (3.6)

He has a bunch of {friends/hotels/investments} scattered about. (cited in Langacker to appear: 12)

Corpus data bear out restrictions, though there may be regional differences. With regard to layering and persistence it can be noted that “there is no requirement that the lexical source be lost after the grammaticalized option becomes available through reanalysis” (Campbell 2001: 28) (cf. Hopper’s 1991 layering). This goes against Langacker (2009 and to appear), who claims that the reanalysis of a lot of as a quantifier is incomplete as long as its original lexical sense persists, which seems an untenable position.

3.4.2.2 Subjectification and intersubjectification The precise nature of (inter)subjectivity and the relation between (inter)subjectification and grammaticalization have received much attention over the last few years, e.g. Traugott and Dasher (2002), De Smet and Verstraete (2006), Athanasiadou, Canakis & Cornillie (2006), Davidse, Vandelanotte & Cuyckens (2010) and Brems, Ghesquière & Van de Velde (forthcoming). The two main interpretations of subjectification in the literature are those of Traugott (1989), (1995a), (2003) and (2010b), on the one hand, and Langacker (1985), (1990), (1998) and (1999) on the other hand. Very generally speaking, subjectification pertains to how expressions come to index speaker-relatedness, in terms of expressing speaker attitude or viewpoint, and how this meaning comes to be the one conventionally associated with the expression in question.

106 Chapter 3: Grammaticalization, delexicalization and subjectification Traugott’s main interest lies with pragmatic (inter)subjectivity versus coded (inter)subjectified meaning in terms of the general shifts from propositional to subjectified meaning. By contrast, Langacker’s concept of subjectification in the NP is explicitly construal-related and refers to the development of ground-related meanings, i.e. identifiers and quantifiers. It involves a detailed account of how the speaker and hearer are construed in subjective versus objective expressions, both of which are speaker-related in his interpretation. The differences between Langacker’s and Traugott’s approach pertain to how the notion of speaker-relatedness is interpreted (cf. De Smet and Verstraete 2006). For Langacker (1985), subjectification refers to construing the speaker-subject less explicitly in an expression. For instance, The man next to me is James is less subjective than This man is James, because in the former expression the speaker is explicitly coded by me, whereas in the latter reference to the speaker is more implicitly part of the meaning of the demonstrative this and is not separately coded. Both objective and subjective expressions are speaker-related, but differ in whether this is formally coded or not. For Traugott only subjective expressions are truly speakerrelated (cf. De Smet and Verstraete 2006: 369). In her earlier work on the three semantico-pragmatic tendencies in grammaticalization Traugott (1988) and (1989) suggested that subjectification subsumes both developments towards textual meanings and towards attitudinal, evaluative meanings (see Section 3.2.2): Tendency I: Meanings based in the external described situation > meanings based in the internal (evaluative/perceptual/cognitive) described situation. Tendency II: Meanings based in the external or internal described situation > meanings based in the textual and metalinguistic situation. Tendency III: Meanings tend to become increasingly based in the speaker’s subjective belief state/attitude towards the proposition. (Traugott 1989: 3435)

In her later work this third tendency is more exclusively associated with the change towards more subjective meaning, or subjectification, which ties in with intuitive definitions of subjectification as the development of attitudinal or evaluative meanings expressive of a speaker’s opinion towards a specific propositional content (cf. attitudinal stance in Hunston and Thompson 2001). Thus, later explicit definitions of subjectification as the mechanisms by which meanings are recruited by the speaker to encode and regulate attitudes and beliefs. (Traugott 2010b: 35, based on Traugott 2003)

3.4 Factors in grammaticalization

107

seem restricted to attitudinal subjectification. Nevertheless, in concrete case studies subjectification also subsumes the development of the quantifier meaning of bit of for instance (see Traugott 2010b). This twofold approach in terms of textual and attitudinal subjectivity leads to some confusion, both theoretically and descriptively, because the notion of speaker-relatedness is rather different in both subtypes of subjectification. Underlying both textual and attitudinal subjectivity is the idea that the role of the speaker/subject in creating these types of meanings is foregrounded, which makes them both fundamentally “speaker-based” or “speaker-involved” (Traugott 1995a: 32). In organizing text, the speaker imposes structure on the propositional content or states of affairs, for instance by choosing specific connectives, and thereby links it up with the ground (Langacker 1991). Attitudinal subjectification is subjective in that it allows a speaker to express his/her opinion or attitude towards propositional content. Textual and attitudinal subjectification are not always easy to distinguish, as also noted by Traugott (2010b).60 Intersubjectification, then, is a change involving the "development of markers that encode the Speaker’s (or Writer’s) attention to the cognitive stances and social identities of the Addressee" (Traugott 2003: 124) and is dependent on previous subjectification. In Traugott's work this awareness has mainly become restricted to attention to the addressee’s social self, i.e. his/her attitudes and beliefs, as well as the development of metadiscursive functions such as turn-giving or the elicitation of response (Traugott forthcoming: 6). Examples of intersubjectification involve the development of politeness markers, expressions that perform face work such as hedgers (cf. Brown and Levinson 1987) and response-eliciting uses of surely (Traugott forthcoming). Recently, several authors have wondered how the emergence of determiner function fits in with regard to Traugottian (inter)subjectification, e.g. Ghesquière (2010), Narrog (2010) and contributions in Brems, Ghesquière and Van de Velde (forthcoming). This question ties in with the more general problem of delineating subjectivity from intersubjectivity, and the possibility of distinguishing subtypes of intersubjectivity, similar to the attitudinal versus textual distinction that exists for subjectivity, e.g. Brems & Ghesquière (2009), Ghesquière (2010), Davidse, Vandelanotte & Cuyckens (2010), Brems, Ghesquière & Van de Velde (forthcoming). Ghesquière (2010) argues that (post)determiners are intersubjective in nature rather than subjective because they involve the establishment of "joint attention" (cf. Diessel 2006), i.e. the speaker aids the hearer in tracking the identity of referents in the discourse, making sure speaker and hearer are on the same page. This interpretation of intersubjectivity extends the traditional range of hearer-related aspects a speaker can focus on from social self-image to cognitive efforts and information processing. In this way the emergence of

108 Chapter 3: Grammaticalization, delexicalization and subjectification some kinds of phorically textual meaning can be subsumed under intersubjectification rather than subjectification. Alternatively, Narrog (2010) proposes to distinguish a third type of semantic change in addition to subjectification and intersubjectification, namely orientation towards discourse or text. The directionality from less subjective meaning to subjective, intersubjective and beyond is hence currently being (re-)assessed. Traugott (2010b) also distinguishes sharply between (inter)subjective meaning which may be pragmatic in nature, and (inter)subjectified meaning. An expression is (inter)subjectified only if the (inter)subjective meaning is semanticized, i.e. coded as the conventional meaning of the expression. The idea is that (inter)subjective meaning first arises in the form of conversational implicatures through pragmatic enrichment and gradually this (inter)subjective polysemy becomes semanticized. In this view, intersubjective meaning often remains inferential, and intersubjectification is more rarely attested than subjectification. Examples are the development of Japanese honorifics and response-eliciting uses of surely (Traugott forthcoming). I will be somewhat less strict in applying the notion of intersubjectification, as the emergence of the intersubjective values in certain SNand TN-expressions seems to exceed purely pragmatic meaning, even though it may not be fully conventionalized. It appears that the intersubjective value of mitigators such as well and I think for instance in at least some cases does not rely merely on context, but seems largely conventionalized (see Vandelanotte 2009). In addition, I will also allow for textually intersubjective meaning in the sense of Ghesquière (2010), i.e. relating to the emergence of (phoric) determiner function. (Inter)subjectivity remain difficult to disentangle but in some cases need not be distinguished between since several constructions serve dual functions. In the grammaticalization of (S)SN- and TN-constructions both types of Traugottian subjectification are involved, as well as the development of (at least) pragmatic intersubjective values. With regard to SN-constructions, two paths of grammaticalization were proposed in Brems (2003a), namely towards pure quantifier meaning and towards more evaluative quantifier meaning. In the present study, these are argued to relate to textual and attitudinal subjectification respectively.61 The interpretation of textual subjectification here ties in with Langacker’s notion of (deictic) subjectification, as leading to the establishment of grounding elements, i.e. identifiers and quantifiers.62 In SN-constructions, primary textual or deictic subjectification involves the semanticization of quantifier meaning through repeated pragmatic inferencing of size or scalar implications that are part of the lexical semantics of the SNs. A load of wash, a bunch of grapes and a shred of skin for instance, in addition to profiling a specific constellation of entities designated by N2, also imply information about the typical size associated with these SNs. In specific contexts and constructions these size implica-

3.4 Factors in grammaticalization

109

tions can be foregrounded through pragmatic enrichment, to the detriment of the lexical meaning, which is backgrounded. Langacker (to appear) visualizes this shift in profile for a barrel of oil.

(a)

(b)

(c)

In (a) barrel is represented in its sense of a container (i.e. the rounded rectangle in bold) that holds a certain substance (i.e. the oval) with which it is roughly coextensive, represented by the equation mark. In (b) barrel is profiled as a unit of measurement and the profiled entity is no longer a physical object, but an abstract entity. It is still related to the physical object in that the unit of measurement refers to the amount of substance a barrel typically holds. The arrow indicates the scale of measurement that comes about through such a metonymic shift in overall profile. Diagram (c) illustrates the further gradual fading of the physical barrel as the basis for delimiting the scale. These diagrams can be said to visualize Traugott’s notion of pragmatic enrichment. Quantifier uses of SN-expressions, such as (3.7) to (3.9) are similar to diagram (c), with the scalar meaning foregrounded and the lexical meaning backgrounded to a large extent: (3.7)

The Sultans of Ping are pop stars who sell loads of records. (CWUKmags)

(3.8)

Their show has heaps of humour. (CW-OZnews)

(3.9)

There’s now a whole bunch of studies from different cities that show the same thing. (CW-NPR)

This formulation ultimately seems more accurate than saying that the lexical meaning is lost altogether (cf. Bybee, Perkins & Pagliuca 1994 on retention of older meanings), since it can be alluded to even in grammaticalized uses, as in lorry-loads of flattery where lorry functions as an intensi-

110 Chapter 3: Grammaticalization, delexicalization and subjectification fier, but at the same time harks back to the original semantics of loads (see Section 3.4.2.1). In the developments of TN-expressions, textual or deictic (inter)subjectification is involved in the emergence of quantifier uses and determiner-like function in complex determiner uses, both grounding predications. The quantifier use is illustrated by (3.10): (3.10)

[Said about lamps] “They come in all kinds of colours.” (CWTimes)

In (3.10) all kinds of is not a head noun use premodified by a quantifier, referring to many different kinds of colour, but a complex quantifier that in addition to ‘many colours’ stresses the variety in colours. To the extent that complex determiners involve the establishment of joint attention, they will also be argued to involve textually intersubjective values. Attitudinal subjectification in SN-constructions, then, is related to the emergence of valuing quantifying meaning, as illustrated by (3.11). In TNconstructions it is visible in qualifying modifier constructions for instance, as in (3.12): (3.11)

What we show is the truth, Luciano Benetton said. Traditional advertising pictures are a bunch of lies. (CW-Today)

(3.12)

Though overpriced, this slim volume contains great erudition and a fund of common sense; it is a kind of appendix to the history of the 1939-45 war by this eminent Sandhurst lecturer turned journalist-historian. (CW-Times)

In (3.11) the speaker is not primarily quantifying the lies but evaluating the advertising pictures in a very negative way. In (3.12) the speaker is not talking about different subtypes of the superordinate category ‘appendix’, but creatively compares a publication to the notion of an appendix as a subjective approximation. To the extent that this comparison helps the hearer to decode the identity of the referent, qualifying uses can be said to involve an intersubjective function. In addition, qualifying uses often involve hedging meaning, which is typically associated with attitudinally intersubjective meaning, i.e. focus on the hearer's self-image. In (3.13) and (3.14) the speaker is aware of the fact that the hearer might be intimidated or offended and hence uses a SSN (3.13) or TN (3.14) to hedge or mitigate the force of the assertion (cf. face work). (3.13)

I'm going to test you now Gwyneth so you we've got an examination coming up, now bit of an oral examination. (CW-UKspoken)

3.4 Factors in grammaticalization

(3.14)

111

“We are trying to progress as a team and he kind of let it go by the wayside,” Jordan said. (CW-Times)

In the case of valuing quantifier constructions both the SN, as the node, and the collocations are expressive of the speaker's evaluation and the construction is hence attitudinally subjectified, as reflected by specific semantic prosody patterns. In sum, (inter)subjectification, in their twofold interpretations, will be argued to accompany developmental paths of grammaticalization from head to modifier uses in (S)SN- and TN-constructions. The corpus studies in Part 2 will discuss in greater detail how these processes intersect for each meso-construction and the various constructional meanings they can realize. I will use the general distinction between non-subjective, subjective, subjectified and intersubjectified meaning as interpreted by Traugott, complemented with Brems & Ghesquière (2009) and Ghesquière (2010).

3.4.3 Decategorialization: A positive interpretation Heine (2003) sees decategorialization as the factor underlying most of Lehmann’s formal grammaticalization parameters, except erosion. In my view, the term decategorialization, like most of Lehmann’s parameters, suggests too much an attrition model of grammaticalization, and hence to a certain extent misconstrues it. When interpreted more positively and constructionally in terms of functional and formal pairings, the concept of decategorialization encompasses many of the semantic and formal characteristics generally observed for grammaticalization. Such a more positive interpretation of decategorialization would be similar in nature to the lossand-gain-model proposed for the semantic changes involved in grammaticalization. SNs on the one hand gradually lose characteristics of the category they belonged to, i.e. NOUN, and gain in the behavioural properties of the grammaticalizing construction they move towards, i.e. QUANTIFIER. Typical nominal properties are the possibility of being premodified by adjectives and/or determiners and quantifiers, as well as having both a singular and plural form. The loss-and-gain model of decategorialization that I argue for here ties in with the interpretation of reanalysis as involving analogy proposed in Section 3.4.1.3 and with Langacker (2009)'s constructional account of a lot of. Langacker (2009: 61) distinguishes a source-construction, i.e. a lot of, as a specific instance of the more schematic determiner + noun + preposition + noun-construction, which is also instantiated by “a woman from Brazil” for instance. The hypothesized end result or target63 is alotta, which is analogous with monomorphemic quantifiers like many and much. Gram-

112 Chapter 3: Grammaticalization, delexicalization and subjectification maticalization towards a (monomorphemic) quantifier would then require the reinforcement of similarities with the more schematic quantifierconstruction and loss in similarities with the source-construction. Langacker operationalizes this in terms of changes in the functionality of the indefinite determiner, noun and preposition of as the component elements of a lot of. However, according to Langacker, a lot of has not yet become one unanalyzable element, in that its component elements still retain some of their functionality. He considers the fact that lot can be premodified, as in a whole lot, and that it has a plural variant, i.e. lots of, as signs that it is still a noun, with some of the behaviour connected with this category. In addition, lot’s original lexical meaning of ‘group/collection of things (for sale)’ still exists, which makes him decide that at present a lot of is blocked from proceeding to full monomorphemic quantifier status. Langacker concedes that the determiner status of a has been attenuated and the indefinite determiner is virtually rendered “vacuous” with regard to its grounding function. In another paper he admits that the indefinite article in a lot of is “basically just a historical relic” (Langacker to appear: 14). Nevertheless, Langacker (2009: 75) maintains that a still has “discernible meaning consistent with a vestigial grounding function”. Against Langacker (2009), I will argue in Part 2 that premodification patterns in quantifier uses of (S)SNs are very restricted, namely to intensifying expressions such as whole, and do not go against decategorialization and grammaticalization. In a whole bunch of people for instance, whole further reinforces the quantifier interpretation, whereas foot-high in a foothigh pile of manure accentuates the independent head noun status of the SN (cf. Brems 2003a) (see Chapter 4). Considering the fact that all SNs, both as lexemes and in their singular/plural contrast, have grammaticalized to different extents, the general hypothesis is that none has acquired all of the properties prototypical of canonical quantifiers such as many and much. To different degrees, SN-expressions retain properties associated with nouns. This explains the current semi-stable layered system of argued for in Section 3.4.1.2. Rather than concluding that partial decategorialization detracts from grammaticalization claims, I argue that it may help to enrich targetparadigms. In the case of SN-quantifiers the possibility of intensifying submodification as in a whole bunch of studies contributes to the functionality of periphrastic SN-quantifiers within the quantifier paradigm compared to monomorphemic quantifiers such as many/much, which do not allow it as easily (cf. Brems & Davidse 2010b on emergent complex subordinators such as in the (vain) hope of). In keeping with the overall redistributional outcome of the semantic processes, the positive interpretation of decategorialization to be proposed can be called redistributional decategorialization, or decategorialization followed by recategorialization. Denison (2006: 279) also suggests a posi-

3.4 Factors in grammaticalization

113

tive interpretation of decategorialization in the context of gradience and category change in the English determiner system. Starting from a prototype view of grammatical categories, he argues for gradience within and between categories. Decategorialization can then be interpreted positively in terms of an item gradually acquiring properties more typical of another category, while losing features typical of its previous category. This notion of positive decategorialization also gives further support to the earlier argument that analogy is involved more essentially in reanalysis than commonly assumed.

3.4.4 Paradigmaticization: A dynamic interpretation Paradigmaticization refers to the fact that grammaticalizing items become part of a more restricted and internally cohesive paradigm (Lehmann 1985: 309), which is typically of the closed-class type64 (cf. Heine 2003: 588). Bisang and Wiemer (2004) among others have doubted its necessity or validity. I do include it as one of the diagnostics or concomitants of the broader concept of decategorialization, i.e. the loss of properties of the grammaticalizing item’s former grammatical class and the gain in properties of the category it develops towards. A dynamic interpretation of paradigmaticization, which interprets categories as prototype-based, allows us to compare different SN-expressions with each other as well as with more established members of the quantifier system and, possibly, account for these differences in terms of entrenchment (cf. Langacker 2009 and to appear) and paradigmatic enrichment. In recent work Diewald (2010) has emphasized the paradigmatic phase of grammaticalization, i.e. how do new members of a category settle into their target-paradigm as a subparadigm, and interact with previous members?

3.4.5 Syntactic extension A grammaticalizing construction may extend beyond the syntactic context in which it originally developed. Such extension or generalization of the grammaticalized construction to increasingly more syntactic contexts is a common concomitant of grammaticalization processes (e.g. Heine 2003, Hopper and Traugott 2003). Himmelmann (2004) refers to it as syntactic context expansion. He gives the example of emerging articles which extend from core argument positions to adpositional expressions and other syntactic environments they could not occur in as demonstratives. Syntactic context extension, I would argue, is typically a characteristic of a high level of grammaticalization, in the sense that it presupposes re-

114 Chapter 3: Grammaticalization, delexicalization and subjectification analysis, as defined in Section 3.4.1.3, to have taken place in order for the reanalyzed pattern to be extended to other syntactic contexts. Both SN- and TN-expressions have extended towards new syntactic domains. For instance with hedging TN-uses there is a clear extension from nominal (3.15) to adverbial (3.16) and sentential qualifiers (3.17), and to discourse marker uses (3.18) (cf. Denison 2002): (3.15)

Well I don’t like these er…what’s…. for want f a better word these sort of organized paths like the Pennine way. (CWUKspoken)

(3.16)

He kinda sorta maybe loves him. (www.alykat.hispeed.com /unfrozen/fanfic/mooks)

(3.17)

CMG: Sort of like In The Fishtank? Beam: Kinda. Definitely like that but not with all the parameters. Not like you have to go and perform it all in one day. (www.cokemachineglow.com/feature/interview/beam)

(3.18)

As I remember it used to be sort of like fairly common on a Tuesday, that I’d pretend to be sick. (cited in Denison 2002: 4)

In these uses the hedging meaning has extended to new syntactic environments: in (3.15) the nominal type specification these organized paths is hedged, whereas in (3.16) a verbal, processual type specification is hedged which gives the TN-expression an adverbial function. In example (3.17), then, the TN-expression has scope over the entire preceding sentence and in (3.18) over even larger stretches of discourse. As noted in Section 3.2.2, Traugott (1995b), against Lehmann (1985), also argues that the development towards a discourse marker use typically involves an increase in scope. With some SNs, there is extension towards degree modifier uses with adjectives, verbs and adverbs (3.19-3.22), as well as to an independent answer use for bit (3.23) for instance: (3.19)

I'm heaps better than I was, although I have headaches 24 hours a day which drains me a lot. (CW-OZnews)

(3.20)

“Perhaps you could suggest you're too old,” he offered sullenly. “Thanks a bunch!” (CW-UKbooks)

(3.21)

She complains a lot. (cited by Langacker 2009: 66)

3.4 Factors in grammaticalization

115

(3.22)

Just as every seeker who comes to me for a prophecy believes what he wants to believe and not one jot else. (CW-UKbooks)

(3.23)

Q: Did you like the movie? A: A bit. (cited in Traugott 2005: 8)

Langacker (2009: 66) describes adverbial uses of a lot of as either quantifying the intensity or frequency of a process. In the case of the SNexpressions focused on in this study, mainly intensifying adverbial uses seem to be involved, except for loads, which allows frequency readings, as in the following example (3.24)

I see her loads 'cos we've got a good view down the road. (CWUKspoken)

It may be that in such examples loads has to be interpreted as elliptical for the frequent collocation loads of times.

3.4.6 Coalescence and phonetic erosion Coalescence is connected with chunking and routinization, typical of grammaticalizing constructions in slightly later stages. It is also connected with decategorialization. When component elements lose in individual functionality and decategorialize, their substance may erode, since they no longer contribute individual meaning to the periphrastic structure which now expresses the new grammatical meaning. This ties in with Lehmann’s (1985: 308) point that autosemantic signs become synsemantic, i.e. the formerly independent meanings of the component elements are absorbed in the more grammatical meaning of the coalesced expression as such. In the case of TN-expressions this is often reflected in spellings such as kinda/sorta/typa and in a pronunciation in which of is weakened to schwa. SN-expressions less commonly appear in such coalesced spellings, but in informal spoken parts of the Collins Wordbanks corpus as well as on the Internet spellings such as a lotta/alotta and buncha can be found: (3.25)

Alotta people dont know about this band that billie, jason and mike all participate in... (www.absolutepunk.net/printthread. php?t=137604)

(3.26)

We got a lotta catchin' up to do. (CW-USbooks).

116 Chapter 3: Grammaticalization, delexicalization and subjectification (3.27)

I don't know where I got the feelin', but I think there's more to it than this shit, a buncha people runnin' around thinkin' they got it made. (CW-USbooks)

In the Collins Wordbanks corpus alotta appears only twice and a lotta appears 74 times, mostly in American English. Buncha occurs 8 times, almost exclusively in American English. However, as noted by Joseph (2007) and De Smedt, Brems and Davidse (2007), coalesced spelling as such is no proof of grammaticalized status, but typical of any routinization process. It was also mentioned earlier that Langacker (2009 and to appear) considers arriving at a monomorphemic status, reflected graphologically, important in claiming a reanalyzed quantifier status for a lot of. However, several grammaticalized constructions do not display it and are generally accepted as being grammaticalized, e.g. the to infinitive. The split to infinitive, whether fully accepted now or not, might be argued to be possible precisely because of the highly grammaticalized status of such expressions, as in the famous Star Trek motto To boldly go where no one has gone before.65 Functional unity hence does not have to be reflected graphologically, but the two often go together. The premodification patterns referred to in Section 3.4.3, such as a whole bunch of people, will not be seen as intruding on the functional unity or coalescence of grammaticalizing SN-expressions.

3.4.7 Frequency Frequency and the many guises it can take on in grammaticalization processes have been referred to by means of several concepts, e.g. routinization, habituation, automatization, conventionalization of pragmatic inferences, and context generalization. High frequency and lack of lexical specificity are generally considered as either important indicators or even conditions for delexicalization and grammaticalization, especially in usage-based models of grammar. Increase in frequency has also been considered “an active force in instigating the changes that occur in grammaticization” (Bybee 2003b: 602), as well as a consequence of grammaticalization (e.g. Heine et al. 1991, Bybee 2003a and 2003b, Hoffmann 200466 and Himmelmann 2004). Already this provides frequency with an ambivalent dual role as both indicator and facilitator. The problematic role of frequency becomes even clearer when we are dealing with the grammaticalization processes of infrequent and rare lexical items, as discussed in Hoffmann (2004) and Brems (2007) with regard to low-frequency complex prepositions and infrequent small size nouns respectively. The main question in this context is how these infrequent items

3.4 Factors in grammaticalization

117

can acquire grammaticalized status if they cannot have engaged in pragmatic inferencing, reanalysis, attrition and routinization, so typical of gradual grammaticalization (cf. Bybee 2003a and Haiman 1994). Other explanations need to be sought and the exact role of frequency explored (see Chapter 5 on SSNs). In order to assess more accurately the importance of frequency for the entrenchment or saliency of certain lexical items and its influence on grammaticalization, Hoffmann (2004) for instance introduces a distinction between (absolute and relative) lexical and conceptual frequency, which is similar to Bybee’s (2003b) distinction between type and token frequency. In addition, he argues for grammaticalization by analogy with more frequent and structurally similar complex prepositions, a proposal also considered in Brems (2007), but accepted only with major adjustments (see Chapter 5). Bybee (2003b: 603) discusses frequency mainly in terms of repetition and suggests a new definition of grammaticalization, incorporating the twofold role of frequency, characterizing grammaticalization as “the process by which a frequently used sequence of words or morphemes becomes automated as a single processing unit.” Repetition, based on Haiman’s (1994) ideas on ritualization, then becomes the process directly underlying several key factors in grammaticalization, such as rebracketing, coalescence, semantic and phonological erosion processes and context extension. Habituation, through high frequency, seems to lie at the basis of other processes in that it makes a unit lose in semantic force and specificity and subsequently this semantic generalization leads to extension of pragmatic and syntactic contexts of use. Bybee (2003b: 605) especially considers high token frequency to be a trigger of many of the changes associated with grammaticalization, since it affects the nature of cognitive representations in important ways. Without wanting to deny the intricate interaction between grammaticalization and frequency which can certainly be attested, I think Bybee (2003b) overstates the active influence of frequency as a truly causal factor in grammaticalization. I will treat it more generally as an interesting factor in grammaticalization processes.

3.4.8 Cyclical processes of renewal It is well known that expressive functionality wears off through frequent use and routinization. For instance, intensifying expressions such as very, awfully, really and pretty are susceptible to wear and tear and fashion (cf. Lorenz 2002). Therefore, they often give rise to onomasiological processes of renewal, i.e. “the process whereby existing meanings may take on new forms.” (Hopper and Traugott 2003: 122, cf. Meillet 1915/16 [1958]). Just like intensifiers, the hyperbolic expressivity of SN-expressions is affected

118 Chapter 3: Grammaticalization, delexicalization and subjectification by fashion, in that different types of expressions were in vogue in different decades. In the 1960s for instance things typically came in barrels, e.g. barrels of fun, or in bags(ful) (p.c. M.A.K. Halliday and Eirian Davies), expressions that now seem superseded by other expressions. The cyclicity of such renewal processes and renewal itself have often been considered as epiphenomena of grammaticalization processes (e.g. Hopper and Traugott 2003 and Heine 2003). These renewal processes themselves then take the shape of grammaticalization processes, recruiting new, (more) lexical material to replace older forms, or to exist alongside these. However, renewal processes are not exclusive to grammaticalization processes, nor is the cyclicity attributed to them. Hence, the observed cyclicity in the development of SN-expressions such as heaps of and piles of can only give indirect support to a grammaticalization claim, since it is neither a sufficient nor a necessary condition for it. The potential for renewal of SNs also relies on the fact that the SN-construction as a mesoconstruction has become established and may coerce new SNs into quantifier meaning.

3.4.9 Lexicalization and grammaticalization Articles such as Wischer (2000) and Himmelmann (2004), and monographs such as Brinton and Traugott (2005) address the interrelationship between grammaticalization and another type of linguistic change, namely lexicalization. Lexicalization in its most general sense refers to the process that leads to new conventionalized lexical items or phrases.67 However, even more so than grammaticalization, it has been used to encompass conceptually rather disparate phenomena, such as idiomatization, splits, blends, syntactic conversion and the onomasiological encoding of semantic features in lexical items (see Wischer 2000 and Himmelmann 2004 for a more extensive survey and discussion). Grammaticalization and lexicalization are often presented as two extremes which are in radical opposition to each other, but in actual practice they are often mistaken for another. Part of this confusion arises from the various ways in which the lexicon and grammar have been defined.68 Rather than constituting two opposites, Himmelmann (2004) for instance observes that grammaticalization and lexicalization have several mechanisms and formal reflexes in common. Since grammaticalization and lexicalization are both processes of conventionalization, they share reflexes common to it, such as erosion or fusion effects (Himmelmann 2004: 3538). Himmelmann (2004) restricts lexicalization to the “univerbation of a frequently recurring collocation of two or more lexical items”, also called idiomatization for longer phrasal units. Ultimately, he sees its non-

3.4 Factors in grammaticalization

119

directionality and abrupt host class reduction as the main differences with grammaticalization, which is directional and gradual. In lexicalization one frequent collocation from the total set of collocations a node can have acquires special status and is interpreted as constituting one single processing unit, rather than a compositional expression. In Trousdale’s construction grammar (2008c) terms, lexicalization leads to idiosyncratic microconstructions, whereas grammaticalization leads to more schematic constructions at meso- or macro-level. Himmelmann (2004: 22) seems to allow for the fact that both grammaticalization and lexicalization can be involved in one and the same change. Wischer (2000: 355) makes this question a central focus, when she looks at the “idiomaticization of syntagms which take on a modal or discourse function”, such as methinks. She decides that the two processes are not contradictory, but “operate on different levels of the language”. As I will show later on, the grammaticalization of SN- and TNexpressions involves lexicalization in the sense that certain SNs and TNs come to participate in semi-lexicalized strings which are listed in dictionaries, and can be argued to be stored in the mental lexicon as a whole entity. An example is the quantifier TN-string all sorts of, which is listed as a separate phrase in the entry of sort in the Collins Cobuild English Dictionary for Advanced Learners. Importantly, I do not see lexicalization here as displaying an increase in lexical meaning with regard to the compositional starting point, as Wischer (2000) does. The shift of all sorts of from referring to ‘all the different types of something’ to a quantifier meaning ‘many’ can hardly be called a shift towards more lexical meaning. On the contrary, it displays deictic subjectification and hence a shift to more grammatical meaning, which is typical of grammaticalization. The lexicalization-like effects referred to in this study pertain primarily to the idiomatization and fixation of frequently occurring collocations within grammaticalized functions.

3.5 Conclusion: Grammaticalization as functional and formal approximation of a target construction The grammaticalization of SN- and TN-constructions centres on the mechanism of reanalysis, from head to modifier uses, in which the analogic pull of models plays an important role (cf. Fischer 2007). In the proposed constructionist approach, reanalysis works both on the semantic and formal level, which interlock, as do their reflexes in the specific grammaticalization processes of SN- and TN-expressions. Some factors were interpreted as parameters, such as paradigmaticization, syntactic extension and coalescence, whereas delexicalization and decategorialization for instance were

120 Chapter 3: Grammaticalization, delexicalization and subjectification hypothesized to have a potentially more active role in addition to diagnostic validity. In such an approach, a loss-and-gain understanding of decategorialization, interpreted functionally as well as formally, can be said to underlie many of the other grammaticalization factors. The grammaticalization of SN- and TN-expressions (overall) leads to new members of grammatical categories that have already been established in the language, e.g. to new quantifiers, determiners and hedges. Such processes of grammaticalization can be interpreted more explicitly in terms of CxG. The grammaticalization path is then hypothesized to constitute a trajectory from a construction in which the item has its full lexical value to a construction in which it contributes to the expression of more grammatical meaning. Incorporating schematicity in the vein of Traugott (2006) and Langacker (2009 and to appear) into this, grammaticalization can be interpreted as the functional and formal approximation of a source-construction to a target-construction through increasing functional analogization and recategorialization. This target-construction is at a higher level of schematicity than the concrete source-construction, namely at the macro or meso-constructional level, for example many and much as targets for SNexpressions such as heaps of. The schematicity of these models or targets pertains to their degree of decategorialization and delexicalization, as well as to the fact that the properties of such concrete quantifiers can be generalized over in a schematic quantifier-construction. The source-construction and end result of a grammaticalization process are substantive constructs (as per Traugott 2006 and Fried and Östman 2004). In some grammaticalization processes multiple models can come into play. In keeping with a non-deterministic interpretation of grammaticalization, such a functional approximation need not go on to so-called completion (see e.g. Fischer 1999). A question that needs answering in this context is when approximation of the target is sufficient to talk about grammatical reanalysis and what is completion in grammaticalization? Should the various SN-quantifiers over time become monomorphemic lexemes in order to be accepted as true quantifiers (cf. Langacker 2009 and to appear)? Do SN-quantifiers need to extend beyond their present microconstructional collocational profiles in order for delexicalization and syntactic context expansion to be sufficient? De Smet & Ghesquière (2010) discuss what they call hesitant grammaticalization in verb particle constructions and the development of intensifying adjectives such as mere and utter from lexical sources. Because both of these construction types at present show collocational constraints, e.g. will you hear me out versus *will you talk me out, and have not extended to increasingly more contexts beyond their source-construction, they are interpreted as instances where grammaticalization is hesitating. However, De Smet & Ghesquière (2010) do not pay sufficient attention to the fact that the restricted contexts both construc-

3.5 Conclusion

121

tions associate with at present are also the contexts that enabled the enrichment with new grammatical inferences in the first place. Collocational restrictions are hence not just hindrances which at present block the constructions from going on to full grammaticalization. In my view it seems only natural that once a grammaticalized function has become established it will not always be easy to shake off the construction and collocates that gave rise to that grammatical meaning in the first place. Statements like "when grammaticalization hesitates" (De Smet & Ghesquière 2010) furthermore run the risk of snatching grammaticalization from its rightful owners, i.e. language users (cf. Joseph 1992). It is they who, given the appropriate syntactic, semantico-pragmatic and collocational context enrich statements beyond their purely propositional or lexical content. Ultimately, neither the reanalysis/analogy concept, nor the label grammaticalization in itself explains why the changes making up grammaticalization occur in the first place. The motivations behind the grammaticalization paths attested for SN- and TN-constructions, and probably for most grammaticalizing constructions, have to with expressivity and the possibility of language users to analogize, which may lead to alternative analyses or interpretations of stretches of language. Interacting with this is the fact that it is their perceived similarities with canonical quantifiers that makes it possible for SN-expressions (with their scalar implicatures) to be recruited as new means of quantification. Very often existing forms are reused for this purpose (Hopper and Traugott 2003: 73), and it is typically the more substantive extant structures that are put to new purposes, since they are more accessible to the hearer. Human language does not consist of static one-to-one correspondences between form and meaning, but is essentially open to constant meaning negotiation and re-negotiation in view of its function as a means of communication, and considering the fact that in language, for reasons of parsimony and memory restrictions, “a small set of units and constructions must serve a much larger set of functions” (Hopper and Traugott 2003: 42). Language is hence characterized by very dynamic systematics (cf. Hopper 1987 and 1991). Expressivity allows for increased informativeness on the part of the speaker, abductively inferenced by the hearer.69 SN-expressions allow adding information on speaker attitude, mostly in the form of hyperbolic quantifier information. The importance of indirectness, in the sense of euphemisms and understatement, and extravagant expressivity in humour, irony, mix-up of styles, hyperbolism and enhanced emotion, although often referred to in sociofunctional accounts of language change, have received little explicit attention with regard to their actual influence on reanalysis. I argue that they, as facilitating pragmatic factors, are enabling factors in grammaticalization processes and continue into secondary subjectification and intersubjectification processes. The constant interaction or competition between clarity,

122 Chapter 3: Grammaticalization, delexicalization and subjectification extravagance and routinization of forms keeps grammaticalization and renewal processes going, especially in the semantic domain of intensification (cf. Lorenz 2002). For SN- and TN-expressions, changes in collocational patterning and collocational reclustering are important syntagmatic relations that mark the reanalysis, and have received too little explicit attention in actual grammaticalization studies. Semantic prosody of grammaticalizing constructions has tended to be seen as an inhibiting factor in grammaticalization processes, in that it indicates the incomplete status of the delexicalization and grammaticalization processes. It has been associated with marking a restricted environment of grammaticalization, whereas grammaticalization is fundamentally associated with extension and generalization (cf. Himmelmann 2004 and De Smet & Ghesquière 2010). However, with regard to SN-constructions in particular, the hypothesis will be put forward that it is a positive and facilitating factor of grammaticalization, which promotes emergent alternative readings in and hence encourages actual reanalysis. The corpus studies in Part 2 will discuss all of these notions in greater detail for SN- (Chapter 4), SSN- (Chapter 5) and TN-constructions (Chapter 6).

Part II Synchronic and diachronic corpus studies Part 1 of this study set out the main grammaticalization and subjectification argument put forward in this thesis with regard to SNs and TNs moving from head to modifier status. The corpus studies in this second part will propose a detailed characterization of the various modifier uses of SNs and TNs, based on qualitative and quantitative corpus analyses of primarily synchronic data, complemented with diachronic case studies. In the case of SN-constructions synchronic case studies of English SNs will also be complemented with a comparative case study of English and Dutch SNs. For TN-constructions my synchronic description is based on the corpus study by De Smedt, Brems and Davidse (2007); the diachronic study draws in observations from Brems and Davidse (2010a).70 The synchronic data for the English case studies were extracted from the Collins Wordbanks corpus, in addition to COLT for the TNs. Internet data will be included in the case studies for illustrative purposes. The Dutch data were extracted from the 38 million word corpus of the INL or Instituut voor Nederlandse Lexicologie (‘Institute for Dutch Lexicology’). The diachronic data derive from various corpora: The Penn-Helsinki Parsed Corpus of Middle English (PPCME) and Early Modern English (PPCEME), the Innsbruck Middle English Prose Corpus Sampler (IMEPCS), the Corpus of English Dialogues (CED), the Corpus of Early Modern English texts (CEMET) and Corpus of Late Modern English Texts, extended version (CLMETEV), and the Old bailey Corpus (OBC). In addition the OED and the Middle English Dictionary (MED) were used. Corpora predating the Middle English period are generally not included in this study, since the NP + of + NP-syntagm, which is the main focus of this study, only appears from Middle English onwards (cf. Denison 2002). On the basis of close study of these synchronic, diachronic and comparative data I will argue that the functional categories distinguished in the synchronic classifications of the various attested SN- and TN-uses can be interpreted as steps in diachronic processes of delexicalization, grammaticalization and (inter)subjectification. See the reference section on corpora for more information on the nature of the corpora uses in this book. This book subsumes three case studies of specific binominal structures, i.e. of SNs, SSNs and TNs. The SN-expressions will be further subdivided into regular SNs and small SNs, similar to canonical quantifiers, e.g. many/much versus few/little. For the regular ones, data for bunch(es), heap(s), pile(s), load(s) and lot(s) have been analyzed, as well as Dutch stelletje and zooitje (based on Brems and Kimps 2005). The case studies for the SSNs look at bit, jot, flicker, scintilla, scrap, shred, skerrick, smidgen

124 Part II: Synchronic and diachronic corpus studies and whiff. The largely semantic distinction between regular SNs and the set of SSNs correlates with a number of variables the influence of which seemed interesting to study in the context of grammaticalization and synchronic variation. Firstly, whereas the regular SNs looked at in this study include many common items, the set of SSN-expressions, except bit of, consists of infrequent and rare ones, e.g. jot, scintilla and skerrick. Secondly, if these two sets display differences in pragmatic and collocational features, these differences can be assumed to be maximally manifest in these two contrastive sets, especially since the SSNs seem to be polarity sensitive (cf. Israel 2004). The Dutch SNs, then, with their diminutive forms and the pragmatico-semantic values associated with these, may further clarify the impact of certain pragmatic factors on the grammaticalization and subjectification of SNs. For the case study of TNs, the three most common expressions were chosen, which internally contain the opposition between sort/kind versus type, as also noted by Denison (2002). It will be argued that SN-constructions, SSN-constructions and TNconstructions all constitute layered semi-stable systems of lexical, grammaticalized and ambivalent uses, incorporating different degrees of delexicalization, decategorialization and grammaticalization. Within a CxGframework, the constructional semantics of SN-constructions and TNconstructions can be compared among each other as well as with regard to their more or less schematic models, such as Quantifier-constructions and Hedge-constructions. Ultimately, the case studies will first and foremost reveal the very intricate and often inextricable interaction between lexical and grammatical meaning in SN- and TN-expressions, and how this ties in with the grammaticalization and subjectification of SNs as emergent means of quantification with greater expressivity than such quantifiers as much/many, and the various (inter)subjective motivations for TNexpressions.

Chapter 4 SN-constructions 4.1 Introduction to SNs and issues overarching SNs and SSNs SN-expressions, just like canonical quantifiers, can be scaled up into subsets. I have made the main distinction between regular SNs versus small SNs (SSNs), similar to many/much versus few/little. Each subset introduces its own problematic issues and displays characteristics specific to it, but they also share processes and mechanisms on higher levels of mesoconstructional schematicity. I will argue that the features in the grammaticalization of both subsets have theoretical significance for grammaticalization research as such. For instance, with respect to the specific operationalization of delexicalization in terms of collocational extension/reclustering, the role of semantic (or emotive) prosody, lexical persistence, idiomatization and renewal. As defined earlier, semantic prosody refers to the fact that the collocates of a word may be characterized by a specific emotive colouring, e.g. utter has a predominantly negative semantic prosody as it typically combines with negative nouns: utter idiot/fool/bastard/chaos. The assessment of semantic prosody of SN-expressions is always context-dependent and takes into account the emotive colouring of verbs and other items in the specific context, in addition to the noun following SN + of. Furthermore, notions such as constructional analogy and frequency issues are addressed within the overarching framework of grammaticalization as generally motivated by the expressive needs of language users (cf. Hopper and Traugott 2003). It will be argued that SNs and SSNs, in addition to the similarities they display, each use different systems for expressing these basic needs. On the basis of qualitative and quantified synchronic and diachronic corpus analysis, two grammatical uses will be identified for regular SNs specifically, i.e. a quantifier use (heaps of people walked in) and a more evaluative (quantifier) use, referred to as valuing (quantifier) use in this study (that's a load of crap!). The term valuing quantifier refers to its evaluating meaning as well as to the values of the N2 it typically combines with, as in bunch of morons. On the basis of semantico-pragmatic, syntactic and collocational evidence head and modifier uses of SNs will be argued to constitute separate constructions. Section 4.2 will introduce the data set and methodology uses in the synchronic corpus study of SNs. In Section 4.3, I will describe SN-expressions that have developed quantifier uses from an original lexical head noun use, i.e. heap(s), pile(s) lot(s) and load(s), as involving (deictic) subjectification, and I will compare these expressions in terms of their synchronic degree of

126Chapter 4: SN-constructions grammaticality. In Section 4.4, the valuing (quantifier) construction, involving deictic and attitudinal subjectification, will be explored by means of bunch of and load of. Section 4.7 presents a diachronic corpus study of the SNs heap(s) and lot(s). In this study the diachronic dimensions of the synchronically co-existing SN-uses will be discussed, focussing on head, partitive and quantifier uses of SNs. This section will also address such issues as the confusion between synchronic degrees of grammaticality and diachronic steps in grammaticalization processes (cf. Traugott and Trousdale 2010). As cross-linguistic evidence for the grammaticalization of SNs into (valuing) quantifiers, Dutch data will be brought in of stelletje and zoo(i)jte as functional equivalents of bunch in Section 4.8. Even though the Dutch SN-expressions display differences in (the degree of) their specific delexicalization and grammaticalization processes and layering effects, I will argue that they have followed a grammaticalization-cum-subjectification path that is ultimately the same as that of bunch and driven by similar motivations. The influence of the diminutive –je seems to be very important in this.

4.2 Description of the synchronic data set, selection of the SNs and method of analysis Data were extracted from the Collins Wordbanks corpus (CW). There is a predominance of British English subcorpora and hence of the amount of British English data in this corpus.71 SNs were selected that were sufficiently frequent to guarantee comprehensiveness. For each SN the query consisted of SN + of, which allowed for all kinds of premodification of the SN and the noun following of. Singular and plural variants of each SN were extracted as separate expressions. By not specifying NOUN after of, I allowed for non-nominal elements. No restrictions were made in terms of register (formal/informal), medium (spoken/written), or region (British, Australian, American English). All of CW’s subcorpora are hence included and indicated in abbreviated form following the cited examples. Even though this study is not intended as a regional or stylistic study as such, significant tendencies with regard to such variables will be discussed when necessary. Statements on regional preferences are based on the assumption that texts deriving from a British, American or Australian English subcorpus are mostly written in those specific variants. Table 4.1 lists the total number of occurrences of the SNs looked at throughout the corpus, both within the SN-construction and irrespective of it. The first row of numbers indicates the token frequency of the SN in a

4.3 Head and quantifier construction

127

pertinent nominal use throughout the corpus, via the query SN/NOUN, which extracts all noun uses of the SN. The second row shows normalized frequencies per million words. The next row indicates the total number of tokens of an SN within the SN-construction, searched via SN/NOUN+of and the final row again displays these frequencies per million words. Wrongly tagged or irrelevant uses were filtered out manually. Clear references to piles in the sense of haemorrhoids for instance were taken out of the extraction. Wrongly tagged verbal uses of heaps, loads, piles and bunches were taken out of the data set, as well as extractions that included fixed expressions such as Cup/get/grab a load of this or British English slang bunch of fives, referring to a hand or fist. Examples with The Brady Bunch and The Wild Bunch referring to the television series, film or gang name were equally filtered out of the data set. Their frequent occurrence could skew the quantified results of the corpus study. Table 4.1. Token frequencies of ten SNs bunch bunches heap heaps pile piles

lot

lots

load

loads

overall

1240

77

353 149

824

239

40118 5125

1165

1152

per 1 million words

21.6

1.3

6.1

2.6

14.4

4.2

698.7

89.3

20.3

20.1

within SN-con- 816 struction

49

106

90

364

165

24734 4353

594

760

per 1 million words

0.9

1.8

1.6

6.3

2.9

430.8

10.3

13.2

14.2

75.8

For each SN-expression the data set used for the analysis is either exhaustive (i.e. bunches of, heap of, heaps of and piles of), or constitutes a random sample of 250 analyzable examples (i.e. bunch of, pile of, lot of, lots of, load of and loads of). Extractions of the SNs as such, without of, were also analyzed to explore extended uses of these expressions outside of the NP-environment and in other syntactic contexts. However, the results of these analyses are only used for illustrative purposes and are not included in the quantified

128Chapter 4: SN-constructions study. The analysis of SNs in this study is primarily concerned with the (binominal) source-construction and the target-construction, which makes it possible to study collocational changes very closely as SNs approximate a more schematic quantifier model. The synchronic corpus data for each of the SNs + of only showed syntagms of the form SN + of + N2, with no non-nominal elements following of. The data were categorized as head, quantifier, valuing (quantifier) and ambivalent uses. For each of the SNs, N2-collocates were categorized semantically, as either concrete or abstract nouns and generalizations were made over semantic subsets of collocates within those more general sets. I use the semantic terms concrete and abstract in the traditional sense of Quirk et al. (1985: 247), but also Langacker (1991: 27–30). Referents of concrete nouns are typically instantial and can be registered by the senses in the here and now, whereas abstract nouns describe concepts and pertain to "type space" instead of physical space (Langacker 1991: 64). Naturally, the categories of concrete and abstract noun have fuzzy boundaries, and assignment to either is context-dependent. An important separate semantic category is that of animate nouns (typically referring to human referents, and more rarely to animals). This set will mostly be discussed separately, even though it is a subset of the larger set of concrete nouns. Animate concrete nouns are allowed by some uses and disallowed by others. As we will see, animate nouns, like abstract nouns, signal a clear move away from lexical head status of the SN. Syntactically, collocates of SN-expressions were characterized in terms of countability, as either (plural) count or uncount noun depending on the specific context. Since collocational changes are argued to be of central importance in distinguishing SN-uses as separate constructions, such detailed collocational profiles of the various SNs are essential to the corpus analysis. For each SN-expression a quantified collocational profile will be presented. In classifying SNs as head or non-head and ambivalent, two main types of interlocking evidence were used, i.e. semantic and syntactic. The semantic criteria pertain to the degree of lexicality of the SN, as well as to the degree of coextensiveness between the SN and the element following of. All SNs have an original lexical meaning that typically designates a specific constellation, container or shape. The more literal and lexically specific in meaning the SN is, the more likely it is that it will constitute the head noun of the SN-construction. It then is coextensive with the element in the of-phrase, i.e. they both have the same referent in that the postmodifier merely specifies what the SN refers to. For instance, in the case of a bunch of grapes, grapes specifies what the bunch consists of and they both refer to one and the same thing in reality. In other words, the SN in its head noun use has appositive meaning or shows coextensiveness between both nominals (cf. Langacker 2009 and to appear). The degree of lexicality and

4.3 Head and quantifier construction

129

coextensiveness of course is reflected by the semantic generality of the N2collocates; literal and fully lexical SN-uses generally predict a restricted set of collocates in the of-phrase, namely those that typically appear in the constellation or shape designated by the SN or can be profiled as such, e.g. bunch of grapes, bananas, keys. Hence, as the collocational range of an SN increases and more collocates are allowed to team up with it, the probability of a quantifier analysis likewise increases. Syntactic criteria pertain to number concord between the SN or N2 and finite verbs and phoric pronouns and determiners, but this is rather difficult to check systematically. In keeping with the idea of grammatical concord, the hypothesis is that the element that determines the number of the verb or the pronominal element is the head noun of the binominal construction. However, subject-verb concord has limited applicability, since it only applies when the binominal SN-construction has subject function, and can only be empirically tested whenever the SN and N2 differ in number. In the past tense only be marks singular and plural number, i.e. was versus were. These restrictions prevent systematic syntactic corroboration of the status of the SN within the SN-construction. Furthermore, as noted in the reference works reviewed in Section 1.2, verb concord can be motivated by factors other than grammatical number, e.g. proximity concord and notional concord. Moreover, fluctuating verb agreement can be indicative of vacillating grammatical and semantic behaviour of the SNs as such. Finally, even though corpus data derive mostly from native English sources, they can contain mistakes with regard to verb concord. Nevertheless, if syntactic criteria like number concord converge with semantic indications about SNstatus, they are taken into account and often indicated in examples by underlining of verbs or pronouns. In addition to concord patterns, substitution of the SN-expression by a (monomorphemic) quantifier such as many/much is a test both for formal constituency and functional unity. Another formal test is the possibility to prepose the of-phrase. Finally, besides grammatical concord and substitutability, selection restrictions on the choice of verb or other elements in the co(n)text can be a marker of head versus quantifier meaning. Such markers too will often be underlined in the examples cited. In the case of head or ambivalent status only the SN itself will be put in bold, whereas in the case of quantifier status of will also be put in bold, reflecting their functional unity. The (indefinite) determiner will not be put in bold in order to indicate the partially decategorialized status of the SN in that respect. Examples (4.1) and (4.3), on the one hand, and (4.2) and (4.4), on the other hand, show head and quantifier status respectively, confirmed by (indirect) verb concord:

130Chapter 4: SN-constructions (4.1)

The fox, unable to reach a bunch of grapes that hangs too high, decides that they were sour anyway. (CW-USbooks)

(4.2)

Although I should add that today a whole bunch of top Clinton economic advisers flew in to Chicago. (CW-NPR)

(4.3)

Six plane loads of food are also being flown today to the city of Mogadishu. (CW-NPR)

(4.4)

Both children and adults have been seriously attacked and I think there’s loads of evidence that something fairly drastic should be done. (CW-UKspoken)

Both semantic and syntactic assessment criteria yield two obvious categories, namely head and grammaticalized modifier uses, but also cases which are ambivalent between a head and a modifier use. As discussed in Section 3.4.1.2, these subsume both ambiguous and vague cases, which are two different concepts. Ambiguity refers to those cases in which two separate readings can be distinguished, relying on two distinct ways of contextualization. An example is loads of stuff and furniture, discussed earlier as (3.4). Vague examples, by contrast, are irreducible blends, combining elements from two functions in metaphorical contexts, but do not allow disambiguation. An example is compost heap of nostalgia, discussed earlier as (3.5). Synchronically ambiguous SN-constructions are interesting in that they may represent transitional contexts of grammaticalization and delexicalization and may point up factors that diachronically allowed the lexical SNmeaning to be pragmatically enriched with quantifier meaning. Vague constructions are more likely end results of grammaticalization processes, consciously exploiting lexical and grammatical meaning layers of SNs. Both ambiguous and vague instances are proof par excellence of the fact that the grammaticalization of SN-expressions is driven by language users’ search for expressivity in communication, since they both display in different ways the tension between lexical and grammatical meaning present in SNs. It will become clear that SNs at present differ in the relative proportions of their different layered uses and that some SNs are closer to such semantically more schematic quantifiers as many/much than others. In Section 4.6 a scale of grammaticality will be drawn representing this comparison. Relative frequency counts of the different uses of each SN-expression will allow us to draw up quantitative profiles of the SNs in terms of marked and unmarked use (cf. Halliday and James 1993), with high frequency interpreted as a marker of unmarkedness (but see Section 3.4.7 on frequency). Halliday and James (1993) distinguish between equi-probable and skewed systems of choice in grammar. In equi-probable systems the options are equally

4.3 Head and quantifier construction

131

probable and neither option is marked. Skewed systems, on the other hand, display a disproportion between the options, which, according to Halliday and James (1993), in the ideal case approximates a 90%-10% distribution. The less frequent choice then is the marked one. Statistical markedness is hence taken to reflect functional markedness. In general, the grammaticalization of SNs synchronically has given rise to two main modifier uses in addition to a lexical head noun use, i.e. quantifier and valuing quantifier, or valuing uses as such. Extensive corpus analysis shows that all three uses associate with distinct collocational patterns in addition to their being distinct pairings of form and meaning, i.e. they qualify as constructions which are collocationally constrained. From a usually restricted set of concrete collocates in the head use, in which ideational features are shared between the nodal SN and its collocates, the collocational range of SN-quantifiers is much more extended and includes more types of concrete, abstract nouns as well as animate nouns. The path leading to valuing (quantifier) uses involves interpersonal collocational reclustering, which is based on the sharing of affective or emotively coloured values between the nodal SN and its collocates. This constructional use involves attitudinally subjective values manifested in semantic prosody patterns. As mentioned earlier, the term valuing (quantifier) captures reference to the emotive values of N2 as well as points out the overall evaluative character of these expressions. All of the SNs looked at have to different extents developed microconstructional quantifier uses or valuing (quantifier) uses within NP of NPsyntagms. Some SNs, such as heaps of, loads of, lot of, lots of and piles of have only followed one path of grammaticalization, namely that leading to quantifier uses. Others, such as pile of and heap of have mainly followed the path from head to quantifier and only to a limited extent moved beyond pure quantifier status. Load of and especially bunch of, on the other hand have a prominent valuing quantifier use in addition to a quantifier use and also appear in almost purely valuing uses, in which the quantifying aspect is backgrounded very significantly or may be lost. In the case of bunch of both modifier uses are in fact more or less equally strongly attested and intertwine in the sense that they almost overlay each other, as manifested by semantic prosody patterns with certain types of nouns. In addition, the various SN-expressions synchronically show different degrees of grammaticality. Some SNs still have a predominant head use, while others have developed a predominant quantifier use. These relative frequencies will be assessed and discussed in the following sections.

132Chapter 4: SN-constructions 4.3 Head and quantifier constructions: heap(s) versus pile(s) and lot(s) Section 4.3.1 will discuss heap(s) of and pile(s) of as two semantically similar expressions, and will compare them in terms of degree of grammaticality, i.e. the relative frequencies with which they can be used as quantifiers. Section 4.3.2 will discuss lot(s) of, which are the most frequently used SNs in present-day English. Possible differences in the collocational range, selectional restrictions and semantic prosody patterns of the SNs will be assessed and discussed with reference to lexical persistence and singular versus plural forms of the SNs, among other factors.

4.3.1 Heap(s) versus pile(s) Working with exhaustive extractions from the Collins Wordbanks corpus, I analyzed 106 instances of heap of, 90 of heaps of and 165 of piles of. For pile of a random sample of 250 instances was analyzed. I will compare pile(s) and heap(s) in terms of what Lehmann (1985: 303, 305) has called “synchronic variation”, i.e. ordering synchronically co-existing “structural devices” for expressing a given grammatical function by setting them out on a scale which shows their respective degree of grammaticality.

4.3.1.1 Head uses of pile(s) and heap(s) The lexical uses of heap(s) of and pile(s) of synchronically belong to the same semantic domain and can often be used interchangeably to refer to the same sort of referent. Nevertheless, definitions in dictionaries, as well as corpus data, make clear that a pile is a more specific kind of heap in several respects. The OED defines both concepts as follows: A heap is a collection of things lying one upon another so as to form an elevated mass often roughly conical in form. (A heap of things placed regularly one above another is more distinctively called a pile.) (original italics in pile, other italics are mine)

The Concise Oxford Dictionary more explicitly defines heap as “a collection of things lying haphazardly one on another” and pile as “a heap of things laid or gathered upon one another” (Italics mine). I describe the difference in conceptual semantics by saying that pile as a concept typically incorporates notions of an intentional agent constructing the pile, as well as a feature of layered regularity, rather strong verticality and constructional solidity. On the other hand, 'heap’, as a concept, is in

4.3 Head and quantifier construction

133

itself semantically more vague in terms of its precise constellation, boundedness and internal composition. Instead of a strong vertical dimension ‘heap’ incorporates the notion of potential horizontal expansion, since it mainly seems to profile an undifferentiated mass. Overall, ‘heap’ connotates nonchalance or unintentionality on the part of the person creating a heap. A heap can also have a non-volitional cause or instigator, such as forces of nature, whereas this seems unlikely for a pile. These differences between ‘heap’ and ‘pile’ can be referred to by saying that they invoke different frames. These differences in conceptual specificity between heap and pile do not appear to play a very important role in the original head uses of both SNexpressions. In fact, corpus analysis shows that heap and pile share a large number of collocates in their head uses and generally impose few restrictions on N2, other than that their referents have to be stackable and concrete. This means that in the head noun construction heap(s) of and pile(s) of have a collocational range that is restricted to concrete and truly stackable nouns, but is fairly unrestricted in terms of countability, as the figures in Tables 4.2 to 4.5 will show. Both count and uncount concrete nouns are attested, e.g. bricks, papers, pancakes, rubble, ironing, as well as animate nouns, though the latter are much less frequent and typically refer to (dead) bodies in an objectified way. Abstract nouns are excluded, which is in keeping with the lexical meaning of the head uses, which designate concrete shapes taking up space in the real world. In their head noun use, heap(s) and pile(s) are coextensive with the referent of N2, i.e. they have the same referent in the real world. The of-phrase specifies what the heap or pile consists of. Adjectival, numeral or other premodifiers of the SN apply to the entire binominal construction, e.g. a foot-high pile of manure (CW-Today), great steaming heaps of dung (CWOZnews), a delicious heap of almond brittle (CW-Times), two large piles of filed correspondence (CW-UKbooks). Premodifiers of N2 indicate which subtype of the type designated by N2 is involved or attribute a quality to these entities, e.g. the untouched pile of morning papers (CWToday), a large pile of dead penguins (CW-BBC), untidy heaps of muddy or wet footwear (CW-UKephem). Example sets a, b and c illustrate head noun uses of heap, heaps, pile and piles with concrete count, animate and concrete uncount nouns respectively. The contextual markers of head status as well as adjectives, verbs, determiners, modifying the SN or N2 are underlined. These emphasize the prototypical semantic features of heap(s) and pile(s) respectively. In the head noun use, only the SNs themselves have been put in bold, while N2 is in italics.

134Chapter 4: SN-constructions a) Concrete count N2s (4.5)

The production was marred by a mountainous unitary set by Lucio Fanti, a dreary heap of stone slabs scattered helter-skelter. (CWTimes)

(4.6)

“My first impression was not that it was an earthquake”, said Heinz Hermanns, standing by a heap of bricks that had fallen from his 100 year-old house. (CW-Today)

(4.7)

Can’t afford a 12ft by 12ft portrait of your country estate? Well, how about a sensitively arranged pile of bricks instead. Artists are wage slaves too, changing their product to fit the market. (CWTimes)

(4.8)

Apart from the occasional sniff, the real subject of the animal’s resentment –Her majesty the Queen, no less – maintained an uncomfortable silence as she stared at the untouched pile of morning papers that Benson had placed on a side table. (CW-Today)

(4.9)

There is a small pile of Unpronounceable German Beer bottles stacked on the table. (CW-UKmags)

(4.10)

However, many of the tunnels had been completed and were now filled with untidy heaps of military hardwear-helmets. (CWUKbooks)

(4.11)

If you’re fed up with coats thrown over the banisters and piles of dripping brollies in your hall, bring a bit of order to the chaos with a hall stand. (CW-UKmags)

b) Animate N2s (4.12)

A 55-year-old Muslim survived a massacre near the town of Karakaj by staying motionless under a pile of bodies. (CW-OZnews)

(4.13)

While we were attempting to disentangle ourselves from the heap of crumpled bodies, Alain coolly skied off into the distance, expecting us to follow. (CW-UKmags)

(4.14)

Emaciated bodies. Piles of bodies. On the ground. Covering carts. Falling from boxcars. (CW-UKmags)

4.3 Head and quantifier construction

(4.15)

135

The soldiers in their grubby brown, blood-soaked uniforms have been arranged into heaps of two or three bodies. (CW-Times)

c) Concrete uncount N2s (4.16)

Some stairs were the only thing left standing. As if in a dream, Pearle circled the enormous pile of smoking rubble looking for a sign of Bud, rather than for Bud himself. (CW-USbooks)

(4.17)

A jilted girlfriend got revenge on the boyfriend who dumped her by dumping a foot-high pile of manure in his bed. (CW-Today)

(4.18)

The synagogue of the Hadassah Hospital is lit by the glow through Marc Chagall’s 12 stained-glass windows representing the Sons of Jacob. When they were shattered in the 1967 war he had the heap of glass sent to his studio and put them together again, leaving a single, symbolic bullet hole. (CW-Times)

(4.19)

Each home belonging to a Croat or Moslem was reduced to a heap of rubble. (CW-Today)

(4.20)

One day I was in Donald Baverstock’s office and he pointed to two large piles of filed correspondence. (CW-UKbooks).

(4.21)

Many parts of Kuwait city are still without electricity, and huge piles of garbage still litter the streets. (CW-NPR)

Heap and pile team up with the same semantic sets of collocates, e.g. bricks, stones, rubble, paper, etc. Helter-skelter in (4.5) indicates the more chaotic concept of heap, just like untidy in (4.10). The choice of verbs may also do this, e.g. scattered in (4.5), which also shows the horizontal expansion potential of heap. This dynamic horizontal dimension is also found in shattered in (4.18) and reduced to in (4.19). In example (4.21) litter also shows concord with piles and not with garbage. In the head noun use pile, piles, heap and heaps all seem to have a predominantly neutral or negative semantic prosody, with many collocates referring to rubbish, the sparse remains of something or paperwork sitting on someone’s desk. The head noun construction allows for rather unrestricted premodification patterns, which shows that the SNs are still fully categorial nouns in this construction. For pile(s) we find premodification stressing its conceptual orderliness, e.g. sensitively-arranged pile of bricks in (4.7) and a foothigh pile of manure in (4.17), which also illustrate the volitional agent typi-

136Chapter 4: SN-constructions cally present in creating a pile. This volitional agent also shows up in (4.8) and (4.9), in placed by Benson and stacked respectively. By contrast, the possibility of a non-volitional agent associated with the formation of a heap is illustrated by (4.6), in which an earthquake is the cause of the heap of bricks. Spatial markers point out the concrete location and spatial dimensions of the entities referred to, e.g. under a pile of bodies (4.12), from the heap of crumpled bodies (4.13), foot-high pile of manure (4.17) and two large piles of filed correspondence (4.20). This is also expressed by the choice of verbs, e.g. arranged into (4.15), circled (4.16), litter (4.21) and pointed to (4.20). The definite determiners in examples (4.8), (4.13), (4.16) and (4.18) have their full referential function with regard to the referent of the binominal NP and the overall presence of determiners, quantifiers and other premodifiers of SNs or N2s all indicate the non-decategorialized nominal status of the SN. Premodification patterns will be discussed in more detail in Section 4.5, in which the premodification in head noun constructs is compared to that in grammaticalized and ambivalent constructs in terms of their respective restrictions. The quantified results of the qualitative analysis of heap, heaps, pile and piles are summarized in Tables 4.2 to 4.5. Token frequency is indicated by # and percentages by %. The latter indicate the overall relative proportions of head, quantifier, valuing quantifier, and ambivalent uses for each SNexpression. Table 4.2. Quantified collocational profile of heap of

Heap of (106) Animate Concrete/count Concrete/uncount Abstract/count Abstract/uncount Total # Total %

Head

Quantifier

# 2 23 24 / /

# 6 2 1 11 14

Valuing quantifier # / / / / 2

49 46.2%

34 32.1%

2 1.9%

Ambivalent Ambiguous Vague # # / / 2 4 / 6 / 1 / 8 2

19 19.8%

4.3 Head and quantifier construction

137

Table 4.3. Quantified collocational profile of heaps of

Heaps of (90) Animate Concrete/count Concrete/uncount Abstract/count Abstract/uncount Total # Total %

Head

Quantifier

# 1 11 15 / /

# 8 8 5 14 25

Valuing quantifier # / / / / /

27 30%

60 66.7%

0 0%

Ambivalent Ambiguous Vague # # / / / 1 / / / / 1 1 1

2 3.3%

Table 4.4. Quantified collocational profile of pile of

Pile of (250) Animate Concrete/count Concrete/uncount Abstract/count Abstract/uncount Total # Total %

Head

Quantifier

# 6 146 82 / /

# 1 / 1 1 /

Valuing quantifier # / / / / /

234 93.6%

3 1.2%

0 0%

Ambivalent Ambiguous Vague # # / / 4 / / / 5 / / 4 9

4 5.2%

Table 4.5. Quantified collocational profile of piles of

Piles of (165) Animate Concrete/count Concrete/uncount Abstract/count Abstract/uncount Total # Total %

Head

Quantifier

# 2 77 69 / /

# / / 1 1 /

Valuing quantifier # / / / / /

148 89.7%

2 1.2%

0 0%

Ambivalent Ambiguous Vague # # / / 4 1 6 1 1 1 1 / 12

3 9.1%

138Chapter 4: SN-constructions The quantified results for heap(s) of and pile(s) of reveal several significant differences with regard to head/quantifier proportions for both SNs, singular and plural. Firstly, pile of as well as piles of display significantly fewer quantifier uses than heap of and heaps of. Especially for the plural variants the discrepancy stands out very clearly, with the quantifier use of heaps of accounting for almost 70% of the attestations, versus a mere 1.2 % of quantifier uses for piles of. Singular heap of has 32.1% quantifier uses, which is significantly more than pile of which only has 1.2%. Both singular expressions can express valuing quantifying meaning, although mostly in ambivalent uses and not very productively. The observed differences between pile(s) and heap(s) will be explained in terms of differences in the conceptual semantic make-up and delexicalization potential of these SNexpressions. Since the developments of heap(s) and pile(s) differ in terms of extent of grammaticality, the rest of their discussion will be split up. Section 4.3.1.2 will discuss heap(s) of in its quantifier use.

4.3.1.2 Quantifier uses of heap(s) of Tables 4.2 to 4.5 above show that the collocational range of heap of and heaps of has extended significantly, whereas pile of and piles of show hardly any systematic extension of collocational range, which ties in with a low number of grammaticalized uses. When used as a quantifier, the constellatory semantics of heap are backgrounded to a very large extent, whereas its size implications are brought to the fore through pragmatic strengthening. In the quantifier construction heap(s) no longer serve to categorize N2 as an instance of the type 'heap(s)' (cf. Langacker 1991; Bache and Davidsen-Nielsen 1997: 350). Heap of and especially heaps of can now unambiguously denote mere quantity of a variety of concrete things, including the collocates that are also found in the head use. However, their quantifier use has extended most significantly with regard to abstract nouns, both count and uncount, which were not allowed in the head noun use. Another major collocational extension is towards animate nouns. I will first discuss quantifier uses with concrete nouns under a. Quantifier constructs with abstract and animate nouns will then be discussed under b and c respectively. a) Extension to more kinds of concrete nouns With regard to concrete nouns, collocates are allowed in the quantifier construction that no longer form literal heaps. This collocational set includes pipes (CW-OZnews), stalls (4.24) and salmon (4.26) for heap of and cars

4.3 Head and quantifier construction

139

(4.22), hair products (CW-OZnews), giveaways (4.23), paddocks (CWOZnews), prawns (CW-UKephem), footnotes (CW-Times), games (CWSunnow) and nosh (4.25), garlic (CW-OZnews), hair (CW-OZnews), machinery (CW-UKmags), meat (CW-UKspoken): (4.22)

A police spokesman said more than a dozen cars have been stolen with “heaps of cars broken into”. (CW-OZnews)

(4.23)

Doors will open at 10.30pm for a great night out, featuring the somewhat bizarre antics of “The Vibemaster”, Evan Fletch mystery musicians, heaps of giveaways and prizes. (CW-OZnews)

(4.24)

There's also a heap of stalls, selling everything from clothes, arts and crafts to junk. (CW-OZnews)

(4.25)

There'll be heaps of good nosh and champagne to keep occupied sweetie. (CW-OZnews)

(4.26)

Bloke in Townsville tired of maintaining his swimming pool winter filled his with fresh water at the end of last summer and tossed in a heap of infant fresh-water salmon. (CW-OZnews)

In their quantifier uses heap of and especially heaps of have hyperbolic expressive value and are often used in registers that invoke this, e.g. commercial writing, advertisements, intense evaluative writing in reviews and stretches of text that are on the whole very expressive and emphatic in nature. Heap(s) of designates an abundant quantity of something, in the sense of more than average. As Tables 4.2 and 4.3 show, heaps of has grammaticalized to a much larger extent than heap of, which has a predominant head use in synchronic data. Plural heaps of is also more hyperbolic than the singular form, which seems a natural effect of the plurality marker: it increases the intensity of the size implications. See Section 4.7 for a detailed discussion of the diachronic processes and changes leading up to the synchronically attested quantifier uses of heap and heaps of. In heaps of cars (4.22) the cars do not literally constitute heaps; the police officer wants to say that (too) many cars are broken into. Similarly in a heap of infant fresh-water salmon (4.26) the salmon does not form a literal heap, but refers to a lot of salmon. In heaps of giveaways (4.23) and a heap of stalls (4.24) the hyperbolic quantifier uses of the SN-expressions are in keeping with the purpose of emphatically advertising an event or place to shop. Heaps of good nosh and champagne (4.25) describes the abundant presence of food and drink at a high-society event. In this quantifier use SN + of constitute one functional unit that jointly expresses grammatical quan-

140Chapter 4: SN-constructions tifier meaning. For heaps of and heap of no graphological reflections of this were found in the Collins Wordbanks corpus, but internet examples extracted via Google do show a coalesced spelling heapsa and heapa. This will be discussed in more detail in Section 4.3.1.5. Semantic prosody seems to be mostly positive or neutral in these examples, with neutral N2-collocates like cars (4.22), stalls (4.46) and salmon (4.68) and more positive ones such as giveaways (4.23) and good nosh (4.25). In these contexts verbs and other elements equally point towards positive emotive colouring. b) Extension to abstract nouns (4.27)

They went through my bags, searched me and asked a heap of questions. (CW-OZnews)

(4.28)

This, very suddenly, was the man who had once regularly outdriven Niki Lauda, who had taken a heap of pole positions, commanded and won races from the front. (CW-UKbooks)

(4.29)

Sure, I would like to have been in the position of Tony mccoy or David Bridgwater (the current leading jockeys), and had things worked out better for me after I won the National and a whole heap of other good races in 1988 and '89, I might have been," Powell said. (CW-Times)

(4.30)

My brother Shane was a State swimmer and represented Queensland at heaps of tournaments. (CW-OZnews)

(4.31)

Tickets for the program's largest concert, Mahler's Symphony of a Thousand, have already been snapped up and tickets for the heaps of other events are selling faster than the lap time at today's Indy. (CW-OZnews)

(4.32)

With busking and street theatre and heaps of surprises and lots for kids including a children's fairground and special performances. (CW-UKephem)

(4.33)

He does a heap of charity work and from the moment I first shook his hand it's been a case of “if there's anything I can do to help you out, then just give me a ring.” (CW-OZnews)

4.3 Head and quantifier construction

141

(4.34)

For a fast-thinking, strong-minded woman with an insatiable curiosity, Anna does manage to get herself into and out of a heap of trouble. (CW-USephem)

(4.35)

Providing the German Krauser company can supply, the newly released xj900 will also feature with the luggage gear which has attracted heaps of demand. (CW-OZnews)

(4.36)

Miranda has won heaps of praise for her roles since she left National Institute of Dramatic Art in 1990. (CW-OZnews)

(4.37)

It gets fairly quiet up here, even though there's heaps of noise underneath. (CW-OZnews)

(4.38)

The wine, from low-yielding Barossa Valley vines, has heaps of peppery flavour, made more complex and softened by ageing in American oak. (CW-OZnews)

(4.39)

Their show has heaps of humour, and you can detect the cheeky influences of Circus Oz. (CW-OZnews)

(4.40)

Michael's requirements are simple: heaps of room, a window with a view and an assistant to do the cleaning. (CW-OZnews)

(4.41)

Leading the charge for privately-owned establishments was emphatically excellent Grissini's, positioned on the eastern edge of Paradise Waters, a supremely reliable Italian eatery with tons of class and heaps of scrumptious cuisine. (CW-OZnews)

Within this quantifier use, heap of and especially heaps of again have hyperbolic expressive value and designate an abundant quantity of something, i.e. ‘so much/many of something that it can be too much/many’, e.g. a heap of questions (4.27) and a heap of trouble (4.34). In examples such as a heap of pole positions (4.28), a whole heap of other good races (4.29), heaps of demand (4.35) and heaps of praise (4.36) the SN-expressions function as intensifying quantifiers with regard to the descriptions or enumerations of various kinds of achievements. The larger discourse contexts of these examples contain many emphatic verbs, nouns, adjectives or other qualifications, such as outdriven and commanded in (4.28), the SNpremodifier whole in (4.29) reinforcing the quantifier meaning of heap of, and the presence of other quantifiers such as lots in (4.32) and tons of in (4.41). Heap(s) of also appears in commercial writing trying to sell or convince people to buy certain items or participate in certain events, e.g. buy-

142Chapter 4: SN-constructions ing wine in (4.38), an event in (4.32), a car in (4.35). Heaps of is sometimes repeated in order to further increase the hyperbolic effect, e.g. Heaps and heaps of dope smoking (OZnews). Additional premodifiers of N2 can reinforce the emphatic value of heaps of as a quantifier, e.g. scrumptious in (4.41), a heap of hyperbolic press (CW-UKmags), heaps of junk-sponsored bad vibes (CW-UKmags). (4.42)

PUT together six planning aid staff, 180 volunteers, heaps of enthusiasm, hard work and dedication, and what do you get? Planning Aid For London! (CW-UKmags)

(4.43)

Island hopping is hard to do in the winter unless you have heaps of time. (CW-USephem)

(4.44)

The graphics are very polished, with pitch detail, markings and the like adding heaps of atmosphere. (CW-Today)

(4.45)

The “surrogate mum” to princes William and Harry shared heaps of fun with them at a fair yesterday while father Charles was otherwise engaged. (CW-Today)

Another discourse context in which SN-expressions occur as hyperbolic quantifiers is that of the formulation of quantities within a metaphoricallike recipe for something, usually something good. Often these formulations can be somewhat hackneyed. Examples are Put together six planning aid staff, 180 volunteers, heaps of enthusiasm in (4.42) and throw in heaps of Urdu dialogue (CW-Times). We will see this context with other SNexpressions as well, often combining several of them in one example. Not only do the SN-expressions no longer refer to stackable concrete things, more abstract implications of their lexical meaning, such as spatiotemporal contiguity further loosen collocational requirements. For instance in a heap of pole positions (4.28) the pole positions are neither spatially nor temporally contiguous, just like the races in a whole heap of other good races in (4.29), tournaments in heaps of tournaments in (4.30), trouble in heap of trouble in (4.34), praise in heaps of praise in (4.36), and so on. In other cases there still is temporal or spatial proximity between the referents, e.g. heap of questions in heaps of questions (4.27). Semantic prosody in these examples is predominantly negative for heap of from the perspective of the speaker, e.g. a heap of questions in (4.27), a heap of trouble in (4.34), a heap of bad debts (CW-OZnews), a heap of contradictions (CW-UKspoken), paying a heap of tax (CW-OZnews), a heap of bad memories (CW-Times) and a heap of incidents and litigation (CW-OZnews). Heap of can appear with positive collocates as well, e.g. a

4.3 Head and quantifier construction

143

heap of pole positions in (4.28), a whole heap of other good races in (4.29). Heaps of, which has intensifying hyperbolic reference as a quantifier, has an overall positive semantic prosody, e.g. heaps of surprises and giveaways in (4.32), heaps of praise (4.36), heaps of humour in (4.39) and heaps of scrumptious cuisine (4.41), heaps of time (4.43), heaps of fun (4.45), heaps of treats (CW-OZnews), heaps of potential (CW-OZnews), heaps of energy (CW-UKspoken). However, heaps of can also quantify N2s with negative emotional colouring, e.g. heaps of noise in (4.37), heaps of junk-sponsored bad vibes (CW-UKmags) and heaps of criticism (CW-Today). These quantifier constructs also show reflexes of partial decategorialization of the SN-expressions in the sense that the use of determiners and premodifiers displays restrictions. Overall, as quantifier expressions, heap(s) of appear with indefinite articles or the zero article, except in the heaps of other events in (4.31), where the precedes heaps of, but quantifiers like many also allow this, e.g. Harding was not approached to answer the many questions put in the article (CW-OZnews). Heap(s) of are rarely premodified, except by means of intensifiers that reinforce the quantitative function of the SN-expression, as in a whole heap of other good races in (4.29). In sum, there is some variability in the determiners that can precede heap(s) of, but the indefinite article is predominant with singular heap of and the zero article with plural heaps of. The indefinite determiner no longer seems to have its full deictic function here and has to some extent coalesced with heap of, which as a whole can be substituted by a monomorphemic quantifier such as many/much. The substitution test is a test of functional as well as formal unity of the SN-expressions. Section 4.5 will argue in greater detail that the premodification patterns attested for grammaticalized SN-expressions need not detract from the decategorialization claim, since they systematically differ from those attested for head noun uses. In quantifier uses they support the hyperbolic quantifying meaning, as in a whole heap of other good races (4.29). c) Extension to animate nouns The other major collocational extension of heap(s) of as a quantifier is towards animate nouns. Both heap(s) of and pile(s) of appeared with nouns referring to human beings or animals in head uses, with pile even having 6 attestations. However, in these head uses reference is hardly ever to living animate creatures, but to dead bodies in a rather objectified way, e.g. (4.12)-(4.15). In the quantifier use, however, we see that heap and especially heaps of have systematically extended to animate nouns, without reference to a specific constellation that these people or animals take up. Just like the quantifier use with concrete nouns and abstract nouns, heap of and heaps of here too have a quantifier meaning that is generally devoid of

144Chapter 4: SN-constructions its original lexical semantics and functionally similar to more schematic monomorphemic quantifiers such as many and much, apart from the additional hyperbolic and expressive features they typically express. Examples (4.46) to (4.56) illustrate quantifier uses of heap(s) of with animate nouns: (4.46)

Modesty was eventually restored, but not before a heap of people in Pine Rivers had become privy to intimate knowledge what Santa wears beneath those baggy pants. (CW-OZnews)

(4.47)

“A whole heap of people step on the gas when them see me and run back to tell people me gone mad[e]”, he recalls wryly. (CWUKmags)

(4.48)

The Madder story began two years ago in New York City: there was this guy called Billy Cote who was making music with a whole heap of people (CW-UKmags)

(4.49)

Scouring the cast list for stars, you find Eleanor Bron plus a heap of nobodies. (CW-Times)

(4.50)

Reviewed in The Sunday Times is Science in the Bedroom A History of Sex Research. What's interesting is how many sexual researchers and observers were driven by self-interest. Heaps of them at least, the dead ones with no access to legal redress were, well, different. (CW-OZnews)

(4.51)

I often wonder where respect for others has gone. Certainly, heaps of kids are aware of it, carry it out and are brilliant but others seem to have lost the plot. (CW-OZnews)

(4.52)

There are heaps of good players coming through there now who can take over. (CW-OZnews)

(4.53)

I bet there are heaps of people in Hexton who didn't use Marcus as their vet. (CW-UKbooks)

(4.54)

He said, “You need not talk so slack about it; there may be heaps of Ku-Klux in the State, and they might get hold of your talk by the Chairman.” (CW-USbooks)

(4.55)

She is practical, sexy and a fabulous cook. And then there's all the cute stuff, sitting surrounded by heaps of butterflies, and how he

4.3 Head and quantifier construction

145

likes to go walking in the woods and get up to things you wouldn't expect of the middle-aged (CW-Times) (4.56)

There are no prizes for guessing that I’m talking about Sky Trackers, the Children’s Television Foundation program attracting heaps of young viewers to its 8am Sunday timeslot on Channel 7. (CW-OZnews)

Again, heap of and especially heaps of have a hyperbolic value as quantifiers, which can be reinforced by whole, e.g. a whole heap of people in (4.47) and (4.48). Heap of and heaps of express a large quantity of something without calling up the original lexical semantics of heap, nor its more abstract implications of contiguity. For instance, sexual researchers referred to in heaps of them (4.50) are scattered in time and place. In heaps of KUKlux in the State (4.54), the Ku-Klux clan are similarly at different places within the State, just like the heaps of young viewers in (4.56) in Australia. Additionally, from examples such as (4.47) and (4.50) it becomes clear that people need not be temporally contiguous either. In (4.47) the whole heap of people do not all step on the gas at the same time; rather, their braking is spread in time. This again shows that besides its specific constellation meaning, the more abstract implications of the original concept of heap such as the idea of spatial or temporal contiguity can be lost or at least backgrounded to a very large extent. A final remark about the quantifier use of heap of and heaps of pertains to the region in which it is most frequent. Used as quantifiers, both expressions are typical of Australian English, with 35 of the 60 quantifier uses of heaps of and 16 of the 34 quantifier uses of heap of deriving from the OZnews-subcorpus. (cf. Peters et al. 2009: 159).72 The remaining grammaticalized attestations are divided over UKephem (2), UKmags (4), UKspoken (4), USephem (1), NPR (1), UKbooks (2), USbooks (1), Times (8), Today (3) and Sun (1). Heap(s) of is thus least frequent in American English, but not uncommon in British English.

4.3.1.3 Valuing quantifying uses of heap of There is one more grammaticalized use of heap of that has remained undiscussed, even though at present only two attestations are found. It is a type of use that is similar in some ways to the emotively coloured uses of load of and bunch of that will be discussed in Sections 4.4.1.7 and 4.4.2.3, though the systematicity and frequency with which load of and bunch of have developed valuing (quantifier) uses cannot be compared to that of this use of heap of here. The examples in the corpus data are:

146Chapter 4: SN-constructions (4.57)

Australia's greatest double act laid into Fletcher's comparison between England's current pace crop and Lillee and Thomson in their prime. Lillee blasted: “Sounds like he's been talking a heap of shit to me. I've never seen anyone as fast or as frightening as Thommo”. (CW-Today)

(4.58)

“We have a ranking of 92nd in the FIFA world lists. That's a depressing record.” Jim Boyce Cliftonville) was re-elected IFA president for the third successive year. “What a heap of shit.” (CW-Sunnow)

In both cases N2 is shit in the sense of ‘complete nonsense’ or ‘worthless information’. In (4.57) there is also quantitative information involved in the sense of ‘a lot of nonsense’, but what is mostly at stake is the evaluation of the information as being completely useless or untruthful. This is clearer in (4.58), where the premodifier what intensifies the emphatic value of heap of shit. This emotive quantifier use is restricted to these two attestations and to the collocate shit, which gives the entire expression idiomatic and semi-lexicalized features. This use can be explained as a metaphorization of the frequently attested set of concrete negative collocates in the head noun use of heap of referring to junk and rubbish. Concrete rubbish is then metaphorized to verbal junk, so to speak.

4.3.1.4 Ambivalent uses of heap(s) The only category left to discuss for heap and heaps is the ambivalent category, which contains ambiguous and vague instances. Both categories are very important in order to grasp the specific semantics of the quantifier uses of these SN-uses. Notice that heap of, pile of and piles of have significantly higher percentages for the ambivalent category than heaps of. The latter is by far the most grammaticalized SN-expression of this set and seems to dichotomize rather clearly into either head or quantifier uses. This can be seen as additional evidence for its high degree of grammaticalization and delexicalization, as discussed earlier. Heaps of has only three ambivalent cases, two of which are vague (4.60)-(4.61) and one is ambiguous (4.59) between head and quantifier reading: (4.59)

Thanks so much for the cards and letters and faxes regarding mango mania. We received heaps of mango chutney recipes, all different, and all sorry to say now that the season is finished, for green mango cooking. (CW-OZnews)

4.3 Head and quantifier construction

147

In example (4.59) the recipes sent in can constitute actual heaps, or there might just be large quantities of them. The semantic mechanism at work here is metonymy between paper and the text on it. As Section 4.3.1.2 has shown, such collocational extension via metonymy is no longer necessary to promote the quantifier use of heaps of, which is already firmly established. We will see in Section 4.3.1.6 that pile(s) of is much more dependent on such mechanisms and contextual clues for its quantifier uses and ambivalent uses. Vague cases featuring heaps of are typically metaphorical and generally extremely expressive. They exploit both the grammatical and lexical potential of the SN-expression and hence typically include contextual markers for both head and hyperbolic quantifier readings in the form of verbs, adjectives, or other clues. They constitute an irreducible mix of both types of meaning of the SN-expression, and in this sense are intensified illustrations of the kind of semantic and structural tension inherent in SN-expressions as such: (4.60)

Surely their non-sexist Prince Charming might be out there somewhere? “Are you trying to apply for it?” smirks Anjali before collapsing into quaking heaps of raucous laughter. Rather too raucous, in my opinion. (CW-UKmags)

(4.61)

Many other viruses are highly malignant reducing the priceless words of your aforementioned seminal novel, letter to bank manager or phd thesis to amorphous heaps of molten letters. (CWUKmags)

In (4.60) Anjali is clearly laughing very loudly and very much, but this is described by comparing the person laughing, and its effect on that person, to quaking heaps of raucous laughter, which adds information to the quantifying and intensifying elements by suggesting that Anjali is robustly built for instance, and possibly that the intensity of the laughter is to a certain extent uncalled for. Collapsing into ties in with the potential for horizontal expansion present in ‘heap’ and referred to earlier. In example (4.60), there is a throwback of heaps of to its literal meaning, but the hyperbolic quantifier function and the abstract uncount N2 disallow simply classifying the examples as head noun uses. Example (4.61) also sets up a metaphorical comparison, likening virus-affected texts to amorphous heaps of molten letters. Here too the separate premodification of heaps of has a literalizing effect on heaps of, but at the same time heaps of molten letters refers to many letters too. Heap of has more ambivalent uses than its plural variant, with 2 ambiguous ones and 19 vague ones. An example of ambiguity is (4.62):

148Chapter 4: SN-constructions (4.62)

A spokeman for her London agents, Storm, confirmed yesterday: “At the moment she's looking at a heap of movie scripts. Another possibility Kate has toyed with recently is becoming a fashion photographer.” (CW-Today)

The movie scripts in (4.62) may constitute a literal heap or there might be many of them. It is not entirely clear whether size implications or constellatory semantics are foregrounded most here. Vague cases include metaphorical comparisons, often with evaluative undertones, in which a person is likened to a heap of something very negative. In these vague examples, heap of draws a comparison with the human body and N2 specifies the negative quality to which the speaker reduces the person talked about, often reinforced by premodifiers accentuating the negative semantic prosody, e.g. (4.63). In other cases reference is not so much negative as ironic, for instance in (4.64) and (4.65), or even positive, as in (4.67): (4.63)

Heartbroken mother Kay Potts attacked the magistrates who bailed sex offender Andrew Hagans before he raped and strangled her daughter. “They are imbeciles who released a lethal weapon,” she said. “Hagans is a heap of woman-hating filth.” (CW-Today)

(4.64)

Now that Liberace has climbed the sugar loaf in the sky we can repeat the defamation in full: “That deadly, winking snuggling chromium-plated, scent-impregnated, luminous, quivering, giggling fruit-flavoured, mincing, ice-covered heap of mother-love.” The Mirror howled for its money back when it was revealed Liberace died of AIDS. (CW-OZnews)

(4.65)

“I've spent almost as many hours trying to change men as I have sleeping, no wonder I'm so behind on everything. Nothing more romantic than a steaming heap of male potential just beyond your reach.”(CW-NPR)

(4.66)

Cross marveled at the self-possession of this man who, twentyfour hours ago had mercilessly reduced Bob's enthusiasm for organizing to a heap of whimpering flesh, and Cross was certain that no regretful memories lingered on in Hilton's mind as he coldly grappled with this new crisis. (CW-USbooks)

(4.67)

That's David. He's one heap of muscle. He doesn't smoke or drink and he is in bed most nights by half-past nine. (CW-Times)

4.3 Head and quantifier construction

149

Examples like (4.63) may seem similar to examples (4.57) and (4.58) with heap of shit, discussed earlier as emotively coloured quantifiers, but (4.63)(4.67) are not idiomatized or routinized. Instead they set up very specific metaphorical comparisons for one specific occasion. Vague uses are also found with reference to vehicles, as in (4.68) and (4.69), work (4.70) and a city in (4.71) for instance. Again mostly the evaluation tied to this use of heap of is negative but need not always be. In these ambivalent examples, the often extensive premodification patterns typical of the head noun use are combined with abstract nouns in the ofphrase, which are typical of grammaticalized uses. (4.68)

You know he is still a worthy sailor despite the heap of junk he calls a boat. (CW-OZnews)

(4.69)

While Mary mccarthy was in Newport with her then-husband Bowden Broadwater, Edmund Wilson and his son by Mary, Reuel, were in Wellfleet. mccarthy had sold Dwight her Wellfleet heap of a car for one dollar. One day on the street of the village, Wilson accused Dwight of driving his car. Dwight replied that he had the bill of sale and that he was the proud owner of the humble vehicle. (CW-USbooks)

(4.70)

Eating, washing and dressing, two hours. Work and travel, another ten hours. Which leaves just four hours a day for all those other things, including watching TV and that creeping, self-renewing slag heap of duty we call housework. (CW-UKmags)

(4.71)

God, it felt great to be alive in this seething great shit-heap of a city, every window a bloodshot eye, every alleyway a shelter for broken deadbeats whose only crime was dreaming themselves out of the gutter. (CW-UKmags)

In addition to this set of examples, other vague cases include the compost heap of nostalgia (CW-Times), referred to earlier.

4.3.1.5 Syntactically extended uses of heaps As discussed in Section 3.4.5, extension to other syntactic contexts usually indicates a high level of grammaticalization (cf. Himmelmann 2004). For heap, heaps, pile and piles extractions were made on the bare noun without of in order to check whether a grammaticalized quantifier use had extended to other syntactic contexts. Only heaps has developed such an extended

150Chapter 4: SN-constructions use, namely into what I will call adverbial functions, quantifying verbs, adverbs or (comparative forms of) adjectives: (4.72)

“My first Opals tour was really good and a real eye-opener," Alexander said. I learned heaps although I didn't get much courttime. (CW-OZnews)

(4.73)

Those big leather-covered seats (electrically adjustable front and rear very rare), airbags for both driver and front passenger, power sunroof, compact disc player and heaps more goodies. (CWOZnews)

(4.74)

I miss the old Valley Pool. The new one is heaps better, of course, but I pity the children who have to attend school carnivals there. (CW-OZnews)

(4.75)

For 900 quid we're not talking about BCCI, are we? Yes, they will cost heaps to insure and sure, a big V12 will eat fuel, but let's be honest, the biggest single cost with any new car is depreciation, and you will not lose much sleep about that. (CW-Times)

(4.76)

“Now I get to travel all over, be my own boss, earn heaps more money and gain more confidence than I ever thought I had.” (CWSunnow)

In keeping with Traugott (2010b) this adverbial use can be called a degree modifier use. As with the quantifier uses within the binominal SNconstruction, the extended uses of heaps are mainly attested in Australian English data. Only 2 of the ten attestations derive from British English sources and none from American English ones. The semantics of this adverbial construction clearly display a similar hyperbolic expressiveness as the original quantifier use. In this extended use of is lost. In the following example from the Internet, however, we see the coalesced form heapsa, the second of which can modify a non-comparative adjective. This use is very much restricted to very informal and spoken contexts: (4.77)

It looks like an original scan! heapsa good work for the effort u put into it! turned out heapsa cool. (http://gallery.minitokyo. net/view/40937)

4.3 Head and quantifier construction

151

4.3.1.6 Quantifier uses of pile(s) of I will now turn to the quantifier developments of pile of and piles of. From the percentages in Tables 4.4 and 4.5 it is clear that piles of and pile of have not nearly grammaticalized to the same extent as heaps of, or even heap of. Indeed, pile(s) of have predominant head noun uses that display a skew of about 90%. In the data set of the singular and plural variant, respectively three and two quantifier uses were attested, with both count and uncount nouns and an animate noun, listed as (4.78)-(4.82): (4.78)

After all, crooked political machines managed to get their candidates elected time after time, and ordinary citizens had a disturbing taste for vices that made piles of money for unscrupulous men. (CW-USbooks)

(4.79)

The results reflect the 13th consecutive year of record profits from Tomkins which also owns a pile of businesses, including guns group Smith Wesson, and Murray Ohio, the giant US lawnmower concern. (CW-Today)

(4.80)

He knows he’ll be one of the first players shunted out in Dick Advocaat’s Gers revolution. He said: “Never mind Walter’s last game- there’s about eight of us in the same boat! I can just see a whole pile of the boys walking out after the final and saying bye bye". (CW-Sunnow)

(4.81)

He recovered well after a nervous start and put in a pile of strong defensive work in the second-half especially. (CW-Sunnow)

(4.82)

Mike Atherton has been warned he must score piles of runs for Lancashire to keep his England Test place. (CW-Sunnow)

These uses have been labelled quantifier uses because they functionally behave as such. However, as a quantifier, pile of and piles of are rather strongly dependent on contextual clues, metonymy or other very specific contexts. In (4.78) piles of money relies on a shift from concrete stackable banknotes to the more abstract concept of money as such. This example makes clear that plural piles of does have potential hyperbolic value. The implied meaning really is ‘lots of money’, in which the lexical semantics have been backgrounded to a very large extent. In a pile of businesses in (4.79) pile of equally functions as a quantifier. In a whole pile of the boys in example (4.80), pile of quantifies an animate noun. Pile of seems rather odd with animate nouns, in that it typically

152Chapter 4: SN-constructions calls up layered verticality when used with animate nouns, as it does in its head use (e.g. a pile of bodies). Maybe a non-native football player or coach holds the floor here and uses a less appropriate SN-expression, for instance Dutch football coach Dick Advocaat referred to in the discourse? It is not clear from the context whether he is quoted. A pile of strong defensive work in (4.81) and piles of runs in (4.82) both relate to sports, i.e. rugby and cricket respectively, and the choice for pile as a quantifier may be a pun on characteristics of the sport, referring to tackles and scrums, or the verb to pile on runs, which is sometimes featured in reports of cricket, e.g. Aussie openers pile on runs; a headline on the BBC-website reporting on a game of cricket (http://news.bbc.co.uk/sport1/hi/ cricket). In sum, the quantifier uses attested for pile of and piles of are not really context-independent in the way that heaps of and even heap of are and hence do not exhibit a systematic loosening of collocational requirements. Instead, these quantifier uses are dependent on metonymic and metaphorical mechanisms at work in the context and in N2. However, since these uses are not vague or ambiguous in the sense used in this study, they are classified as quantifiers and not in any of the latter categories.

4.3.1.7 Ambivalent uses of pile(s) In addition to rare and unsystematic quantifier uses, piles of and pile of display a rather large number of ambivalent uses. Typical examples of ambiguous uses of piles are similar to those attested for heaps, although ambiguity with the piles of micro-construction is more frequently attested. There are 12 ambiguous attestations for piles, spread over concrete (un)count and abstract (un)count nouns. The N2s largely fall within the semantic sets of either ‘paper/money’ or ‘vegetables/spices’: (4.83)

At 4pm Margo has another sandwich and a slice of cake. Dinner is a large rump steak or chicken fillet with piles of vegetables and a packet of rice. Margo said: “I don't pig out. I just need to eat because of my size.” (CW-Sunnow)

(4.84)

A: “Are you a garlic fan in general? Because I am”. B: “Yeah. Yeah I really love it. I mean piles of it”. A: “There's three cloves in there”. B: “That's not very much”. (CW-UKspoken)

(4.85)

The real fun starts when you start receiving piles of property details. It's just like Christmas. Certain houses can be discarded immediately, others look interesting. (CW-UKmags)

4.3 Head and quantifier construction

153

In piles of vegetables (4.83) the vegetables may be stacked as literal piles or they may refer to large quantities as such. Likewise, in piles of it (4.84) it may be that speaker B would not object to literal piles of chopped garlic ending up in the dish talked about, or he/she may be referring to large quantities. The uncount nature of garlic especially triggers off the ambiguity. In piles of property details (4.85) the property details are most likely printed out on paper, and hence may constitute real stacks of documents, but the abstract, though count, N2 details, on the other hand, suggests a quantifierlike reading. Similar metonymies are found for other paper referents and their contents so to speak, e.g. cash, bills, resumes, literature or mail. These ambiguous instances of the piles of micro-construction hence again show that plural piles of does have the potential for expressing hyperbolic quantitative meaning, because this is part of the constructional meaning in these examples. Vague cases of piles of are different in that they first and foremost involve metaphors and integrate literal and grammatical quantifier meaning more inextricably or exploit them both explicitly: (4.86)

Leshan emphasizes a remark by George Bernard Shaw that Lourdes is the most blasphemous place on the face of the earth: “mountains of wheelchairs and piles of crutches exist, but not a single wooden leg, glass eye, or toupee!" This is evidence, Shaw maintained, that God's power is limited; there are things he apparently cannot do, and this is blasphemy. (CW-USbooks)

(4.87)

This silly hospital story has piles of jokes (yes, that's one of them) when a past indiscretion walks back into the life of smoothy Dr Mortimore: namely the nurse he seduced on the sluice-room floor 18 years ago. (CW-Today)

In example (4.86) Shaw’s quotation is hyperbolic throughout, describing Lourdes graphically as hills made up of wheelchairs and crutches that have become useless due to so-called healing. In piles of jokes (4.87), piles of is used to express that the series contains many jokes, but at the same time it is mentioned explicitly that using piles of as a quantifier is a pun as well, namely by referring to another term for haemorrhoids. Pile of also has ambiguous and vague attestations. Typical examples of ambiguity is (4.88), where pile can be literal or quantifying depending on the contextualization: (4.88)

The actor works part-time as a chef in the town’ hotel after answering an advert for trainees. He said: “My family bought me a

154Chapter 4: SN-constructions pile of cooking utensils last Christmas so I decided to go for it.” (CW-Sunnow) At present quantifier meaning in the pile(s) of micro-constructions remains context-dependent or ambiguous. However, in the future pile of and especially piles of may become detached from specific metonymies or contextual markers activating the literal meaning of pile(s), as in piles of property details in (4.85). Piles of has hyperbolic value in these ambiguous instances, as examples (4.83)-(4.85) show, and may hence become new recruits in processes of renewal, so typical within the field of intensifying expressions, particularly within more informal contexts (Hopper and Traugott 2003) (see Section 3.4.8). Pile of also has a rather systematic vague use that might lead to a routinized use if it becomes less context-dependent and more productive. It is functionally very similar to the emotively coloured quantifying uses I described for heap of in a shit-heap of a city (4.71) for instance. What is involved is negatively evaluating meaning, applied mostly to what one could generalize over as artistic/creative products, such as albums, art projects, or other performances: (4.89)

Nothing I can say will stop the Shamen’s new album Boss Drum (One Little Indian) also topping the charts. But be warned nevertheless: it’s the biggest pile of pretentious tosh I’ve ever heard this year. (CW-Today)

(4.90)

I saw them at Brixton Academy and I’ve never seen such a display of sub-Zodiac Mindwarp shit in my fucking life. It was the biggest pile of want ever. Wolfsbane with samples. (CW-UKmags)

(4.91)

In fact, give four students three million quid and I bet they’d come up with something ten times better than this pile of arty old toot. Don’t even let wild animals drag you to see this. (CW-Sunnow)

(4.92)

[About King Kong] If this glorious pile of horror-fantasy hokum has lost none of its power to move, excite and sadden, it is to no small measure due to the remarkable technical achievements of Willis O’Brien’s animation work. (CW-UKephem)

These instances can be said to have come about through metaphorization of reference to concrete junk to classifying artistic products as junk or nonsense. In these instances the function of pile of is not so much to quantify N2, but to value it. In this respect this meaning is similar to that expressed by bunch of in its valuing quantifier use. However, the latter, is much more

4.3 Head and quantifier construction

155

systematically used and also displays other premodification patterns. In examples (4.89) to (4.92) pile of is typically preceded by a definite determiner, often demonstrative, even though reference is to a metaphorical pile. All the evaluative and emphatic premodifiers in these examples intensify the emotive value of these vague constructions. The conspicuous metaphorical context, the determiner and premodification patterns contribute to the vague character of these examples.

4.3.1.8 Conclusion: Comparison of heap(s) and pile(s) In conclusion, heaps of has developed a systematic quantifier use that is characterized by hyperbole, but is in many other respects very similar to semantically more schematic quantifiers such as many and much. This quantifier use came about through delexicalization of heap and foregrounding of its size implications, to the detriment or even loss of its constellation meaning. This is paralleled by a systematic loosening of collocational requirements on the ideational level. In one way many and much are each more restricted than heap(s) of, in that many can only team up with count nouns and much with uncount nouns (cf. Langacker 2009 and to appear). I have argued earlier that the difference in degree of grammaticality between the singular and plural variant of heap can be explained by the additional plurality morpheme, which enhances the size implications inherent in ‘heap’ and increases its hyperbolic expressivity, which is one of the main motivations behind its functioning as a quantifier. Grammatical number adds to the hyperbolic value. In addition, the intrinsic mass meaning of plural nouns (cf. Langacker 1991) likewise enlarges the magnitude already expressed by the plural SN. Singular heap of still has a predominant head noun use, but its vague cases subsume an emotively coloured quantifying use, based on metaphorization, which might become more productive, even though at present it is rather restricted in terms of absolute frequency, contexts of use and collocates in the of-phrase. Still, heap of has grammaticalized and delexicalized more than pile of and piles of. The fact that plural piles of has grammaticalized slightly more than its singular counterpart can again be explained by the influence of the plurality morpheme. However, the overall lead of heap over pile as SNs deriving from a similar semantic domain needs explaining. Since delexicalization plays such a fundamental role in the grammaticalization of SNs to quantifiers, differences in degree of grammaticalization are argued to be linked to differences in the delexicalization processes. There seems to be a fundamental difference in the delexicalization potential of heap and pile, in its turn dependent on certain lexico-semantic features inherent in the frames of both SNs, referred to at the beginning of this dis-

156Chapter 4: SN-constructions cussion in the definitions of both SNs. Compare the following decontextualized constructed SN-expressions, which alternatively have heaps of and piles of combined with the same set of collocates: (4.93)

Piles of stones/paper/rubbish/dust/ money/ people

(4.94)

Heaps of stones/paper/rubbish/dust/ money/ people

In (4.93) piles is analyzed as a head noun irrespective of the specific N2 following, whereas in (4.94) a quantifier reading of heaps of is just as natural as a literal head noun reading with regard to stones, paper, rubbish, dust and money, even though their referents are prototypically stackable or often appear in the shape of a pile or heap. Moreover, the quantifier reading is even more unmarked with regard to the animate noun people, whereas piles of people tends to call up dead bodies that are placed one on the other. We have seen in the discussion of the rare quantifier uses of piles of that the quantifier reading always relies heavily on context, metonymy or other mechanisms, which makes a pure and context-independent interpretation of piles of as a quantifier almost impossible. I argue that the specific differences in conceptual-semantic make-up in pile and heap discussed at the very outset of this section are either blocking or facilitating factors in the delexicalization of the two SNs. Pile calls up a vertical dimension and the idea of being intentionally constructed by a volitional agent, whereas heap lacks this idea of constructional solidity and layered verticality. Instead, heap is conceptually more vague and unbounded as a concept, and profiles an undifferentiated mass with a more chaotic internal consistency. Heap also has a horizontal meaning component, which makes it possible to use it as a quantifier for things that are scattered in place and/or time. Pile is hence more specific and contains more lexical features than heap, which makes it less conducive to further delexicalization, necessary for systematic grammaticalization. This ties in with general claims in grammaticalization theory that lexical items serving as input for grammaticalization processes typically lack lexical specificity (e.g. Bybee 2003a and 2003b, Heine et al. 1991 and Hoffmann 2004). It is easier to isolate a mere quantity meaning from the less specific heap than from the much more specific pile. Hence, the quantifier uses of pile(s) of do not constitute systematically productive uses supported by delexicalization and at present the lack of semantic generalization attested for pile is matched by a rather restricted collocational extension. We have seen in the ambivalent cases as well as in the head uses that plural piles of has a clear potential for hyperbolic expressivity. Hence, despite its not having a systematic and context-independent quantifier use, piles of may put this potential for hyperbolic meaning to use in the future,

4.3 Head and quantifier construction

157

for instance through the paper(work)-metonymy. In this way piles of may one day be recruited in a process of renewal of hyperbolic quantifier heaps of. I have noted the odd emotively coloured quantifying uses of heap of and pile of, which were not attested for the plural variants. Except for two almost idiomatized examples with heap of shit (e.g. 4.57 and 4.58), this use is still very infrequent and unproductive. In addition, it is mostly based on a readily detectable metaphorical comparison with literal junk. This is why most of these instances with an emotive quantifying function were classified as vague. However, the second developmental path within the general grammaticalization processes affecting SN-constructions precisely leads to a systematic and grammaticalized emotively coloured quantifier use. A more general explanation for why this path mainly includes singular variants of the SN-expressions will be given at the end of Section 4.4.

4.3.2 Lot of and lots of: Head, modifier and ambivalent uses In this section we will discuss lot/lots of. The discussion will be shorter than the previous sections. Tables 4.6 and 4.7 display the relative frequencies of head, modifier and ambivalent uses for lot and lots respectively. Contrary to the other SNs discussed in this book, lot(s) of are more generally acknowledged as emergent means of quantification and especially a lot of is considered to act as a quasi-canonical quantifier across various registers (cf. Biber et al. 1999, Peters and Westerståhl 2006, Langacker 2009 and to appear). Table 4.6. Quantified collocational profile of lot of

Lot of (250) Animate Concrete/count Concrete/uncount Abstract/count Abstract/uncount Total # Total %

Head

Quantifier

# / 1 1 / /

# 72 34 82 18 41

Valuing quantifier # / / / / 1

2 0.8%

247 98.8%

1 0.4%

Ambivalent Ambiguous Vague # # / / / / / / / / / / 0

0 0%

158Chapter 4: SN-constructions Table 4.7. Quantified collocational profile of lots of

Lots of (250) Animate Concrete/count Concrete/uncount Abstract/count Abstract/uncount Total # Total %

Head

Quantifier

# 1 / 1 / /

# 51 96 51 30 20

Valuing quantifier # / / / / /

2 0.8%

248 99.2%

0 0%

Ambivalent Ambiguous Vague # # / / / / / / / / / / 0

0 0%

From these figures it immediately becomes clear that both SNs have an almost exclusive quantifier use. The lexical head noun use was attested only twice for each SN-expression and refers to "an article or set of articles for sale" or the semantically related sense "a group of people or things of the same kind" (cf. OED): (4.95)

The minister said a lot of blood sold in this way was infected (CW-UKbooks)

(4.96)

I used to fetch three lots of milk and I had sixpence a week (CWUKspoken)

This head noun use allows for quantifiers such as three in (4.96) and typically teams up with concrete N2-collocates, similar to the head noun constructions discussed with pile(s) and heap(s). N2-referents are also typically spatio-temporally contiguous in the head noun construction. The predominant use of lot(s) of synchronically is the quantifier use. In this use the collocational range is fairly unrestricted with a preference for animate and inanimate concrete nouns. (4.97)

"Making babies is going to be a whole lot of fun" (CW-UKbooks)

(4.98)

It was a typical English farce with lots of running in and out of doors (CW-UKbooks)

(4.99)

A lot of pensioners now want an inside lavvy (CW-UKmags)

(4.100) Quite a lot of earthquakes are shallow (CW-UKspoken)

4.3 Head and quantifier construction

159

In examples (4.99) and (4.100) the verb agrees with the plural N2collocate, which further confirms the modifier status of lot of. In the quantifier-construction N2-collocates no longer need to be contiguous, but can be. Lot(s) of do not have specific semantic prosodies in their quantifier microconstructions. As referred to earlier, Langacker (2009) and (to appear) discusses a lot of and lots of as becoming important members in the quantifier paradigm. However, he claims that they are not yet reanalyzed as quantifiers because they still appear in a (rare) lexical use, can be premodified by whole in their quantifier use, and have a singular and plural form. In other words, they still behave too much like nouns. Against this, I have argued that layering and partial decategorialization need not detract from grammaticalization claims. Lot(s) also have syntactically extended uses, in which the quantifier meaning functions outside of the NP as a degree modifier of adjectives (4.101), verbs or adverbs (4.102): (4.101) SoHo was a lot different then (CW-UKbooks) (4.102) The white blood cells will recognize it and produce lots more antibodies very quickly (CW-UKbooks) Similar to heap(s), these extended uses of lot(s) appear without the particle of. In addition to extended uses lot(s) also appear in elliptical NPs in which of + N2 are left out but can be retrieved or deduced from the context, e.g. Dr Cheung had asked him if he had a photo of Mom. He'd got lots (CWUKbooks). Lot(s) of also appear in coalesced spellings in synchronic corpus data and Internet data, e.g. alotta, a lotta, a lodda and lotsa. Whereas for most SN-expressions, except bunch of, there is a strong association between the plural form and a predominant quantifier function, lot(s) of function equally strongly as quantifiers. This loss of functionality in the singular/plural contrast can be seen as a kind of partial decategorialization. As also observed by Langacker (2009) and (to appear), lot(s) of can quantify both count and uncount nouns, whereas many and much are restricted to the former and the latter type of noun respectively. This is a quality that all of the SN-expressions have vis-à-vis canonical quantifiers. Unlike the SN-expressions discussed so far, no ambivalent uses were attested for lot(s) of, which can be taken to reflect the high degree of grammaticality associated with these SN-expressions. In Section 4.7 a diachronic case study of SNs will be presented focussing on lot(s) and heap(s). This case study will provide further support for the overall grammaticalization claims for SNs and will compare lot(s) of and heap(s) of in terms of the rate at which they grammaticalized into quantifiers. This diachronic case study will add some complexities to the synchronic story of SN-

160Chapter 4: SN-constructions expressions, i.e. a partitive construction in addition to a lexical head noun and quantifier construction. In addition to quantifier uses, the lot of data set yielded one use that can be interpreted as valuing: (4.103) "Bonus points are a lot of rubbish" (CW-OZnews) In this use the speaker does not so much quantify the amount of rubbish relative to a scale of measurement, but makes an evaluative statement about bonus point being nonsensical. For plural lots of no valuing uses were attested, which is in keeping with the constructional features of this use.

4.4 Towards valuing (quantifier) uses Quantifier constructions with SNs were characterized mainly by an extended collocational range on the ideational level vis-à-vis head noun constructions, and deictic subjectivity. This second grammatical modifier construction is characterized by collocational reclustering on a more interpersonal level and attitudinal subjectivity. This attitudinally subjective character is manifested by specific semantic or emotive prosody patterns that determine the value of these valuing quantifiers. SN-expressions that systematically display valuing quantifier uses, in addition to quantifier uses, are load of and bunch of. Their plural variants have either only developed quantifier uses, in the case of loads of, or do not show grammaticalized attestations in their data set, in the case of bunches of. For bunch of the two modifier constructions overlay to a rather large extent and both meanings are often difficult to separate. The valuing uses of load and bunch differ in certain respects, for instance with regard to the preference for specific sets of collocates and the strength of their valuing use. The valuing (quantifier) uses will also be compared to the (vague) valuing uses of heap of, pile of and lots of discussed earlier. Following the discussion of bunch(es), I will address the issue of premodification patterns in SN-constructions and will compare the SNs looked at in terms of degrees of grammaticality. 4.4.1 Load and loads of For both load of and loads of a random sample of 250 examples was extracted from the Collins Wordbanks corpus. These data sets were quantified according to the same principles as heap of, heaps of, pile of, piles of, lot of and lots of. Data were classified as head, quantifier, and valuing (quantifier) or ambivalent. Within these categories, relative frequency counts were

4.4 Towards valuing (quantifier) uses

161

made of concrete (un)count, abstract (un)count and animate N2s, in order to draw up collocational profiles. Tables 4.8 and 4.9 present these quantified results of the qualitative analyses. Table 4.8. Quantified collocational profile of load of

Load of (250) Animate Concrete/count Concrete/uncount Abstract/count Abstract/uncount Total # Total %

Head

Quantifier

# 6 15 32 / 2

# 26 33 20 16 7

Valuing quantifier # 8 / / / 77

55 22%

102 40.8%

85 34%

Ambivalent Ambiguous Vague # # / 1 1 1 / / 1 / 1 3 3

5 3.2%

Table 4.9. Quantified collocational profile of loads of

Loads of (250) Animate Concrete/count Concrete/uncount Abstract/count Abstract/uncount Total # Total %

Head

Quantifier

# 3 4 8 / /

# 48 50 35 51 48

Valuing quantifier # / / / / /

15 6%

232 93%

0 0%

Ambivalent Ambiguous Vague # # / / 1 / 2 / / / / / 3

0 1%

4.4.1.1 Head uses of load and loads The lexical meaning with which load is attested synchronically is defined in the OED as “that which is laid upon a person, beast, or vehicle to be carried; a burden.” Load has a rather vague constellation, in that it refers to the volume of a load as enclosed by some container and as such does not impose restrictions on N2 in terms of countability. Load of is used predominantly as a quantifier, with pure quantifiers and valuing ones almost equally attested, i.e. 40.8% versus 34%. The colloca-

162Chapter 4: SN-constructions tional patterning in the head noun use is similar for the singular and plural variant, and will therefore be discussed together here. Examples of this are (4.104) to (4.110): (4.104) A plane-load of holidaymakers had to wait nine hours for take-off so turtles on a Greek beach could enjoy a night of moonlit romance in peace. (CW-Today) (4.105) A driver looked death in the face yesterday when her car was crushed by a lorry-load of tombstones. They tumbled from a truck as Josephine Roberts drove through rush-hour traffic in Derby. (CW-Today) (4.106) Residents are worried about increased numbers of heavy transports carrying loads of dangerous materials. (CW-OZnews) (4.107) Thousands of commuters were stuck in a traffic nightmare yesterday after a lorry spilled its load of glue. (CW-Sunnow) (4.108) Six plane loads of food are also being flown today to the city of Baidoa. (CW-NPR) (4.109) Instead he instructed his pilots to concentrate on shooting down the bombers. For it was these aircraft, carrying their deadly load of destruction, that formed the main threat. (CW-BBC) (4.110) It can be unbearable for daughters who inherit a double load of angst- their mother’s doomed expectations, and their own. (CWUKmags) In the head noun use the N2 in the of-phrase specifies what the load consists of, which, as illustrated by examples (4.104)-(4.108), can be expressed by animate, concrete count and uncount nouns. In their head use load or loads can form a compound with nouns depicting the unit transporting the load referred to, e.g. plane-load of holidaymakers (4.104), lorry-load of tombstones (4.105) and Six planeloads of food (4.108). The contexts may also contain other contextual clues that point to concrete loads, e.g. verbs like carrying (4.106), and (4.109), quantitative premodifiers like six (4.108) and double (4.110), and definite or indefinite determiners that function independently and are fully referential with regard to the binominal NP. Verb concord is also systematically with load(s), e.g. (4.108). Load of also appears with two abstract nouns, i.e. destruction in (4.109) and angst in (4.110). Example (4.109) has a concrete referent within the

4.4 Towards valuing (quantifier) uses

163

context, i.e. bombs. Example (4.110) is a metaphorical use of load, referring to a proverbial burden one has to carry, but this does not affect its structural status. The description of the grammaticalized uses and ambivalent cases has to be split up for load of and loads of.

4.4.1.2 Quantifier uses of loads of In its quantifier use the collocational range of loads of extends to all kinds of concrete nouns, both count and uncount, but most significantly, it systematically includes abstract nouns, count and uncount, as well as animate nouns. I will first discuss examples with concrete (un)count nouns under a, then examples with animate nouns in set b and with abstract nouns in set c. a) Extension to more kinds of concrete nouns (4.111) Trump Tower. Donald Trump’s Midtown monument to opulence, extravagantly presented in a pink marbled atrium, with waterfalls and loads of upscale shops. (CW-USephem) (4.112) “Everything was fantastic at first when we met two years agoloads of little love notes and plenty of ‘I love you’.” (CWSunnow) (4.113) I got plenty of chocolates; loads of chocolates. (CW-UKspoken) (4.114) Christmas time I was in the British Home Stores and I tried loads and loads of dresses on. (CW-UKspoken) (4.115) Sixteen natural lakes and loads of mature trees have given them plenty to work on after an un-named businessman put up the money. (CW-Sunnow) (4.116) All the time the local people were trying to sell us cannabis. They think because you’re white folks you must have loads of money. I just wasn’t interested. (CW-Today) (4.117) “The Prodigy movie will feature last year’s live gig in Brixton and loads of footage of the lads messing about.” (CW-Sunnow) (4.118) I bought the house for the simple reason cos it’s got loads of ground to it. (CW-UKspoken)

164Chapter 4: SN-constructions (4.119) “If you’re determined to get rid of your ‘Little Shop of Horrors’ then plaster loads of this thick, creamy mayo all over your hair when it’s dry and before you wash it. Wrap your head in a burban [turban sic] and settle down to look completely silly for about twenty minutes. (CW-UKephem) Within the set of concrete count nouns, the collocational extension is very broad and fairly unrestricted, including shops (4.111), love notes (4.112), chocolates (4.113), dresses (4.114), trees (4.115), but also money (4.116), footage (4.117), sarnies, presents, Easter eggs, syringes, streets and cakes. No specific phenomena of collocational reclustering in the sense of favoured sets of collocates are attested. With regard to concrete uncount nouns there is one favoured collocational set, namely one to do with money or synonyms, e.g. loads of money (4.116), which accounts for 13 of the 35 concrete uncount N2s attested in the data. The semantic prosody is mostly neutral and loads of also has hyperbolic meaning, which can be reinforced by repetition of the SN as in loads and loads of dresses (4.114). Example (4.113) rephrases the quantifier plenty of as loads of, which then suggests that loads of denotes a larger quantity or at least profiles the quantity more emphatically. b) Extension to animate nouns In the head noun use too animate N2s were attested, e.g. a plane-load of holidaymakers (4.104) and four bus-loads of Iraqis (CW-NPR), but unlike these head uses, the quantifier use of loads of does not invoke a literal load of people, carried by a means of transportation (often compounded with the SN-expressions). Moreover, in quantifying animate nouns, loads of shows very clearly that it has delexicalized to a very large extent. Whereas literal loads are always spatio-temporally contiguous, people quantified by loads of need not be contiguous in space or time, though they often may be. This is similar to the quantifier use of heaps of and lot(s) of. The NPs with SNs in examples (4.120)-(4.122) refer to temporally, spatially or spatio-temporally bounded referents, whereas examples (4.123) to (4.125) have referents that are not spatio-temporally bounded. The latter examples especially show the high degree of delexicalization of loads of in its quantifier use, not only backgrounding the more obvious lexical specifications, but also their more abstract implications. Clues to the spatial and temporal boundedness are underlined: (4.120) I’ve always fancied the idea of emigrating as I’ve loads of relatives in Australia. (CW-Sunnow)

4.4 Towards valuing (quantifier) uses

165

(4.121) Seconds later it was surrounded by loads of police and the man was ordered out and searched. (CW-Sunnow) (4.122) I’ve just come home from work. I want a bit of peace and quiet. I don’t need loads of kids and loads of families. (CW-UKspoken) (4.123) I’m sure there are millions of people, millions of bikers, well not millions, but loads of bikers and people wearing exactly what Marlon Brando was wearing. (CW-UKspoken) (4.124) At the other end there was just like a big shelf that you could sit on and loads of people had like scratched their name into their rock. (CW-UKspoken) (4.125) You’ve got loads of people that you can have conversation with, but not many people that you can have communication with. (CWUKspoken) In examples (4.120) to (4.122) verbs or adjuncts of place and time make clear that the referents of the quantified NPs are bounded in time or place or both. In (4.121) there is surrounded by for instance. In (4.120) spatial boundedness is less strict, but still confined to Australia. In examples (4.123) to (4.125), on the other hand, referents are not bounded in space or time. In loads of bikers (4.123) a statement is made about bikers and people all around the world copying Marlon Brando’s sense of style, in a time-span probably starting with Brando’s appearance in films such as The Wild One (1953) and continuing into the present and future. In loads of people (4.124) different people have scratched their name in the rock over time and in (4.125) the people you can have either conversation or communication with are not contiguous in space or time. In these examples too SN-quantifiers are juxtaposed to other means of quantification, e.g. millions of (4.123), many (4.125). c) Extension to abstract nouns The second major collocational extension is towards abstract nouns, a type of noun that was not attested in the head noun use of loads of. In its quantifier use loads of can team up with a wide variety of abstract nouns, both count and uncount. Examples (4.126) to (4.130) have abstract count nouns and examples (4.131) to (4.133) quantify abstract uncount nouns.

166Chapter 4: SN-constructions (4.126) I went off and did loads and loads of interviews. (CW-UKmags) (4.127) How many times have Al Pacino and Robert De Niro acted opposite one another? Loads of times, right? (CW-Times) (4.128) He’s really intellectual and tells loads of clever jokes. (CWUKmags) (4.129) I’ve applied for loads of jobs, including one in a flower shop, but they wanted a woman. (CW-Times) (4.130) I don’t know except she’s been very poorly lately cos she’s had loads of asthma attacks and things. (CW-UKspoken) (4.131) “I’ve been going to bed early, I’ve been getting loads of rest and I’m eating well” (CW-Sunnow) (4.132) Those who must deal with them on a regular basis and wish to get the better of them must employ much patience, deviousness and duplicity, as well as lorry-loads of flattery. (CW-Today) (4.133) Glasgow today has loads of appeal, historic buildings, music, theatre and a rich mix of contemporary galleries and museums. (CW-OZnews) Except for the collocation loads of times, which recurs about ten times throughout the data with abstract count nouns, there are no favoured collocational sets. Loads of also has hyperbolic quantifier semantics, as is very obvious with the repeated use loads and loads of interviews in (4.126). Example (4.128) is interesting to compare with the ambivalent use of piles of jokes in example (4.87), which also featured jokes as N2. Whereas piles did not function as a straightforward quantifier in a systematic way, and was dependent on metonymy and puns, loads of does function as a true quantifier devoid of its original semantics. Lorry-loads of flattery (4.132) might seem to contradict this claim, since it includes a noun referring to a means of transportation, a mechanism also found in head noun uses (e.g. a planeload of holidaymakers in 4.104). However, the meaning here is not that the flattery has to be transported by means of a lorry; rather, lorry functions as an intensifier for the hyperbolic meaning of quantifier loads of (see Section 4.5 on premodification patterns). Hence, even in grammaticalized uses the original lexical semantics of the SN can be alluded to or rekindled, without affecting the grammaticalized status.

4.4 Towards valuing (quantifier) uses

167

Abstract N2-collocates are on the whole negative, e.g. loads of asthma attacks (4.130) and lorry-loads of flattery (4.132), including contextually induced negativity such as an implied large effort on the part of the speaker, as in (4.126)-(4.129). On the other hand, loads of can also appear with positive collocates, as in loads of appeal (4.133) and loads of rest (4.131). Abstract collocates as well as animate nouns clearly show the delexicalization of loads of in its quantifier function. Loads of is sometimes attested in a coalesced form in the Collins Wordbanks data as loadsa. This spelling is also frequently found in Internet data, though only in very informal or spoken data.

4.4.1.3 Ambivalent uses of loads In addition to clear head and quantifier uses loads of also displays ambivalent constructions, more specifically ambiguous ones, though only a very limited number of 3. No vague uses were attested for loads of. The ambiguity in these constructions is typically brought on by the presence of contextual clues foregrounding either the size implications of loads of, or its lexical specifications: (4.134) The program consists of a staff and many volunteers who gather up loads of art supplies and go to transitional housing sites and shelters in Marin, San Francisco, and San Mateo counties to spend time with the young people and provide a safe, creative environment in which kids can express themselves freely. (CWUSephem) (4.135) On his last trip here, just over a month earlier, he had witnessed the murder of an acquaintance, another man who had brokered loads of cocaine, Juan Villegas. Villegas had made a mistake few men in his trade in Medellin made. (CW-USbooks) In loads of art supplies (4.134) the size implications and lexical-conceptual meaning of loads seem to be on a par. The art supplies may constitute actual loads that can be transported, supported by gather up, but loads of may also be interpreted here in its hyperbolic quantifier meaning. In loads of cocaine (4.135) the cocaine probably comes in real loads in ships transported by the Medellin cartel in Columbia, but loads of can also be interpreted as a quantifier in the absence of explicit contextual clues emphasizing the transportation of these quantities of drugs.

168Chapter 4: SN-constructions 4.4.1.4 Syntactically extended uses of loads In the discussion of heap(s) of and lot(s) of, it became clear that their grammaticalized quantifier use has extended beyond the syntactic environment of the binominal SN-construction and into adverbial constructions. Loads of has also extended its quantifier use to other syntactic contexts. In these uses of is typically left out or coalesced into loadsa: (4.136) The Sultans Of Ping are having an off night. This is probably the wrong place and wrong time, and the sound wobbles loads. (CWUKmags) (4.137) With Crass they failed to notice it totally and we're like f ing loads softer than that. (CW-UKmags) (4.138) All of which is tosh, really, but some kind of sign of what The Sultans of Ping have become and how they've gripped on ailing pop-town by its eyeballs, and shaken it halfway to death. Still, I'm sure that The Sultans Of Ping laughed loads. They might be pop stars, but they've always had the best handle in irony. (CWUKmags) The extended uses of loads involve adverbial uses expressing intensity or frequency of a verbal or adjectival predicate, as in wobbles loads (4.136), loads softer (4.137) and laughed loads (4.138). Examples (4.136)-(4.138) were the only adverbial uses attested in the data set, which in this case involved an exhaustive extraction on loads. A quick search on the Internet yields more of these adverbial uses, e.g. Sweden is loads better than Britain (http://answers.yahoo.com/question), he cries loads if you leave him on his own (http://profile.myspace.com/index.cfm) and you dont really wanna go out with some boring old woman who reads loads and acts frigid (www.digitalspy.co.uk/forums). In addition to adverbial uses, the corpus data yielded a set of what seem to be elliptical NPs, which feature a speaker assessing the quantity denoted by loads: (4.139) I don't think there's that many people having loads and loads really. Well I mean loads say nine ten. (CW-UKspoken) (4.140) I hate getting spots 'cos you get a time that you just get loads of spots and I had them ages ago and I've got loads round my face now. (CW-UKspoken)

4.4 Towards valuing (quantifier) uses

169

(4.141) I've been in loads. Yeah. Well tell me all about that then. Well like I've been in Annie. (CW-UKspoken) (4.142) You get so many cards and you can get something. Oh yeah. Mm brand that my dad smokes and he's got loads. He's got two thousand four hundred points and you get ten points in each card. (CW-UKspoken) In these examples of + N2 are elided, but N2 is either explicitly mentioned in the context or can be deduced from it. Canonical monomorphemic quantifiers such as many can function similarly. Finally, whereas the quantifier use and extended uses of heaps of are typically, though certainly not exclusively, associated with Australian English and rather uncommon in American English, loads of is often associated with American English, though not uncommon in British English (see also Biber et al. 1999). The highest percentages for loads of, irrespective of its grammatical status, can be found in UKspoken, Sunnow and UKmags; all British English subcorpora. The highest rate of loads of in an American English subcorpus is USephem, which ranks fifth; the other American English corpora rank tenth and eleventh. If we take a look at just the quantifier uses of loads of, my analysis equally shows a predominance of British English uses, alongside many American English uses and Australian English uses. Regional preferences are not as pronounced as with heaps of it seems. The development of a systematic quantifier use for loads of might be reinforced by its graphological and phonetic resemblance to lots of. The fact that the vowel in loads of can be prolonged makes it a more expressive variant of the latter, with vowel length iconically paralleling the hyperbolic quantification expressed by loads of (p.c. Brian Joseph). We will see in Section 4.7 that lot(s) developed quantifier uses rather early on. The following sections will turn to load.

4.4.1.5 Quantifier uses of load of As mentioned earlier, load of synchronically has quantifier as well as valuing quantifier uses. Its head noun use has already been discussed in Section 4.4.1.1. As we have seen in the discussion of the quantifier uses of heap(s) of, lot(s) of and loads of, the development of a quantifier SN-use typically involves a broadening of collocational range towards abstract nouns and animate nouns compared to head noun constructs. The same goes for the quantifier use of load of, as the numbers in Table 4.8 show. As a quantifier, a load of teams up with all kinds of concrete nouns, both count and un-

170Chapter 4: SN-constructions count, without imposing the collocational requirements it does when it functions as a head noun. Set a subsumes concrete (un)count nouns. Abstract and animate N2s will be discussed under b and c respectively. a) Extension to more kinds of concrete nouns (4.143) I took a load of sleeping tablets and a bottle of vodka and I passed out. (CW-Today) (4.144) As a business you end up sending out a load of Christmas cards which takes a lot of time. (CW-UKspoken) (4.145) Well auditing is just checking the books and records. So I could spend a couple of days picking a load of items on their books tracing them back to find the invoice. Phew. You know. (CWUKspoken) (4.146) This win means I can throw out a lot of my old stuff and get a whole load of new clothes. (CW-Sunnow) (4.147) It’s good because I get Christmas and birthdays at once so I get a load of presents. (CW-OZnews) (4.148) I glued a load of patches to a jeans I had that were totally destroyed. (CW-UKspoken) (4.149) What are you going to write about when you get success and a load of money and get comfortable? (CW-UKmags) (4.150) Make sure you have a room where you can easily set up your machine, otherwise you might keep putting it off and the wife won't want a load of machinery in the front room. (CW-Sunnow). (CWSunnow) Overall the semantic prosody of load of is to some extent negative. Load of is often used in contexts that refer to a big effort, often not painless, which is emphasized, or something is exaggerated to be a large effort, e.g. (4.143) to (4.145). Contextual markers reinforcing this negative semantic prosody are underlined. Load of has hyperbolic meaning, but often there is an inference that the entity quantified by load of is too much, in a negative way, e.g. an overdose in (4.143) and the job of auditing in (4.144). In example (4.150) the amount of machinery is presented as too much or cumbersome. As examples such

4.4 Towards valuing (quantifier) uses

171

as a whole load of new clothes (4.146) and a load of presents (4.147) show, load can also be used in positive contexts, in which the hyperbolic meaning of abundance is appropriate. In (4.146) load of is premodified by whole which further reinforces this hyperbolic meaning. In a load of money (4.149) the speaker sketches a future perspective to his hearer, in which he/she becomes too rich and too comfortable, but possibly in an ironic way. Overall, the semantic prosody of load of skews towards negative semantic prosody, especially with uncount nouns, whereas loads of had a more neutral hyperbolic semantic prosody, except with certain abstract nouns. A possible explanation for this will be given in Section 4.4.1.7. Again, the systematic development of a quantifier use of load of may be supported by its phonetic and graphological similarities with lot of, for which it can function as a more expressive variant (cf. loads of and lots of, discussed in the previous section). b) Extension to abstract nouns (4.151) Stirling blew a barrel-load of chances in a clash they totally dominated. (CW-Sunnow) (4.152) You’ll really want this storming single which blends a nursery rhyme chorus with classic R ‘n’ B beats. It’s a bouncy, catchy pop tune with lively lyrics and a whole load of funk thrown in. (CWSunnow) (4.153) There are a whole load of reasons bandied for why this is happening (CW-UKmags) (4.154) She discovered he'd left her with a five grand phone bill, not to mention a load of other debts (CW-UKbooks) (4.155) There'll be a load of compromising (CW-OZnews) Again we see a predominantly negative semantic prosody in examples (4.151) and (4.153) to (4.155), though a few examples of positive semantic prosody were attested, e.g. a whole load of funk in (4.152). In (4.152) a new single is positively reviewed because of the large variety of genres it displays and a whole load of funk is one of its main qualities. In example (4.151) with a barrel-load of chances we again see that grammaticalized quantifier uses can be compounded in a way that is reminiscent of the kind of premodification found in head uses of load of and loads of e.g. plane-load of holidaymakers (4.104), lorry-load of tombstones (4.105) and Six plane loads of food (4.108), which indicated the means of

172Chapter 4: SN-constructions transportation carrying the literal load. In barrel-load of chances (4.151), however, this does not trigger off a literal head noun reading, but rather works as an expressive intensifier for the quantifier reading of load of. As I did for loads of, I argue that such premodifiers, as well as the occurrence of whole for instance in whole load of funk (4.152) and (4.153), does not detract from the decategorialization claim for these grammatical quantifier uses. Nor does it go against the argument that load and of function as one unit. c) Extension to animate nouns (4.156) I still had a load of friends there. (CW-OZnews) (4.157) When our image first went out of control, we played to a load of skinheads, and this guy came up to me and spat on me. (CWUKmags) (4.158) You’ve got a load of mums who are moaning and groaning about the disadvantage of being left-handed. I’ve been left-handed all my life and it has never been a problem to me. (CW-UKspoken) (4.159) So it's a three three-year waiting list. The doctor does it there must be a load of people waiting for the chop. (CW-UKspoken) Semantic prosody is typically negative with animate nouns, e.g. skinheads (4.157), or the context makes clear that N2 is negatively valued, e.g. mums (4.158). As example (4.156) shows, load of can also be relatively neutral or positive. The examples with negatively coloured N2s are classified as quantifier uses since load of refers to a multitude of (negatively evaluated) N2referents. Just as I noted for loads of, it seems that load of does not have very pronounced regional preferences in its quantifier use. It predominantly occurs in the British English subcorpora of Collins Wordbanks, but again the overall predominance of British English data in the this corpus has to be kept in mind.

4.4.1.6 Ambivalent uses of load In addition to clear head uses and quantifier uses, load of also occurred in ambivalent constructions. Three of these are ambiguous between a head reading and a quantifier reading. A typical example of this is:

4.4 Towards valuing (quantifier) uses

173

(4.160) You can pick up a load of plant pots at a sale very cheaply, and Sylvia would buy her seed at the less popular times of the year. There are bargains around in plants if you know where to look. (CW-UKmags) In (4.160) the verb pick up partly triggers the ambiguity. In the meaning of ‘stumbling upon something’, load of can have a quantifier reading. However, if the meaning of pick up is ‘transporting something’, a load of plant pots gravitates towards a head reading. The data for load of also contained vague constructions. (4.161) It would have to yield to the raven, but few other birds carry a heavier load of sombre myth and legend, or a longer list of local names. (CW-Times) (4.162) When I described her funeral as “a revolting orgy of emotional incontinence and exhibitionism” I braced myself for a containerload of condemnation and abuse. It never materialised. Your letters have been running nine out of ten in support of what I said. (CW-Sunnow) In a heavier load of sombre myth and legend (4.161) both quantifier and head noun meaning are activated by means of contextual cues. A longer list of local names relates to the quantifier meaning, meaning 'the raven has many names', whereas carry and heavier activate the literal head meaning. The global interpretation of the sentence, however, incorporates both these values. In a container-load of condemnation and abuse (4.162), then, the speaker was expecting to be abused and condemned to a very large degree, which is expressed using the compound-mechanism in its intensifier sense. This is combined with the collocational behaviour of a quantifier use, i.e. with abstract nouns as N2. Again, both readings cannot be disambiguated. The global meaning of these sentences involves both layers of meaning at the same time, which distinguishes them from ambiguous examples (cf. Willemse 2007: 26).

4.4.1.7 Valuing (quantifying) uses of load of We have seen in the previous sections that load of as a quantifier often radiates a negative semantic prosody. It was noted that there is some disagreement in the literature as to what semantic prosody is and how it comes about. Many studies on semantic prosody cling to the idea that the affective colouring moves from a specific set of N2-collocates to the node which in

174Chapter 4: SN-constructions itself has a neutral meaning. However, if one digs deeper into the lexical meaning of items, semantic prosody patterns can be hypothesized to be enhanced by the persistence of particular lexical features. In the case of load of these pertain to the fact that the weight constituted by a literal load can be a burden. This evaluative inference is clear in the figurative use of load as a psychological burden or hurdle, exemplified by a double load of angst (4.110). However, this does not explain why this negative semantic prosody is manifested much more strongly in the data for singular load of than it is in the data for loads of. My hypothesis is that singular load of accentuates that all of the referents involved form one group, which facilitates statements generalizing over a group of things or people to the extent of almost equating them with a particular (negative) characteristic or feature. The singularity of load of allows singling out more graphically one set, emphasizing their bondedness in terms of this one negative quality, which gives it more of an emphatic condemning power than plural loads of. This mechanism is probably also at work in the rare (vague) valuing quantifying uses discussed for singular pile of and heap of, e.g. talking a heap of shit to me (4.57), What a heap of shit (4.58) and the heap of junk he calls a boat (4.68). Whereas in the quantifier uses discussed up to now the main function of load of was to quantify things or people, which themselves often proved to be negative in some way, this second grammatical use is more strongly concerned with evaluating its referents. In this valuing quantifier use attitudinal subjectification is at work, which is manifested most strongly in the semantic or emotive prosody patterns, which lead to new collocational clusterings. The valuing use of load of teams up with negative animate N2s on the one hand, e.g. (4.163) to (4.165), and most predominantly with negative abstract uncount nouns, e.g. (4.166) to (4.172): (4.163) “What a load of killjoys. What’s their problem?” The mum-of-two said: “Tell the workers to keep flying the flags- we’re on their side.” (CW-Sunnow) (4.164) "You load of wankers!" (CW-UKbooks) (4.165) A: “Is there anything else they need to know?” B: “No, I don’t think so. Apart from the fact that we’re a load of horrible old drunks.” (CW-UKspoken) (4.166) Do you honestly believe that if somebody kills people, you honestly believe that as a society and God-fearing person, we should look after him. What a load of crap. (CW-UKspoken)

4.4 Towards valuing (quantifier) uses

175

(4.167) A: “Where er I started from was that the notion of capitalism could never give social or environmental justice. Right, therefore you have to explain to people why capitalism is a load of old rot.” (CW-UKspoken) (4.168) A: “I don’t worry, because God‘ll save me.” B: “Right, I mean, to me that’s a load of codswallop.” (CW-UKspoken) (4.169) He asked me to come over to Mayfair Studios. But I didn't interview David Bowie. That's a load of bollocks. We sat and talked, about all sorts, but I couldn't interview him. (CW-UKmags) (4.170) Many people, myself included, were duped into thinking that if you supported the lottery, you were helping charities. What a load of rubbish. (CW-Today) (4.171) Do you believe in the paranormal? I think it is a load of mumbojumbo and if you spend your life thinking about it, you are sad. (CW-Sunnow) (4.172) I don’t believe in Saint Valentine’s day. It’s a commercial load of crap ploy. (CW-UKspoken) The category of valuing quantifier uses is an internally graded category. Some of its members combine the function of evaluation with some quantification, while others are almost purely evaluative with little or no quantificational function left. These functional differences depend in part on the nature of the N2-collocate. When load of evaluates animate concrete nouns, which are typically plural, the reference to quantification is less backgrounded than when it evaluates abstract uncount nouns. Nevertheless, in these valuing examples the SN-expression cannot just be substituted by the schematic quantifier many, which shows that the SN functions differently here. In most cases load of can be left out, which it cannot in the quantifier uses, and serves to intensify the negative value of N2. In we're a load of horrible old drunks (4.165) the speaker uses a predicative construction to describe a group of people in a negative way. The idea of generalizing over an entire group and reducing them indiscriminately to a label such as horrible old drunks ties in with the hypothesis I proposed as to why the singular form load of is especially apt for expressing such evaluative meaning. Examples (4.165), (4.167), (4.168), (4.169), (4.171) and (4.172) all use identifying copular constructions to evaluate something, not, in the first place, to quantify it. In example (4.163) what a emphasizes the negatively valued meaning of killjoys. The same goes for what a load of

176Chapter 4: SN-constructions crap in (4.166) and what a load of rubbish in (4.170). Additional evaluative or emphatic modifiers, such as horrible old in (4.165), old in (4.167), commercial and crap in (4.172) add to this effect. I argue that the negative semantic prosody, already characteristic of the grammaticalized quantifier use and possibly even the head noun use, is systematically exploited in a process of attitudinal subjectification, leading to valuing (quantifying) uses. Attitudinal subjectification, as defined by Traugott (2010b: 35) refers to the semantic change whereby "meanings are recruited to encode and regulate attitudes and beliefs”. Whereas quantifier uses of heaps of and loads of are used to describe “the external described situation” in a usually hyperbolic way, valuing quantifying uses of a load of describe meaning that is “increasingly based in the speaker’s subjective belief state/attitude towards the proposition” (Traugott 1989: 34-35). In its valuing quantifier use load of is not primarily concerned with expressing the quantity of the referent following of, but with evaluating or valuing it in an emphatically negative way. The inferred meaning components which led to quantifying uses have developed their own further inferences to do with evaluation in appropriate contexts, e.g. the what a-construction and identifying copular constructions. This can motivate additional developments, such as this attitudinally subjectified use (cf. Sweetser 1988). Especially in the examples with abstract uncount N2s, load of has an almost exclusive valuing function. The lack of plural marking on the N2-referent defocuses a scale of measurement and instead may aid to shift load of from indicating evaluation of size to evaluation as such.

4.4.2 Bunch and bunches of Just like load of, bunch of has developed a valuing quantifier use in addition to its pure quantifier use. Bunch of is a particularly interesting SNexpression for a number of reasons. Firstly, and contrary to heap, load and even pile, bunch has a more specific lexical meaning, which seems at odds with the general prerequisites of lack of lexical specificity for engaging in delexicalization and grammaticalization (e.g. Bybee 2003a and b and Hopper and Traugott 2003). We have seen that the relatively more specific conceptual make-up of pile versus heap might explain the lack of delexicalization and concomitant lack of grammaticalization of pile. Such lexical persistence does not seem to affect bunch of. Hence other motivations for the grammaticalization of bunch of have to be sought that are sufficiently strong to overcome such lexical-conceptual persistence. Secondly, contrary to the other SN-expressions, it is the singular variant which has grammaticalized the most; the plural variant bunches has an almost exclusive head

4.4 Towards valuing (quantifier) uses

177

noun use. Thirdly, bunch of does not have any ambivalent cases, similar to lot(s) of, whereas bunches of has several. Tables 4.10 and 4.11 display the quantified results of the qualitative analysis for bunch of and bunches of respectively. For bunch of 250 instances were analyzed and for bunches of the extraction was exhaustive. Table 4.10. Quantified collocational profile of bunch of

Bunch of (250) Animate Concrete/count Concrete/uncount Abstract/count Abstract/uncount Total # Total %

Head

Quantifier

# / 26 3 / /

# 83 16 6 19 7

Valuing quantifier # 86 / / 2 2

29 11.6%

131 52.4%

90 36%

Ambivalent Ambiguous Vague # # / / / / / / / / / / 0

0 0%

Table 4.11. Quantified collocational profile of bunches of

Bunches of (49) Animate Concrete/count Concrete/uncount Abstract/count Abstract/uncount Total # Total %

Head

Quantifier

# / 38 8 / /

# / / / / /

Valuing quantifier # / / / / /

46 93.3%

0 0%

0 0%

Ambivalent Ambiguous Vague # # 3 / / / / / / / / / 3

0 6.1%

4.4.2.1 Head uses of bunch and bunches The etymology of bunch is to some extent unclear. The OED hypothesizes that it has onomatopoeic roots, similar to the synonymous bulch, and is related to hunch and lunch. Bulch itself is related to bulge which ultimately derives from Old French boulge, also bouge or Latin bulga, which is a

178Chapter 4: SN-constructions leather knapsack or bag and is of Gaulish origin. Bulge is attested in English from the Middle English period onwards. The oldest meaning of bunch is “a protuberance, esp. on the body of an animal; a hump on the back (of a human being, a camel, etc.); a goiter; a swelling, tumour.” The earliest example of this meaning dates from c1325. The meaning involved in the synchronic head uses of bunch is that of “a collection or cluster of things of the same kind, either growing together, or fastened closely together in any way.” The earliest example of this meaning occurs in 1570. Allan (1977), cited in Lehrer (1986: 118), states that “a bunch is more untidy and less focused than a cluster.” The head noun use and its collocational characteristics are similar for bunch and bunches. They will be discussed together here and separated again for the discussion of quantifier, valuing quantifying and ambivalent cases. The lexical semantics of bunch profile a very specific constellation for the N2 following of. In its synchronic head use bunch(es) are collocationally restricted to a specific subset of concrete (un)count, including grapes, bananas, carrots, flowers (and co-hyponyms), herbs (and hyponyms such as parsley) and keys. In each collocation, bunch(es) are coextensive with these N2s and designates the standard portion of this referent or its prototypical constellatory appearance. Corpus attestations for the head noun use take up 11.6% for bunch, whereas it is the near-exclusive use for bunches with 93.9%: (4.173) A fox, unable to reach a bunch of grapes that hangs too high, decides that they were sour anyway (implied moral: it is easy to spurn what we cannot obtain). (CW-USbooks) (4.174) All stop a moment when Linda, in clothes of mourning, bearing a little bunch of roses, comes through the draped doorway into the kitchen. (CW-USbooks) (4.175) There’s always a fresh bunch of locally grown dill at the market. (CW-UKmags) (4.176) Could you get me three bunches of white spray chrysanths please. (CW-UKspoken) (4.177) The greens all look particularly fresh and succulent: furled, pale green Chinese leaves, deeper green, spoon-shaped bok choi with crisp, ivory stems, floppy bunches of Chinese water spinach and the elegantly fluted leaves of Chinese kale. (CW-Times)

4.4 Towards valuing (quantifier) uses

179

(4.178) Thin developing bunches of grapes by lifting the bunch carefully with a forked stick and nipping out unwanted fruit with long scissors. (CW-Sunnow) Bunch and bunches are often premodified in the head noun construction, and this premodification applies to the binominal NP as such. Premodifiers of N2 make it more specific by referring to a subclass of N2 for instance, as in Chinese water spinach (4.177), or adding a quality to N2. Other markers stressing that bunch refers to a specific constellation of items that takes up a certain amount of space are also underlined. For instance, in (4.178) the bunches of grapes have to be lifted in order to thin them.

4.4.2.2 Quantifier uses of bunch of Table 4.10 shows that the quantifier use of bunch of involves major collocational extensions towards abstract nouns, and most prominently towards animate nouns. In its quantifier use bunch has delexicalized in that its size implications are foregrounded and its specific constellation meaning is backgrounded. Whereas its collocates were restricted to a very specific subset of tangible concrete nouns in the head use, the quantifier bunch of can appear with concrete nouns of various semantic kinds. Examples (4.179) to (4.183) illustrate this quantifier use with concrete nouns, count and uncount. a) Extension to more kinds of concrete nouns (4.179) Ned wanted to give me a bunch of suits. (CW-NPR) (4.180) I can go down any street in virtually any city in America and show you a whole bunch of help wanted signs. There is no shortage of jobs. (CW-NPR) (4.181) When flooded roads on the Nullarbor stranded a bunch of trucks at Norseman the operator of the local servo reckoned his magazine rack was in for a bit for a pounding. (CW-OZnews) (4.182) They got a whole bunch of money from the agencies. (CWUKspoken) (4.183) People come and see a whole bunch of work, and it looks as though I work quickly. CW-UKmags)

180Chapter 4: SN-constructions Collocates such as suits (4.179), help wanted signs (4.180) and work (4.183) do not appear in a literal bunch. Instead the speaker wants to indicate that there are a number of them. As example (4.180) shows, the spatial contiguity between the various help wanted-signs need not be very strong, though the street does impose boundaries. Overall, referents quantified by means of bunch of seem to be relatively more spatially and temporally bounded than heap(s) of and loads of, for instance, which were discussed earlier. A heap of pole positions (4.28), a whole heap of other good races (4.29), heaps of tournaments (4.30), heaps of praise (4.36), loads of bikers (4.123), though these earlier examples mostly involve either animate or abstract N2-referents. There also seem to be regional differences in the degree of grammaticalization of bunch of. American English appears to allow a more delexicalized use than other English variants. American linguist Langacker (2009 and to appear), for instance, argues for a strongly delexicalized use of bunch of, whereas Australian Paul Thibault argued for a less delexicalized use. As discussed earlier, according to Paul Thibault (p.c.) the example a bunch of rivers can only be used felicitously when the rivers form a sort of delta of rivers that are spatially contiguous and bounded. Reference to rivers scattered across the continents would be infelicitous. Independently from this statement, Langacker (to appear: 12) argues for the opposite when he considers He has a bunch of {friends/hotels/investments} scattered about as acceptable. b) Extension to abstract nouns (4.184) There’s now a whole bunch of studies from different cities that show the same thing. (CW-NPR) (4.185) The Malaysian-built Proton is the newest in a bunch of marques trying out for a slice of the Australian market. (CW-OZnews) (4.186) It’s extremely discouraging when we find the key witness in a whole bunch of our cases is an officer who has been dismissed from the force because he is a thief. (CW-OZnews) (4.187) Hey, it’s not about heroin, okay? It’s about going through a serious bunch of feelings. It’s what we do, man; we get a lot of feelings out. (CW-UKmags) (4.188) I don’t just teach them a bunch of facts; I’ll work with them on how to write a grant proposal because if you don’t get grant money you’re sort of out of a job real fast. (CW-UKbooks)

4.4 Towards valuing (quantifier) uses

181

(4.189) Here's a whole bunch of aid so you can buy you can grow cotton and coffee and tobacco as opposed to growing wheat and getting your agriculture and that kind of thing sorted out. (CWUKspoken) (4.190) I spent a bunch of time when I was visiting the country talking to his neighbors and so forth including a man named Alan Jepson, who was the link between the family Weaver and the community. (CW-NPR) The collocates in these examples indicate the high degree of delexicalization in the SN-expression. These examples with abstract N2s show that the features of spatial and temporal contiguity in the meaning of bunch in its quantifier use can be backgrounded to a certain extent, though not as strongly as with heaps of and loads of. In a whole bunch of studies (4.184) the studies are scattered spatially over several cities. In a whole bunch of our cases (4.186) the cases are probably temporally and spatially noncontiguous. Just as the quantifier uses with concrete nouns, the SN-expression is sometimes premodified, mostly by whole, as in a whole bunch of studies (4.184), a whole bunch of our cases (4.186) and a whole bunch of aid (4.189), or serious in (4.187), which is rephrased as a lot of feelings. This premodification pattern serves to reinforce the quantifier meaning of the SN-expression, though serious in (4.187) may also pertain to the feelings. Examples (4.179) to (4.190) tend to attach an air of nonchalance or cynicism to the description of a quantity of something that is casually dropped or posited somewhere, e.g. a bunch of trucks in (4.181) and a whole bunch of aid in (4.189). Examples (4.188) and (4.189) are most explicit in this regard. In a bunch of facts (4.188) the speaker stresses that he/she teaches people important and pertinent knowledge and not merely a random set of facts; just stresses this. In a whole bunch of aid (4.189) the speaker feels that aid offered to third world countries is not always well thought-out or explained to those who need it. This kind of slightly querulous or casual tone is present in most of the quantifier uses, though not always as clearly and not always with negative implications. The second major extension in the quantifier use of bunch of is towards animate nouns. This extension is also numerically the most important one, with 83 of the 131 attested quantifier uses teaming up with animate nouns. The casual air associated with bunch of, discussed with regard to the examples cited above, is borne out more strongly with an animate noun as N2:

182Chapter 4: SN-constructions c) Extension to animate nouns (4.191) A bunch of drunken, brain-dead louts seem determined to disgrace our team and our country. (CW-Sunnow) (4.192) We guarantee the noble young lord will complain about having to spend time with such a boring bunch of geriatrics. (CW-Today) (4.193) I turned things around, despite taking over a bunch of players who were clearly not fit or organized enough. (CW-Sunnow) (4.194) Deng was pictured taking a dip with a bunch of his beaming buddies at a summer resort in the north of the country. (CW-Today) (4.195) Now, anybody passing through Dario’s hands is taught the game thoroughly; besides the fact that he has turned a bunch of nohopers into a proud club, with growing support. (CW-Today) (4.196) Isn’t that pretty painful stuff to express in front of a bunch of strangers? (CW-UKmags) (4.197) It’s the most talented side in Europe and I feel I can still win more big prizes with a bunch of skilful youngsters. (CW-Today) (4.198) Andy Roxburgh bumped into a bunch of UEFA officials here yesterday. (CW-Today) (4.199) I should add that today a whole bunch of top Clinton economic advisers flew in to Chicago. (CW-NPR) (4.200) It was alongside the Isar that a few hundred years ago a bunch of wandering monks found a perfect spot to picnic and then stayed to found the city of Munich. (CW-UKephem) (4.201) The only people on this street are a bunch of kids playing marbles in an empty lot. (CW-NPR) The semantic prosody is mostly strongly negative, e.g. louts (4.191) and no-hopers (4.195), and may be emphasized by the inclusion of negative evaluative elements, as in drunken brain-dead in (4.191), boring in (4.192), not fit or organized enough in (4.193). A bunch of wandering monks (4.200) is more ironic than negative in its humorous recounting of a rather serious event like the foundation of a city. In a boring bunch of geriatrics

4.4 Towards valuing (quantifier) uses

183

(4.192) and a bunch of his beaming buddies (4.194) the expressive power of bunch is further reinforced by alliteration. Bunch of can also occur with more neutral collocates and even some positive ones, e.g. a bunch of skilful youngsters (4.197), a bunch of UEFA officials (4.198), a whole bunch of top Clinton economic advisers (4.199) and a bunch of kids (4.201). The predominantly negative semantic prosody seems to derive from a feature of the original cluster meaning referred to earlier, namely its unruly nature. Whereas in the literal head use the unruliness was part of the constellation, it pertains to people’s morals or character in this quantifier use. It is only with regard to children that this unruliness is positively evaluated as youthful vivaciousness. In its quantifier uses with animate referents, bunch of, just like the quantifier uses with concrete and abstract nouns, still seems to impose some spatial or temporal contiguity, e.g. a bunch of his beaming buddies (4.194), a bunch of UEFA officials (4.198), a whole bunch of top Clinton economic advisers (4.199), a bunch of wandering monks (4.200) and a bunch of kids (4.201). Some spatial or temporal latitude is possible, however, as in turned a bunch of no-hopers into a proud club (4.195), which suggests progress through time. The fact that bunch of still shows restrictions in most contexts and dialects of English is probably due to its source semantics, which are much more specific than those of load or heap and include the feature of ‘clustering’.

4.4.2.3 Valuing quantifying uses of bunch of Especially with regard to animate N2s the distinction between quantifying and valuing quantifying uses is not always clear-cut for bunch of. The main functional difference is that the valuing use does not serve to quantify (negatively coloured) N2-referents, but to value typically people, mostly in a negative way, without explicitly quantifying them. The valuing use is based on the semantic prosody discussed for the quantifier use, foregrounded here through attitudinal subjectification to the point that the quantificational semantics are backgrounded or lost. The predominantly negative semantic prosody also fuels a collocational reclustering in keeping with this. As Table 4.10 shows, the valuing use is most frequent with animate nouns, i.e. 86 of the total 90 attestations. In addition, it occurs with a few abstract count and uncount nouns. In the valuing uses with animate nouns the semantic prosody is mostly negative, except in several uses referring to children or a group of friends. In such cases of positive semantic prosody, another feature than unruliness is foregrounded, namely the feature of constituting a close-knit group. Examples (4.202) to (4.214) feature negatively

184Chapter 4: SN-constructions coloured animate nouns in specific valuing contexts, whereas (4.215) to (4.217) are examples with positively valued collocates: a) Animate nouns (4.202) In Newcastle a 42-year-old saleswoman summed up her colleagues' resentment about British Gas Management: “I think they are a bunch of greedy bastards lining their own pockets.” (CWToday) (4.203) Who said Americans were a jingoistic bunch of rednecks who know or care nothing about what happens beyond their shores. (CW-Today) (4.204) Your editorial calling for warnings in the brochures of non-ABTA travel agents makes holidaymakers look like a bunch of morons. (CW-Today) (4.205) Warts-only copperdom, presenting the police as a bunch of hamfisted dimwits. (CW-Today) (4.206) They’re behaving like a bunch of idiots. (CW-Times) (4.207) Money was what was rotting the moral fiber of this country, turning all those yuppies into a bunch of pansies. (CW-USbooks) (4.208) He was dismayed by the foot dragging of the chiefs, whom he likened to a bunch of fucking municipal bond salesmen. (CWUSbooks) (4.209) He suspected that all the politicians were a bunch of phonies. (CW-USbooks) (4.210) You might think we sound like a bunch of arseholes, but we have a lot of faith and commitment in everything we do. (CW-UKmags) (4.211) CIA, KGB, MI6. The ideologies may be different, but you’re all a bunch of lying, treacherous bastards when it comes down to it. (CW-UKbooks) (4.212) When I’m in England, I get a go at their snobbery, the Irish get it for their drunkenness and the Welsh are just a bunch of sheep. (CW-Sunnow)

4.4 Towards valuing (quantifier) uses

185

(4.213) I’ve met quite a few Lotto winners over the years. And most of them are a right bunch of misery guts. (CW-Sunnow) (4.214) This time I was stuck in my car at a junction in town for nearly half an hour whilst the usual hair-brained bunch of sweaty saddos staggered past on a one-way trip to an early grave. (CW-Sunnow) (4.215) Look we’re reforming, aren’t we a bunch of great guys? (CWNPR) (4.216) We have another three games and that will be a real test but they are a smashing bunch of lads to work with and they keep on going. (CW-Sunnow) (4.217) I’m racing with a great bunch of guys and it makes racing fun again. (CW-UKmags) As I noted for load of, the valuing use of bunch of typically occurs in a predicative or comparative construction of the form: ‘X is/behaves (like) Y’ or ‘X is presented like Y’, where X refers to an institution, or group of people and Y equates them with negative characteristics, often reinforced by adjectives premodifying the SN or N2. In they are a bunch of greedy bastards (4.202) people in the British Gas Management are evaluated as being greedy bastards, in a jingoistic bunch of rednecks (4.203) Americans are characterized as jingoistic rednecks, in a bunch of morons (4.204) holidaymakers are likened to morons, etc. The valuing use can be reinforced by intensifying premodification of the SN, as in a right bunch of misery guts (4.213). It is only in examples (4.215) to (4.217) that people are presented in a positive way, emphasizing that they form a close-knit group with good internal relationships. This use was only found with regard to male bonding. Premodifying adjectives are accordingly positive here and can occur in front of the SN-expression or N2. Great can occur in both positions as shown by (4.215) and (4.217). This is in keeping with Halliday’s (1985a: 174) observation that qualitative adjectives can be transferred in expressions such as a strong cup of tea, where the adjective strong still pertains to N2. When such adjectives are in front of N2, as in a bunch of great guys (4.215), the quantifier semantics are less backgrounded than when they precede the SN, as in smashing bunch of lads (4.216) and great bunch of guys (4.217).

186Chapter 4: SN-constructions b) Abstract count and uncount (4.218) What we show is the truth, Luciano Benetton said. Traditional advertising pictures are a bunch of lies. (CW-Today) (4.219) Trouble was, the funds were able to neatly hide all but the most conspicuous of their charges in a bunch of legalese. (CWUSbooks) (4.220) What a bunch of gobbledygook. (CW-OZnews) There are not many attestations of valuing quantifying uses with abstract N2s, but they are very clearly valuing. A bunch of lies (4.218) is not about quantifying an amount of lies, but aims at evaluating traditional advertising pictures as lies. A bunch of serves as an evaluative element here and can as such can be left out. For instance in legalese has roughly the same value as bunch of legalese in (4.219). It serves to enhance the emphatically negative value of N2. These examples feature the same constructional environments as observed for the valuing uses with animate nouns, namely predicative constructions, e.g. traditional advertising pictures are a bunch of lies (4.218), or emphatic constructions such as what a bunch of gobbledygook (4.220). The semantic prosody is again negative, with a majority of N2s referring to nonsense, as in a bunch of legalese (4.219) and a bunch of gobbledygook (4.220). If we compare the valuing use of bunch of with that of load of, as well as with vague valuing uses of pile and heap of, it is bunch of and load of that show most similarities in terms of their extension of collocational range and the productivity and frequency of this use. Heap of is restricted to the collocate shit in my data set and pile is restricted to metaphorical comparisons with creative or pseudo-artistic referents and only occurred 4 times in my data set. Bunch of and load of can both team up with various categories of N2s, though there are several differences. Bunch of can evaluate animate nouns as well as count and uncount abstract nouns, whereas load of only appears with animate and abstract uncount nouns in my data. Within these sets of collocates, bunch of moreover predominantly evaluates animate nouns (i.e. 86 out of 90 attestations), e.g. (4.202)-(4.217), whereas load of significantly favours abstract uncount nouns (i.e. 77 out of 85 attestations), e.g. (4.166)-(4.172). There is one sentence in my data set that contained both bunch of and load of in their valuing use with an animate noun. Bunch of killjoys is rephrased as load of killjoys (partly cited earlier as 4.163):

4.4 Towards valuing (quantifier) uses

187

(4.221) “They’re a bunch of killjoys”. Locals were outraged at their council’s ban. West Ham fan Sonia Davies, 27, said: What a load of killjoys. What’s their problem?” (CW-Sunnow) No grammaticalized examples were attested for bunches.

4.4.2.4 Ambivalent uses of bunches Table 4.11 shows that no quantifier uses were attested for plural bunches in the exhaustive extractions from the Collins Wordbanks corpus. Bunches does display several ambiguous uses that are in between a pure head and a true (valuing) quantifier reading. These examples involve animate nouns: (4.222) They had bunches of erm they had bunches of skinheads standing round on the stage going “Sod off”. (CW-UKspoken) (4.223) That two different bunches of hunters, two thousand years apart, would work the same quarry and then carry the stuff two hundred miles to work on it. (CW-USbooks) (4.224) What a long, endless, teeth-grinding, nostril-flaring, fistthumping-into-palm-of-other-hand-like-Robert-De-Niro week of listening to those two bunches of spoiled, pampered brats whinge their brains out about which is the hardest done by. (CW-Sunnow) In two different bunches of hunters (4.223) and those two bunches of spoiled, pampered brats (4.224) the cluster meaning seems to be the strongest meaning. The examples refer to distinct groups of people and not just a ‘lot of people’, which is emphasized by the premodifiers two different in (4.223) and those two in (4.224). Yet, other contextual clues, such as evaluative premodification of N2 in (4.224), as well as the animacy of the N2s, which is typical of (valuing) quantifier uses, render these examples ambiguous. In bunches of skinheads (4.222) the skinheads are primarily profiled as forming groups of people, but by implicature this also refers to ‘many skinheads’. Whereas the human referents in bunches of skinheads (4.222) are spatio-temporally contiguous, the total amount of hunters in (4.223) are not, as the temporal adjunct two thousand years apart shows, but each group separately is. The same seems to go for (4.224). We have seen that the loss of such more grammatical implicatures of the original lexical semantics is typical of straight quantifier readings of SNs. Rather than referring to ‘a large heap or load’, as the plural variants heaps and loads do, the cluster meaning of bunches appears to persist and designates

188Chapter 4: SN-constructions ‘many different clusters’ instead of ‘one big cluster’. This difference will be returned to in Section 4.6, where the degree of grammaticalization of the various SNs in this study will be compared.

4.4.2.5 Syntactically extended uses of bunch (of) and buncha The grammaticalized quantifier meaning of bunch of has extended outside the binominal construction, though only to a very limited extent in the Collins Wordbanks data. I found two adverbial uses with the formula thanks. (4.225) There's even a new language, which in Jeremy Sams' translation includes such poetic phrases as ‘Thanks a bunch’. (CW-Today) (4.226) “Perhaps you could suggest you're too old,” he offered sullenly. “Thanks a bunch!” She was about to laugh until she realized he might have a point. (CW-UKbooks) It should be noted that these extended uses were not found in the exhaustive data set extracted on bunch as singular common noun, i.e. bunch/NOUN, which did not yield any extended uses. Instead they emerged through queries searching for specific environments, such as VERB+a+bunch. Queries looking for adverbial uses attested for other SN-expressions such as heaps and loads used with adjectives and adverbs, did not yield any results. By contrast, on the Internet several adverbial uses were found, for instance with more, as in (4.227) or with comparative adjectives, such as nicer in (4.228), and with other verbs such as help in (4.229): (4.227) Here’s a bunch more MOSS-based Internet facing websites that have recently gone live (go here for a list of others mentioned in the past) (http://blogs.msdn.com/ecm/archive) (4.228) Now the game configurations have a much nicer format, and also I've added support for custom ui skins ;) (not to mention the code is a whole bunch nicer this time around. (http://slade.mancubus. net/) (4.229) SQL code to join the tables would help a bunch. (http://www. programmersheaven.com) In example (4.229) whole intensifies this adverbial degree modifier use, as it often does in the regular quantifier uses in the binominal NP.

4.4 Towards valuing (quantifier) uses

189

Five attestations of coalesced spellings for bunch of were found in the Collins Wordbanks data, i.e. buncha. Four of these derive from the USbooks subcorpus and one from UKmags. This may tie in with the regional differences observed for the degree of delexicalization of bunch of. The attested coalesced spellings all seem to express the valuing use of bunch of: (4.230) Of course, none of this matters to the packed house, who whoop and yell from the rafters with unbridled enthusiasm and, even when singer Whitfield Crane informs them that they are a buncha f in' dicks they still cheer approvingly and crush even closer to the front. (CW-UKmags) (4.231) “No objection from Fish and Game?” “Not since my last citation for jumpmeat venison.” “What a laugh.” “How'n hell should I know? Buncha gobbledegook, you ask me.” (CW-USbooks) A buncha f ing dicks (4.230) has a human N2, whereas Buncha gobbledegook (4.231) has an abstract uncount noun. Both examples, as well as the other examples not cited here, clearly derive from very informal registers, or cite spoken language, as in (4.230). In conclusion, it can be noted that bunch of has grammaticalized to a very large extent. 88.4% of the attestations are grammaticalized uses, either quantifier or valuing (quantifier) uses. Quantifier uses prevail slightly over the valuing uses, with 52.4% versus 36% uses respectively. Bunch of did not have any ambivalent cases in this Collins Wordbanks data set. This dichotomization between lexical head uses and grammaticalized uses can be seen as a sign of strong grammaticalization and delexicalization. Bunch of is, moreover, also attested in coalesced spelling, which is usually associated with advanced grammaticalization. Its grammaticalized meaning also shows up outside of the binominal SN-construction, which, with Himmelmann (2004), is taken to be another sign of grammaticalization. Specific to the developments attested for bunch of is the fact that grammaticalized quantifying and valuing semantics often overlay each other, especially with human N2s. The valuing use was argued to arise through attitudinal subjectification of the unruliness component in the source semantics of bunch and to be reflected in the collocational reclustering of specific semantic prosodies. Plural bunches shows hardly any grammaticalization in my data set: 93.9% of its attestations are head uses and only 6.1% are non-heads, all of which are ambivalent uses. This goes against the trend observed for heaps of and loads of, namely that the plural variant is typically more grammaticalized than its singular counterpart. For the latter, this was explained by the plurality marker being conducive to a hyperbolic quantifying meaning.

190Chapter 4: SN-constructions In fact, in the case of bunch two facts need explaining, which are potentially interrelated. On the one hand, there is the fact that singular bunch has grammaticalized considerably, and more so, than plural bunches. On the other hand, one might wonder how bunch, with its very specific source semantics, came to grammaticalize so strongly, since lexical specificity is typically considered to be a blocking factor for grammaticalization processes (e.g. Bybee 2003a and b, Hoffmann 2004). One reason why bunches of might be less suited to grammaticalize has to do with its prosodic features, which differ from the other plural SNexpressions, namely the fact that it consist of two syllables, rather than one (p.c. M.A.K Halliday). The precise relation between syllable structure and grammaticalization has to my knowledge never been studied explicitly and may differ depending on the function a particular word or strings of words grammaticalizes towards. The influence of syllable structure on grammaticalization and delexicalization is an avenue of research that merits more attention in future grammaticalization studies. Bunch of developed a quantifier use with a predominantly negative semantic prosody, in the sense that it at least refers to a quantity of something in a casual way, and often refers to it as being quantité negligible, frequently preceded by just/merely/nothing but/etc. It does not really have hyperbolic quantifier semantics in the way that heaps of or loads of have, but does occur in intensifying or emphatic contexts. This is particularly clear in its attitudinal or valuing (quantifying) uses (with animate referents). The actual quantity designated by bunch is not hyperbolic as such, and varies, but seems to refer to a good-sized amount of something, ‘somewhat more than could be reasonably expected’ and ‘no lack of’. (4.232) is a metalinguistic example that illustrates what the speaker at hand considers to constitute ‘a bunch’: (4.232) “She kissed you, huh?” “A bunch of times.” Upon consideration, it seemed to Peter that four times could be termed ‘a bunch’. “What was it like?” “I don't know.” (CW-USbooks: Barn Blind by Jane Smiley 1994) The reason why singular bunch might be better suited for expressing the grammaticalized meanings it does, is probably also due to its source semantics. It was noted for the (ambivalent) valuing uses of a heap of, a pile of and a load of that the singular form allowed one to demarcate a group of people and apply a negative evaluation to that group. In the case of bunch this also comes into play. The semantic prosody of bunch of in its valuing use seems to have facilitated this development. By contrast, plural bunches seems to have developed only the meaning of different separate clusters, which do not give way to one large cluster

4.4 Towards valuing (quantifier) uses

191

and, by implication, one large amount. The source meaning seems to persist more in the plural bunches, whereas one bunch in the sense of ‘a cluster of things’ does seem to allow the inference of ‘a number of things’. I hence want to argue that pragmatic factors and semantic prosody are very important factors that facilitate the grammaticalization and delexicalization of bunch of, despite its lexical specificity. This probably also explains why the quantifier and valuing uses of bunch of overlay each other so much. The plural forms of heap and load add plurality to the semantic generality present in both concepts and intensify their size implications, unhampered by specific lexical properties that might block delexicalization. Plural bunches, however, seems to accentuate the original cluster meaning and pluralizes the latter, rather than being conducive to the inference of size implications.

4.5 Premodification patterns and decategorialization: A case for collocationally constrained constructions In the previous discussions of the grammaticalized and nongrammaticalized uses of SN-expressions it was pointed out that the SNexpression as well as N2 can be premodified. In the head noun uses this is unproblematic, but in grammaticalized uses such premodification might be considered to go against the decategorialization argued for grammaticalized SN-uses. One of the aims of this book is to show the importance of systematically studying premodification patterns in arguing for the grammaticalization status of changes in binominal constructions. Can SNs be argued to have lost in nounness, when they can still be premodified by adjectives, premodifying nouns, various determiners, as well as appear in the plural? However, with regard to the latter it was noted in the individual discussions that quantifier uses of SN-expressions showed significant differences in degree of grammaticalization between the singular and plural variant of each SN. For SN-expressions with a predominant quantifier use, the plural variants have grammaticalized the most, e.g. heaps of and loads of. The only exception is bunch of, discussed earlier, which has grammaticalized only in its singular form and is restricted in that sense. Lot(s) of have grammaticalized equally. Hence, variability in number has become restricted for most SNs looked at in this study, which is a sign of at least a degree of decategorialization, since the quantifier use is restricted to constructions with either the plural or singular variant. The valuing quantifier construction, on the other hand, is restricted to singular load and bunch of. Considering the fact that SN-constructions participate in an ongoing process of grammaticalization, it is not surprising that grammaticalization diagnostics such as decategorialization or coalescence do not apply in an absolute way. Hopper (1991), Heine (2003) and Hopper and Traugott

192Chapter 4: SN-constructions (2003) all note that many of the traditional parameters as described by Lehmann (1985) pertain to characteristics of very advanced or final stages of grammaticalization and fail to diagnose ongoing or initial grammaticalization. Grammaticalization processes that are ongoing show variation with regard to these parameters (cf. Hopper 1991). With regard to the choice of determiner, it can be noted that in the quantifier use there is a strong preference for the indefinite article. The definite article in the SN-string is mainly used with specific premodifiers or with certain pointing uses of the definite article, which are contexts in which more schematic quantifiers such as much/many/few/little also allow a definite determiner. For instance in tickets for the heaps of other events (4.31) heaps of can easily be replaced by many, i.e. the many other events. In most of the examples the definite article or demonstrative determiner was restricted to head noun uses of the SN, in which case the definite determiner has phoric reference, as in (4.233), cited earlier as (4.18), or fulfils other functions that the definite determiner can fulfil: (4.233) The synagogue of the Hadassah Hospital is lit by the glow through Marc Chagall's 12 stained-glass windows representing the Sons of Jacob. When they were shattered in the 1967 war he had the heap of glass sent to his studio and put them together again, leaving a single, symbolic bullet hole. (CW-Times) The definite article also showed up in a number of ambivalent constructions, in which this determiner is precisely part of the ambivalence, e.g. (4.234), cited earlier as (1.19): (4.234) The British have forged a fine tradition of gardening and cannot afford to sit on their well-clipped laurels. Striding past the compost heap of nostalgia comes Christopher Lloyd, now in his seventies and still coming up with new ideas for his garden at Great Dixter in Sussex. (CW-Times) There are some valuing uses of load of and bunch of for instance that also take the definite article, but this is because of the superlative premodifier in front of the SN-expression as in (4.235) and (4.236), or a definite genitive, as in (4.237): (4.235) PATRICK'S letter was perhaps the biggest load of issue-avoiding toss it has been my displeasure to read in a long, long time. I wonder if by any chance he practised it regularly in sociology class? (CW-UKmags)

4.5 A case for collocationally constrained constructions

193

(4.236) The zaniest bunch of dog-like kids, the Warner Brothers -Wakko and Yakko -- and Warner Sister, Dot, are the number one cartoon show in the US and now they're coming to your home. (CW-UKephem) (4.237) On no, it's that time again. Michael Ball sings the UK's load of old tosh, while assorted Europeans drag their nations' reputations through the mire with their preposterous concoctions. In other cases, such as (4.238), which contains a quantifier use of bunch of, the definite article seems to serve as an anaphoric determiner, referring back to the bunch of musicians referred to at the beginning of the sentence. Again this is a use that is allowed for more schematic quantifiers or collectives as well. (4.238) A bunch of musicians larking about creating merry havoc in a studio is usually only of interest to the bunch of blokes who were larking about in the studio. (CW-UKmags) In sum, there are restrictions with regard to the determiners that can occur in the quantifier use of SN-expressions. In the case of singular SNexpressions, the indefinite determiner prevails by far, except for contexts with premodifiers demanding a definite article or in ambivalent constructions. For the plural SN-expressions, the zero-determiner is most frequent, except again in specific contexts demanding a definite article, or in ambivalent constructions. Several more schematic quantifiers, such as many/much, few/little also allow uses with both definite and indefinite determiners, though in a more restricted way, e.g. a great many friends; the many/few people that came; a little research has been done; the little research that has been done; a few people were arrested. More important perhaps than restrictions on determiner use are the premodification patterns attested for SN-constructions. Premodification patterns include the potential premodification of both the SN and N2. These patterns have to be discussed in detail in terms of which kinds of premodification are allowed in the various uses of SN-constructions. An important question in this context is the difference between premodification patterns in non-grammaticalized head, ambivalent and grammaticalized uses. An additional question is the difference in premodification patterns attested for the quantifier uses and valuing uses of SN-expressions and potential microconstructional, i.e. SN-specific, premodification patterns. I will argue that the head and two modifier uses of SNs are each characterized by distinct pre- and postnominal collocational patterns between the SN and premodifi-

194Chapter 4: SN-constructions ers and the SN and N2s, and hence have to be studied as collocationally constrained constructions. Some observations have already been made in the individual discussions of the SN-expressions, but I will now more systematically discuss the premodification patterns attested for head, ambivalent and grammaticalized uses, both quantifier and valuing-quantifier. With regard to the latter two, I argue that there are two types of premodification pattern, each of which supports the grammaticalized function of the SN-expression and hence does not go against the decategorialization claim. Let’s first look at the premodification patterns observed for the head noun uses of SN-constructions, as in (4.239) to (4.245): (4.239) Could you get me three bunches of white spray chrysanths please. (CW-UKspoken) (4.240) All stop a moment when Linda, in clothes of mourning, bearing a little bunch of roses, comes through the draped doorway. (CWUSbooks) (4.241) Ford the Frugal fretted over the mounting heaps of wood scraps from his car factory. (CW-OZnews) (4.242) A jilted girlfriend got revenge on the boyfriend who dumped her by dumping a foot-high pile of manure in his bed. (CW-Today) (4.243) Mike says he's always been domesticated. Spotless kitchen surfaces and neat piles of freshly-washed laundry bear testimony to this. (CW-Today) (4.244) There’s always a fresh bunch of locally-grown dill at the market. (CW-UKmags) (4.245) I happened to be on the Kuwait-Iraq border late Tuesday evening when I saw four bus loads of Iraqis, Yemenese, Sudanese, Palestinians expelled into Iraqi territory. (CW-NPR) In the head noun construction both SN and N2 can take all kinds of premodification, mostly qualitative or classifying adjectives, quantifiers or classifying nouns. The modifier in front of the SN applies to the entire binominal NP and hence also to the coextensive N2 in the of-phrase. In example (4.239) three quantifies the number of bunches of chrysanths; similarly in (4.240) little applies to the size of the bunch of roses, not the roses

4.5 A case for collocationally constrained constructions

195

themselves. In the case of load(s) we find compounding nouns designating a means of transportation, e.g. bus loads in (4.245). Modifiers in front of N2 describe more precisely the entities making up the bunch, heap, pile, lot or load and apply to N2 specifically. They usually cannot be paraphrased with the premodifier applying to the SN when the premodifier is a classifying element as in: (4.246) Three bunches of white spray chrysanths vs. *Three white spray bunches of chrysanths A bus-load of Canadian majorettes load of majorettes

vs. *A Canadian bus-

In the case of qualitative adjectives such a transfer might seem more felicitous (see Halliday 1985a: 174), yet it often changes the semantic interpretation of the entire construction: (4.247) Untidy heaps of muddy or wet footwear muddy or wet heaps of footwear A bunch of red carnations bunch of carnations

vs.

*Untidy

vs.

*A red

Such transfer from pre-N2 to pre-SN position is possible in other cases, such as a strong cup of tea (Halliday 1985b: 174). In sum, the premodification of SNs and N2s in head SN-constructions is fairly unrestricted, which shows that the SN still functions very much like a noun. Both quantifying and valuing uses can appear with modifiers in front of the SN-expression. All of the premodification phenomena attested in the data can be reduced to two types of systematic premodification patterns, parallel with the two grammaticalized uses of quantifier and valuing quantifier. Moreover, those two patterns serve to reinforce the interpretation of both grammaticalized uses and are part of the benefits of using an SNexpression for expressing quantification, i.e. they aid to enrich the quantifier paradigm (see Section 3.4.4). Firstly, and parallel with the grammaticalized quantifier use, there is a premodification pattern of adjectives that reinforces the quantitative information supplied by the SN-quantifier as in (4.248) to (4.252): (4.248) There’s now a whole bunch of studies from different cities that show the same thing. (CW-NPR)

196Chapter 4: SN-constructions (4.249) The world that awaits me is a bit scary there is going to be a whole bunch of idiot journalists waiting to ask me stupid questions-but then that's life. (CW-UKmags) (4.250) There was this guy called Billy Cote who was making music with a whole heap of people, chiselling out chunks of noise. (CWUKmags) (4.251) I actually left there to go up to the middle school whereas a whole load of people from the middle school came back down to our school because they suddenly found themselves too young to be there any more. (CW-UKspoken) (4.252) Hey, it's not about heroin, okay? It's about going through a serious bunch of feelings. It’s what we do, man; we get a lot of feelings out. (CW-UKmags) Whole can reinforce the quantifier meaning of bunch of, heap of, lot of and load of and makes the quantifier semantics more hyperbolic.73 For the plural variants no examples premodified by whole were attested in Collins Worbanks corpus, by means of the query whole+0,3heaps|loads|lots|bunches. This query translates as attestations of whole heaps/loads/lots/bunches, allowing for up to three submodifiers in between whole and the plural SNs. For the singular variants as well additional extractions were made in COBUILD using the query whole+0,3bunch|load|heap|lot. Serious in (4.252) above seems to function similarly, in the sense that it is rephrased as a lot of feelings. Still, it also incorporates the notion that the many feelings they go through are not trivial and hence also acts as a qualitative adjective modifying N2. In addition, there is a rather productive set of quantifier uses with load(s) of that are premodified by an element referring to a means of transportation or an object that can hold a literal load in it. In other words, the premodifier rekindles the SN’s original semantics; however, rather than referring to the actual container, it functions as an expressive intensifier: (4.253) Mccoist made them pay for their failure to convert a hat-load of chances. (CW-Sunnow) (4.254) Stirling blew a barrel-load of chances in a clash they totally dominated. (CW-Sunnow)

4.5 A case for collocationally constrained constructions

197

(4.255) Those who must deal with them on a regular basis and wish to get the better of them must employ much patience, deviousness and duplicity as well as lorry loads of insincere flattery. (CW-Today) In valuing SN-constructions, then, the SN-expression can be premodified by qualitative adjectives further accentuating the grammaticalized valuing meaning. These pre-SN evaluative adjectives seem to be the result of a transfer of a qualitative adjective actually applying to the referent of N2 (cf. Halliday 1985a: 174). The following set of examples also subsumes vague valuing/head uses with pile and heap: (4.256) Who said Americans were a jingoistic bunch of rednecks who know or care nothing about what happens beyond their shores. (CW-Today) (4.257) Most of them are a right bunch of misery guts. (CW-Sunnow) (4.258) They are a competent bunch of bankers. (CW-Sunnow) (4.259) What a load of pretentious waff, i.e. the people who are into Pearl Jam buy ripped jeans and sneakers cos it’s f ing rad, dude, what kind of childish excrement is that, may I ask? (CW- UKmags) (4.260) There is, however, no truth in the rumour that Chinese film chiefs considered calling last year’s Batman movie A Complete Load of Prawn Balls. (CW-Sunnow) (4.261) His Romanian employers dismissed his excuse as a whole load of rubbish. (CW-Sunnow) (4.262) “’I don't believe in Saint Valentine's day’. He claims it's a commercial load of crap ploy. You know that one?” “Yes, that is one from people who haven't sent you anything for Valentine's Day.” (CW-UKspoken) (4.263) It’s the biggest pile of pretentious tosh I’ve heard this year. (CWToday) (4.264) What a heap of shit. (CW-Sunnow) In examples (4.256) to (4.264) the adjective preceding the SN-expression may apply only to N2 or to N2 and the SN, depending on the extent to which a quantifying meaning is still present. If it is, premodifiers of the SN

198Chapter 4: SN-constructions reinforce the quantity and the evaluation, e.g. a whole load of rubbish (4.261). This is different from the premodification pattern observed for the head SN-construction, in which adjectives premodifying the SN apply to the SN as well as its coextensive referent designated by N2. In uses in which the valuing meaning prevails, evaluative or emphatic adjectives describe the referent of N2 or emphasize the degree to which it is part of the class designated by the referent, e.g. A Complete load of Prawn Balls (4.260). Relocating the adjectives to pre-N2 position does not change the meaning of the entire construction drastically, but mostly tones down the effect of the valuing SN-construction and foregrounds the quantifier reading more. Emphasizing adjectives, such as complete, whole or right and superlative adjectives do not really allow such relocation without altering the meaning of the construction, as in: (4.265) A jingoistic bunch of people vs. A bunch of jingoistic people They are a competent bunch of bankers vs. They are a bunch of competent bankers It’s a commercial load of crap ploy vs. It’s a load of commercial crap ploy A Complete Load of Prawn Balls vs. *A Load of Complete Prawn Balls They are a right bunch of misery guts vs. *They are a bunch of right misery guts The biggest pile of pretentious tosh vs. *The pile of biggest pretentious tosh When the adjective submodifies the SN-expression, its evaluative semantics are reinforced and bunch of is to some extent absorbed in the evaluative meaning expressed by the evaluative adjectives, which it further intensifies. A jingoistic bunch of people means more or less the same as jingoistic people, but the former adds emphatic and quantitative meaning. The difference in meaning between valuing constructions with the adjective in front of the SN-expression and those with the evaluative adjective in front of N2 can be illustrated by the following examples with great, attested in the corpus data:

4.5 A case for collocationally constrained constructions

199

(4.266) Look we’re reforming, aren’t we a bunch of great guys? (CWNPR) (4.267) I’m racing with a great bunch of guys and it makes racing fun again. (CW-UKmags) In (4.266) great applies directly to guys, whereas in (4.267) the unity or group aspect of the guys is highlighted and then further evaluated as great. This is in keeping with the notion of diagrammatic iconicity (see Haiman 1985). In this view, syntactic constellations of words are interpreted as diagrammatic signs. Iconicity is then based on a similarity between the relationship of these component signs and the relation between the objects or actions that they designate. For instance, in Caesar's dictum "veni, vidi, vici" a temporal sequence of actions is reflected in the sequence of the three verbs. In examples (4.266) and (4.267) the specific syntactic structures and relations between the elements of structure mirror the meaning relationships expressed by them. In keeping with diagrammatic iconicity, then, a different location of the evaluative adjective triggers a different profiling of constructional meaning in terms of backgrounding and foregrounding of valuing and quantifying semantics. No real systematic preference of evaluative adjectives for either pre-SN or pre-N2 position could be observed in the data set. Two further observations have to be made. Firstly, as mentioned earlier, only qualitative adjectives allow positional transfer in grammaticalized SNconstructions, whereas classifying adjectives do not, as exemplified by the ungrammaticality of the alternations in (4.268): (4.268) Andy Roxburgh bumped into a bunch of UEFA officials (CWToday) *Andy Roxburgh bumped into a UEFA bunch of officials A whole bunch of help wanted signs. (CW-NPR) *A whole help wanted bunch of signs Secondly, whereas qualitative premodification of the SN-expression in grammaticalized constructions seems to be the result of positional transfer of the premodifier to pre-SN position, quantitative premodification is not. Instead, it seems that in examples (4.248) to (4.255) the quantitative or intensifying adjectives are the result of adding an extra piece of quantitative information directly to the quantifying SN-string, almost forming it into a functional quantifier complex. Finally, let us have a look at the kind of premodification observed for ambivalent SN-constructions, especially in the vague subcategory. In the

200Chapter 4: SN-constructions vague uses we see a mixed kind of premodification, which is partly similar to the pattern found in the head SN-uses, but also incorporates intensifying or other modification typical of grammaticalized quantifying or valuing SN-constructions: (4.269) Eating, washing and dressing, two hours. Work and travel, another ten hours. Which leaves just four hours a day for all those other things, including watching TV and that creeping, self-renewing slag heap of duty we call housework. (CW-UKmags) (4.270) God, it felt great to be alive in this seething great shit-heap of a city, every window a bloodshot eye, every alleyway a shelter for broken deadbeats whose only crime was dreaming themselves out of the gutter. (CW-UKmags) (4.271) The British have forged a fine tradition of gardening and cannot afford to sit on their well-clipped laurels. Striding past the compost heap of nostalgia comes Christopher Lloyd, now in his seventies and still coming up with new ideas for his garden at Great Dixter in Sussex. (CW-Times) In example (4.269) creeping and self-renewing do not describe a real heap taking up an amount of space in reality, but describe the concept of a job that is never finished, i.e. the duty called housework. Slag in slag-heap, which literally refers to a hill with waste material, is used as negative emphasizer; thus activating elements at both the lexical and grammatical layer of meaning in the SN-expression. As these are metaphorical examples, a blend of both types of premodification patterns is unsurprising. In (4.270) seething applies to city rather than heap and great can apply to both. In (4.271) then, compost also rekindles the literal semantics of heap, but at the same time evaluates nostalgia. In ambiguous constructions, premodifiers equally allow for either a literal interpretation or a reading as intensifier, but these can be disambiguated depending on specific contextualizations: (4.272) There's a great pile of work that builds up because they haven't got anybody to get everything back in the system. (CW-UKspoken) In sum, the premodification patterns attested for quantifying and valuing uses can be reduced to two systematic systems, each reinforcing the functional semantics of their respective construction. For this reason, I do not see the possibility of such systematic, and also restricted, premodification of the SN as counterevidence of decategorialization or as proof of the

4.5 A case for collocationally constrained constructions

201

nounness of the SN, as Langacker (2009 and to appear) argues with regard to a lot of. Moreover, the possibility of such functional premodification of the quantifying and valuing use of SN-expressions also constitute advantages compared with semantically more schematic quantifiers such as many/much. The latter are also restricted to either count or uncount nouns, whereas SN-expressions allow both, as also noted by Langacker (to appear: 1). The distinct collocational restraints for each SN-use are summarized in Table 4.12, which also lists the semantic and syntactic features of the various constructions (cxns): Table 4.12. SN-uses as collocationally constrained constructions Semantics Head cxn

Syntax

'SN consists of N2' [SN]+[of+N2]

- subsets of concrete N2s

Head modifier

- unrestricted premodification of SN

Quantifier cxn 'a quantity of N2'

[SN+of]+[N2] Modifier head

Valuing (quantifier) cxn

Collocational patterns

'valuing N2'

[SN+of]+[N2] Modifier head

- concrete, abstract, animate N2s - restricted premodification of SN (i.e. quantification- reinforcing) - (mostly uncount) abstract, or animate N2s - negative semantic prosody - restricted premodification of SN (i.e. evaluation- reinforcing)

4.6 Degrees of grammaticality By way of conclusion to the study of ten English SNs, Table 4.13 summarizes the degrees of grammaticality of the various singular and plural expressions by comparing the relative frequencies of head, ambivalent and grammaticalized uses. The table includes separate percentages for quantifying and valuing quantifying uses, as well as an overall percentage for the total number of grammaticalized uses.

202Chapter 4: SN-constructions

Table 4.13. Degrees of grammaticality Head use

Quantifier use

Valuing Q use

Total grammaticalized uses

Ambivalent

Heap

45.3% 32.1%

1.9%

34%

20.7%

Heaps

30%

/

66.7%

3.3%

Pile

93.6% 1.2%

/

1.2%

5.2%

Piles

89.7% 1.2%

/

1.2%

9.1%

Load

22%

40.8%

34%

74.8%

3.2%

Loads

6%

93%

/

93%

1%

Lot

0.8%

98.8%

0.4%

99.2%

/

Lots

0.8%

99.2%

/

99.2%

/

Bunch

11.6% 52.4%

36%

88.4%

/

/

/

6.1%

66.7%

Bunches 93.9% /

When we compare the relative frequencies of these SNs, it becomes clear that lot(s) of, bunch of, loads of and load of have grammaticalized the most, each displaying a clearly skewed system in which the grammaticalized use is the most frequent. With heaps of the grammaticalized quantifier use also predominates, but in a slightly less pronounced way (66.7%). The other SNs, i.e. heap, pile, bunches and piles have a predominant head use. Heap to some extent has an equi-probable system in that the percentages for its head use and quantifier use do not differ very much, whereas pile and piles show very clear skews between the head noun use and quantifier use. The data set of heap turned up a high percentage of ambivalent constructions and this blurs the equi-probability of the entire system somewhat. Visualizing Lehmann’s (1985) concept of synchronic grammaticalization, we can draw a scale that represents the percentages of grammaticalized uses for each SN:

4.6 Degrees of grammaticality 203 Pile Bunches Piles 0% 1,20%

Heap 34%

Heaps Load 5 0%

Figure 4.1. Scale of grammaticalization

66,70%

75%

Bunch Loads A lot of/Lots of 88,40%

93% 100%



Explaining differences in degree of grammaticality, in my view, boils down to explaining differences in the delexicalization (potential) of the SNs. It was argued earlier that the higher degree of grammaticalization displayed by plural variants of SNs, such as heaps of and loads of, is due to the fact that the plurality marker enhances the hyperbolic quantifier value of these SN-expressions and is therefore conducive to grammaticalization towards this function. The only exception to this pattern is bunches of, which did not display any grammaticalized uses in my data set, whereas singular bunch of has grammaticalized very extensively. One of the explanatory factors are prosodic features which disfavour polysyllabic bunches to construe hyperbolic quantifier meaning, and favouring monosyllabic expressions such as loads and heaps. Another is the strong pragmatic motivation for developing valuing quantifying uses, in addition to quantifier uses. This valuing quantifier use was attested only with singular SN-expressions. This was explained functionally in terms of what these singular forms express conceptually and how this is conducive to construing evaluative meaning. It was suggested that bunch of, load of and pile of, as singular forms, have the advantage of singling out one set of entities in terms of a shared qualitative feature and of keeping their focus on that one bounded set which is then usually negatively valued. Plural SNexpressions, such as *bunches of morons or *loads of bull cannot convey such emphatic evaluation it seems, but instead foreground hyperbolic quantifier meaning, i.e. ‘many groups of morons’ and ‘a lot of nonsense’. It can be hypothesized that bunch, with its original cluster meaning, shows this effect even more strongly, in that singular bunch refers to one group of entities or people, whereas plural bunches does not refer to one gigantic cluster, but to several separate clusters. Pluralization hence achieves a different result here than with semantically more schematic concepts expressed by heap and load. I have also argued that the lack of grammaticalization of pile was due to its lack of delexicalization, because pile is semantically more specific. Bunches can only be assumed to have strong conceptual image persistence as at present it does not give way to hyperbolic quantifier meaning such as ‘a lot of’, except in a rare ambivalent case, but not in any systematic and context-independent way. The high degree of grammaticality attested by load(s) of may be enhanced by the existence of the very frequent and relatively register-neutral SN-quantifiers lot(s) of, to which load(s) are phonetically and graphologi-

204Chapter 4: SN-constructions cally similar. The idea is that the load-variant offers a better means of hyperbolic expressiveness because of its prolongable vowel. Loads of, with this prolongable vowel, seems to be much more hyperbolic in meaning than lots of, which just means something like ‘many’, without transgressing a presumed standard. By contrast, loads of always surpasses this standard. The same goes for heaps of, which also has a prolongable vowel and plural marker. An example from the Internet demonstrates how a specific speaker assesses the referential quantifier meaning of heaps of relative to a lot of. The example pertains to an online New-Zealand survey where students have to express agreement with several statements by rating it as heaps, quite a lot, some or very little (http://nemp.otago.ac.nz/science/ 1999/surveys). In Section 4.7 the diachronic developments leading to quantifier uses of lot(s) of will be discussed. Traugott (2008c) argued that a lot of developed its quantifier use relatively late. In addition to lot(s) of the diachronic case study also traces the developments that heap(s) has gone through. In order to check the precise analogical pull between lot(s) and load(s), but also between other SNs, more diachronic studies of course need to be done. The diachronic corpus study discussed in Section 4.7 is a first detailed case study included to show the potential diachronic complexities that have led to the synchronic percentages of head, grammaticalized and ambivalent uses.

4.7 Diachronic corpus study of heap(s) and lot(s) 4.7.1 Introduction In this section, I will study the diachronic dimensions of the synchronic variation attested for SN-expressions. I will do this by means of a diachronic case study of four SNs that were part of the synchronic analysis, i.e. heap(s) and lot(s). Lot(s) of are singled out because they are synchronically the most generally used SN-quantifiers in terms of the syntactic and semantic kinds of N2s they can quantify and the fact that they almost exclusively function as quantifiers in synchronic data. Heap(s) have 34% and 66.7% (valuing) quantifier uses each in the present-day English data sets, which makes them interesting for a comparative analysis. They are also central SNs from a cross-linguistic perspective. In several Romance and Germanic languages SNs with these source semantics can start grammaticalizing, e.g. hoop/hopen in Dutch (see Section 4.8), Haufen in German, tas de in French and montón and montones de in Spanish (see Brems & Verveckken 2008 on English and Spanish SNs). The diachronic study will

4.7 Diachronic corpus study of heap(s) and lot(s) 205

focus on the move from head to quantifier constructions. Valuing uses are largely left out of the present study. First the diachronic data sets and methodology used for the case study will be presented. Section 4.7.3 then presents the case study itself and a discussion of heap(s). Section 4.7.4 will then discuss lot(s). Section 4.7.5 will bring together the main findings of the diachronic case study and Section 4.7.6 will round off with some conclusions and prospects for future research. It will be argued that synchronic clines, such as Figure 4.1, above cannot be assumed to reflect the direction or rate of historical change in a straightforward manner (Kroch 1989, Andersen 2001, Lehmann 2004, Traugott 2008c). Items farthest on the cline thus did not necessarily start grammaticalizing first. Corpus analysis indeed shows that diachronically the grammaticalization of heap(s) and lot(s) involved more complex paths of development with partitive constructions in addition to lexical head noun constructions and quantifier constructions. Collocational patterns have not unidirectionally extended from inanimate concrete nouns to all kinds of (in)animate concrete and abstract nouns. In addition, I will trace the chronology of lot(s) and heap(s) in developing quantifier uses. In the purely synchronic study of Brems (2003a) it was hypothesized that lot(s) of may have been the model for other SNs to develop quantifier uses with modifier status. Traugott (2008c), which is a diachronic study of a set of singular SNs, argues that heap was the first SN to develop a quantifier use, i.e. around 1350, while lot only developed one several centuries later, i.e. around 1800. In the present diachronic study, just like in the synchronic study, SNuses are distinguished by means of careful constructional analysis on syntactic, semantic and collocational grounds. Only on the basis of such analysis can it be decided whether heap(s) or lot(s) was the first to develop quantifier uses in the strict sense. It will be shown that quantifiers in the strict sense for heap(s) and lot(s) of emerge around the same period. Heap has a relatively early size-related use designating a multitude or host (of people), but this still has syntactic head status, is collocationally restricted and semantically similar to collective expressions such as a bevy of swans. This use will be argued to be a partitive construction that is distinct from the quantifier construction in which heap(s) of and lot(s) of have modifier status and are collocationally open. The latter have the grammatical semantics of canonical quantifiers such as many and much which assess size with respect to a scale of magnitude (cf. Langacker 1991: 81), i.e. they are subjective in nature. Within the proposed constructional approach, the early multitude sense thus cannot simply be compared to the quantifier uses of load(s), lot(s) and bunch of which have modifier status. The synchronic cline represented above as Figure 4.1 represents modifier quantifier uses of the SNs involved. When Traugott (2008c) assesses how this synchronic

206Chapter 4: SN-constructions cline matches up with diachronic developments and chronologies, she compares early partitive uses of heap with quantifier uses in the full sense of lot.

4.7.2 Data sets and methodology used Exhaustive extractions were made on heap, heaps, lot and lots from various corpora, using Abundantia Verborum (AV)74: The Penn-Helsinki Parsed Corpus of Middle English (PPCME) and Early Modern English (PPCEME), the Innsbruck Middle English Prose Corpus Sampler (IMEPCS), the Corpus of English Dialogues (CED), the Corpus of Early Modern English texts (CEMET) and Corpus of Late Modern English Texts (CLMETEV) (De Smet 2005), and the Old bailey Corpus (OBC). Corpora predating the Middle English period are not included in this study, since the NP + of + NP-construction, which is the main focus of this study, only appears from Middle English onwards (cf. Denison 2002). In order to net all spelling variants listed in the OED and MED, singular and plural forms, and all possible case markings, the following queries were used in AV for heap(s) and lot(s) respectively: hy?\+?[aei]\+?[aeiyo]?\+?a?\+?pp?[usnaer]?[maesnu]?[eam]? and h?le?oo?h?\+?tt?h[usnaer]?[maesnu]?[eam]?. These queries capture all occurrences of heap, heaps, lot and lots irrespective of the syntagmatic environment in which they occur. Occurrences within the NP + of + NPconstructions were labelled manually using AV. Irrelevant tokens or multiple attestations of concordances were filtered out manually. Table 4.14 indicates the number of tokens for heap(s) and lot(s) within the NP + of+ NP-syntagms per corpus. The numbers in bold represent absolute frequencies, the second rows consist of normalized frequencies per million words in each respective corpus.

4.7 Diachronic corpus study of heap(s) and lot(s) 207 Table 4.14. Token frequencies of heap, heaps, lot and lots in diachronic corpora

IMEPCS PPCME 1100 1150 1500 1500

PPCEM E 1500 1710

CED 1560 1760

CEMET OBC CLMET CLMET CLMET 1640 1674 1 2 3 1710 1834 1710 1780 1850 1780 1850 1920

Heap

24

4

12

5

4

108

13

51

45

pmw

5.1

3.5

6.7

4.2

2.1

2.1

4.3

8.9

7.2

1

1

3

2

6

6

7

39

16

pmw

0.2

0.9

1.7

1.7

3.1

0.1

2.3

6.8

2.6

Lot

2

11

2

2

3

230

13

68

220

pmw

0.4

9.5

1.1

1.7

1.6

4.5

4.2

11.9

35.5

Lots

0

0

1

2

1

37

0

7

53

pmw

0

0

0.6

1.7

0.5

0.7

0

1.2

8.5

Heaps

The low frequencies indicate an overall dearth of diachronic data for SNs. This might be linked to the intrinsic lack of more informal and/or colloquial data for the older periods of English. Halliday (1978), Chafe (2003) and Du Bois (2003) have singled out casual conversational language as an important locus of language change and innovation because it is less subject to overly conscious forms of monitoring or engineering. It is also the type of language that SNs today still typically occur in. The OBC, which consists of the proceedings from London’s central criminal court, profiles itself as reflecting more colloquial, spoken language, but it too yielded relatively few data. The CED contains Early Modern English speech-related texts such as constructed dialogues and authentic speech situations, and amounts to almost 1.2 million words. It too yielded few attestations. In addition, the normalized frequencies do not always show a very clear directionality of increase or decrease. The diachronic corpus study will be largely qualitative in nature because of this.

4.7.3 Heap and heaps As we have seen in Section 4.3.1, in present-day English, heap of has 34% grammatical quantifier uses and heaps of has 66.7%. By comparison, load of, loads of and bunch of are much more frequently used as (valuing) quantifiers, i.e. respectively in 75%, 93% and 88.4% of the attestations. Lot(s) of are almost exclusively used as grammatical quantifiers in present-day Eng-

208Chapter 4: SN-constructions lish. It is tempting to correlate the relative positions of the SNs on the synchronic scale with the direction and the rate at which the individual SNs have come to be used as quantifiers (cf. Brems 2003b). Judging from Figure 4.1, one might deduce that heaps of has simply extended its collocational range further than heap of, which stopped short at 32.1%, and lot(s) are the only ones to have extended to all kinds of collocates and move all the way down the cline. However, as observed earlier, synchronic clines of grammaticality do not straightforwardly reflect actual diachronic change, but may obscure more complex paths of development (cf. Kroch 1989, Andersen 2001, Lehmann 2004, Traugott 2008c and Traugott & Trousdale 2010). The earliest sense of heap and heaps attested in my diachronic data set refers to "a collection of things lying one upon another so as to form an elevated mass often roughly conical in form" (OED, sense 1.a.; MED sense 2.a.). The earliest example in the OED is dated c725. Examples (4.273) and (4.274) illustrate this lexical constellation sense both inside (4.273) and outside of the of-construction (4.274): (4.273) He led hym to a hep of stonys. (IMEPCS, Richard Rolle of Hampole, a1349) (4.274) dong hepys comely and it ha3t vertu. (IMEPCS, Agnus Castus. A Middle English Herbal, a1450) In this use the SN is the syntactic and semantic head of the binominal syntagm, similar to synchronic head noun uses (4.5)-(4.21). N2-collocates in this use are restricted to concrete, stackable referents. A figurative use of this lexical use is illustrated by (4.275). (4.275) He yzi3þ þane greate heap of his zennes. (IMPECS, Ayenbite of inwyt, or Remorse of Conscience, 1340) 'He sees that great heap of his sins' In examples such as (4.275), N2-collocates are restricted to abstract nouns sins, vices, virtues and often associate with metaphorical hyperbolic contexts where sins for instance are said to be piled up and weigh heavily on the sinner. In such a use the SN is still the head noun, which is shown by the fact that the of-phrase can be fronted and topicalized when NP1 is definite (cf. Traugott 2010b). A second early use, which seems restricted to singular heap, is illustrated by examples (4.276) - (4.278):

4.7 Diachronic corpus study of heap(s) and lot(s) 209

(4.276) The Bear sprang up between the bush and the river among a Heap of wives (IMEPCS, The History of Reynard the Fox, 1481) (4.277) That egle that lighted amonge the hepe of swannes, sygnyfieth our enmye stranger (IMEPCS, Caxton’s Eneydos, 1490) (4.278) Among an heep of sterris fixe (IMEPCS, A Treatise on the Astrolabe, c1391) In this sense singular heap refers to a host or multitude, especially of people, but as example (4.278) shows, also to a cluster of stars, or a collective of angels. Just like the figurative use of the lexical sense, the multitude sense has a restricted collocational range of N2s, i.e. people, animals, angels and stars. In examples (4.276) to (4.278), reference is to a collective of wives and swans or a constellation of stars in which the referents of N2 are spatio-temporally bounded. The frequent presence of adverbials such as among in (4.276) to (4.278) further supports such a paraphrase. I will label these uses partitive constructions (cf. Traugott 2007). With Traugott, I argue that heap syntactically functions as the head noun in this construction, rather than a modifier, as the of-phrase can be fronted, e.g. in And of vpholders an hep (MED). Even though the partitive construction has to do with indicating size, unlike canonical quantifiers, it does not rely on a semanticized scale of absolute quantification for doing so (cf. Langacker 1991). By implication an indefinite number of stars or swans is referred to, but the semantics of heap in such partitive constructions are equivalent to collective nouns or idiomatic measure noun expressions, e.g. a bevy of swans, a constellation of stars, rather than to canonical quantifiers such as many and much. The oldest example of this sense of heap is dated 971 in the OED. It is said to have become obsolete around 1600. Examples (4.279) to (4.282), then, exemplify quantifier uses of heap and heaps similar to synchronic examples such as (4.22)-(4.56). N2collocates can be concrete (in)animate, abstract and both count and uncount nouns. (4.279) "I hope you won't go yet, for I expect my son home soon, and I've a heap of things to talk to you about" (CLMETEV, Cecilia, 1782) (4.280) I should produce heaps of sonnets (CLMETEV, Stories from the Italian poets: with lives of the writers, 1846) (4.281) He had also, by Gan's advice, brought heaps of wine and good cheer to be set before his victims in the first instance.

210Chapter 4: SN-constructions (CLMETEV, Stories from the Italian poets: with lives of the writers, 1846) (4.282) There were heaps and heaps of mothers in the world, of course (CLMETEV, The Extra Day, 1915) In these examples heap(s) of have modifier status and serve to quantify N2, which is the head. The modifier status is shown by the fact that verb concord is systematically with N2, e.g. there were a heap of breeches [...] (OBC, 1801), there were a heap of young men beating Mr. Bradley (OBC). As quantifiers, heap of and heaps of serve to assess the size of N2-referents with respect to a scale of magnitude (cf. Langacker 1991: 81), i.e. they are subjective in nature, as opposed to heap in examples (4.276) to (4.278). The SNs heap and heaps together with of, as chunks, can therefore be substituted by canonical absolute quantifiers such as many and much rather than ‘a collection/group of’. This quantifier use does not rely on metaphorical contexts as in example (4.275) and N2-collocates are fairly unrestricted in nature. In addition, N2-referents do not need to be spatio-temporally contiguous as they did in the partitive construction. The sonnets in (4.280) may be written over an extensive period of time and the mothers in (4.282) are located all over the world. Discourse contexts are typically hyperbolic, especially with plural heaps of. In my data sets first occurrence of heap(s) of in the quantifier construction occur around the end of the 17th century.

4.7.4 Lot and lots Similar to heap and heaps, the diachronic data sets for lot and lots yielded three main uses: lexical head uses, partitive uses and quantifier uses. In the Middle English data sets lot(s) outside of the NP of NP-construction almost exclusively occur as part of composite predicates referring to the drawing or casting of lots as a method of decision-making: (4.283) But lete us draw lottes whose it shall be (PPCME, In Die Innocencium, 1497) In NP of NP-constructions in this period reference is typically to what is assigned to someone by lot or fate: (4.284) For our Lord ne shal nougt forsake te penaunce of te singers up te lot of te rightful (PPCME The earliest complete English prose psalter, c1350)

4.7 Diachronic corpus study of heap(s) and lot(s) 211

These lexical senses have been around from 950 onwards (cf. OED). Lot(s) appear in NP of NP-constructions from Middle English onwards. In addition, there is an early partitive use referring to a portion or part, which also occurs in NP of NP-constructions and seems largely specific to The Ormulum.75 N2-collocates in such partitive constructs can be inanimate concrete nouns (4.285), but also animate (4.286) or refer to more abstract concepts, as in (4.287): (4.285) An lott off manne fode (PPCME, The Ormulum, c1200) 'a portion of food for men' (4.286) tat tegg [LB: Farisewisshe menn] wisslike warenn an lott off tatt Judisshenn follc (PPCME, The Ormulum, c1200) 'that they certainly were part of the Jewish people' (4.287) For to forwerrpenn anig lott Off Moysœsess lare [...] (PPCME, The Ormulum, c1200) 'For to reject any part of Moses' teaching' In the Early Modern English data sets lot(s) are attested in the sense of 'an article or set of articles for sale' (4.288) and the semantically related sense 'a group of people or things of the same kind' (4.289): (4.288) There the merchants and gentlemen of the country going on board, to demand those lots of slaves they had already agreed on (PENN2, Aphra Behn, 1668) (4.289) "No, miss, not all, You know when a lot of servants gets together, they like to talk about their betters" (CLMETEV 2, The Tenant of Wildfell Hall, 1848) In examples (4.285) to (4.289) lot(s) have head status. The SNs refer to spatio-temporally contiguous groups of people or units of things, similar to rare synchronic head uses with lot(s) (see examples 4.95 and 4.96). In these examples the finite verb agrees with lot(s) rather than with N2, as illustrated by (4.289). In the 1780-1850 data set, examples emerge that are ambiguous between the 'group of people' or 'set of articles' reading and a quantifier reading in which lot(s) of have modifier status. Ambiguities typically arise with concrete N2-collocates, both inanimate (4.290) and animate (4.291). (4.290) There was a whole lot of people behind (OBC, 1790s)

212Chapter 4: SN-constructions (4.291) I bought a lot of clothes of the shopman, and took them to the stable (OBC, 1820s) In these examples semantic paraphrases with either 'a group of' or many/much are plausible. Even in sales contexts, pragmatic inferences of mere quantity are allowed (4.291) via the pragmatic implicature that a set of articles (for sale) refers to a number of them. As with synchronic data, diachronic data offer few contexts to corroborate the head status of SNs in NP of NP-constructions syntactically. Verb agreement patterns can only be checked when the binominals appear as subjects, and the SN and N2 need to differ in number. With past finites only were and was are marked for number. Furthermore, there is the possibility that fluctuation in verb concord in itself may be a sign of synchronic variation and possible diachronic change. Syntactic and pragmaticosemantic ambiguity or underspecification may have enabled the reanalysis of lot(s) of as a quantifier (cf. reference to the "pragmatic slack" of bridging contexts in Traugott 2007 following Lasersohn 1999). In these contexts the originally partitive or collective uses of lot(s) can be associated with mere quantity and similarities to canonical (monomorphemic) quantifiers such as many and much are reinforced. Around the time of ambiguous examples such as (4.290)-(4.291), the first unambiguous quantifier uses appear. The first attestations are of plural lots of quantifying room. The spelling in (4.292) reflects the phonetic erosion of the quantifier: (4.292) What, there's lots o' room! (CLMETEV 2, Agnes Grey, 1847) (4.293) Clear away, my lads, and let's have lots of room here! (CLMETEV 2, A Christmas Carol, 1843) Quantifier uses occur with all kinds of (un)count (in)animate concrete nouns that cannot appear with lots in the lexical head sense. Money is a frequent collocate, just like things, but lot(s) of also collocates with abstract nouns, e.g. dignity in (4.296), thinking in (4.299), animate nouns (4.294) and (4.300). Verb concord is systematically with N2: (4.294) A lot of fellows have crotchets (CLMETEV 3, The English Constitution, 1867) (4.295) 'But I mean a lot of money - tens of thousands, hundreds of thousands?' (CLMET 3, The Grand Babylon Hotel, 1902)

4.7 Diachronic corpus study of heap(s) and lot(s) 213

(4.296) And she had her dignity too--and a lot of it (CLMETEV 3, The Old Wives' Tale, 1908) (4.297) "There are lots of places we could go to" (CLMETEV 3, The Old Wives' Tale, 1908) (4.298) A name, yes, and lots of triumphal processions (CLMETEV 3, The Christian, 1897) (4.299) It'll save us [sic] lot of thinking! (CLMETEV 3, The Prisoner of Zenda, 1894) (4.300) They ought then to have heaps of work, or they ought to have a lot of children to look after (CLMETEV 3, Clara Hopgood, 1896) Spatio-temporal contiguity of N2-referents is often loosened in quantifier constructions compared to partitive constructions. Lot(s) of can be substituted by many or much. As observed by Langacker (to appear), lot(s) of as quantifiers are unrestricted with regard to the countability of the noun they quantify whereas many and much are, i.e. to count and uncount nouns respectively. In expressing size, lot(s) of as quantifiers "no longer construe their quantity in relation to a physical entity", i.e. a specific set of articles, a batch or a group, but vis-à-vis a schematic quantity scale, similar to canonical quantifiers (Langacker to appear: 13). In the same period adverbial uses can be attested, where lot(s) of grade comparative adjectives: (4.301) all the unlicensed scoundrels from the Tower to Gravesend, and a lot further (CLMETEV 3, The Grand Babylon Hotel, 1902)

4.7.5 Discussion of findings Both heap(s) of and lot(s) of seem to have passed through a partitive construction on their way to a real quantifier construction with modifier status. How do the lexical head, partitive and quantifier construction relate diachronically? In the case of lot(s), there seem to be clear metonymic and semantic links between the lexical head, partitive and quantifier use. The historical lexical head construction with lot of collocationally differs from that with heap of in that the fate meaning typically co-occurs with animate concrete nouns, whereas the latter almost exclusively occurs with inanimates. The partitive construction with lot(s) was collocationally already open to animate and inanimate concrete nouns and more abstract nouns

214Chapter 4: SN-constructions such as lare ('teaching') in (4.287), whereas the partitive use of heap of was much more restricted. For heap(s) the interrelations between lexical head, partitive and quantifier construction are less straightforward and give rise to a number of interlocking questions. Are the early partitive construction, referring to a multitude of people, and the later quantifier use with modifier status connected? Did the quantifier use as we know it today evolve out of the partitive construction or the lexical head construction? Is the partitive sense of heap in itself an original use, unrelated to the concrete lexical meaning, or is it based on an, admittedly, early metaphorization of the constellation sense? As we have seen, the partitive use of heap has very specific collocational preferences, mostly animate nouns, which is almost at right angles to those of the lexical head use of heap which prefers inanimate concrete nouns. This makes it hard to see the partitive use as related to the lexical head use, in which case one would expect collocational overlap of some sort. Interestingly, the MED, which only reflects the Middle English period, lists the multitude sense of heap first, and examples of this meaning predate the lexical head meaning, i.e. c1175 versus c1330 respectively. In the OED it is the other way around. In my data the literal constellation meaning is attested slightly earlier than the partitive sense. If the quantifier use developed from the partitive use, where does the early partitive use in which heap has head status stop and where does the quantifier use, in which heap(s) of have modifier status start? The OED simply mentions that the multitude sense persisted in some way in the delexicalized quantifier sense that was illustrated by synchronic examples (4.22)-(4.56) and diachronic examples such as (4.279) to (4.282). The most recent example of the now obsolete multitude sense is dated 1594 in the OED and the earliest of the “later colloquial use” is dated c1661. If the present quantifier use is a continuation of the partitive sense, how, when and why did the status of heap of shift from a head followed by an of-phrase to a modifier? How do we explain the extension of N2-collocates from animate to inanimate concrete and abstract nouns? Can we interpret it simply in terms of invited inferences that enabled the collection sense to bleach and SN-status to shift from head to modifier in a process of subjectification and collocational loosening? In the case of lot(s) this seems more straightforward to accept since the partitive construction already allowed (in)animate concrete and abstract nouns and only the semantics of lot(s) needed to delexicalize so to speak. As mentioned earlier, my diachronic data do not allow verifying systematic changes in verb concord, thereby precluding syntactic corroboration of the status of the SN-expressions. Collocation-wise, the partitive sense of heap, as opposed to lot(s), has restrictions in terms of animacy and the association with metaphorical contexts, which the later quantifier use has not, but this of course cannot be used as a diagnostic. The substitution

4.7 Diachronic corpus study of heap(s) and lot(s) 215

test, by either ‘a collection/group of N2’ for the partitive construction or ‘quantity of N2’ for the quantifier use, is likewise tricky since the first, more lexical, replacement also has scalar implicatures which can make replacement judgments rather subjective. Given the fact that the partitive sense is specific to heap, one may wonder whether singular heap and plural heaps have followed different pathways in the development of quantifier uses or were influenced by other constructional uses. It could be hypothesized that the singular quantifier use of heap of is a more direct continuation of the partitive construction, whereas quantifier uses of plural heaps of arose predominantly in contexts that were ambiguous between a literal constellation reading and a quantifier one. What separates both the constellation sense and the partitive sense, as two types of head use, from quantifier uses with modifier status is that the latter are subjective in nature. They call up a scale and require a subjective assessment of quantity relative to that scale. Traugott (2008c) only looks at singular forms of a set of SNs and for heap seems to follow the chronology proposed by the MED. She considers the multitude sense of the partitive construction to be the original meaning of heap and a distinct one from the lexical head sense. Within her semantic analysis, this partitive use of heap is labelled a quantifier and considered to be the earliest quantifier use of the set of SNs she looks at while a lot of is the most recent, emerging only around 1800. In doing so, she seems to compare what in my constructional analysis are partitive constructions of heap with quantifier uses of lot. Using a constructional approach to define head, partitive and quantifier constructions, my diachronic data showed that quantifier uses for heap(s) of and lot(s) of arose more or less around the same period, i.e. 1780-1850. Partitive constructions with heap and lot(s) are both attested from around 1200. It thus seems that heap and lot have developed much more on a par than assumed in previous studies.

4.7.6 Conclusions The case study of heap(s) shows that synchronic clines of grammaticality, such as Figure 4.1, cannot be translated into diachronic pathways of change in any straightforward way. Synchronic corpus analysis has shown that at present singular heap (of) has a predominant head use that designates a specific constellation; only in 32.1% of the examples does it function as a quantifier with modifier status. Its position on the synchronic cline suggests that heap has not delexicalized enough to function as a general quantifier or started grammaticalizing later than elements that are located towards the

216Chapter 4: SN-constructions right hand side of the cline. Plural heaps of by contrast has a predominant quantifier use of 66.7%. In Late Old English/Middle English singular heap had a common multitude sense within a partitive construction that for some reason petered out and may or may not have given way to the new but infrequent quantifier use of heap of with modifier status in Modern and present-day English. Syntactic head uses of heap could hence at one point realize the lexical head meaning or the partitive sense, which complicates matters, particularly because the link between these two uses is unclear for heap. In the synchronic data either no partitive uses are attested in the head noun construction, as in the case of heap(s), or the only lexical use is a partitive sense, as in the case of lot(s). In further research the case studies of heap(s) and lot(s) should be supplemented with diachronic corpus studies of other SNs, such as load(s), bunch(s) and pile(s). These will yield more insight into the relative chronology of their respective developments into quantifiers, and potential analogical pull between them. Extensive diachronic corpus analysis might also shed more light on how the SN-construction as a meso-construction, instantiated by load(s)/bunch/heap(s)/lot(s) of, relates to semantically more schematic quantifiers such as many/much. This will involve looking at the influence of discourse schemata on the development of hyperbolic quantifier meaning, as well as differences between singular and plural forms of SNs and in-depth study of the collocational histories of SNs. Section 4.8 will present a comparison between Dutch SNs hoop, hopen, zooitje and stelletje and the (functionally) equivalent English expressions heap, heaps and bunch. These Dutch case studies will further validate the two kinds of grammatical modifier uses posited for English SNs, as they are also displayed by the Dutch data. With regard to valuing uses, it is furthermore confirmed that semantic prosody patterns are a facilitating factor. The main difference between the two languages is that in English the valuing (quantifying) uses form a subset of the layered uses of some SNs, whereas for Dutch zooitje and stelletje it is the exclusive meaning.

4.8 Comparative study: English heap(s) and bunch versus Dutch hoop, hopen and stel, stelletje, zooi and zooitje76 4.8.1 Head and (valuing) quantifier constructions of English and Dutch SNs Dutch SNs have developed the same two modifier uses as the English ones, but, in a slightly different way and to different degrees. English heap(s) will be compared with hoop and hopen, the literal translations of the latter. As I observed in the diachronic case study SNs with 'heap' as their source

4.8 Comparative study 217

semantics crosslinguistically seem to have engaged in grammaticalization processes. For the valuing quantifier construction, English bunch will be compared to stel(letje) and zooi/zoo(i)tje, which are offered by translation dictionaries as functional equivalents. The literal translation of bunch in Dutch is tros, as in een tros druiven ('a bunch of grapes'), which has not engaged in grammaticalization and is hence not used in this case study. Stel literally translates as ‘a number of items that form a set’ or ‘a couple, i.e. 'two people’ and stelletje is its diminutive form. Zooi literally means ‘a quantity of food that is cooked together or is enough for one meal’ and zooitje is its diminutive form (cf. Van Dale Groot Woordenboek van de Nederlandse Taal). Section 4.8.2 will first offer an introduction to the Dutch data and a discussion of several methodological issues involved in the comparative study. Section 4.8.3 will present both a qualitative and quantified corpus study that classifies and quantifies the various lexical and grammatical uses of Dutch SN-expressions in their specific binominal construction, which differs somewhat from the English SN-construction.

4.8.2 Introduction to the data and methodological issues The English data for this comparative study are those that were used for the English synchronic study. The Dutch data were extracted from the Dutch 38 million word corpus of the INL, i.e. Instituut voor Nederlandse Lexicologie (‘Institute for Dutch Lexicology’). The synchronic part of the INL corpus consists of a miscellaneous subcorpus (1970-1995) with 5 million words and a newspaper subcorpus (Meppeler Courant 1992-1995) comprising 27 million words. For each Dutch SN-expression an exhaustive sample was made. The Dutch and English data show several discrepancies in that not all of the Dutch counterparts are literal translations of the English SNs, nor are the number of attestations the same. The composition of the INL-corpus furthermore differs from that of Collins Wordbanks. Finally, the lexicogrammatical pattern in which the SNs occur differs in both languages. Dutch SNs occur in the following generalized construction: determiner + (adjective) + SN + (adjective) + N2, e.g. een hoop verdorde bladeren (‘a heap of wilted leaves’). The construction rarely occurs with a preposition or particle in between the SN and N2, which is typically the case in English with of (which would be van, i.e. ‘of’). In my data the juxtaposed construction is almost exclusively found. The potential premodification slots are similar to those in English. The query in the INL in each case consisted of the SN with no further specifications to allow all kinds of premodification and capture all of the

218Chapter 4: SN-constructions uses in the source construction as well as extended uses. As with the English SNs, we are mainly interested in the binominal source-construction and the target-construction, because this allows me to study collocational differences more closely.

4.8.3 Quantitative and qualitative data analysis The tables in this section display the qualitative and quantified results of the corpus analysis of hoop, hopen, stel, stelletje, stelletjes and zoo(i)tje. The functional analysis distinguishes between head, quantifier, valuing quantifier and ambivalent uses. Collocates of the various uses are quantified in terms of semantic type (i.e. abstract versus concrete) and countability (i.e. count or uncount). The numbers and percentages pertain to attestations within the NP of NP-construction. Uses outside of this construction are discussed separately. The total number of extractions per SN-variant was rather small, which is probably due to the nature of the corpus, but no other Dutch corpora were available at the time.

4.8.3.1 Hoop and hopen: From head to quantifier uses Table 4.15. Quantified collocational profile of hoop

Hoop (105) Animate Concrete/count Concrete/uncount Abstract/count Abstract/uncount Total # Total %

Head

Quantifier

# / 4 8 / /

# 8 6 19 13 40

Valuing quantifier # / / / / 3

12 11.6%

86 83.4%

3 3%

Ambivalent Ambiguous Vague # # / / / / 2 / / / / / 2

0 2%

4.8 Comparative study 219 Table 4.16. Quantified collocational profile of hopen

Hopen (9) Animate Concrete/count Concrete/uncount Abstract/count Abstract/uncount Total # Total %

Head

Quantifier

# / 1 4 / /

# 1 1 2 / /

Valuing quantifier # / / / / /

5 55.6%

4 44.4%

0 0%

Ambivalent Ambiguous Vague # # / / / / / / / / / / 0

0 0%

When we compare the singular and plural variants, we notice that the corpus data for hoop show a predominance of quantifier uses and a very small portion of valuing quantifier uses (3%), while plural hopen displays an equi-probable system of 55.6% head uses, 44.4% quantifier uses and no valuing (quantifier) uses. These relative frequencies have to be put into perspective considering the small data set for plural hopen. Hoop is very similar to English heap. The only difference is that heap has 45.3% head uses versus 11.6% for hoop. Just as with English heap, the quantifier use displays a much broader collocational range than the lexical head use, extending towards abstract collocates, which were not found in the lexical head use. An example of a head use, with an uncount and count concrete N2 is (4.302):77 (4.302) Nu staat er niks meer overeind. Alleen een hoop puin en stenen. De mensen leven tussen de resten van hun woning, die verwoest zijn. (INL) ‘Nothing is upright anymore. Just a heap of rubble and stones. People live in the ruins of their houses, which have been destroyed.’ Collocates in the head noun use are typically negative, referring to the debris of something that was formerly one whole. The semantic prosody is hence predominantly negative. The main use of hoop is that of a quantifier (i.e. 83.4%), which has a hyperbolic value that is very similar to English heap of and heaps of. The following examples illustrate the quantifier use:

220Chapter 4: SN-constructions (4.303) Honderden Koerden hebben door dit alles de bergtocht niet overleefd. Een hoop ellende; de mensen hebben hard hulp nodig. (INL) ‘Hundreds of Kurds did not survive the journey through the mountains because of all this. A heap of misery; those people need help badly.' (4.304) Een oude vliegtuigbom uit de Tweede Wereldoorlog heeft voor een hoop gedoe gezorgd in Eindhoven. (INL) ‘An old airplane bomb from the Second World War caused a heap of fuss in Eindhoven.’ (4.305) Maar het neemt niet weg dat de PvdA een hele hoop overspannen pretenties heeft laten varen. (INL) ‘This does not detract from the fact that PvdA has let go of a whole heap of overstrained pretensions.’ The main difference with the English SNs is that hoop seems to have a more pronounced negative semantic prosody, e.g. ellende (‘misery’) (4.303), gedoe (‘fuss’) (4.304) and pretenties (‘pretensions’) (4.305). Hoop also has valuing quantifier uses with abstract uncount nouns (3%): (4.306) Ik moet zeggen dat ik het een hoop retoriek vind. Als u het een keer fatsoenlijk beredeneerd zou kunnen neerzetten, dan hebben wij er wat aan en kunnen wij er iets mee doen. (INL) ‘I must say that I find it to be a heap of rhetoric. If you could expound your views in a well-reasoned manner, it could be of use to us and we could do something with it.’ Plural hopen shows the same patterns in its head and quantifier uses as singular hoop. Example (4.307) illustrates a head noun use and example (4.308) illustrates the quantifier use. The percentages as they are now show a less high degree of grammaticalization than English heaps. Whereas Dutch hopen has 44.4% grammaticalized quantifier uses, English heaps has 66.7%. (4.307) Ook zag men hier grote hopen cokes liggen die verkocht werden als brandstof. (INL) ‘One also saw large heaps of cokes lying there to be sold as fuel.’ (4.308) Hij moet - om tussen de reuzen niet te verstikken - zijn lot zelf in handen nemen. Met hopen geld en erg ambitieuze plannen. (INL) ‘In order not to suffocate between the giants, he has to take control

4.8 Comparative study 221

of his own destiny. With heaps of money and very ambitious plans.’

4.8.3.2 Stel, stelletje, zooi and zoo(i)tje: Head and valuing (quantifier) uses Table 4.17 represents the quantified results of the qualitative analysis of stel. Table 4.17. Quantified collocational profile of stel

Stel (106) Animate Concrete/count Concrete/uncount Abstract/count Abstract/uncount Total # Total %

Head

Quantifier

# / 16 1 / /

# 63 13 / 6 /

Valuing quantifier # 7 / / / /

17 16%

82 77.4%

7 6.6%

Ambivalent Ambiguous Vague # # / / / / / / / / / / 0

0 0%

Stel has a predominant quantifier use and a very restricted set of valuing uses. In the head noun use it refers to a complete set of something, e.g. (4.309) and (4.310), or to the two members of a pair, e.g. (4.311): (4.309) De kinderen leefden bijna als dieren. Het was er vies en ijskoud. En ze hadden vaak maar één stel kleren. (INL) ‘The children were practically living like animals. It was dirty and ice cold in there. And they often had only one set of clothes.’ (4.310) De aankoop van een volledig stel symbolen voor alle mogelijke godsdienstige en levensbeschouwelijke overtuigingen vergt volgens de schepen een drastische meeruitgave. (INL) ‘According to the alderman the purchase of a complete set of symbols for all possible religious and philosophical persuasions entails a drastic additional cost.’ (4.311) Een stel brede schouders is echt nog geen garantie je hier te kunnen handhaven. (INL) ‘A pair of broad shoulders does not guarantee that you can hold your own here.’

222Chapter 4: SN-constructions In its quantifier use stel can team up with concrete nouns too, e.g. (4.312), but it can also quantify abstract nouns, e.g. (4.313)-(4.314), and animate nouns, e.g. (4.315)-(4.316): (4.312) Er zijn al diverse kostbare ramen vernield, waarvoor in de plaats een stel planken is getimmerd. (INL) ‘Various valuable windows have already been demolished; they are replaced by a bunch/couple of boards’ (4.313) Er komen een stel regels, waar alle speeltoestellen in Nederland aan moeten voldoen. (INL) ‘There will be a bunch of rules that all playground equipment in the Netherlands has to meet.’ (4.314) Dat roept vanzelf een stel vragen op, die ik nu aan de minister van Buitenlandse Zaken wil stellen. (INL) ‘This inevitably raises a number of questions, which I am now going to ask the Minister of Foreign Affairs.’ (4.315) Het is de dag bij uitstek om te laten zien wat voor een goede en aangename dingen er tot stand kunnen komen als een stel mensen zich ergens voor inzet. (INL) ‘It is the best day to show how many good and pleasant things can be established when a number of people dedicate themselves to something.’ (4.316) Toen Joschka het spreekgestoelte beklom, bestormde een stel jongeren in uniform het podium en hield affiches omhoog met het opschrift: “Soldaten zijn moordenaars – nog altijd.” (INL) ‘When Joschka mounted the platform, a bunch/couple of youngsters in uniforms rushed to the stage and held up posters saying: “Soldiers are killers- they still are.’ In its quantifier use, stel has an overall neutral semantic prosody, e.g. planken (‘boards’) (4.312), mensen (‘people’) (4.315), but the contexts often have an air of nonchalance. Valuing-quantifying uses of stel are mainly restricted to animate nouns: (4.317) Sommige mensen denken onterecht dat wij een stel anarchisten zijn. (INL) ‘Some people wrongly assume that we are a bunch of anarchists.’

4.8 Comparative study 223

Just like bunch, stel has a negative semantic prosody in these uses. Let us now look at the diminutive form stelletje. Table 4.18 presents its relative frequencies. Table 4.18. Quantified collocational profile of stelletje

Stelletje (30) Animate Concrete/count Concrete/uncount Abstract/count Abstract/uncount Total # Total %

Head

Quantifier

# / / / / /

# 5 4 / / /

Valuing quantifier # 20 / / 1 /

0 0%

9 30%

21 70%

Ambivalent Ambiguous Vague # # / / / / / / / / / / 0

0 0%

The data set, which is again small, does not display head uses within the NP of NP-construction. In its quantifier use stelletje co-occurs with concrete count nouns, e.g. (4.318), and animate nouns, e.g. (4.319) and (4.320). The SN itself is never premodified, but N2 is often premodified by either qualitative or classifying adjectives, e.g. (4.318) and (4.319). (4.318) Ze waren dan ook verbaasd toen ze een stelletje oude kiezen vonden. Het bleken heel bijzondere exemplaren. (INL) ‘They were surprised then when they found a couple/bunch of old molars. They turned out to be very special specimen.’ (4.319) Het gebeurt niet elke dag dat je met een stelletje Amerikaanse soldaten een partijtje kunt hinkelen. (INL) ‘It does not happen every day that you can play hopscotch with a bunch of American soldiers.’ (4.320) In Frankrijk kwam een stelletje windsurfers op het idee om een groot basin in een sporthal te bouwen. (INL) ‘In France a bunch of windsurfers came up with the idea to build a large reservoir in a sports centre.’ As a quantifier, stelletje suggests that the speaker quantifies N2 in a casual way and expresses slight irony or disapproval with regard to the entities referred to. Nevertheless, the main aim of the SN-expression remains quan-

224Chapter 4: SN-constructions tification of N2. In the case of human beings it is often more difficult to tease the quantifier meaning and the valuing meaning apart, as was the case with the English SN bunch of. In the valuing use the main aim of the SN is to evaluate the referents of N2. Hence, this use mainly appears in comparative contexts, indicated by verbs and markers such as ‘portray as’ or ’look like’. Examples of valuing uses involve animate N2s, which are negatively evaluated: (4.321) …Greenpeace naar voren brengen als een volwassen, positieve actiegroep en niet als een stelletje linkse terroristen- het beeld dat de Franse pers zo graag schetst. (INL) ‘…portray Greenpeace as a mature, positive action group and not as a bunch of leftist terrorists- the image the French press likes to sketch.’ (4.322) De verwende vedetten bewogen zich als een stelletje vrijetijds voetballers op een zwoele avond in mei en dat hautain gedrag werd afgestraft. (INL) ‘The spoiled stars moved about like a bunch of amateur football players on a sultry evening in May and this arrogant behaviour was punished.’ (4.323) Met behulp van de nonnen wordt het stelletje ongeregeld omgetoverd tot een swingend ‘hip hop’ gospelkoor. (INL) ‘With the help of the nuns the motley crew is turned into a swinging hip hop gospel choir.’ (4.324) …. Omdat er een stelletje van die onsmakelijke, zwaar gesponsorde cyclisten aankomt… (INL) ‘…because a bunch of those distasteful, heavily sponsored cyclists is arriving...' The semantic prosody is emphatically negative here, e.g. linkse terroristen (‘leftist terrorists’) (4.321), and contexts involve several negatively coloured cues, e.g. verwende vedetten (‘spoiled stars’) and hautain gedrag (‘arrogant behaviour’) (4.322). Again there is a difference between the premodification patterns in the Dutch and English data. Whereas English valuing quantifying SNexpressions are very often themselves premodified by evaluative adjectives that strengthen the valuing function, Dutch valuing SNs are not, or at least not in these Dutch data. Evaluative modifiers are typically situated in front of N2, e.g. (4.321), (4.322) and (4.324).

4.8 Comparative study 225

In examples such as het stelletje ongeregeld (‘the motley crew’) (4.323), the N2-slot is filled in by the (deverbal) adjective ongeregeld (‘disorderly’), which forms an idiomatic-like expression together with stelletje. Plural stelletjes, just like plural bunches, did not show any grammatical uses. The fact that stelletje, with its very specific original lexical meaning, has managed to develop a quantifier use with a casual tone, or an emphatically negative valuing quantifier use, may be explained by the need for such expressive means of expression, with semantic prosody as a facilitating factor. Let's switch to zooi(tje). There were no attestations of zooi(tje) in the original head noun sense, where it refers to food that is boiled or cooked to constitute one meal. Zooi as such did not appear in the data. Let us take a look at the diminutive form zoo(i)tje. Table 4.19 sums up the relative frequencies of its ʊfewʊ attested uses. Table 4.19. Quantified collocational profile of zoo(i)tje

Zoo(i)tje (15) Animate Concrete/count Concrete/uncount Abstract/count Abstract/uncount Total # Total %

Head

Quantifier

# / / / / /

# 3 / / / /

Valuing quantifier # 12 / / / /

0 0%

3 20%

12 80%

Ambivalent Ambiguous Vague # # / / / / / / / / / / 0

0 0%

Just as with stelletje, the data set for zoo(i)tje does not contain any head noun uses, but only grammatical ones. It also typically teams up with animate N2s in both quantifying and valuing uses. Just as with stelletje, it is often difficult to distinguish strictly between quantifying and valuing uses. Examples (4.325) and (4.326) illustrate more purely quantifying uses of zootje: (4.325) Huismussen, ringmussen, heggemussen en drie soorten mezen verdringen zich om het staand buffet. Een zootje merels en een enkele lijster mengen zich ook nog in het gewoel. (INL) ‘House sparrows, tree sparrows, hedge sparrows and three kinds of titmice crowd around the standing buffet. A bunch of blackbirds and a single thrush join the bustle.’

226Chapter 4: SN-constructions (4.326) Je koopt voor honderd gulden een zootje van die bekers en je hebt de hele avond muziek. (INL) ‘For 100 guilders you buy a bunch of these cups and you can listen to music all night long.’ In these examples zootje is ultimately concerned with quantifying N2. In (4.325) the notion that the birds are being quantified is reinforced by een enkele lijster (‘a single thrush’). At the same time, zootje contributes to the casual style of these examples. Zootje proves to be semantically even more similar to bunch of than stelletje, with contexts referring to disorderly behaviour, such as gewoel (‘bustle’) in (4.325). In the valuing use zootje co-occurs only with animate nouns. The most frequent collocate, however, is ongeregeld, which accounts for 61% of all the valuing uses. This collocate also appeared with stelletje, though only one time, i.e. in (4.323). Zootje ongeregeld constitutes an idiomatic expression and is included in the dictionary as such. As remarked earlier, translating dictionaries translate the expression as a motley bunch/crew (cf. Van Dale Groot Woordenboek Nederlands-Engels). Examples (4.327) to (4.329) illustrate this use. Een zootje ongeregeld is also used to refer to a situation that is a mess or to a mess as such, as in (4.330) (4.327) Orde en discipline zou eigenlijk beter thuishoren bij een konservatieve partij dan bij het linkse zootje ongeregeld, vinden ze. (INL)‘Order and discipline would suit a conservative party more than the left-wing motley crew , they think.’ (4.328) Jazz-liefhebbers genieten van de experimenten en ergeren zich aan die gemiste noot van het optredend zootje ongeregeld. (INL) ‘Jazz lovers enjoy the experiments and are annoyed at the note missed by the performing motley crew.’ (4.329) De stemming van dit zootje ongeregeld is en blijft stijgen, de crew zakt nimmer af naar een bedenkelijk peil. (INL) ‘The spirits of this motley crew are rising and continue to rise. The crew never slips down to a questionable level.’ (4.330) “Soms was het een zootje ongeregeld, maar we deden er van alles en hielden elkaar aan de gang.” (INL) ‘At times it was a mess, but we did all kinds of things there and kept each other going.’ In these examples quantification is backgrounded to a very large extent which makes these uses predominantly valuing. The semantic prosody is

4.8 Comparative study 227

negative or the contexts at least imply great casualness or disapproval. Again, the fact that zooitje, like stelletje, despite its very specific source meaning displays such a high number of especially valuing (quantifying) uses may be explained by the emotive prosody which promotes it as a valuing expression. The main difference with the valuing use of stelletje is that zootje does occur with premodifiers in front of the SN, just like English bunch, e.g. linkse (‘left-wing’) (4.327), optredend (‘performing’) (4.328). The expression zootje ongeregeld is mostly preceded by a definite determiner, when it refers to people, and preceded by an indefinite determiner in the sense of a mess, as in (4.330). The following table sums up the relative frequencies of the nondiminutive and diminutive Dutch SNs. Table 4.20. Degrees of grammaticalization of Dutch SNs Hoop

Hopen

Stel

Stelletje Zooi Zoo(i)tje

Head

11.6%

55.6%

16%

0%

/

0%

Quantifier

83.4%

44.4%

77.4%

30%

/

20%

Valuing (Q)

3%

0%

6.6%

70%

/

80%

Ambivalent

2%

0%

0%

0%

/

0%

If we compare these relative frequencies, we see some significant differences. In all cases the diminutive form displays significantly more valuing uses than the non-diminutive ones, which only have a very minimal percentage of the latter uses. Both diminutive forms furthermore only display grammatical uses in the data sets and these grammatical uses are very much connected with a specific semantic prosody and a general air of casualness, particularly in the valuing uses. We have also seen that non-diminutive stel too has an air of nonchalance in its quantifier use and a negative semantic prosody in its valuing-quantifying use. Yet, stelletje and zooitje display these valuing undertones much more clearly and in many more attestations. I argue that this effect is mainly due to the influence of the diminutive suffix which further attracts negative semantic prosody as a facilitator of attitudinally subjectified SN-uses. In what way can the diminutive be said to enhance or bring about this valuing use then? Geeraerts (1993) and Bakema (1998) have both studied the semantic structure of the Dutch diminutive analyzed in terms of a proto-

228Chapter 4: SN-constructions type category with central senses and extended senses. In Bakema (1998) these extended senses are argued to come about through metonymy and metaphor, leading to concretizing and evaluative nuances respectively. The latter are of interest here and involve derogatory, appreciative, relativizing and approximative subsenses, based on the conceptual metaphors ‘small is minor’, ‘small is beautiful’, ‘small is not bad’ and ‘small is not much’ respectively. In the case of stelletje and zootje the diminutive mainly seems to add derogatory and relativizing meaning, which also gives rise to the negative semantic prosody of the valuing uses. Stelletje and zootje in presentday Dutch can only be used in a valuing quantifier meaning and no longer display layering with head noun uses or pure quantifier uses. This is different from English where SN-expressions that can function as valuing quantifiers generally have layering with another grammatical uses, as well as lexical head uses and ambivalent uses.

4.8.4 Discussion and conclusion In conclusion, the comparison of SNs in English and Dutch shows that in both languages two similar kinds of modifier uses can be found, i.e. a quantifier and a valuing (quantifier) use. There are, however, a number of differences between the two languages. Firstly, in contrast with English, Dutch SNs display specialization in the sense that one set of SNs realizes mainly the quantifier construction, e.g. hoop and hopen, while the other is almost exclusively restricted to the valuing quantifier construction, i.e. stelletje and zootje. The English data did not show such a functional crystallization, except that the valuing use was restricted to the singular variants of the English SNs. Secondly, while in English the valuing use is mainly a result of semantic shifts within the meaning of the SN-expressions themselves, such as attitudinal subjectification, paralleled by semantic prosody patterns, Dutch valuing uses primarily come about through the (subjectified) semantics of the diminutive suffix, which in its turn attracts a specific semantic prosody. In addition, we again see how semantically rather specific elements, such as zootje and stelletje in Dutch and bunch in English, despite their lexical specificity can grammaticalize because of strong pragmatic motivations related to the need for expressivity. The case studies with the Dutch SNs validate the valuing construction. Micro-constructions with stelletje and zooitje also confirm that semantic prosody patterns are a facilitating factor in the emergence of the constructional use. Taking into account the main findings of the synchronic and diachronic English and synchronic Dutch case studies, we can now make a general

4.8 Comparative study 229

conclusion about the SNs looked at in this study. Section 4.9 will present such a conclusion in terms of synchronic layering.

4.9 Synchronic layering of SN-constructions Based on extensive synchronic and diachronic data I argue that English SNexpressions display a number of synchronically coexisting uses, i.e. lexical head uses, grammatical quantifier and valuing uses, as well as ambivalent uses. The main dividing line between these uses is head versus modifier status of the SN. In the head use the SN functions as the head noun of the binominal NP, while the of-phrase acts as a postmodifier with which it is coextensive, e.g. a bunch of grapes. In the quantifier and valuing construction, however, the SN together with of forms a string or SN-expression which has modifier status, and N2 is the head, as in a bunch of people or a bunch of crap. Tests for (non-)head status include verbal and pronominal concord patterns with either the SN or N2. As mentioned earlier, verbal concord is difficult to check since it requires the SN-construction to have subject function and the SN and N2 to differ in number. Moreover, as discussed in earlier, verb concord can also be determined by other principles than strict grammatical concord. Another test for the status of the SN is substitutability by a monomorphemic functional equivalent, e.g. Heaps/loads/a bunch of people arrived can be substituted by ‘many people’. Another constituency test is the potential preposing of the of-phrase, which is possible when SNs have head noun status, but not in the modifier uses, e.g. of bananas a bunch is possible, whereas *of crap a bunch is not (cf. Traugott 2010b). In the modifier uses the SN-expression can also be phonetically less salient, with the vowel in of often reduced to schwa. Head versus quantifier and valuing status is furthermore signalled by different collocational sets, as visualized in Table 4.12. In addition to describing these collocationally constrained constructions as synchronically attested SN-uses, I have also argued that the head versus modifier status they display are the result of diachronic shifts, i.e. (ongoing) processes of grammaticalization, delexicalization and subjectification. The diachronic case study in Section 4.7 illustrated this. Synchronic variation is taken to reflect diachronic developments, with present layering as a result. Layering as such is a sign of ongoing grammaticalization (Hopper 1991). The semantic and syntactic reanalysis from head to quantifier is not just a metaphorization of the collocates of the head SN-construction. Such metaphorization is attested in the corpus data, but pertains to instances of the ambivalent category or the head noun category. An example of metaphorization is (4.331), cited earlier:

230Chapter 4: SN-constructions (4.331) Striding past the compost heap of nostalgia comes Christopher Lloyd, now in his seventies and still coming up with new ideas for his garden at Great Dixter in Sussex. (CW-Times) Example (4.331) has an abstract uncount N2, but, within the vehicle of the metaphor, the SN remains head and does not shift to quantifier status. This is opposed to examples such as (4.332), cited earlier: (4.332) The wine, from low-yielding Barossa Valley vines, has heaps of peppery flavour, made more complex and softened by ageing in American oak. (CW-OZnews) In grammaticalized example (4.332), with an abstract N2, the status of the SN-expression is not head, but quantifier, as shown by the fact that heaps of can be replaced by a schematic quantifier such as much, which is not possible in example (4.331). As discussed in Section 4.5, the synchronic quantifier and valuing (quantifying) uses furthermore display reflexes of decategorialization, interpreted in a positive way as proposed in Section 3.4.3. The SNs lose properties typical of the noun category and develop properties typical of the quantifier category. Quantifying and valuing SN-constructions, for instance, show restrictions in determiner use and are restricted to either the plural or singular variant of the SN. The premodification patterns in grammaticalized uses are moreover restricted to two types, each reinforcing the constructional semantics of either the quantifier or valuing use, e.g. whole in a whole bunch of people further marks the quantifier meaning and complete in a complete load of bullshit intensifies the evaluation. Such premodifiers differ significantly from premodification patterns in lexical head uses, which do accentuate the independency and nounness of the SN, e.g. a foot-high pile of manure. Decategorialization also implies paradigmaticization of the SN-expressions, which come to be part of a cohesive paradigm of quantifier and valuing-quantifiers respectively. We have seen that for several SN-expressions coalesced spellings are attested, both in the Collins Wordbanks corpus and on the Internet, though these are always restricted to very informal registers and are typical of many routinization processes (see Joseph 2007). In addition, several SN-expressions show syntactic context expansion, in that their grammaticalized meaning can occur outside of the NP of NPconstruction (cf. Himmelmann 2004), most frequently in adverbial or degree modifier constructions, as in thanks a bunch, as well as some elliptical uses, e.g. [Musicals]. I've been in loads (CW-UKspoken). Furthermore, some renewal effects may be at work, e.g. loads of and load of as more expressive variants of lots of and a lot of. As observed earlier, more in-

4.9 Synchronic layering of SN-constructions

231

depth diachronic case studies are necessary to determine analogical pull between various SN-expressions. The emergence of quantifier and valuing uses is accompanied by delexicalization and subjectification. The latter involves deictic and attitudinal subjectification, as described by Traugott (1989: 34-35). Deictic subjectification consists of the semanticization of scalar implications in SNs through pragmatic enrichment in specific contexts that foreground these quantitative implicatures to the detriment of the original lexical semantics. This entails a shift from the SN contributing to propositional content to expressing meaning that indexes speaker-relatedness, in that quantifier meaning involves a speaker assessing size relative to a scale (cf. Traugott 2010b). Other semantic changes bleach the SN of its lexical and more concrete meaning and enrich it with grammatical and more abstract quantifier meaning. Whereas the original meaning of load for instance refers to something that is transported and by implication pertains to a mass or set of entities that are spatially and temporally contiguous, the quantifier meaning of load(s) of not only lacks this ‘being transported’ meaning, but can also quantify items that are spatially and/or temporally non-contiguous, e.g. I have loads of friends all over the world. In the valuing (quantifying) uses, in addition to deictic subjectification, attitudinal subjectification comes into play, i.e. a shift to evaluative meaning that is expressive of the speaker’s stance on or attitude towards the propositional content. From (subjectively) assessing the relative positioning of elements on a scale of measurement, we move to an evaluation of the entities referred to. Valuing (quantifying) uses involve collocational reclustering based on affective or evaluative features shared between collocate and node, manifested in specific semantic prosody patterns. The English as well as the Dutch case studies make it clear that semantic prosody can be a facilitating factor in the development of the grammaticalized valuing use of SN-expressions, rather than a blocking factor of delexicalization and grammaticalization. Yet, this is how semantic prosody is generally interpreted, since it imposes restrictions, while grammaticalization and delexicalization are mostly associated with extension and generalization. I argue that the main motivation behind the delexicalization and grammaticalization of SNs is the language user’s search for expressive means of quantification and valuing-quantification. The individual discussions of the SN-expressions contain ample reference to emphatic and creative contexts. This expressivity is especially clear in the vague examples, in which speakers (consciously) exploit the tension between the lexical and grammatical layers of meaning of SN-expressions in a very visible way and tend to combine the premodification patterns and other behaviour of head and grammaticalized uses, e.g. compost heap of nostalgia. Grammaticalized uses themselves can rekindle the original lexical semantics of SN-

232Chapter 4: SN-constructions expressions, as in a barrel-load of chances, hat-load of chances and lorryloads of flattery. In addition, one might say that the (expressive) power of SNs as a means of quantification and valuation lies in the tension between their lexical and grammatical semantic potential. This tension can differ in intensity depending on the specific use of the SN, but in theory it is always there, even though it is strongest and metalinguistically the most visible in many of the vague SN-constructions. This, in my opinion, seems to suggest that the semi-stable layered SN-systems may never go on to completion, in the sense of all of these individually-layered SNs developing an exclusively grammatical quantifier meaning. This brings us back to Traugott’s levels of schematicity in grammaticalization and the interpretation of degrees of grammaticalization as functional approximation of a target construction. At the micro-constructional level it can be argued that all SNs display collocational profiles specific to them. Individual developments of SN-expressions have also given rise to a SN meso-construction. The various SN-expressions have approximated macroconstructional quantifiers such as many/much to different extents in terms of semantic schematicity, usage contexts, etc. Pile of and bunches of differ significantly from bunch of and loads of for instance in their overall degree of grammaticality. On the meso-constructional level the SN-expressions share their periphrastic outlook and several traits of so-called incomplete decategorialization, such as the singular/plural contrast and potential (restricted) premodification as in a whole lot. More locally, some SNexpressions may have benefited from analogical pull by other SNexpressions. Some SN-expressions approximate behaviour typical of macroconstructional quantifiers such as many and much. For instance, heaps of and loads of allow relative quantifier uses in addition to their absolute quantifier use. Example (4.333) illustrates an absolute quantifier use of heaps of: (4.333) Czar Oak, a late maturing New Zealand-bred stayer with heaps of potential but also a major leg problem (CW-OZnews) where heaps of merely indicates the inherent magnitude of the potential Czar Oak has. Examples (4.334) and (4.335) operate differently: (4.334) What’s interesting is how many of the sexual researchers and observers were driven by self-interest? Heaps of them at least. (CWOZnews)

4.9 Synchronic layering of SN-constructions

233

(4.335) Certainly, heaps of kids are aware of it, carry it out and are brilliant but others seem to have lost the plot. (CW-OZnews) In both examples heaps of construes relative quantification, i.e. it designates the quantity of N2 by comparing how this profiled mass stacks up to a reference mass. In (4.334) heaps of them is the answer to a question about a proportion and in (4.335) others explicitly refers to the remainder of the reference set of kids. At least loads of and lot(s) of also show this potential for relative quantification. Relative quantifier uses were not as clearly attested for bunch of or other SNs that developed a strong valuing use, which may be explained by the constructional interference of that valuing meaning, which seems to relate more typically with direct assessment of degree, e.g. They are a bunch of morons. The same would then go for load of. Probably, SN-quantifiers will rely somewhat more on contextual clues and factors like stress on the SN-expression than more schematic quantifiers do in order to function as a relative quantifier in an unambiguous way. Elliptical uses of SN-expressions, as in [musicals]. I’ve been in loads (CWUKspoken) and substitutability by a schematic quantifier are further signs of approximation of schematic quantifiers. In sum, the SNs looked at in this study have all to various extents approximated quantifiers at higher levels of schematicity, while also retaining micro-and meso-constructional particularities. This partial approximation may be interpreted as a limitation, but it is also one of the strong points of SN-quantifiers and valuing expressions and ties in with the functional motivation behind their grammaticalization and subjectification.

Chapter 5 Small SN-constructions 5.1 Introduction: Similarities and differences between regular SNs and SSNs as sets of meso-constructions In addition to the SN-expressions discussed in Chapter 4, this study also looks at a semantically more specific set of SN-expressions which all denote a small amount of something, i.e. bit of, flicker of, jot of, scintilla of, scrap of, shred of, skerrick of, smidgen of and whiff of. As noted in previous chapters, I refer to them as small SNs. As was the case with the regular SNs, SSNs will be looked at mainly in the context of the (binominal) source construction, which consists of a SSN, potentially preceded by a determiner, quantifier or (adjectival) premodifiers and followed by of and N2, which can also be preceded by adjectives or determiners. Examples with bit are (5.1) and (5.2), in which the SSN-expression is in bold, N2 is in italics and modifiers are underlined: (5.1)

Maybe we just suffered a wee bit of bad luck today because we led for most of the match before giving a goal away late on. (CWSunnow)

(5.2)

The nice thing about my colleagues is that they do want to be kind to the candidates so this is leading them to a little bit of a misleading economy with the truth or perhaps overstating of the truth. (CW-BBC)

In terms of the constructional framework proposed in Chapter 2 the relationship between the set of SNs studied in Chapter 4 and the SSNs discussed here can be said to be meso-constructional in nature. The various regular SNs and the constructions they occur in form one set of similarly behaving constructions and so do the SSN-expressions. The TNconstructions, discussed in Chapter 6, are also meso-constructionally related to both sets, though the semantic link between the two sets of SNexpressions is obviously greater than that between either of the two sets of SN-expressions and TN-constructions. The focus of the present chapter is on the description of SSNs as such, but also on the relation between the two sets of SN-expressions. Parallels with TN-constructions will be discussed in Chapter 6. This chapter's main focuses is on synchronic corpus analysis. Diachronic evidence derives from queries from the CEMET and CLMETEV

5.1 Regular SNs and SSNs

235

corpora spanning 1640-1920, the OED and MED, and Traugott (2008b) and (2010b), who discusses a piece of, bit of and shred of. Apart from bit of, my queries in CEMET and CLMETEV yielded few attestations. Smidgen/skerrick/scintilla of did not yield any attestations. Table 5.1 represents the token frequencies of the attested SSN-expressions: Table 5.1. Frequencies of SSNs in CEMET and CLMETEV (1640-1920) Bit of

Jot/iota of

Flicker of

Scrap of

Shred of

Whiff of

522

11

14

42

6

18

I argue that SSN-constructions with the SSNs listed above have all at least partially grammaticalized. The grammatical reanalysis involves a rebracketing from [SSN] + [of + N2], e.g. a scrap of paper, to [SSN + of] + [N2], e.g. (not) a scrap of evidence. The original head noun function of the SSN is hence grammatically reanalyzed as a quantifier and in some cases further extended uses have developed in which the grammaticalized meaning is exported to other syntactic contexts (cf. Himmelmann 2004). The primary grammaticalization towards quantifier status is related to deictic subjectification, i.e. the semanticization of the scalar implicatures of SSNexpressions, similar to regular SN-expressions. In addition, some SSNstrings show signs of attitudinal subjectification as well as the development of intersubjective values. Interestingly, precisely the subjectified and intersubjective uses of the SSN-expression bit of for instance constitute potential zones of constructional-semantic overlap with a set of uses attested for TN-constructions, e.g. mitigating uses such as I’m a bit of a gypsy (CWOZnews) and well, it was kind of a miracle (CW-Today). The grammaticalization of these SSN-expressions means that they, just like other quantifying SN-expressions, fulfil the same basic function of indicating size as semantically more schematic quantifiers, such as few and little, without, however, necessarily having acquired the full syntactic behaviour of the latter (cf. Langacker 1991, 2009 and to appear). In this grammaticalized use N2 has head status (cf. Brems 2003a and Brems 2007). Examples (5.3) to (5.7) illustrate quantifier uses of SSNexpressions: (5.3)

Sources say there has been a bit of hold-up assembling a cast. (CW-OZnews)

(5.4)

It's all I look forward to; at the end of the day it gives me a flicker of comfort. (CW-UKbooks)

236Chapter 5: Small SN-constructions (5.5)

Conquest had shown not a skerrick of remorse. (CW-OZnews)

(5.6)

This is all done without the slightest, merest ghost of a scintilla of irony. (CW-Times)

(5.7)

If our attention lapses for even one generation, a smidgen of time on any geological scale, transformations in the land that are truly dangerous could catch us unawares. (CW-USbooks)

With the exception of bit of, all of the SSN-expressions in this study have a low frequency, both within the SSN-construction and outside of this construction (see Table 5.2 in Section 5.2). Table 5.1 shows that they have also been infrequent in earlier periods of English. They are hence not common in the English lexicon. This raises some important theoretical questions, since, especially in usage-based approaches, high frequency and lack of lexical specificity are generally considered as either an important indicator of or prerequisite for delexicalization and grammaticalization. Increase in frequency is moreover often considered a consequence of grammaticalization (e.g. Bybee 2003a and 2003b, Heine et al. 1991 and Hoffmann 2004). How can these infrequent items have acquired grammatical quantifier status, if they cannot have engaged in processes of routinization, semantic attrition and grammatical reanalysis, characteristic of gradual, default grammaticalization (cf. Haiman 1994, Bybee 2003a and Hopper and Traugott 2003)? Other explanations hence need to be sought. One such alternative is offered by Hoffmann (2004), who explains the grammatical status of low-frequency complex prepositions, such as in presence of and at cost of, as being licensed by analogy between the infrequent patterns they occur in, and very frequent structurally similar sequences, such as in front of. At first blush this approach could work rather well for low-frequency SSNs too. In the 18th century a whole set of binominal quantifiers emerged, including the SSNs discussed in this chapter (cf. Traugott 2010b). Of these bit of started developing a partitive and quantifier use about a century earlier than the other, less frequent SSNs (cf. Traugott 2010b). In keeping with the proposals made by Hoffmann (2004) one could hypothesize that bit of acted as an attractor model for the infrequent SSNs which then attained their grammatical quantifier status by analogy with bit of. However, qualitative and quantitative corpus analyses of infrequent SSN-constructions suggest a more complex picture than simple structural analogy with bit of. Synchronically, infrequent SSNs individually display very specific restrictions with regard to polarity sensitivity (cf. Israel 2004), as well as collocational patterns and semantic prosodies (cf. Louw 1993 and Bublitz 1996). For instance, shred is typically associated with negative polarity contexts, as in there is not a shred of evidence.

5.1 Regular SNs and SSNs

237

By contrast, a bit of is not restricted in these ways; hence, how, or to what extent, can one posit that it functions as analogical model for infrequent SSNs, which do display these constraints, which moreover appear to be vital for their quantifier function? As with the regular SN-expressions, SSNs as quantifiers provide new expressive means of quantification and evaluation. However, the case studies of SSNs show that slightly different linguistic systems and factors come into play in the grammaticalization of these meanings, mainly polarity sensitivity. Analogy between constructions as a driving factor in grammaticalization was referred to also in Chapter 3, where analogical models were argued to be situated on the various constructional levels (i.e. the macro-, meso- and micro-level). In addition to each (S)SN displaying microconstructional particularities in behaviour, a more schematic mesoconstructional SN-construction has emerged and some SNs have even acquired features of macro-constructional quantifiers such as many and much or little and few or some in the case of SSNs. In the case studies on SSN-constructions, the issue of analogy and its precise nature is a very central one in that it might be criterial to the question of whether we can talk about a grammaticalization process as such for infrequent SSN-expressions, or instead have to explain their grammatical status as merely copied from some model construction. These case studies will also draw attention to the fact that one has to be very careful in identifying what served as a model for analogical extension, both in terms of describing all the specific factors that come into play and that have to be incorporated into the model construction, and in terms of deciding on the specific level of schematicity at which analogy operates. SSN-constructions also confront us more explicitly with the question of the difference between grammatical reanalysis and simple metaphorization of collocates without grammaticalization. In this chapter I will argue that simple structural analogy with a high frequency pendant cannot be the sole factor in the emergence of infrequent SSN-quantifiers, and that more complex analogies with different quantifier models are at work. Instead of claiming that only one highly schematic structural schema, a + SSN + of, as instantiated by a bit of, has analogically pulled the infrequent SSNs towards quantifier status, I will argue that the infrequent SSN-expressions synchronically divide into two groups according to their polarity sensitivity, which also seem to have been part of their onset constructions in developing a quantifier function. In this respect, polarity sensitive quantifiers some and any also have to be brought into the picture as distant analogical models. However, in contrast with some and any, which can be used generally in quantifying contexts, the infrequent SSNs are synchronically typically restricted to very specific discourse contexts. Pragmatico-semantic values such as downtoning versus amplifying

238Chapter 5: Small SN-constructions (cf. Quirk et al. 1985), and the establishment of new collocational sets and semantic prosodies interact with the specific polarity preferences of infrequent SSNs, relative to each SSN. These pragmatically enhanced polarity contexts seem to function as highly "constrained pragmatic and morphosyntactic contexts” for the delexicalization and grammaticalization of the SSNs at hand (Traugott 2003: 645). The split in the pragmatic and collocational properties of SSNquantifying expressions gives support to Noël’s (2005) partial criticism of Hoffmann (2004) that constructions serving as direct input and output of grammaticalization are not highly schematic in nature, as suggested by Hoffmann (2004), but semantically very specific and (at least partially) substantive. In this chapter, questions of constructional schematicity and analogy, as discussed by Hoffmann and Noël, are looked at in the light of the empirically observable properties of SSN-constructions in corpus data. The structure of this chapter is as follows. First, Section 5.2 will discuss the data set used for this study, the choice of precisely these nine SSNexpressions and several methodological problems. In Section 5.3 I will present in more detail the proposals made by Hoffmann (2004) in his study of low-frequency complex prepositions, with a discussion of his claim that grammaticalization, rather than applying to substantive constructions may also apply directly to constructional schemas. After a discussion of these more general theoretical questions, I will move to SSN-specific issues in the next sections. In Section 5.4 I will introduce the notion of polarity sensitivity as a potentially very specific feature of SSNs in their functioning as a quantifier and for other extended uses. Taking into account this basic polarity distinction, Section 5.5 will discuss the results of the corpus analysis of nine SSN-expressions, paying attention to their main semantic, pragmatic and syntactic features. Bit of is discussed first and will be considered as a possible analogical model in the vein of Hoffmann (2004); then the infrequent SSNs of the two polarity groups are presented. Section 5.6, then, will tie the descriptive and theoretical observations into a conclusion, integrating Hoffmann’s ideas on constructional analogy with the more specific pragmatics of polarity sensitive contexts, collocational changes, semantic prosody and the (deictic subjectification of the) scalar semantics of SSNs (cf. Israel 2004, Traugott 2010b, Brems 2003a and 2007).

5.2 Description of the corpus data and methodological issues The synchronic corpus used for the analysis is again the Collins Wordbanks corpus, which contains some 57 million words from spoken and written registers and from different regional variants, i.e. British, American and

5.3 Frequency and analogy

239

Australian English. As mentioned in Section 5.1, I will also bring in diachronic corpus data and evidence from various sources. Table 5.2 shows the token frequencies of the nine SSNs within the NP of NP-construction and absolute frequencies of nominal uses throughout the corpus. The latter indicate the number of tokens of the SSN-expressions in a size noun-related meaning, using the query SSN/NOUN, which extracts all nominal uses of the SSNs. First the raw frequencies are given and then frequencies per million words. Verbal uses of shred wrongly included in these extractions were filtered out manually, as well as nominal meanings that are not related to the SSN-meaning and are hence irrelevant for this study. Instances of scrap with the meaning of ‘fight’ for instance are excluded. In the case of smidgen, spelling variants mentioned in the OED, such as smidgeon, yielded 4 hits, but these do not feature in the SSNconstruction or lack context and are hence not included in the study. Spelling variant smidge remained unattested. In the case of skerrick, none of the spelling variants mentioned in the OED yielded attestations in the Collins Wordbanks corpus. Table 5.2. Token frequencies of nine SSNs bit

jot

flicker

scintilla

scrap

shred

skerrick

smidgen

whiff

Within the SSNconstruction

4976

8

62

5

85

46

4

12

118

Per million words

86.68

0.13

1.08

0.08

1.48

0.80

0.07

0.21

2.06

Nominal uses

17 829

36

87

5

280

48

5

15

133

Per million words

310.61

0.62

1.52

0.09

4.88

0.84

0.09

0.26

2.32

For the analyses, extractions were made for each SSN by means of two queries: SSN and SSN + of, with no further specifications. This netted all types of premodification of the SSN as well as the element following of, so as to capture potential variety in the determiner and polarity markers preceding the SSN. By not specifying a word class after of or the SSN in the queries, I allowed non-nominal elements to follow the expressions. The aim

240Chapter 5: Small SN-constructions of the query without of was to capture uses outside of the SSN-construction so that these could be incorporated into the qualitative analysis. For the low-frequency SSNs the data analysis uses exhaustive data samples; for bit the total amount of 4976 attestations was restricted to a random sample of 200 analyzable concordance lines. No register or regional restrictions were incorporated into any of the queries, since the SSNs looked at, except for bit, are too infrequent to impose such restrictions. The lowfrequency SSNs entail several obvious methodological problems for corpus analysis. Most importantly, the lack of corpus data necessitates putting the observations and conclusions made from the corpus data into sufficient perspective. The choice of precisely these eight low-frequency SSNs and bit as highfrequency SSN is partly explained by some of the observations first made in Traugott (2005) and published as Traugott (2010b), who discusses "brief case studies" of shred of and bit of in the larger context of the interaction between grammaticalization and (inter)subjectification. Following Israel (1996), these SSN-expressions78 are referred to as polarity (sensitive) items (henceforth P(S)Is)79 as well as quantifiers. The eight low-frequency SSNs looked at in the present chapter were chosen primarily because they were infrequent, but still yielded a workable set of data. In addition, they were chosen because of their suspected polarity sensitivity, which, following Traugott (2005), was hypothesized to be a potentially important factor in their grammaticalization. Bit, jot, shred and smidgen were chosen from Traugott’s (2005: 6) list of PSIs (which is “drawn from Israel’s extensive list of PSIs in English”), with bit and smidgen classified as positive PIs and shred and jot as negative PIs. Whiff comes from Israel (2004: 716) where it is classified as a positive PI, and flicker, scintilla, scrap and skerrick were added because they all incorporate small size implications and are rare. Given its early grammaticalization and high frequency, bit was considered to be the only true candidate for a high frequency SSN with sufficient semantic schematicity.

5.3 Hoffmann (2004): Frequency and analogy in low-frequency complex prepositions Hoffmann (2004) discusses the theoretical and (corpus-)methodological problems in positing grammaticalization processes for low-frequency complex prepositions that consist of a preposition-noun-preposition string (henceforth P-N-P strings), such as in presence of, at cost of and by dint of.80 Because of their low discourse frequency it is disputable that they can undergo the processes of routinization and coalescence, with semantic and phonological attrition and grammatical reanalysis, so typical of gradual

5.3 Frequency and analogy

241

grammaticalization (e.g. Hopper and Traugott 2003). Importantly, Hoffmann (2004) at all times remains cautious in claiming that the infrequent PN-P strings indeed have grammatical(ized) status, admitting that this hypothesis is largely intuition-based (e.g. Hoffmann 2004: title; 204). Interestingly, his efforts to confirm this intuition lead to the conclusion that the grammatical status of the infrequent complex prepositional patterns can only be partially supported by such established factors as (relative) internal invariation, decategorialization, and wide distribution of the infrequent patterns across genres as well as relative lack of combinatorial restrictions in the form of collocating verbs or specific contexts of use. In addition, a distinction between conceptual frequency and (absolute and relative) lexical frequency is introduced that more accurately assesses the importance of frequency for the entrenchment or saliency of certain lexical items, and its influence on grammaticalization (Hoffmann 2004: 180, 190, 199 and further, 202, 203, 204 and 206). Nevertheless, all of these are in their own way inconclusive, mostly because low-frequency items per definition do not yield sufficient results to confirm these tendencies in a decisive manner. In his search for a frequency and routinization-independent criterion that can account for the grammatical(ized)81 status of low-frequency complex prepositions Hoffmann arrives at the notion of grammaticalization by analogy (Hoffmann 2004: 195). Infrequent complex prepositions have acquired grammatical status because of their analogy to structurally similar variants with a high frequency, such as in front of and in view of. Infrequent P-N-P strings then copy the grammatical status of the frequent ones, which do seem to have grammaticalized gradually, with the P, N and P progressively coalescing and losing in semantic and formal autonomy, so as to become one complex unit. Hoffmann (2004) is cautious about the general applicability of his analogy proposal to other infrequent constructions, saying that this “would require further analysis” (Hoffmann 2004: 195). Still, it is explicitly proposed for “low-frequency units of language” beyond the ones studied in Hoffmann (2004: 171, 172 and 195), and, as hinted at in Section 5.1, it could work well for infrequent SSNs too. Nevertheless, a few important differences between the case of low-frequency SSNs and low-frequency complex prepositions are apparent and have to be addressed before the detailed discussion of the corpus analysis is presented in Section 5.5. In addition, there are some consequences of Hoffmann’s proposal that are problematic for important strands in grammaticalization research as such (see for instance Hopper and Traugott 2003 and Traugott 2003 as reviewed in Section 3.2.2), as well as for the specific grammaticalization of SSNs. Firstly, there is a difference between infrequent P-N-P strings and the infrequency of the SSNs. Hoffmann’s 132 complex P-N-P strings are infre-

242Chapter 5: Small SN-constructions quent as sequences, but the nominal elements in them need not be infrequent at all and in fact only sometimes are. In presence of, under mistake of, in proof of for instance are among the most infrequent P-N-P-strings analyzed in his study (with only 5 attestations each), but the nouns incorporated in the complexes are not themselves infrequent or lexically specific and uncommon. Some infrequent P-N-P sequences are merely uncommon variants of more frequent complex prepositions, e.g. in face of is a variant of the much more frequent in the face of. Others are claimed to be unrelated to more common complex prepositions, and do not contain rare lexical items either, e.g. in presence of and in awareness of (Hoffmann 2004: 182185). Infrequent nouns only appear in by dint of and in contradistinction to according to Hoffmann (2004: 182).82 Just like the infrequent SSNs, the rare nominal elements occurring in P-N-P-sequences are largely restricted to that specific construction (see Table 5.2). In contrast with Hoffmann’s low-frequency P-N-P strings, not only the SSN-patterns as such are infrequent, but almost all of the SSNs in them are infrequent too, as well as rare (e.g. scintilla and skerrick). Hoffmann (2004) does not distinguish between infrequency of the P-N-P pattern and infrequency of the nominal item it incorporates (Hoffmann 2004: 170 and 172173). This is nevertheless an essential difference, because it is the infrequency and semantic specificity of the lexical item that can preclude it from engaging in typical grammaticalization by gradual routinization. In principle, the infrequent SSN-expressions relate to frequent ones, such as bit, just like an infrequent P-N-P string such as by dint of relates to an established frequent complex preposition such as in front of. The other infrequent P-NP strings should, strictly speaking, be taken out of the equation in comparing Hoffmann’s case with that of the infrequent SSNs. Secondly, for infrequent P-N-P strings straightforward analogy with a frequent complex prepositional model can be envisaged, if only because most of them are largely variants of more frequent ones, even in cases where Hoffmann does not claim this. The similarity between frequent and infrequent P-N-P strings is very strong. The infrequent P-N-P patterns have not only copied grammatical status, but also seem to have copied all of the semantic and formal features that go with it. Hoffmann does not point out any specific differences in the functioning of infrequent P-N-P patterns in comparison with the frequent complex prepositions. The only difference seems to be the frequency with which they are attested in corpus data. Thirdly, even though Hoffmann’s analogy proposal is tentative in its formulation and range of application, it nevertheless leads to some farreaching consequences for certain conceptions of grammaticalization. Hoffmann for instance hypothesizes that, if his particular analogy proposal is sound,

5.3 Frequency and analogy

243

[c]ertain aspects of grammaticalization rely much less on the nature and context-dependent use of individual content words than previously assumed. In such an approach, grammaticalization would result in the establishment of constructional schemas whose slots can be filled with suitable lexical items. (Hoffmann 2004: 195).

Firstly, as discussed in Section 3.2.3, this statement is at right angles to an important tendency in grammaticalization theory, advocated by Traugott (2003) and Hopper and Traugott (2003), who explicitly argue that historically grammaticalization operates in specific contexts that trigger off delexicalization and grammaticalization in the first place. Consider Hopper and Traugott (2003: 18), cited earlier, who define grammaticalization as the change whereby lexical items and constructions come in certain linguistic contexts to serve grammatical functions, and once grammaticalized, continue to develop new grammatical functions.

Several studies of grammaticalization have stressed that the specific semantics of grammaticalizing lexical items can play an important role in their concrete development within a specific construction, for instance through lexical persistence (Hopper 1991 and also Brems 2003a; see Section 4.3 on pile versus heap). Secondly, and tying in with the previous remark, Hoffmann’s proposal entails that grammaticalization can lead directly to the establishment of non-substantive constructional schemas (Hoffmann 2004: 195). In fact, the potential interaction between schematic constructions and grammaticalization is currently a subject of debate (e.g. Haspelmath 2004, Himmelmann 2004, Hopper and Traugott 2003, Lehmann 2005, Noël 2005 and Traugott 2003; see Section 3.2.3.2) and whether fully schematic constructions can be the output of grammaticalization processes has not yet been established (e.g. Bisang, Himmelmann and Wiemer 2004 and Trousdale 2008b). It seems that grammaticalization studies so far have mainly looked at cases in which it works on and results in a specific substantive construction (but see Lehmann 2008 on information structure for instance). Current parameters for grammaticalization moreover might only be attuned to such processes. The question whether schematic constructions can serve as the direct output of grammaticalization processes can hence not easily be answered positively but should be addressed by constructional approaches.

5.4 Polarity sensitive contexts in SSN-constructions In order to see how polarity sensitivity can be an important factor in these studies, let us first look at a definition of the concept as provided by Israel

244Chapter 5: Small SN-constructions (2004).83 Polarity items are defined as “a class of constructions which do not themselves express negation or affirmation, but which are restricted to sentences of one or the other polarity” (Israel 2004: 701-702; italics mine). Polarity sensitivity is hence a distributional phenomenon that distinguishes between negative PIs and positive PIs, depending on the polarity they are most sensitive to, since PSIs “vary in their sensitivities” and polarity sensitivity does not have to be absolute (cf. Israel 2004: 713). With regard to how certain linguistic units become sensitive to polarity in the first place, Israel (2004: 709) rather unhelpfully notes that “[i]intuitively, they are constructions whose use or interpretation is sensitive to polarity.” The main question with regard to the corpus study is whether polarity sensitivity might systematically create constrained environments that trigger off specific processes of grammaticalization for the various SSNs (cf. Traugott 2003 cited earlier). If so, the emergence of such polarity restrictions has to be accounted for by explaining its impact on the grammaticalization of the individual SSNs and it will have to be built into a model of grammaticalization by analogy. The following types of negation markers are taken into account in the corpus study: negative adverbs such as, not, without, never; near-negative adverbials, such as scarcely and hardly; negative quantifiers such as not a single; and negative verbs such as fail to, deprive of any, etc, as illustrated in examples (5.8) to (5.11), with the negation markers underlined. Eckardt (2004) would call several of these expressions markers of weak negative polarity sensitivity, e.g. near-negative adverbials and verbs: (5.8)

One lady commented approvingly that she could not detect a skerrick of grease or fat in the dish. (CW-OZnews)

(5.9)

You've never shown me a scrap of kindness, so why would I expect you to now? (CW-UKbooks)

(5.10)

But inquiries by the highest-ranking police in the State have failed to produce a shred of evidence. (CW-OZnews)

(5.11)

Though they have enjoyed scarcely a jot of Bath's success, there should be a fellow feeling between the clubs. (CW-Times)

Israel (2004: 713) stresses that “a comprehensive theory of sensitivity must be attuned to the needs of individual polarity items” In his own way, and from the specific point of view of polarity research, Israel hence refers to the importance of the individual properties of lexical items within the larger contexts they function in, i.e. either negative or positive polarity sensitivity. This ties in with Hopper and Traugott (2003) about specific conditions for

5.4 Polarity sensitive contexts in SSN-constructions

245

grammaticalization. Goldberg (2006), in the context of usage-based construction grammar, similarly argues for the incorporation of item-specific knowledge, in addition to generalizations, and stresses the interaction between these two. All of this then gives support to the intuition expressed in Section 5.3 that the context, substantive constructions and the lexical items in them all contribute significantly to the changes they may be involved in.

5.5 Individual discussions of SSN-constructions The corpus analysis of nine SSNs presented in this section, distinguishes between SSNs that have head status and quantifier status. As in Chapter 4 and Brems (2003a), grammaticalization refers to the actual grammatical reanalysis of (S)SNs from (lexical) head noun status to (grammatical) quantifier status, in which the original structure [(S)SN]+[of+N2] is rebracketed as [(S)SN+of]+[N2]. Similar to SNs, this (covert) structural reanalysis of SSNs is argued to be brought about by delexicalization, i.e. loss of lexical content of the (S)SN, which allows the specific lexical semantics of (S)SNs to make way for a more generalized quantifier meaning and hence combine with a larger range of N2-collocates (cf. Traugott 2010b). As concluded at the end of Chapter 4, the functional approximation of SNs towards more schematic quantifiers has at present resulted in a layered semi-stable system of SN-constructions which includes head and quantifier uses, as well as some ambivalent cases, with some SNs synchronically displaying a higher degree of grammaticality than others. Basing myself on the synchronic and diachronic corpus studies presented in the next sections, I argue that SSNs have also been reanalyzed as quantifiers. Synchronically, SSN-expressions display the following semantic and formal reflexes of grammaticalization. Firstly, the main formal reflex of the reanalyzed status of SSNs is substitutability by such more schematic quantifiers as a little/a bit of, in the pattern (indefinite) determiner + SSN + of + N2, e.g. a smidgen of trust (‘a little trust’). This is impossible for the head use of SSNs, without changing the meaning of the latter, e.g. The shred of skin (*’the little skin’). This substitutability test formally reflects the functional unity that developed from the chunking of previously “autosemantic” signs, very typical of grammaticalization (e.g. Lehmann 1985: 308 on coalescence; cf. Bybee 2003a, see Chapter 3). Secondly, one can clearly observe semantic generalization, implemented among other things, as in Chapter 4 and Brems (2003a), as extension of collocational range, e.g. a whiff of rotten eggs/garlic in the head use versus a whiff of sentiments/economic slowdown in its quantifier use, where N2 can be concrete as well as abstract, count and uncount (cf. Himmelmann’s

246Chapter 5: Small SN-constructions 2004 host-class expansion). Traugott (2010b) also clearly illustrates the process of subjectification in bit and shred from what she calls pre-partitive over partitive, extended partitive and quantifier stages. Her category of extended partitive sometimes also contains what I would label as quantifiers, e.g. I'll tell you a bit of a secret (1833 Clifford, Traugott 2010b: 47). Examples of quantifiers in Traugott (2010b) are Why yes it is a bit of a kind of a sword (1794 Morton, Traugott 2010b: 47) and can be parahrased as 'somewhat of/rather/quite'. In my classification such examples go beyond pure indication of small size, but have additional (inter)subjective values to do with hedging. The boundaries between the various categories we use vary slightly, but this does not detract from the similarities between our analyses and lines of reasoning. Thirdly, the synchronic SSN-quantifiers show decategorialization, in the sense of the loss of features prototypically associated with the (more) lexical category the grammaticalizing unit formerly belonged to (cf. Hopper 1991) and the development of characteristics associated with the target construction. This is very similar to the decategorialization described for regular SNs in Chapter 4. SSNs lose in nounness and are typically attested only in the singular in their quantifier use, whereas in the head noun construction SSNs may be singular or plural. They are furthermore mainly restricted to indefinite determiners, e.g. not a shred of evidence, except in constructs such as not the remotest jot of nostalgia, in which the definite article is triggered off by the intervening superlative adjective. The occurrence of adjectival premodifiers again might, at first sight, appear to be counterevidence for the decategorialization claim, but as with the regular SNs this need not be so. Expressions such as not the remotest can be considered to constitute a single negative polarity item. Israel (2004: 716) includes several PIs that are multiple-word strings, e.g. in the slightest and be all that. Furthermore, Brems (2003a) and Chapter 4 of this study have argued that grammaticalized SN-uses allow for two restricted types of premodification, parallel to the two types of grammaticalized uses, i.e. quantitative and valuing uses. Such premodifiers hence do not intrude upon the internal cohesion of the grammaticalized SSN-structure either, but rather reinforce or echo it, as in a little whiff of insurgency (CW-NPR) or a little bit of soppy sentiment (CW-OZnews). These premodification patterns are clearly different from the ones attested in head uses, which do emphasize the independent head noun status of the SSN, e.g. the first whiff of cranberry sauce or A sixteenth century scrap of parchment (CWUKephem). Finally, for some SSNs syntactic extension of the grammaticalized meaning to other contexts is attested, which Himmelmann (2004) lists as a parameter for grammaticalization. Concord patterns, such as subject-verb concord and pronominal concord (with either SSN or N2) are again difficult to check since the SSN and N2

5.5 Individual discussions of SSN-constructions

247

always agree in number in my data. Furthermore, subject-verb concord is not just necessarily a reflection of grammatical status, but can be influenced by notional concord or even proximity concord. In addition to head uses and quantifier uses, the tables in Section 5.5 also represent ambivalent uses of SSNs, i.e. ambivalent between a head and quantifier reading, often involving metaphorization and/or lexical persistence. As with the regular SNs, this category subsumes both vague and ambiguous instances. In Section 5.5 the corpus results of the nine SSNs are discussed systematically in terms of the following features: collocational range, with possible collocational reclustering, semantic prosody, amplifier and downtoner uses of the quantifying SSN-expressions and polarity sensitivity. In addition to these systematic features, SSN-specific characteristics such as lexical persistence are discussed when relevant. As with the regular SNs, collocational reclustering refers to the fact that some SSNs in their quantifier use may display specific and sometimes rather restricted collocational patternings in the sense of significantly favoured collocates, or, in the case of bit of, a specific semantic prosody. Pragmatically, SSNs can function either as downtoners or as amplifiers, the two main types of pragmatic marker distinguished by Quirk et al. (1985): “Amplifiers scale upwards from an assumed norm”. They are further subdivided into maximizers and boosters. Maximizers “denote the upper extreme of the scale”, e.g. absolutely, completely, extremely, in all respects, etc. (Quirk et al 1985: 590). Boosters “denote a high degree, a high point on the scale” (Quirk et al 1985: 590), e.g. badly, deeply, highly, terribly, a great deal, etc. The second type of pragmatic markers are downtoners, which “have a lowering effect, usually scaling downwards from an assumed norm” (Quirk et al. 1985: 590). They further divide into minimizers and diminishers. Minimizers “are negative maximizers”, meaning “(not) to any extent” (Quirk et al. 1985: 597), e.g. barely, hardly, at all, in the least, etc. Diminishers “scale downwards and roughly mean ‘to a small extent’” (Quirk et al. 1985: 597), e.g. only, merely, quite, slightly, partly, etc. Downtoning seems the pragmatic function most in keeping with the original semantics of all SSNs, but we will see that, due to polarity effects and ironic reversals of meaning, amplification may be involved as well, especially in subjectified uses of SSNs. Section 5.5.1 starts off with bit of, which in an approach such as Hoffmann’s could be regarded as the frequent construction serving as analogical model for the infrequent SSNs. Sections 5.5.2.1 and 5.5.2.2 deal with synchronically positive polarity SSNs whiff of and smidgen of, and sections 5.5.3.1 to 5.5.3.6 deal with synchronically negative polarity SSNs jot of, flicker of, scintilla of, scrap of, shred of and skerrick of.

248Chapter 5: Small SN-constructions 5.5.1 Bit of Table 5.3. Relative frequencies of bit of Head use

Quantifier use

Ambivalent use

Total

#

%

#

%

#

%

#

0

0

199

99.5

1

0.5

200

Bit clearly is a deverbal nominal derivation from to bite; hence the first meaning of a bit (of) is ‘a bite of’ or ‘a mouthful of’. From denoting a small morsel of food, the expression generalized to non-food: "a small piece, a fragment (of something); a small portion or quantity, a little" (cf. New Shorter Oxford English Dictionary, henceforth NSOED and Traugott 2010b). Unlike the other SSNs looked at, bit is not uncommon and was never infrequent in the English lexicon, which makes it more conducive to delexicalization and ensuing grammaticalization through routinization. Bit has a high text frequency in synchronic data, with 4976 attestations of the SSN-construction bit of and an overall frequency of 17 829 tokens. Except for one ambivalent case, all instances of bit of included in the sample are grammaticalized. In its strict quantifier use bit of typically has downtoning quantifier semantics. Through delexicalization the collocational range of N2 has become very wide, with a predilection for (often abstract) uncount N2s: (5.12)

Despite the African violet's reputation as a somewhat difficult houseplant a bit of inside knowledge does wonders. (CWOZnews)

(5.13)

Nature is not averse to a bit of wife or children bashing and certainly has no conscience in that regard. (CW-OZnews)

(5.14)

And yet, on reflection perhaps a little bit of soppy sentiment might be just what sport needs at the moment. (CW-OZnews)

(5.15)

There’s a little bit of horseradish sauce if you want it. (CWUKspoken)

In this quantifier use bit of is usually preceded by the indefinite article a, and together they form a coalesced string, though premodifiers can be inserted between the article and bit, as in (5.14) and (5.15).

5.5 Individual discussions of SSN-constructions

249

As shown by the N2s in (5.12)-(5.15), bit of synchronically does not have a specific semantic prosody. Inside knowledge in (5.12) for instance is positive, but wife and children bashing in (5.13) is clearly negative. The quantifier use of bit of is synchronically likewise polarity-independent. Both positive and negative polarity contexts occur in my data set, with a preference for positive polarity. Only three examples in the synchronic data sample associated with negative polarity. Here are examples with positive (5.16) and negative polarity (5.17): (5.16)

I wrapped the bogey in a bit of Rizla cardboard. (CW-UKmags)

(5.17)

Lovey, it makes not one bit of difference to me whether you're hungry or not. (CW-USbooks)

In (5.16) bit of has a downtoning value, whereas in (5.17) the negation turns the meaning into an emphatic one. Contrary to most of the other SSNs discussed later on, bit of has engaged in further grammaticalization developments beyond the development of a quantifier meaning, which can be said to be the result of subjectification and intersubjectification (cf. Traugott 1988, 2005) leading to meanings regulating attitudes and ones that are centred on the addressee respectively. Bit of in this respect is similar to regular SNs in displaying a path to more ideational quantifier uses and one to more attitudinally subjective valuing-quantifier uses. The subjectified and intersubjectified uses of bit of I will discuss here concern cases where N2 is preceded by the indefinite determiner a(n), with in some cases an adjective in between the determiner and bit. N2 is typically negatively evaluated. Subjectification and intersubjectification furthermore interact and are not easily distinguished between. I will first discuss the attitudinally subjectified use of bit of. Some typical examples of this are: (5.18)

Century 21, opposite World Trade Centre. A bit of a trek from Soho, but it's worth it. (CW-UKmags)

(5.19)

Feeling as you do it would be unfair to marry your girlfriend. She deserves better than this and that is why you are feeling a bit of a rat. (CW-Sunnow)

(5.20)

It is hard to avoid the feeling that video games are a bit of a con; an electronic circus which transfixes the audience while the ringmaster empties its pockets. (CW-UKmags)

250Chapter 5: Small SN-constructions In these subjective cases the downtoning quantifier semantics are reversed and translate into an emphatic evaluation, e.g. a bit of a con is ‘rather a big con’. In these uses the referent of N2 can be animate (5.19) or inanimate (5.18) and (5.20); in each case N2 is typically evaluated negatively by normal standards. In these examples different mechanisms often interlock, i.e. understatement, irony, hedging and face work (Brown and Levinson 1978, 1987). The latter two also link up with intersubjectivity. The speaker in each case reverses the downtoning quantifier semantics to convey a rather emphatic evaluation. At the same time, s/he uses the hedging capacities of bit of as a safety net in case of possible face damage to the hearer, allowing the ironized meaning to be undone. Despite some intersubjective features, this use can hence be labelled a subjective use in keeping with Traugott’s definition of attitudinal subjectification, because its prime function is to express the speaker’s (negative) attitude (cf. Traugott 2005 based on 2003; Sperber and Wilson 1991) The bit of string in these cases also introduces a valid categorization, e.g. in (5.18) it really is a long way from Soho and in (5.20) video games are a real con according to the speaker. In addition to these subjective reversal cases, bit of can also be used intersubjectively as a hedge, with downtoning semantics. Again we mostly have structures in which a typically negative N2 is preceded by a(n): (5.21)

I'm going to going to test you now Gwyneth so you we've got an examination coming up now bit of an oral examination. (CWUKspoken)

(5.22)

So I'm just wondering whether they're not putting themselves at a bit of a disadvantage by just assuming that they know. (CWUKspoken)

(5.23)

Wright: Isn't there a bit of a double standard there? (CW-NPR)

Examples typically come from the spoken subcorpora and are part of conversations or interaction between various speakers. Following Traugott (2005), I view this hedging use as intersubjective, since it conveys the speaker’s concern for and alertness to the possible impact of his or her words on the hearer. As mentioned earlier, distinguishing between subjective and intersubjective meaning is not always straightforward, since some intersubjective features can often be detected in subjectified uses, but also because the interpretation relies very much on contextual and prosodic clues that are not always available in the corpus data.

5.5 Individual discussions of SSN-constructions

251

Finally, a bit can also function as a degree modifier with adjectives (5.24) and can be used independently as an answer (5.25) (cf. Traugott 2005: 9), e.g. (5.24)

We were a bit slow getting ourselves into gear. (CW-OZnews)

(5.25)

Q: Did you like the movie? A: a bit (cited in Traugott 2005:8)

In these extended uses of is typically left out.84 In conclusion to this section it can be noted that bit of is synchronically used both in negative and positive polarity contexts, which suggests that it is independent from either polarity context in order to function as a quantifier. In the diachronic data sets from CEMET and CLMETEV 43 of the 522 tokens appear in a negative polarity context, mostly in the period spanning 1780-1920. (5.26)

There wasn't a bit of necessity for the toast (CLMETEV, 18501920)

The association with negative polarity hence only emerges around the 18th and 19th century for quantifier uses of bit of. Many instances of negative polarity pertain to the quasi-idiomatic expression not a bit of it, namely 23. (5.27)

As for telling him about anything, not a bit of it! (1850-1920)

The infrequent SSNs synchronically divide into two groups for their functioning as a quantifier based on their polarity sensitivity. One subset typically functions in positive polarity contexts, whereas the other one typically occurs in contexts of negative polarity. I will first discuss the positive polarity SSNs and then the negative ones.

252Chapter 5: Small SN-constructions 5.5.2 Positive polarity SSNs 5.5.2.1 Whiff of Table 5.4. Relative frequencies of whiff of Head use

Quantifier use

Ambivalent use

Total

#

%

#

%

#

%

#

48

40.68

60

50.85

10

8.47

118

Whiff originally means "A slight puff or gust of wind" (cf. OED). Unsurprisingly, the collocational range of whiff in its lexical head use mainly extends to gasses and odours, especially malodorous ones: (5.28)

If you're lucky you'll also catch a whiff of cat's pee! (CWUKephem)

(5.29)

It smells good, too, not one whiff of rhino dung or the scent of the sun on an old canvas tent, as you might expect from the name. (CW-UKmags)

(5.30)

It is as if the turkeys, having voted for Christmas, are now troubled by the first whiff of cranberry sauce. (CW-Times)

When grammaticalized, whiff is typically preceded by the indefinite article, possibly with an adjective in between. In this use it can co-occur with a wide variety of abstract collocates, both count and uncount: (5.31)

But as to what Jerry Brown will do, whether he's still got a little whiff of insurgency left. (CW-NPR)

(5.32)

You gave him a whiff of power and it's turned him into a corrupt megalomaniac. (CW-UKbooks)

(5.33)

Everyone with a whiff of sanity in his or her veins should have no traffic with that way. (CW-UKbooks)

(5.34)

He passes on to his only daughter strong moral values and a whiff of nonconformity. (CW-UK-mags)

5.5 Individual discussions of SSN-constructions

253

Pragmatically, whiff of functions mostly as a downtoner, with whiff often describing a minimal amount of N2, e.g. (5.32), (5.33), or a persistent small amount, e.g. (5.31), emphasized by adverbials, choice of verbs and adjectives, stress, etc. e.g. still left in (5.31), and stress on whiff in (5.33). In addition to clear lexical and grammatical uses, whiff of also has several ambivalent instances, which mostly seem to be vague. In most of these cases N2 is an abstract noun, but construed in such a way that it can evoke concrete olfactory associations in its connotations. These ambivalent instances typically have the original lexical semantics alluded to by the use of verbs, adjectives or other parts of speech in the co(n)text, e.g. (5.35)-(5.38): (5.35)

It's not as if I ever had any emotional attachment to kitchen-sink or rough-stuff Englishness in the first place, but I can still detect the malodorous whiff of adolescence oozing off men whose history is turning into nothing more than a prelude to some misbegotten selfhumiliation. (CW-UKmags)

(5.36)

Most dahlias are tainted with a whiff of vulgarity, their outsized blooms being associated with the obese marrows and inedible onions lined up at a village harvest festival. (CW-UKmags)

(5.37)

In Berlin, where I was the week before last, there's a whiff of Weimar in the air. (CW-NPR)

(5.38)

I yearned for the proximity of the old adobe, the talismanic power of the crucifix and candles, the ancient whiff of miracle. (CWUSbooks)

Most of these cases describe an atmosphere, usually rather nostalgically, as in (6.37) or disapprovingly, as in (5.35) and (5.36). The synchronic corpus data do not include the degree modifier uses or independent answer use that bit has. In the diachronic data set whiff occurs only 18 times and in only one case occurs with negative polarity. The latter use is a head use, with quantitative implicatures, i.e. (5.39)

[...] without a whiff of sulphur to save you from infection (CLMETEV, 1811).

All attestations of whiff in CEMET and CLMETEV are head or ambivalent uses.

254Chapter 5: Small SN-constructions 5.5.2.2 Smidgen of Table 5.5. Relative frequencies of smidgen of Head use

Quantifier use

Ambivalent use

Total

#

%

#

%

#

%

#

0

0

12

100

0

0

12

Smidgen and its spelling variants smidgeon, smidgin, smitchin and smidge are from a colloquial register and originally North American. Its origin is uncertain, as well as the period in which the SN-use arose. As mentioned earlier, the CEMET and CLMETEV corpora did not yield any attestations and the OED is likewise inconclusive. It probably is an alteration of the English dialect word smitch, which means "soiling mark" (cf. OED, NSOED and Webster online). The corpus attestations did not confirm the OED indication American English, since 3 attestations come from the American subcorpus, 8 from the British subcorpus and one attestation comes from the Australian subcorpus. All 12 attestations are quantifier uses, both with concrete nouns and abstract nouns: (5.40)

A dash of love, a pinch of understanding, a smidgen of trust -- all the ingredients for a lasting friendship. (CW-USephem)

(5.41)

Jefferies is terrified of even a smidgen of what he calls “negative thought”‘. If it is not fun, it is not happening. (CW-Times)

Smidgen typically has the meaning of a very minimal amount of something, hence the frequent occurrence of even, e.g. (5.41). The recipe context of example (5.40) is similar to that of examples with heaps of, i.e. "Take a classic of the Western dramatic canon and curry it” (as one waspish critic put it), “throw in heaps of Urdu dialogue and voila, you have a dashing, Indo-Western theatrical experience". Smidgen has a fairly unrestricted collocational range, allowing both abstract and concrete nouns, and it does not have a specific semantic prosody either. Pragmatically, smidgen functions as a downtoner. The corpus data for smidgen display two degree modifier uses, e.g. a smidgen too long (CW-OZnews) and a smidgen more (CW-Sunnow). 5.5.3 Negative polarity SSNs

5.5 Individual discussions of SSN-constructions

255

5.5.3.1 (not a) skerrick of Table 5.6. Relative frequencies of skerrick of Head use

Quantifier use

Ambivalent use

Total

#

%

#

%

#

%

#

0

0

4

100

0

0

4

With only 5 attestations in the entire corpus, and only one of these occurring outside the SSN-construction, skerrick is the most infrequent of all SSNs in this study. Skerrick was not attested in the diachronic data set. All of its synchronic attestations are quantifier uses in the SSN-construction. Skerrick is originally a dialectal form and is now mainly associated with Australian and New Zealand colloquial language. All corpus attestations are from Australian newspapers. Its etymology is unknown (cf. NSOED). The OED relates skerrick to scuddick, which denotes an extremely small coin or amount. The OED also lists as a first, but obsolete, meaning of spelling variant scurrick "a half-penny", illustrated with a quotation from Grose’s 1823 Dictionary of the Vulgar Tongue.85 The earliest examples cited in the OED are from the 1800s, with the example of 1841 possibly hinting at the monetary meaning (i.e. Not worth a skerrick by R.W. Hamilton in the OED). When grammaticalized in the SSN-construction, skerrick means "a very small amount of". N2 is typically an abstract uncount noun; two examples refer to evidential/epistemic notions, i.e. (5.43) and (5.44), and another to an emotional concept, (5.42): (5.42)

Conquest had shown not a skerrick of remorse. (CW-OZnews)

(5.43)

Mr Bullock says there is not one skerrick of evidence to involve this man. (CW-OZnews)

(5.44)

If there is one beach on the Sunshine Coast that can have a skerrick of a claim to being undiscovered, this is it. (CW-OZnews)

In one case N2 is a concrete uncount noun: (5.45)

One lady commented approvingly that she could not detect a skerrick of grease or fat in the dish. (CW-OZnews)

256Chapter 5: Small SN-constructions Looking at the exhaustive corpus examples (5.42)-(5.45), I would argue that the semantic prosody of skerrick is fairly neutral. Skerrick clearly is a NPI and appears with negation and emphatic words, such as one instead of a, which produces an emphatic meaning of zero attestation of N2, i.e. it indicates that no quantity or amount of N2 whatsoever is attested within a specific context, e.g. not one skerrick of evidence in (5.43). No degree modifier uses or independent answer uses are attested.

5.5.3.2 (not a) jot of Table 5.7. Relative frequencies of jot of Head use

Quantifier use

Ambivalent use

Total

#

%

#

%

#

%

#

0

0

7

87.5

1

12.5

8

With eight attestations in its SSN-use, jot of is also very infrequent. In the diachronic data jot is attested 11 times. Originally, jot or iota referred to the smallest letter of the Greek alphabet or the smallest written part of any writing (cf. OED and NSOED). In my diachronic data set the first attestations of quantifier uses are from the period 1780 onwards. In the synchronic data no head uses were attested. In its quantifier use it typically has uncount abstract nouns as collocates, of which the semantics are rather far removed from the original graphological semantics of jot: (5.46)

Yet, although this is a serious cook's kitchen, there's not a jot of stainless-steel professionalism about it. (CW-UKmags)

(5.47)

When he is away from Billingshurst he feels not the remotest jot of nostalgia for the place. (CW-Times)

(5.48)

Though they have enjoyed scarcely a jot of Bath's success, there should be a fellow feeling between the clubs. (CW-Times)

The semantic prosody of jot of is on the whole fairly neutral. This SSN, like skerrick, is used in emphatic, negative contexts, where negation combines with superlatives, e.g. (5.47), or plain negation as in (5.46) or near-negative adverbials, e.g. (5.48). Also note the presence of

5.5 Individual discussions of SSN-constructions

257

adversative and concessive connectors, such as yet, although and though, equally reinforcing the contexts in which jot of appears. Just like skerrick, jot gives rise to an emphatic meaning, which seems slightly stronger than that expressed by skerrick. Jot has adverbial uses that are similar to the degree modifier use of bit. Adverbial uses of jot occur with verbs, as in (5.49) and (5.50), and adverbs, as in (5.51). Jot seems to retain its negative polarity preference and sometimes appears with emphatic markers, e.g. (5.50). Of is dropped: (5.49)

The woman who has given every indication of not caring a jot for her appearance appeared with her hair freshly coiffed and coloured. (CW-OZnews)

(5.50)

It means only that she has not changed one jot. (CW-Today)

(5.51)

Just as every seeker who comes to me for a prophecy believes what he wants to believe and not one jot else. (CW-UKbooks)

I found 22 instances of such adverbial uses, 2 of which with adverbs, the others with verbs. All of the latter belong to one of the following two semantic fields, which suggests strong collocational preferences: either ‘change (for the better)’, i.e. advance, alter, help, improve and change or ‘be/consider important’, e.g. care for, matter, bother, worry about and give for. In the diachronic data set 10 out of 11 tokens associate with negative polarity. Eight tokens are quantifier uses and the other three are ambivalent. Only one quantifier use occurred in a positive polarity context, i.e. his every jot of endurance (CLMETEV, 1850-1920).

5.5.3.3 (not a) scintilla of Table 5.8. Relative frequencies of scintilla of Head use

Quantifier use

Ambivalent use

Total

#

%

#

%

#

%

#

0

0

5

100

0

0

5

Just like skerrick, scintilla only has 5 overall attestations in the Collins Wordbanks Corpus, all of which feature in the SSN-construction. It was not attested in the diachronic data set. Scintilla derives from Latin and origi-

258Chapter 5: Small SN-constructions nally means “a spark” (OED; the earliest example dates from 1570. The earliest example of a quantifier use seems to date from 1734). Again no head uses were attested within the synchronic corpus data. In the grammaticalized attestations scintilla has three epistemic collocates, (5.52)(5.54), and one emotional collocate, i.e. (5.55): (5.52)

If only the loyalist paramilitaries would move, then that would deprive the IRA of any scintilla of justification for holding on to their weapons. (CW-Times)

(5.53)

This is all done without the slightest, merest ghost of a scintilla of irony. (CW-Times)

(5.54)

He was treated as if he was a murderer. Yet there is not one scintilla of evidence to justify his arrest. (CW-Sunnow)

(5.55)

He said he received a “very warm welcome, not a scintilla of resentment”. (CW-Times)

Scintilla has a neutral semantic prosody; only resentment is negative, in addition to such negative verbs as deprive (5.52). Similar to jot of and skerrick of, scintilla of functions in emphatic, negative contexts of zero-attestation, with negative verbs as in (5.52), negation (+superlative), as in (5.53)-(5.55), and contexts with concessive and adversative markers, e.g. (5.52) and (5.54). No degree modifier uses or independent answer uses are attested.

5.5.3.4 (not a) shred of Table 5.9. Relative frequencies of shred of Head use

Quantifier use

Ambivalent use

Total

#

%

#

%

#

%

#

3

6.52

43

93.48

0

0

46

Shred derives from late Old English scread(e), corresponding to Old Frisian skred which means "hair-cutting, clipping of coin" and Middle Low German schrot and schrat, which refers to a "cut-off piece" (NSOED). Its literal meaning in present-day English is "a fragment, a broken piece of

5.5 Individual discussions of SSN-constructions

259

something; specifically (rare) a shard of pottery’; also ‘a finely cut strip of some material, especially paper, the peel of a fruit or vegetable, etc." (cf. NSOED). Shred and scrap (discussed in Section 5.5.3.5) share a number of semantic features in their literal meaning and in fact also share some collocates, both in their literal and quantifier use. Shred is, however, much more grammaticalized than scrap with 93% of the attestations in my corpus data being quantifier uses, while scrap is still predominantly used lexically. In the diachronic data set shred was attested six times. Three of these are head uses. Two head uses and one quantifier use appeared in a positive polarity context; the other quantifier uses and one head use occurred in a negative polarity context. All of the attestations are from the period 1780-1920. In its synchronic quantifier use, shred has one very strong collocate, i.e. evidence, which occurs 17 times (out of the total 43 quantifier uses). Cognate epistemic concepts also occur frequently, e.g. fact, truth, doubt and proof. Other semantic sets are self-respect, self-esteem and decency; anxiety, comfort and dreaminess. Overall, shred has a neutral to positive semantic prosody. Examples of literal head uses are: (5.56)

That accounts for the plaster you saw on his wrist and the shred of skin and dried blood under her fingernails. (CW-UKbooks)

(5.57)

The flame he got by touching a match to a small shred of birch bark that he took from his pocket. (CW-USbooks)

In its quantifier use, shred typically functions in negative and/or emphatic contexts, designating a minimal amount of the noun following of. 5.59 illustrates shred of in an affirmative context. (5.58)

The allegation is unfair and untrue and was made under protection of parliamentary privilege without a shred of supporting evidence. It is plainly an act of cowardice. (CW-OZnews)

(5.59)

I submit the reason she couldn't remember the brownies was that, in order to salvage some shred of self-respect, she had to suppress knowledge of behavior which she had been conditioned to view as vile and inexcusable. (CW-NPR)

Neutral epistemic collocates are particularly frequent in negative contexts, whereas in contexts that affirm a minimal amount positive collocates such as self-esteem and courage are typical. Pragmatically, shred of mostly functions in negative polarity contexts and as an emphatic marker. In combination with negative markers such as not or without or negative verbs, etc. it expresses zero-attestation: the total

260Chapter 5: Small SN-constructions denial of any presence of N2. In those contexts it is often used with the rhetorical effect of signalling the conspicuous absence of something that should be there, for instance evidence in a court case, as in (5.58). When preceded by some, even a, etc. shred of asserts the presence of a minimal amount, sometimes despite adverse factors or in complicated circumstances, as in (5.59). No degree modifier use or independent answer use is attested.

5.5.3.5 (not a) scrap of Table 5.10. Relative frequencies of scrap of Head use

Quantifier use

Ambivalent use

Total

#

%

#

%

#

%

#

56

65.88

29

34.12

0

0

85

Scrap derives from the Old Norse verb skrapa "to scrape" and its plural form originally referred to fragments of uneaten food, especially as constituting the remains of a meal. Another lexical meaning is scrap as a small, detached piece (cf. NSOED). Scrap, despite its relatively high frequency, is still predominantly used lexically, with only 34.12% grammaticalized uses. In its head use, scrap mainly has paper and its hyponyms as collocates, e.g. (5.60)-(5.61); only one collocate relates to food, i.e. (5.62): (5.60)

A sixteenth century scrap of parchment, yellow with age, bears the signature of Guillaume Debande. (CW-UKephem)

(5.61)

He was looking in his pockets for that scrap of paper where he wrote the address of the building they were looking for. (CWUSbooks)

(5.62)

He scooped up the scrap of bread he had been saving, tucked it inside his yellow shirt, and sat down. (CW-USbooks)

In its quantifier use, scrap also has some collocational favourites, i.e. evidence, which occurs 12 times. The remaining 17 collocates range from similarly epistemic concepts such as information and data, over emotional

5.5 Individual discussions of SSN-constructions

261

ones, such as kindness, self-esteem, honour, to more isolated ones such as light, energy, etc.: (5.63)

Yet according to John Studd, “There is not a scrap of evidence that progesterone therapy for PMS works.” (CW-UKmags)

(5.64)

You've never shown me a scrap of kindness, so why would I expect you to now? (CW-UKbooks)

(5.65)

I just feel that somehow we're missing some thing or other that would help us." And they needed every scrap of help they could get, Rhodry realized. (CW-UKbooks)

The predilection for epistemic-like nouns might be explained via a metonymic shift from paper to the information that can be contained on a piece of paper (cf. the paper(work)metonymy attested with pile(s) in Section 4.3). As mentioned earlier, scrap and shred share collocates and collocational reclustering in their quantifier use. The semantic prosody is mainly neutral with a few instances gravitating towards positive prosody, such as kindness and honour. Scrap of almost always occurs in emphatic and negative contexts. Pragmatically, scrap of in combination with the negation functions as a minimizer, e.g. You’ve never shown me a scrap of kindness in (5.64). There are no instances of degree modifier uses or independent answer uses. In the diachronic data set, 10 of the 42 attestations of scrap are quantifier uses. Of these, half combine with negative polarity contexts. Typical collocates of these quantifier uses are similar to synchronic ones, e.g information, news, ardour and ill-feeling. Of the 32 head uses of scrap 10 appear in a negative polarity context.

5.5.3.6 (not a) flicker of Table 5.11. Relative frequencies of flicker of Head use

Quantifier use

Ambivalent use

Total

#

%

#

%

#

%

#

12

19.4

23

37.1

27

43.5

62

Flicker derives from Old English flicorian, flycrian, cf. Low German flickeren and Dutch flikkeren and is originally synonymous with flacker. In its

262Chapter 5: Small SN-constructions original meaning it referred to the sound made by a bird, i.e. "to flutter". Henceforth it came to express "(cause to) flutter rapidly, quiver, undulate, wave to and fro". It further extended to such meanings as, "to flash up and die away rapidly (and repeatedly); (of flame) to burn fitfully" (cf. NSOED). In its lexical meaning flicker is synchronically mainly used with regard to light, eye movements and facial expressions in general: (5.66)

He was aware that Houston was watching every flicker of his eyelids. (CW-USbooks)

(5.67)

Virginia Verran's recent paintings use a skid of brush and a flicker of light against dark to suggest changing speed. (CW-Times)

In quantifier-like uses flicker still carries quite a lot of lexical meaning in that it almost always incorporates an element of suddenness and brevity and mostly relates to a reaction that is visually attestable. Examples (5.68) to (5.70) illustrate ambivalent uses, i.e. ambiguous between a literal head reading and a quantifier reading. We do see that, as opposed to synchronic head noun uses, these ambiguous cases tend to associate with negative polarity. (5.68)

The man looked at him, a flicker of surprise at the fluent German crossing his face. (CW-UKbooks)

(5.69)

She hit the first one with a seven and the second with a king and then she went bust herself without the slightest flicker of emotion. (CW-UKbooks)

(5.70)

Since then Lewis had not shown a flicker of remorse for the injuries she had caused. (CW-Times)

Example (5.71) illustrates a vague ambivalent case, in that the quantifier meaning is bound up here with the metaphor of a romance that extinguishes: (5.71)

A flash of inspiration five minutes from time by Adrian Littlejohn yesterday extinguished the last flicker of romance from an FA Cup second round that had delivered few of its usual emotional highs for the non-League contingent. (CW-Times)

In the true quantifier uses the lexicality of flicker is least apparent, though still there to some extent. The context of facial expressions is virtually absent and the meaning here most closely approximates that of ‘a little’:

5.5 Individual discussions of SSN-constructions

263

(5.72)

It's all I look forward to at the end of the day; it gives me a flicker of comfort and is the only means I have of forgetting all my problems. (CW-UKbooks)

(5.73)

It was warmer inside, though there was hardly a flicker of chilli in the food. (CW-UKbooks)

A flicker of seems to prefer emphatic negative contexts, though affirmative contexts are also represented in the concordances, e.g. (5.72). When combined with negation (and in most cases a superlative), flicker of functions as a minimizer, e.g. (5.69) and (5.70). In rare affirmative sentences it functions as a downtoner, but with a boosting value, e.g. in (5.72) there is no comfort, but it does provide a positive contrast with the remainder of the speaker’s day. In the diachronic data set all 14 examples are head noun uses and only one occurs in a negative polarity context.

5.6 Discussion and conclusions If we link the observations from the diachronic and synchronic data sets, we can try to answer the following interlocking questions. How and when did SSNs come to associate with either negative or positive polarity contexts and did these polarity contexts serve as enabling factors in their grammaticalizing into quantifiers? Did bit of serve as an analogical model for the grammaticalization of infrequent SSN-expressions? The corpus samples for bit of tell us that in the diachronic data 43 out of the 522 quantifier tokens are in negative polarity contexts, 23 of which are the quasi-idiomatic expression not a bit of it. Most of these negative polarity sensitive uses of bit date from 1780 onwards. In the synchronic data 197 of the 200 attestations of bit occur in positive polarity contexts and only 3 in negative polarity ones. It thus seems that bit of has always associated predominantly with positive polarity and does not need negative polarity in order to function unambiguously as a quantifier. Traugott (2010b) notes that bit of only became associated with negative contexts towards the 19th century. This observation and my diachronic data hence lead to the hypothesis that for bit of negative polarity contexts did not serve as onset construction in the grammaticalization towards quantifier status. If we then look at the infrequent SSNs, we see that diachronic polarity effects are similar to the synchronic ones. Since smidgen, skerrick and scintilla did not yield any diachronic data, no hypotheses can be made about these infrequent SSNs. Some synchronically negative polarity SSNs such as shred and jot already had a preference for negative polarity in the dia-

264Chapter 5: Small SN-constructions chronic data sets in their quantifier and ambivalent uses, while head noun uses feature predominantly in positive polarity contexts. Scrap diachronically has 5 quantifier uses that are negative polarity sensitive and 5 which are positively polarity sensitive. Its head noun uses diachronically occurred in positive polarity contexts (i.e. 22 tokens), while 10 occur in negative contexts. Synchronically, head noun uses of scrap almost all appear in positive polarity contexts, whereas almost all quantifier uses occur in negative polarity contexts. Flicker, finally, only has head uses in the diachronic data set, all but one occur in positive polarity contexts. In the synchronic data, we see that head uses favour positive polarity, while quantifiers occur in negative polarity. With regard to synchronically positive polarity SSNs we see that whiff diachronically and synchronically typically appears in positive polarity contexts. All of the diachronic attestations are head noun uses. Only 1 out of 18 tokens has negative polarity. In the synchronic data only 3 quantifier uses of whiff of appear in negative polarity contexts and 6 head uses. The remaining uses occur in positive polarity contexts. SSNs that are negative polarity sensitive in their synchronic quantifier uses turn out to associate with negative polarity in their diachronic quantifier and ambivalent uses, while their diachronic head uses associate predominantly with positive polarity contexts. On the other hand, SSNs that are positive polarity sensitive synchronically, diachronically too occur with positive polarity contexts both in their head and quantifier uses. These observations give rise to the following hypotheses. For synchronically negative polarity sensitive SSN-quantifiers, negative polarity seems to have been an enabling factor in their grammaticalization towards quantifier status, since it is also attested in their ambivalent uses. The positive polarity sensitive SSN-quantifiers did not seem to have gone through a phase in which they could associate with negative polarity, as bit of did in the late 18th and 19th century. All of these hypotheses make it difficult to see bit as a direct model for other SSNs. With every case study of a specific grammaticalization process, it becomes more and more clear just how variable, obviously within very specific bounds, critical factors in grammaticalization can be. In the case of (S)SNs, I argue that collocational patterns are essential in marking head and modifier uses (cf. Brems 2003a and Brems 2010). For SSN-constructions polarity preferences add to this, by providing a highly constrained context in which the infrequent SSNs can develop quantifier uses. Not all SSNs have acquired a quantifier use to the same extent. Highly frequent bit of for instance has grammaticalized to a very high degree, extending its grammaticalized use into new syntactic contexts and going on to further (inter)subjective uses. Within the set of infrequent SSNs there are also differences. Whiff, scrap and flicker display only 50% or less quantifier uses, even though whiff is one of the most frequent SSNs. Scrap of has

5.6 Discussions and conclusions

265

34.12% quantifier uses versus 65.88% head uses and is also quite frequent, even though there is a relatively big discrepancy between its overall frequency (i.e. 280 attestations) and frequency within the SSN-construction (i.e. 85 attestations). For some reason, scrap is not as constructionally fixed as some of the other SSNs (cf. Hunston and Francis 2000: 25; see Section 2.2.4.2). Still, in the case of jot there is also a difference between these two frequencies (8 out of 36 attestations are within the SSN-construction), which does not seem to have influenced the development of a prevalent quantifier use and even an extended adverbial use. Smidgen, on the other, hand is strongly associated (i.e. 12 out 15 attestations) with the SSNconstruction and synchronically only appears in quantifier uses and some rare adverbial uses. In the case of flicker the predominant use is the ambivalent one, with many ambiguous instances, suggesting little delexicalization on the part of the SSN, and instead mostly relying on metaphorization of N2-collocates. Flicker has hence only grammaticalized to a very limited degree and as such should perhaps be taken out of the grammaticalization argument. How can it be explained then, that the very infrequent SSNs developed quantifier uses to these various extents? There is partial compliance with several grammaticalization reflexes, which seem to hint at a normal path of grammaticalization or, rather, which are normally only attributed to traditional gradual processes of grammaticalization, e.g. substitutability of SSN+of by canonical quantifiers, semantic shifts to more abstract meaning, extension to other syntactic environments, etc. (cf. Lehmann 1985, Bybee 2003a and Traugott 2010b). However, considering the infrequency of the SSNs it is not at all clear whether the infrequent SSNs themselves engaged in gradual grammaticalization processes. It is also possible that they grammaticalized by a kind of analogy with more frequent models that did grammaticalize gradually, the grammaticalized behaviour of which they copied so to speak. If so, their analogy model is probably not merely the polarity-independent and semantically rather schematic a bit of, but a more specific, polarity-defined constructional schema. Infrequent SSNs at some level seem to follow quantifier models that are also PIs, such as some and any, but at the same time they rely on specific collocational and discourse values. Infrequent SSNs are restricted to very specific discourse schemas, e.g. not a SSN of in reactive or adversative contexts, also often used as a rhetorical strategy (e.g. shred/scrap), where reference is made to the absence of a quantity of something that should have been there, and which may therefore involve annoyance or disbelief, e.g. (5.43), (5.44), (5.46), (5.54), (5.58), (5.63), (5.70). The relationship with polarity is hence much further schematized in some and any, than it is in the SSN-constructions, which may hint at relatively low grammaticalization (cf. Traugott 2010b and Himmelmann 2004 on highly constrained con-

266Chapter 5: Small SN-constructions texts that first set off grammaticalization and later make way for generalization to various contexts). Infrequent SSNs seem to display a distant modelling with partially polarity-dependent quantifiers as well as a more distant modelling with polarity-independent ones, i.e. with the most schematized small quantifier construction as instantiated by few and little. Since bit of started grammaticalizing before the other SSN-expressions, it probably also exercised some analogical pull as an early member of the periphrastic SSN meso-construction. All of the constructional models involved can be translated in terms of the different levels of schematicity involved in the grammaticalization of the SSNs, i.e. Traugott’s (2006) distinction between “macro-constructions, i.e. the highest level of schematization”, e.g. the Quantifier-construction with many/little; meso-constructions: “sets of similarly-behaving constructions, e.g. kind/sort of vs. a bit/lot of”; and microconstructions: “individual construction-types, e.g. a lot of vs. a bit of.” (Traugott 2006: 2; the examples are Traugott’s). A question is whether the SSN-expressions will ever approximate quantifiers such as bit of or few/little more fully. As also noted by Traugott (2010b), grammaticalizing elements can sometimes congeal in their initial facilitating, but constrained, contexts for grammaticalization, similar to idiomatization. It has also been amply argued that grammaticalization is a non-deterministic process that does not have to go on to completion (e.g. Hopper and Traugott 2003, Fisher 1999). I therefore tentatively venture that these infrequent SSNs will not go on to fully macro-constructional status in their semantic and formal properties, precisely because of the very specific expressive purpose they serve as quantifying expressions. A similar point was made with regard to SN-constructions in Chapter 4. An important question then is how the specific polarity-quantifier alliances arose and how they motivate the emergence of quantifier uses for these infrequent SSNs. How do the semantics, pragmatics and syntax of SSNs interact with the semantics, pragmatics and syntax of either polarity? Israel (2004) does not provide many generalizations over the specifics of both polarity contexts and concludes that PIs as a whole are inherently scalar and polarity sensitivity is a lexical property of PIs. All PIs simply seek out contexts where they can discharge these already inherent lexical properties (Israel 2004:717). This does of course not explain why certain PIs prefer negative contexts, while others prefer positive ones; polarity sensitivity in its entirety is explained away as an inherent lexical property. Traugott (2010b) seems to propose the following generalizations about the prototypical quantifier meanings in positive and negative polarity contexts respectively, which tie in with the corpus observations of this chapter. In positive polarity contexts SSNs typically have a downtoning, diminishing quantifier meaning, i.e. they designate a very small quantity, a minimal amount of something which is nevertheless asserted and affirmed, as in

5.6 Discussions and conclusions

267

(5.74) and (5.75), emphasized by even in (5.74) and stress on whiff in (5.75): (5.74)

Jeffries is terrified of even a smidgen of what he calls negative thoughts. (CW-Times)

(5.75)

Everyone with a whiff of sanity in his veins should have no traffic with that way.(CW-UKbooks)

Since affirmation and positive polarity contexts are the unmarked case (see Israel 2004: 706-707), the question in fact mainly centres on the NPIs and what gears them so strongly to negative polarity. It seems that the appropriate use of a negative sentence requires a context in which the information it negatively conveys is somehow particularly relevant and the negative formulation is meaningful in conveying this message (Israel 2004: 706). When a speaker decides to formulate an affirmative predicate ‘X’ and symbolize it as ‘not “antonym of X”’, two effects can be obtained, depending on the semantics of ‘X’. On the one hand, the possibly crude semantics of ‘X’ may be backgrounded and the intended meaning can thus be hedged behind two words, e.g. not the best idea is more hedged than the worst idea. On the other hand, using a complex predicate with a negative marker can also do the opposite, i.e. draw attention to it and put emphasis on the symbolized meaning, e.g. a statement like I am not uninterested is a marked formulation unless the context would demand such a negative formulation. This ties in with how Israel (2004: 708) sums up the two main pragmatic functions of negation as expressing indirection or attenuation (e.g. hedges, euphemism) on the one hand, and strengthening or emphasizing meaning on the other. I want to argue that in negative polarity contexts SSNs prototypically combine two effects: on the one hand minimizing quantifier meaning, and on the other hand strengthening or emphasizing that minimal quantity pragmatically. The minimizing meaning is typically that of zero-attestation, which often translates into an emphatic evaluation, as in (5.76) and (5.76’) (5.76)

There was not a trace of corruption in him. (CW-OZnews)

(5.76’) ‘He is very honest’ In (5.76) quantification of the feature ‘corruption’ in the person results in zero-attestation, which is conveyed by the emphatic statement in (5.76’). But how do SSNs become associated with their specific polarity preferences in the first place? What is it in their lexical semantics that makes them click with either positive or negative polarity contexts? The pragmatic

268Chapter 5: Small SN-constructions effect of indicating zero-attestation is one that all SSNs could in principle lay claim to, in that all SSNs by definition denote a small quantity. What is it that makes skerrick, jot, shred, scintilla and scrap more suitable for expressing this zero-attestation meaning? As pointed out by Israel (2004: 719): “Every polarity item has its own history” and the lexical semantics of the PIs contribute greatly to this individuality. Jot of course originally referred to the smallest letter in the alphabet and skerrick to an extremely small coin; scrap and shred refer to remnants and hence might stress surprise at the fact that there is something left at all, which all helps the zeroattestation meaning. Yet, can a jot, scrap, shred, etc. truly be said to refer to smaller things than a smidgen for instance? Or does the fact that most of these NPIs consist of only one syllable have an impact on their association with negative polarity contexts? What about skerrick and scintilla, then, and whiff, which also has one syllable, but is a positive polarity item? The fact that whiff is less well suited for expressing zero-attestation might be attributed to the fact that it originally refers to something unbounded and gas-like. All of these questions remain hypotheses and a lot of the nature of polarity sensitivity needs far more explaining than it receives here. In conclusion, the case of the infrequent SSNs focuses attention on the fact that one has to be very cautious in describing what served as analogical source for what and at which level. It also addresses other theoretical issues. Firstly, it addresses the importance of context-dependent grammaticalization and the influence of lexical items in these, as advocated by an important strand in grammaticalization studies, represented by Traugott (2003) for instance. Secondly, the case studies of infrequent SSNs can provide food for thought in the more general discussion held in grammaticalization theory, i.e. whether grammaticalization can operate on and lead to the establishment of schematic constructions (e.g. the emergence of fixed word order) or not (cf. Haspelmath 2004, Himmelmann 2004, Hopper and Traugott 2003, Lehmann 2005, Noël 2005 and Traugott 2003). A specific subpart of this question pertains to the precise level(s) of constructional schematicity grammaticalization operates on. Against Hoffmann (2004) and with Traugott (2003), then, I want to stress the importance of specific contexts interacting with the specific lexical semantics of items. I think the case study of SSNs proves that even in cases where constructional analogy is involved, these factors critically influence and steer grammaticalization paths. Grammaticalization hence is seen as “centrally concerned with the development of lexemes in contextspecific constructions (not merely lexemes and constructions).” (Traugott 2003: 627; italics Traugott) and “it may be more accurate to say that a construction with particular lexical items in it becomes grammaticized, instead of saying that a lexical item becomes grammaticized” (Bybee 2003b: 602). This ties in with recent attempts to rephrase or refine the notion grammati-

5.6 Discussions and conclusions

269

calization in terms of (grammatical) constructionalization (e.g. Traugott & Trousdale 2010).

Chapter 6 TN-constructions 6.1 Introduction The corpus studies presented so far were devoted to SN-constructions. Regular SNs and SSN-expressions each brought in different issues and constructional subtypes within the larger picture of constructional grammaticalization, as reflected by synchronic layering of functions. For both sets of SNs, the attested variation in constructional meanings was argued to be linked to distinct collocational patterns and to be motivated by the language user’s needs for expressivity or extravagance in communication (cf. Joseph 1992, Haspelmath 1999, Hopper and Traugott 2003, Heine 2003) The present chapter turns to TN-constructions and proposes a description of the various uses attested in synchronic and diachronic corpus data as a further study into variation and the mechanisms of grammaticalization processes within the (binominal) NP. In present-day English, three nouns expressing the general meaning of ‘type’ are very frequently used: kind, sort and type. These TNs may fulfil different functions in the NP, as illustrated in (6.1) and (6.2) in which the TNs function as categorizing and quantifying elements respectively. (6.1)

Funny Bones was based on the premise that there are two sorts of comedian. (CW-Times)

(6.2)

The problem was that the Bush administration said the funding should come from the cities and states. The cities and states said they didn’t have this kind of money. (CW-UKspoken)

The TNs also have uses in which their meaning has been extended and modified via increased subjectification (cf. Aijmer 2002, Denison 2002), as in (6.3) and (6.4). (6.3)

She was lookin kinda dumb with her finger and her thumb in the shape of an “L” on her forehead. (http://www.charcards. com/AllStar.htm)

(6.4)

[…] those advert features that they do in the States, they’re like sort of like … half hour adverts, like they’re like shows but they’re […] (COLT)

6.1 Introduction

271

Despite the number of articles and papers devoted to TNs, there is no agreement in the literature about a systematic, formally motivated classification of their different uses (see Section 1.2.2). The aim of this chapter is to work towards such a comprehensive grammatical description based on attested uses in synchronic corpus data. In addition, it will be argued that the synchronic variation is the result of diachronic processes of grammaticalization and (inter)subjectification and hence constitutes a case of synchronic layering. This corpus study aims at accounting for the constructional peculiarities of TN-constructions, as well as differences with regard to (S)SN-constructions. I will also tie TN and SN-expressions together in terms of zones of constructional overlap, i.e. can (S)SN- and TNconstructions construe similar meaning and if so, in similar ways? Such zones of overlap come about through the intersection of the main functional areas within the NP, i.e. categorization, modification and determination in a broad sense, including identification and quantification (cf. Langacker 1991). Regular SNs, SSNs and TNs have all interacted with deictic and attitudinal subjectification, and shifted away from the functional zone of categorization and towards the determination zone. The SSN-expression bit of is a hinge between SNs and TNs in this respect, in which meanings of categorization, quantification and qualification crystallize. The constructional approach to corpus data proposed in this study will allow me to show that all (S)SN- and TN-expressions share certain fundamental constructional and semantic features beyond the similarities in surface manifestations put forward by many traditional English grammars without any motivated structural analysis. Moreover, collocational phenomena will also turn out to play an important role. Interestingly, the parallels between SN- and TN-expressions in part emerge in the context of their NP-internal uses, which in the case of TNexpressions have been neglected in the grammaticalization literature so far and pose the greatest descriptive problems. As reviewed in Section 1.2.2, the present boom in studies of sort of and kind of focuses mainly on describing uses that involve very clear meaning shifts away from the head noun use, e.g. to qualifying uses, discourse markers and other pragmatic markers, which are largely detached from NP structure, e.g. But you know how like sort of like parents sometimes sort of secretly hope that you'll be such-and-such (CW-UKspoken) (e.g. Aijmer 2002, Denison 2002 and 2005 and Margerie 2010). The classification of TN-constructions proposed in this chapter takes a different point of departure in that it focuses on form-function correlates in arguing for distinct categories of TN-uses. This yields additional NPinternal non-qualifying modifier uses which have not been described earlier, but are essential in linking up the more established TN-categories. They help bridge the gap between NP-internal and NP-external uses of TN-

272Chapter 6: TN-constructions expressions, as both seem to involve degrees of (inter)subjectification. In addition to similarities in NP-internal developments, similarities between (S)SN- and TN-constructions are also manifested by uses that have extended beyond the NP. These especially yield close resemblances between the TN-expressions and the frequent SSN-expression bit (of). The structure of this chapter is as follows. Section 6.2 will present the queries and data sets used for the synchronic classification of TNconstructions and the register study. In Section 6.3, I will present my classification of the different functional configurations found in the synchronic data. This classification owes a lot to, but also expands on Denison (2002). It proposes more fine-grained distinctions, particularly for the NP-internal uses of TN-expressions. This is done mainly by pointing out structural, semantic and collocational distinctions between head uses on the one hand and modifier (attributive and semi-suffix) and determiner uses (in the broad sense) on the other. In addition to clear-cut uses, ambivalent contexts were identified, which receive conflated labels, e.g. qualifying/discourse marker use. They are discussed in Section 6.3.8. From the existing descriptions, I retain the distinction between (nominal, adverbial and sentential) qualifiers, discourse markers (Denison 2002 and Aijmer 2002) and quotative markers (Aijmer 2002). Section 6.3.9 then forms a conclusion to this classification by formulating the main fault lines distinguishing between the head and modifier constructions in which TNs may feature. In addition it will discuss areas of constructional-semantic overlap between TN- and SN-expressions. In Section 6.4, the descriptive framework proposed in Section 6.3 will be applied to two British English data sets from opposing registers: the written texts from the quality newspaper The Times (Collins Wordbanks subcorpus) and the spontaneously spoken conversation between teenagers from the COLT corpus. As we will see, the quantification of these analyses reveals strong asymmetries in the relative frequencies of the various TNuses in the two data sets. While TNs are used predominantly NP-internally in The Times, adverbial qualifiers and discourse markers predominate in the COLT data. Section 6.5 will present an extensive diachronic study of sort(s), kind(s) and type(s) spanning the period 1150 - present-day English. In that study the synchronically coexisting uses of TN-expressions will be claimed to constitute synchronic layering, i.e. they are the result of complex and interlocking paths of grammaticalization and (inter)subjectification. Section 6.6, finally, rounds off with a general conclusion.

273

6.2 Description of the synchronic data sets

6.2 Description of the synchronic data sets The data for the synchronic study were extracted from the Times subcorpus of the Collins Wordbanks corpus, as well as from the COLT corpus. Since the focus is on NP-internal uses of TNs, the corpus data include exhaustive extractions on kind(s)/sort(s)/type(s)/of/-a from both corpora, based on De Smedt (2005).This data set was also used in De Smedt, Brems & Davidse (2007). The Times-subcorpus contains text from The Times and The Sunday Times, published in London, and is part of the larger Collins Wordbanks corpus, used for the case studies in Chapters 4 and 5. COLT, on the other hand, focuses on the speech of teenagers. It was collected in 1993 and consists of the spoken language of 13 to 17-year-old teenagers from different parts of London. In this way the broad regional variant of English is kept constant, on the assumption that the majority, if not all, of the material in both corpora comes from native speakers. The frequency of all three TN-expressions is very high in the corpus as a whole.86 The restriction to an exhaustive extraction from two subcorpora should make for internally homogeneous and representative analyses. By the same token, the Times and COLT (sub)corpora also allow comparison between two supposedly maximally different registers of English. The Times contains professional formal writing, whereas the COLT data represent very informal speech. Table 6.1 represents the relative token frequencies of the TNexpressions in these exhaustive data sets. The table shows that some expressions were not attested in the corpora. With regard to the Times subcorpus the first row represents raw frequencies, whereas the second row indicates the normalized frequencies per half a million words so as to make them comparable to those of the COLT corpus. Table 6.1. Token frequencies of the TN-expressions in the Times and COLT (sub)corpora Kind Kinda Kinds Sort Sorta Sorts Type Typa Types Total Times

629

/

62

702

/

75

p0.5mw 54.57

/

5.38

60.9

/

COLT

26

2

/

136

/

1

Total

655

2

62

838

/

76

155

/

93

1716

/

8.07

148.86

3

/

/

168

158

/

93

1884

6.51 13.45

274Chapter 6: TN-constructions For descriptive and illustrative purposes, reference will also be made to a further set of data compiled by Lavrysen, Nys and Vreven (2005), who extracted from the entire Collins Wordbanks corpus all uses of kind(s)/sort(s)/type(s) of/kinda/sorta/typa premodified by adjectives, i.e. adjective + TN + of + N2. These data are particularly enlightening with regard to non-qualifying modifier uses and the collocational patterns related to distinct TN-uses. For examples from this data set I will indicate the specific subcorpus they derive from. Additional data from the Internet have also been looked at. The latter two sets are not incorporated in the quantified results of this study. 6.3 Classification of NP-internal TN-uses and extended uses87 The synchronic characterization of nominal TN-constructions covers the main semantic and formal fault lines attested in the data in terms of head and modifier uses. The general semantic value of these functions will be related to more specific discourse patterns (but also see Section 6.4 for an explicit register study), and the formal features of these functional categories include grammatical structure and intonation, as well as lexical selection restrictions and collocational patterns. Several functional categories are internally graded and boundaries between categories are not always clearcut, with bridging contexts between them as well as dual reading contexts. At first sight, the distinctions between TN-uses did not appear to correlate strongly with differences in collocational patterning, but rather with interaction between the TNs and the functional zones in the prenominal string, i.e. determination, quantification, qualification and classification. However, on closer analysis of the corpus data, the NP-internal uses turned out to correlate more strongly than expected with distinct collocational patterns in the broad sense. It is because of this that collocates have been sorted explicitly in terms of MI and T-scores. The MI-score and T-score are two statistical measures of the significance of association between a node and its collocates, each highlighting different aspects of this relation (see e.g., Church and Hanks 1990, Church, Gale, Hanks, Hindle 1991, Clear 1993 and Stubbs 1995). The MI-score, or mutual information score, measures the strength of their association, in that it expresses the extent to which the observed frequency of co-occurrence differs from the expected frequency. It identifies idiomatic-like or fixed phrases in which the node and collocate predict each other very strongly and are often restricted to this collocation for their occurrence, e.g. knob in a knob of butter. The T-score, on the other hand, takes into account the joint frequency of node and collocate and filters out chance and infrequent co-occurrence of words, which the MI-score does not. The T-score thus brings out significant collocations

6.3 Classification of NP-internal TN-uses and extended uses

275

that occur frequently, and is therefore sometimes considered to be more reliable than the MI-score, especially when collocates have low frequencies. Rather than strength of association, the T-score points out the productivity of collocates, thus often identifying sets of typical collocates, e.g. a pile of stones, bricks, books, novels. It is generally agreed upon in the literature that values equal to or higher than 3 single out very significant collocations (cf. Stubbs 1995: 35). If values are high for both MI- and T-score, we are dealing with very typical, strong and recurrent associations between node and collocate. Both statistical measures will be used to calculate the strength of collocations between the TNs (as nodes) and modifiers and determiners preceding them (as collocates), in order to argue more strongly that distinct layers of TN-uses relate to specific collocational patterns.

6.3.1 Head use construction It is clear from the literature reviewed in Section 1.2.2, that the fact that TNs can function as the head noun of a NP is not in question. Both Denison (2002/2005) and Aijmer (2002) recognize a use of TNs in which they function as the head noun of the binominal NP they occur in.88 The head use is furthermore confirmed by attestations in both the Times and COLT data sets and for all three TNs, in the singular as well as in the plural (the latter occur only in the Times corpus). Table 6.2 summarizes the token frequencies of head TN uses for each attested TN. As announced earlier, for the Times data the first row indicates raw frequencies while the second one translates these to frequencies per half a million words. Table 6.2. Frequencies of head uses of TNs

Times p0.5mw COLT

Sort

Kind

Type

Sorts

Kinds

Types

Total

46

84

86

12

36

84

348

3.99

7.29

7.46

1.04

3.12

7.29

21.51

2

1

1

0

0

0

4

What this section sets out to do then is to define this head use more precisely, and delineate it vis-à-vis the (non-qualifying) modifier, quantifier and postdeterminer use, which will be discussed in Sections 6.3.2 to 6.3.4.

276Chapter 6: TN-constructions This involves identifying the head use’s structural properties and specifying its constructional semantics and most general discourse uses. In head noun constructions the TN is the lexical head of the NP with the of-phrase functioning as a postmodifier. Semantically, the construction is concerned with designating (sub)classes and expressing generic or taxonomic meaning. Prototypically, in these NPs, of + N2 specify the more general, superordinate class that the subclass(es) referred to are included in, as in (6.6) to (6.11).89 Head noun uses can of course also occur without N2, as in (6.5): (6.5)

The action here is mainly of the cerebral kind as he tries to unravel the mind of the man the police believe is the killer. (CWOZnews)

(6.6)

There are 5 types of animals you can have on your farm: your dog, horse, cows, chickens, and sheep. (www.fogu.com/hm4/moo/ index)

(6.7)

Marks [&] Spencer made some of its earliest continental inroads in Belgium, and my local supermarket stocked 19 kinds of shortbread made by Walkers of Aberdeen. (CW-Times)

(6.8)

The late-17th century console has an elaborate top inlaid with agate and pietra paesina, a rare type of marble which in its colours and veining looks like a paesaggio or landscape. (CW-UKmags)

(6.9)

Pountney sees the opera as a study of two kinds of woman, the fleshly and the spiritual. (CW-Times)

(6.10)

New pay, as personnel vanguardists see it, is not about old forms of pay bargaining, but about the new kinds of pay flexibilities: profit-related pay, performance-related pay, team-based pay, and so on. (CW-Times)

(6.11)

The visibility was so poor that it was difficult to know what sort of fish passed by, or even what colours they were. (CW-Times)

(6.12)

There are many kinds of courage, but courage in the face of PR wisdom deserves an award all of its own. (CW-Times)

(6.13)

Funny Bones was based on the premise that there are two sorts of comedian, the born and the made, the physical and the verbal, the classic and the topical. (CW-Times)

6.3 Classification of NP-internal TN-uses and extended uses

277

If we look at the N2-collocates in the head noun construction, there do not seem to be specific restrictions in terms of concrete versus abstract and count versus uncount nouns, as opposed to the head noun construction of SNs. In the head uses here N2 is concrete (plural) count in (6.6), whereas in (6.7) and (6.8) for instance it is concrete uncount. In (6.10) N2 is abstract count and in (6.12) it is abstract uncount. Furthermore, in their function as category labels, concrete count nouns in the singular, such as woman in (6.9) and comedian in (6.13), are construed as denoting what Langacker (1991: 146) calls the "pure type", specifying the type as such, similar to generic cat in cat lover. The countability distinction is hence backgrounded and this might explain why there are no real restrictions on N2 in terms of that parameter. Denison (2002) argues that the TN and N2 typically agree in grammatical number, but examples such as two kinds of woman (6.9) and two sorts of comedian (6.13) as well as two sorts of book (CW-Times) and what sort of Milk Magnesia Tablets (COLT), for instance, show that this need not be so. In (pseudo-)scientific contexts, binominal NPs referring to species, substances and social groups are found. In economic texts or publicity, reference is to brands and types of products. Crucially, if qualitative or classifying adjectives occur before the TN, they apply to the description of the subclass as such. Some examples of this structure are (6.14) to (6.18): (6.14)

In 1981 a doctor found 5 previously healthy young men with Pneumocystis carinii pneumonia, a very rare type of infection. (CW- UKephem)

(6.15)

He put 1 m[illion] of his own (and taxpayers') money into restoring Dairsie Castle in Fife. And what happens? Fined 250 pound; for using the wrong sort of stone. Breach of Schedule Monument consent. (CW-Times)

(6.16)

It's natural that they should provoke so many poetic echoes since, like Dylan, Lennon and Macca were pioneers (and remain masters) of a new kind of minstrelsy. (CW-Times)

(6.17)

George Broadhead suggests a gay alternative to weddings and reckons there is little likelihood of the Danish sort of reform being introduced in Britain in the future (CW-UKmags)

(6.18)

I have a brewery that produces a special kind of beer. It is a very high quality product. We would like to export it. (UKephem)

278Chapter 6: TN-constructions In (6.15) it is the specific subtype of the superordinate type ‘stone’ that is wrongly chosen in the specific context. In (6.16) new pertains again to the categorial status of a type of minstrelsy, almost in the sense of a new version of what is traditionally taken to constitute the category ‘minstrelsy’. In (6.18) the kind of beer is characterized as special. As also noted by Lavrysen, Nys and Vreven (2005), who investigated the lexical sets of adjectives occurring in TN-constructions in an exhaustive extraction from the Collins Wordbanks corpus, adjectives premodifying TN heads consist mainly of a small cohesive set of adjectives that are very frequently used in English as such, e.g. rare, new, right, special, wrong and different. Some of these adjectives are qualitative in nature and others are postdeterminers. Postdeterminer adjectives "help single out or quantify the referent of the construction in relation to some context" (Bache 2000: 235; also see Section 6.3.4 on the postdeterminer use of TN-expressions). In the data set from the Times corpus 70 out of the total 91 head constructions in which the TN is adjectivally premodified include an adjective of this set. To get an idea of the strength of collocation between head noun uses of the TNs and adjectives premodifying them, MI- and T-scores were extracted for six qualitative adjectives throughout the entire Collins Wordbanks corpus and five postdeterminer adjectives, represented in Tables 6.3 and 6.4. The figures are rounded off to two decimal places. A slash indicates that no T- or MI-score was observed, or that their values were lower than 0.5. Table 6.3. T- and MI-scores with qualitative adjectives new

wrong

right

special

rare

strange

sort

/

T: 4.79 MI: 2.85

T: 4.94 MI: 1.19

T: 1.06 MI: 0.63

T: / T: 2.58 MI: 1.19 MI: 2.83

kind

T: 7.61 MI: 2.09

T: 5.96 MI: 3.85

T: 4.95 MI: 1.45

T: 6.29 MI: 3.61

T: 0.92 T: 2.53 MI: 1.45 MI: 3.26

type

T: 7.61 MI: 3.28

T: 2.87 MI: 3.46

T: 1.63 MI: 0.87

T: 3.56 MI: 3.61

T: 1.89 / MI: 4.17

6.3 Classification of NP-internal TN-uses and extended uses

279

Table 6.4. T- and MI-scores with postdeterminer adjectives same

another

other

certain

particular

sort

T:16.59 MI: 4.54

T: 2.92 MI: 2.41

T: 4.73 MI: 1.18

T: 2.29 MI: 1.57

T: 2.77 MI: 1.95

kind

T: 14.68 MI:4.79

T: 5.86 MI: 2.41

T: 7.89 MI: 2.25

T: 7.19 MI: 4.41

T: 5.51 MI: 3.87

type

T: 8.32 MI: 4.57

T: 6.48 MI: 3.95

T: 5.79 MI: 2.78

T: 6.43 MI: 5.65

T: 6.96 MI: 6.03

The overall high values for both scores show the strong association between these adjectival collocates and the TN nodes. The head noun construction hence displays restricted collocational patterns prenominally. These seem to reflect that this construction is associated with specific rhetorical turns, genres and cultural tropes. The latter also explain some of the favoured sets of N2-collocates in the head noun use referred to earlier, i.e. reference to species, substances and social groups for instance. In economic texts or publicity, reference is to brands and types of products The head noun construction is also characterized by particular prosodic patterns. The intonation going with head use has, as noted by Aijmer (2002: 176-177), either primary stress on the TN (or its determiner or premodifier) or on N2. In both data sets90 these two basic options for primary stress seemed to correlate with a contextual emphasis either on the subtype or on the superordinate. Thus, corresponding to the two possibilities of primary stress, two types of discourse context could be distinguished. Firstly, NPs with the TN in head position can be used in contexts in which (aspects of) the subclass are presented as new. The speaker may identify the relevant subclass as such, as in The Danish sort of reform (6.17) above. In this example, some superordinate category such as ‘alternatives to conservative policies’ is given in the discourse, and the NP with the TN brings a new subtype of that superordinate type into the discourse, i.e. the Danish sort of reform. It may also be quantitative or qualitative aspects of the subclass that are presented as new, and hence marked by primary stress. Thus, the speaker may focus on communicating how many relevant subclasses there are, as in 5 types of animals (6.6), 19 kinds of shortbread (6.7), two kinds of woman (6.9), many kinds of courage (6.12), two sorts of comedian (6.13) and example (6.19): (6.19)

My grandfather had a curly moustache and offered two sorts of kisses: an ordinary flat kiss or a tickling kiss reserved for when we had been extra good. (CW-OZnews)

280Chapter 6: TN-constructions Alternatively, the speaker may focus on an attribute of the subclass, as in (6.14), which stresses that Pneumocystis carinii pneumonia is a very rare type of infection, or in the wrong sort of stone (6.15). Secondly, NPs with the TN as head can be used in contexts in which the superordinate category is presented as new information, as in (6.20). (6.20)

The late-17th century console has an elaborate top inlaid with agate and pietra paesina, a rare type of marble which in its colours and veining looks like a paesaggio or landscape. (CW-UKmags)

Such examples instantiate the hyponymic use pointed out by Aijmer (2002: 176). For instance, in (6.20), the rather obscure subclass of pietra paesina is categorized by a rare type of marble as belonging to the better known, more general class of ‘marble’, with stress on the latter. Whether the speaker focuses on the subtype or on the superordinate type is a discourse effect – it does not change the structural relation between the TN in head position and the postmodifier of + N2. It is in virtue of this modification relationship that the structure TN + of + N2 designates specific subtypes of more general types. As such, these structures clearly fall within what Bache (2000: 239) has referred to as the functional zone of categorization in the NP, from which, prenominally, the determination and modification zone have to be distinguished. As pointed out by Langacker (1991: Ch. 2), the categorization elements in the NP contribute to the type specification: together they designate a mere type. In NPs with TNs as head, TN + of + N2 designates the categorization, with which the determiners interact to single out a specific referent. Because the head noun designates a subclass, such NPs always have, in one way or another, generic reference, as in a rare type of marble in (6.20), which does not refer to one spatio-temporal instance, but to the entire subtype as such (cf. Langacker 1991). In conclusion, the construction with TN as head expresses taxonomic meaning by referring to generically construed (sub)classes. Structurally, the TN functions as head of the binominal NP, while the postmodifying ofphrase specifies the nature of the general type. Elements premodifying the TN, such as numbers, classifying or qualitative adjectives, pertain to the subtype designated by the TN and N2, not to a concrete instance. Qualitative adjectives premodifying the description of the subtype were found to be culled mainly from a small set of frequently used adjectives, suggesting rather specialized areas of use for the head construction, e.g. a special kind of beer (6.18) in commercial contexts and the wrong sort of stone (6.15) in the context of regulations for the renovation of buildings classified as monuments.

6.3 Classification of NP-internal TN-uses and extended uses

281

In terms of the functional zones of determination, modification and categorization, head noun constructions with a premodified TN have to be parsed as: [DET: indefinite/definite determiner] [MOD: adjective/number/etc.] [CAT: TN + of + N2]. Example (6.14) In 1981 a doctor found 5 previously healthy young men with Pneumocystis carinii pneumonia, a very rare type of infection has to be analyzed as [determination: a] [modification: very rare] [categorization: type of infection] for instance, with very rare describing the categorial status of Pneumocystis carinii pneumonia as rare. This is a very important difference with the uses that will be discussed in Sections 6.3.3 and 6.3.4, such as an accommodating sort of bloke (CW-UKmags), which yield another structural analysis. This distinction is, however, seldom recognized in the literature on TNs, as shown by the reviewed literature on TNs Section 6.3.2 will first discuss the emergence of a quantifier use of TN-expressions.

6.3.2 Quantifier construction Constructions in which the TN is premodified by a quantifier can express head noun meaning, referring to a number of different types of some superordinate category, as in two kinds of woman (6.9) and many kinds of courage (6.12), discussed in the previous section. However, in some examples with the same surface structure the constructional meaning shifts away from pure head noun status to quantifier(-like) status, via the implicature ‘all the different kinds of’ > ‘instances of many kinds’ > ‘many instances’ (cf. Denison 2005). The boundaries between a head noun construction in which the TN is premodified by a quantifier, and a TN-construction in which the TN functions as part of a quantifier is rather fuzzy and the quantifier category itself is internally graded. Quantifier meaning hence ranges from (more) pragmatic to (more) semantic. The predominant realization of quantifier function with TNs involves a combination of a plural TNexpression preceded by the totality quantifier all. Throughout the corpus, all + kinds has an MI-score of 6.839699 and a T-score of 23.769706. For all + sorts the MI-score is 7.848883 and the T-score is 34.990049. Compared to these very high values, all + types has significantly lower scores, i.e. an MI-score of 4.502197 and a T-score of 11.430415. In addition to all kinds/sorts/types of, there is one example in which the negative totality quantifier no is followed by a singular TN-expression, i.e. no kind of, which was found outside of the Times subcorpus. I will first discuss the variation attested for the pattern all + kind/sorts/types + N2 and then move to the example with no kind of. As we will see in the diachronic study presented in Section 6.5, no kind/sort/type of were more frequent in earlier periods of English than they are now.

282Chapter 6: TN-constructions In (6.21) the speaker is talking about distinct types of socks, eventually opting for one specific subtype, which is underlined. Sorts thus mainly has head noun status here: (6.21)

Having tried all sorts of socks over the years, I now use only the excellent Thor-Lo padded socks. (CW-Times)

In (6.22), on the other hand, the categorization meaning of kinds and the quantitative implicature are in competition: (6.22)

Shopping and shopping until her arms couldn’t carry one more thing. Clothes, gadgets, all kinds of electronic gizmos. (CWTimes)

The examples in which the quantitative implicature is foregrounded to the detriment of the head meaning are analyzed as quantifiers here. Examples are (6.23) and (6.24): (6.23)

His song Cab Driver (from his first album) - a rousing blues invective against cabbies who used to see his dreadlocks and speed on by - elicited all kinds of letters from cab drivers saying “I appreciate your song, but I pick up black people and we're not all bad”. (CW-Times)

(6.24)

Van Liefferinge makes the lights as weird and wonderful ambient light objects rather than as practical lamps. “You wouldn't really want to read by them,” she says of the pearly lights. “They come in all kinds of colours.” (CW-Times)

In (6.23) the context makes it clear that it is not different types of letters that were written, but many letters of one type, namely saying that not all cab drivers are racists. In (6.24) all kinds of colours does not refer to different types of the superordinate category ‘colour’, but to ‘many colours’. One would have to assume reference to technical distinctions into types of colour, such as monochromatic or fluorescent in order to interpret (6.24) as a head use. In addition to all + kinds/sorts/types of, the COBUILD corpus yielded one example in which the negative totality quantifier no combines with kind of: (6.25)

The former Great Britain captain raged: We showed no kind of urgency. Salford could have had 70 points. (CW-Sunnow)

6.3 Classification of NP-internal TN-uses and extended uses

283

In this example no kind of describes the zero-attestation or total absence of urgency in a rugby team. In quantifier constructions incorporating TN-expressions (negative) totality quantifiers all or no together with the TN + of form a quantifier complex that has modifier status with regard to N2, which acts as the head noun. These quantifier complexes seem to express more complex quantifier meanings than simple quantifiers such as many. In the case of all kinds/sorts/types of this function is often similar to various, indicating ‘variety’ in addition to ‘amount’ as well as adding a hyperbolic value. In the case of no kind of the overall meaning is mostly emphatic compared to simple no and may express disbelief, surprise and disappointment. In commercial or hyperbolic contexts these added meanings can be a rhetorical asset, as in all kinds of colours (6.24) where variety adds quality to mere quantity. The relative token frequencies of TNs in quantifier function are represented in Table 6.5. The second row for the Times data again indicates normalized frequencies per half a million words. It can be noted that the quantifier use synchronically occurs mostly with plural TNs. Quantifier complexes in which a singular TN is preceded by no seem to have become unproductive synchronically. The diachronic study presented in Section 6.5 will show that these preferences have changed over time. Table 6.5. Token frequencies of quantifier uses Sort

Kind

Type

Sorts

Kinds

Types

Total

Times

0

1

0

48

9

2

60

p0.5mw

0

0.09

0

4.16

0.78

0.17

5.20

COLT

0

0

0

1

0

0

1

As these relative frequencies show, quantifier complexes with types are very infrequent. The corpus-based Collins Cobuild English Dictionary for Advanced Learners includes all sorts of and all kinds of in the entries of sort and kind respectively, but does not include such a use in the entry for type. All sorts of "things or people means a large number of different things or people" (Sinclair 2001: 1487) and has many as a synonym, but all kinds of "things or people means a great variety of different things or people" (Sinclair 2001: 853). The different entries hence suggest that depending on the specific TN the quantifying meaning ascribed to these quantifier complexes is stronger or weaker. All sorts of is also much more frequent in the

284Chapter 6: TN-constructions Times corpus sample, which can be explained by the fact that sort as such is more typically British than kind (see Biber et al. 1999). The fact that in these quantifier complexes a specific meaning becomes associated with a fixed lexical chunk may lead one to conclude that lexicalization is involved instead of grammaticalization. However, since this new meaning is a grammatical one, I argue that the global shift in meaning and structural parsing reflects grammaticalization (see Section 3.4.9). In Section 6.5 I will discuss in detail the diachronic pathways involving quantifier uses of TNs and potential impact from and on other constructions.

6.3.3 Modifier uses Denison (2002) only posits a head to modifier shift for nominal qualifying uses, as in I suppose it’s as a that’s as a sort of holiday. Denison’s postdeterminer use and the partial lexicalizations he discusses do not receive a conclusive structural analysis in terms of head versus modifier status. However, my corpus data contained NP-internal constructions in which the TN has non-head status and is also non-qualifying, as well as not determiner-like. Such non-qualifying modifier uses have not received much attention in the literature so far. The present study, in agreement with Kruisinga (1932) (see Section 1.2.2.4), argues that there are TNconstructions in which the TN has clearly been demoted from head to modifier status without being a qualifying use. Evidence for this demotion is provided by the semantic and formal characteristics of the modifying constructions at hand as well as collocational properties. This category of non-qualifying modifier uses subsumes a further distinction between attributive and semi-suffix modifier uses, which together make up an internally graded category of descriptive modifiers, with uses that lie in between both subtypes and others that border on some of the other uses distinguished in the general classification. These categories will be discussed in more detail in Section 6.3.8.

6.3.3.1 Attributive modifier use In this modifier use the TN is typically preceded by a lexical item invoking a quality of the instance referred to by the NP. This modifying use occurs with the singular variants of all three TNs, both in the Times corpus and in COLT. Table 6.6 summarizes the relative token frequencies of the attributive modifier use for sort of, kind of and type of as well as normalized frequencies for the Times subcorpus.

6.3 Classification of NP-internal TN-uses and extended uses

285

Table 6.6. Frequencies of attributive modifier uses

Times p0.5mw COLT

Sort

Kind91

Type

Sorts

Kinds

Types

Total

26

14

2

0

0

0

42

2.25

1.21

0.17

0

0

0

3.64

1

2

1

0

0

0

4

Examples (6.26)-(6.35) illustrate the attributive modifier use. Premodifiers of the TN-expressions and determiners in front of N2 are underlined. (6.26)

Being an accommodating sort of bloke, he let me take the car around the paddock at Silverstone. (CW-UKmags)

(6.27)

We swim at the bridge where a woman had lost a buttock in a crocodile attack a few years previously, but it was not a lingering kind of swim. (CW-Times)

(6.28)

He's a very pragmatic kind of a guy. (CW-OZnews)

(6.29)

Quite nice to look at in an arty-farty sort of way. (CW-UKbooks)

(6.30)

A more private sort of happening they were advised, might go down better with opinion-makers in the art world than all those public shenanigans at the Tate. (CW-Times)

(6.31)

A little cheaper Ole Works Inn (495 4837) seems like a laid-back sort of a place and sits handily on Cane Garden Bay across the road from Quito's Gazebo. (CW-Times)

(6.32)

Yes, I could write and direct my own movies, but I haven't gotten around to sitting down and doing that yet. I guess I'm a scattered sort of person. (CW-Times)

(6.33)

Clarke became an odd sort of bod, continually on the football photographers' scene, yet not quite of it. (CW-Times)

(6.34)

I am an uncompromising sort of person and I had an endless supply of awkward questions. (CW-Times)

286Chapter 6: TN-constructions (6.35)

You know... the biggish looking kinda one [friend; LB]. (COLT)

As examples such as an accommodating sort of bloke (6.26), a very pragmatic kind of a guy (6.28), a scattered sort of person (6.32), an odd sort of bod (6.33), an uncompromising sort of person (6.34) and biggish kinda one (6.35) show, this use very often occurs with animate collocates in the context of describing an individual’s characteristics. Besides animate N2collocates, there are the rather frequent, semi-lexicalized strings adj + sort/kind/type of person/way, as in a scattered sort of person (6.32), an uncompromising sort of person (6.34), in an arty-farty sort of way (6.29), in an aggressive kind of way (CW-Times), in a moody, broody sort of way (CW-Times), in an unshaven sort of way (CW-Times), in a flailing, feverish sort of way (CW-Times), in a perverse sort of way (CW-Times). Importantly, in attributive modifier uses the qualities named by the adjective premodifying the TN-expression apply to an instance of N2, not to the type, as was the case in head TN-constructions, such as a very rare type of infection (6.14) and the wrong sort of stone (6.17), which have generic reference. The attributive modifier construction attributes these qualities directly to the instances in question. The adjective accommodating in (6.26) applies to a concrete instance of the category ‘bloke’, not to the category as such. In terms of Quirk et al. (1972: 930), the TN-expressions in examples such as (6.26) to (6.35) are cases of “phrases of the form kind/sort of N […] tak[ing] premodifiers plainly related to N rather than sort, both in semantics and in concord”. This transferred (Halliday 1985a: 174) use of the adjective involves a true semantic and structural reanalysis. With reference to Bache’s (2000) functional zones, an accommodating sort of bloke has to be analyzed as [determination: an] [modification: accommodating sort of] [categorization: bloke], whereas a very rare type of marble was parsed as [determination: a] [modification: very rare] [categorization: type of marble]. The categorial meaning of the TN is hence bleached as well as pragmatically enriched to convey attributive meaning. Semantically, the attributive modifier construction refers to a qualitative variant (Bolinger 1972), rather than to a true subclass of a superordinate class. The attributive modifier use hence does not operate in type space (Langacker 1991), but expresses a constructional meaning that is functionally directly oriented towards a concrete instance. Its main aim is to describe the qualities of a concrete instance of a type and this probably explains why it frequently occurs in predicative constructions, e.g. (6.26), (6.27), (6.28), (6.31), (6.32), (6.33) and (6.34). In this context it is interesting to reconsider the variant of the modifying use of TNs discussed by Kruisinga (1932: 396) with an indefinite article in

6.3 Classification of NP-internal TN-uses and extended uses

287

front of N2, such as what sort of a man and a good sort of a fellow cited Section 1.2.2.4. Kruisinga saw the article in front of N2 as an overt indication of the head status of N2. In my view, it also overtly signals that an instance of N2 is being referred to. Lavrysen, Nys and Vreven (2005) found that the variant with a(n) in front of N2 is now rather uncommon in English usage. It also occurred only one time in my Times data set, i.e. a laid-back sort of a place (6.31). However, in principle this variant is still possible with the modifier uses under discussion here, as shown by the alternation between (6.36a) and (6.36b), but ruled out with NPs in which the TN functions as head, such as (6.37a), without changing the meaning. (6.36a) Only last year, Richards branded Australian team manager Bobby Simpson a bad loser and a rather sour sort of guy (CW-Today) (6.36b) I’m an ordinary sort of a guy. (CW-UKspoken) (6.37a) This week, manufacturer 3-M announced a new kind of floppy disk with more than 80 times the storage capacity of the traditional disk. (CW-Times) (6.37b) *This week, manufacturer 3-M announced a new kind of a floppy disk with more than 80 times the storage capacity of the traditional disk. Thus, the variant with a(n) in front of N2 can be used as a test to recognize modifier uses of TNs, provided N2 is a singular count noun. Phonologically, the modifier status is reflected by the fact that sort/kind/type + of in these uses never have primary stress. It is the premodifying adjective that receives primary stress (as also noted by Kruisinga 1932: 397). The next stressed element in the NP is N2. In between these two stressed elements, TN + of are non-salient, with the TN not stressed and of typically reduced to /∂/. In informal registers, this reduced of may be reflected by spellings such as typa and sorter as in example (6.35).92 As for its discourse usage, specific selection restrictions apply to this modifier use of TNs and it also tends to be associated with specific motifs in present-day English. By far the most prominent motif is that involving an adjective attributing a character trait to a person, as noted earlier for examples (6.26), (6.28), (6.32), (6.33), (6.34) and (6.35). Additional examples from other subcorpora than the Times subcorpus are: clubbable kind of politician (CW-Today), inspiring sort of person (CW-UKspoken), a very dramatic sort of person (CW-UKspoken) and a very pragmatic kind of a guy (6.27) cited earlier. More generally, Lavrysen et al. (2005) observed that, as opposed to the adjectives premodifying head TNs, the adjectives

288Chapter 6: TN-constructions premodifying TNs in attributive modifier constructions constitute a rather large set of unusual and infrequent adjectives, e.g. a niggardly sort of word (CW-OZnews), a pretty torrid sort of programme (CW-OZnews), a peaceful sort of sorrow (CW-OZnews), in a moody-broody sort of way (CWTimes), etc. Again, a distinct collocational set is prenominally associated with this TN-use, but one characterized by statistically low collocational measures. The T- and MI-scores for the collocational strength between peaceful and sort for instance are 0.509802 and 0.000000 respectively. This construction seems to have a rather limited application in presentday English, and is veering towards an interpersonal, meta-linguistic use, in which the speaker may, for instance, be apologetic or ironic about an unusual choice of adjective. In the process, sort/type/kind + of increasingly acquire the feel of a fixed string appended to the adjective, and, in this way, seem to be gradually demoted from core modifier status to a more enclitic position. The latter is typical of the second modifier use, i.e. the semi-suffix use.

6.3.3.2 Semi-suffix use The second modifier use is what Denison (2002) refers to as the semi-suffix use of TNs, as in Quick-reaction type missiles are feasible (Denison 2005: 2). He does not provide a detailed functional or formal characterization of this use and restricts it to -type. This section will attempt to provide a more precise structural and semantic analysis of the semi-suffix use, which was attested in the data sets of all three singular TNs, in both the Times corpus and COLT. Even though TN-expressions are overwhelmingly singular in the semi-suffix use, the data contained one example with a plural TN, i.e. You might find that the people who tend to follow JFK kinds of theories are those who are highly motivated to look for external causes to a negative event (CW- Times). Table 6.7 summarizes the relative token frequencies and normalized frequencies per half a million words of the semi-suffix use.

6.3 Classification of NP-internal TN-uses and extended uses

289

Table 6.7. Frequencies of the semi-suffix use

Times p0.5mw COLT

Sort

Kind

Type

Sorts

Kinds

Types

Total

14

7

5

0

1

0

27

1.21

0.61

0.43

0

0.09

0

2.34

4

2

0

0

0

0

6

The core cases of the semi-suffix construction are readily recognizable, and the creative use made of it is illustrated by examples (6.38)–(6.51): (6.38)

The New Yorker about Andrew Motion's biography of Philip Larkin: “A policeman-in-the-head kind of book.” (CW-Times)

(6.39)

Located a fast three-hour haul down the interstate from San Francisco, it's a blink thrice or you'll miss it kind of town. (CW-Times)

(6.40)

“Oh, yes, and my grandmother had some Omega Workshop curtains,” he says, in an “Oh My Word!” sort of voice. (CW-Times)

(6.41)

And while she was still attractive and up for discreet sexual liaisons in a Mrs Robinson sort of way, publicly, at least, she appeared to have passed the sexual baton to a younger generation. (CW-Times)

(6.42)

Apparently, her mother was a keeping-up-with-the-Joneses sort of person, which made Linda resolve that rich people can be cruel and jealous. (CW-Times)

(6.43)

An Irish family were having a picnic. This was on a terrible day and in a “one hand for the egg sandwich, and one hand for yourself” sort of spot. (CW-Times)

(6.44)

The firm of KPMG announced in October that it would incorporate its audit business, and many firms had been expected to move towards some sort of incorporation in the coming year. But now that a Delaware-type solution is available, that is likely to change. (CW-Times)

290Chapter 6: TN-constructions (6.45)

He was like in prison and he worked himself up in the peak no pain kind of thing. (COLT)

(6.46)

Listen, we have had sudden employment in the nature of developing a European-typa film. (www.hexmaster.com/goonscripts/ s08e16.pdf)

(6.47)

He wears a bright red superhero kinda costume with white-andblack sneakers. (www.whatacharacter.com/g-o/n--age1.htm/)

(6.48)

It is an iMac sorta phone. (www.ciao.co.uk/Reviews/Motorola_ V60_ 5297096)

(6.49)

It’s a Spielberg Kinda Christmas. uk/features/carnal_cinema/96.html)

(6.50)

My lil brother has a superman costume and not just a towel for a cape or underwear on the outside kinda costume either. (sfgate.com/cgi-bin/article)

(6.51)

This is why im crazy about u. in a non-‘worship and serve me forever’ typa way ;-) (www.7thrimofhell.blogspot.com/ 1990/01/blog-post.html)

(www.netribution.co.

Examples (6.38) to (6.51) make it clear that the semi-suffix use of TNs has a number of prima facie similarities and dissimilarities with the attributive modifier use. Both uses are non-head uses, in the sense that the TN is part of a premodifier, while N2 constitutes the head noun. In both modifier uses, TN + of is preceded by a linguistic expression, which receives primary stress, whereas the TN-string is phonetically non-salient. The premodifying expression moreover applies to N2, not to the TN. Again a concrete instance is being referred to and not a type as such in a generic way, as was the case in the head use. Moreover, the attributive modifier use and semisuffix use share semi-lexicalized strings, such as sort/kind/type of thing/way/person, as in in a Mrs Robinson sort of way (6.41), keeping-upwith-the-Joneses sort of person (6.42), the peak no pain kind of thing (6.45) and in a non-‘worship and serve me forever’ typa way (6.51). Additional examples are in a you-old-fusspot kind of way (CW-Times), an amoebic kind of thing (CW-Times), a confused “what’s happening here?” sort of way (CW-Times) and you finding a fiver in the street sort of thing (COLT). There are also noticeable differences. In examples (6.38) to (6.51) the position in front of the TN-string is not associated with the class of qualitative adjectives, as was the case with the attributive modifier construction.

6.3 Classification of NP-internal TN-uses and extended uses

291

Rather, we find classifying adjectives such as European (6.46), nouns such as superhero (6.47), brand names such as iMac (6.48), proper names serving as classifiers, such as Mrs Robinson (6.41), Delaware (6.44), Spielberg (6.49) and JFK kinds of theories (CW-Times); and longish nonce expressions, which may then be hyphenated or surrounded by quotation marks, e.g. blink thrice or you’ll miss it (6.39), “Oh my word!” (6.40), keeping-upwith-the-Joneses (6.42), “One hand for the egg sandwich, and one hand for yourself” (6.43), not just a towel for a cape or underwear on the outside (6.50), non-‘worship and serve me forever’ (6.51). Most of this material is not readily parsed as a premodifier subordinated to the TN; rather, it is the TN-string which is felt to submodify this material, and thus to function as an enclitic. The practice of sometimes linking the TN with a hyphen to the characterization preceding it signals its cliticlike nature particularly clearly. Frequent as such expressions followed by sorta/kinda/typa are on the Internet (see examples 6.46-6.51), the impression may be created that informal registers use this spelling variant very consistently for the semi-suffix use. There is no doubt that this widespread practice reflects that, phonologically, the TN does not receive stress and that of is reduced in this use. By the same token, as already noted, this spelling practice is not restricted to the semi-suffix use – it is even found in the head noun construction in informal registers – and it is not criterial to it, as is shown by examples (6.38) to (6.45). Hence, the common sorta/kinda/typa-spelling should not be taken to imply that the semi-suffix use of TNs is easy to delineate from non-modifier uses – or wholly unrelated to the attributive modifier use of TNs discussed in the previous section. I tentatively propose the following functional definition of the semisuffix construction: the lexical material preceding the TN-string is typically a classifying (rather than qualitative) element, and the TN-string itself always has some sort of interpersonal (rather than purely descriptive) value. If we consider examples (6.38)–(6.51), we see that not only adjectives such as European, but also the other expressions in front of the TN-string appear to be classifiers (Halliday 1985a: 164), i.e. prenominal modifiers indicating a subtype of the general type designated by the head noun (i.e. N2). However, particularly the more creative subclassifications may imply evaluative qualities: for instance, a blink thrice or you'll miss it kind of town (6.39), besides evoking ‘small town’, also suggests qualities such as ‘insignificant’. JFK kinds of theories (CW-Times) suggests ‘conspiracy theories’. The TN-strings function as subjective markers of the classifier function, in that they serve to indicate that the preceding expressions function as ad hoc classifying elements, which might otherwise not be obvious to the listener. This entails that the semi-suffix use has rather strong metalinguistic meaning, especially since the nonce expressions are mostly very expressive and

292Chapter 6: TN-constructions well-considered, rather than off-the-cuff (except for the very informal examples from the Internet perhaps). The TN-strings thus always frame the categorization preceding it in an interpersonal way. Their general metalinguistic value may further convey various interpersonal values such as: •

Hedging the classifier, as in a Delaware-type solution (6.44), a superhero kinda costume (6.47) and an iMac sorta phone (6.48) for instance, where the label used may not be fully correct or applicable. The TN-string then serves as an approximator (Quirk et al. 1972: 452).

•

Downtoning of what might seem a high-flown label, such as potentially highbrow European in (6.46).

•

Humour and irony, as in the extravagant and highly expressive nonce expressions in (6.39), (6.40), (6.42), (6.50) and (6.51).93

It seems that the phonological and structural demotion of the TN-string naturally goes with framing in terms of such interpersonal values. The categorization is foregrounded, but at the same time it is presented from a specific subjective angle. In the light of these observations about the semi-suffix use, it turns out to have more affinities with the attributive modifier use of TNs than appears at first sight. Both involve specific collocational patterns in the shape of unusual lexical items and the TN-strings following them have a metalinguistic flavour. Hence, even though both these features tend to be pushed further in the semi-suffix use, the boundaries between attributive modifier and (classifying) semi-suffix use are fuzzy, as illustrated by a set of examples such as a quirky kinda song, a bluesy sorta song, a luv typa song. Consequently, it seems reasonable to consider attributive and semi-suffix uses as the two ends of one internally graded category.

6.3.4 Postdeterminer use Denison (2002/2005) distinguishes what he calls a postdeterminer or complex (post)determiner use for TN-expressions in structures where a plural demonstrative is followed by a singular TN-expression and plural N2, as in these sort of skills like like driving (quoted in Denison 2002: 3). However, this category has uncertain status in his analysis in terms of head versus modifier status of the TN and with regard to whether it actually is a distinct structure and not just a specific subvariant of the “binominal” TN-use. Moreover, the category is defined almost exclusively on formal grounds,

6.3 Classification of NP-internal TN-uses and extended uses

293

i.e. the specific incongruous concord pattern; the only functional gloss given is that of "anaphoric discourse use". Denison (2002: 2) also notes that it is common only in speech. Before moving to the discussion of a postdeterminer construction for TN-expressions, let us first take a closer look at the definition of the postdeterminer function as such and then, in a second step, apply it beyond its original formal context.

6.3.4.1 The postdeterminer as a functional, dynamic category As discussed in Section 2.1.1, in Halliday’s (1985a/1994) functional description of the English NP the postdeterminer is defined as a secondary deictic which adds ancillary information to the basic meanings expressed by the primary determiner in the NP, i.e. identifiability versus nonidentifiability of the referent. It has been argued by Breban (2002 and 2010) and Breban and Davidse (2003) that it is probably better to think of determiners and postdeterminers as together forming determiner complexes, which can express more complex identifiability statuses and phoric relations than simple determiners. For instance, a determiner and postdeterminer complex such as another (whose functional unity is also reflected in the orthography) can express the introduction of a new instance of a discourse-given type, as in (6.52)

Our people estimate that half the money was lost. We don’t want another big problem of fraud and abuse. (CW-UKspoken)

In this way, as observed by Martin (1992: 116), a NP of which the primary reference value is non-identifiable, may still incorporate an anaphoric relation, i.e. to the instance of fraud described in the preceding discourse. Regarding adjectives used as postdeterminers, it has been pointed out by Breban (2010) that they do not have the meaning of attributing a quality to the instance referred to, but serve to express, together with the determiner, the identifiability status of the referent in the discourse and its relation to other referents. For instance, whereas another in (6.52) indicates that a new instance of fraud and abuse is being referred to, besides those implied in the previous discourse, predicative other in (6.53) (6.53)

Everybody’s trying to be other than what they really are… (CWUSbooks)

refers to the qualities and personality traits of everybody (see also Huddleston and Pullum 2002: 1145). Breban (2010) has argued that an example

294Chapter 6: TN-constructions such as (6.53) constitutes a lexical use of adjectives of comparison, whereas an example such as (6.52) is a grammatical use, resulting from a process of grammaticalization and subjectification. To quote another example of a determiner and postdeterminer complex, the same in (6.54) expresses reference to a generalized concept, abstracting one concept of shape and strength from the two instances talked about. (6.54)

He agreed that at 85kg, his normal weight, he is not heavily-built for a rugby player. Physically, we are not the same shape or strength as British or Australian rugby players. (CW-OZnews)94

It has hence been argued in the literature that the surface structure determiner + adjective can instantiate a determiner complex, consisting of a primary determiner and a postdeterminer, in addition to instantiating a determiner and adjective in its fully lexical meaning, as in (6.53), and different from examples of a head noun construction with the TN premodified by a definite determiner. It has moreover been argued that the postdeterminer use is a dynamic category that can recruit new members through grammaticalization processes. Against Margerie (2010) for instance, Section 6.3.4.2 will argue that a significant portion of the TN-corpus data fulfils this function of postdeterminer, similar in several respects to that observed for the more traditional postdeterminers discussed earlier. The main argument here is that the sequence of a determiner and a TN can function as a determiner complex that as a whole expresses complex information about identifiability status, mostly by construing phoric relations. This function makes it distinct from the NP-internal functions discussed so far, i.e. the head noun, quantifier, attributive modifier and semi-suffix construction. It is especially the difference between head noun constructions with a definite determiner preceding the TN, on the one hand, and determiner complexes of determiner + TN, on the other, that will have to be accounted for.

6.3.4.2 The determiner complex construction for TN-expressions Basing myself on patterns manifested by corpus data and Internet data, I will argue in that a TN can be used to form a determiner complex together with the determiner preceding it. As with adjectives used as postdeterminers, specific meanings of the TNs ancillary to the expression of identifiability statuses and phoric relations are activated. The postdeterminer category will turn out to be an internally graded category, with the basic subtypes depending on what kind of identification relation is being construed, and where the information for identification can be retrieved from. This will

6.3 Classification of NP-internal TN-uses and extended uses

295

lead to distinguishing phoric versus non-phoric postdeterminer uses, with generalizing, size-related and emphasizer uses as further subtypes within these two main types. In the discussion of each subtype, it will be argued that the postdeterminer construction, as instantiated by TN-expressions, is distinct from head noun constructions with a definite determiner in front of the TN. To this end, an essentially functional definition will be put forward with systematic formal corollaries and specific collocational configurations. I will use the phoricity feature as the overarching structure in this section and start with phoric postdeterminer uses, which further subsume both anaphoric and cataphoric reference. Afterwards I will discuss the non-phoric postdeterminer uses. First, consider a number of examples in which the TN forms a determiner complex with the preceding (definite) determiner and construes an anaphoric relation. Crucially, in each case the context includes reference to a very specific instance or instances and generalizes over this or these to construe a generalized higher-order concept. In addition to expressing identifiability and phoric relations, determiner complexes with a TN can add subjective meaning such as hedging and intensification. Examples (6.55) to (6.77) are grouped into subsets, which further subcategorize them in terms of generalizing as such, i.e. (6.55)-(6.70), or sizerelated, i.e. (6.71)-(6.77). The complex determiners are in bold. (6.55)

And to me Orlando Bloom is trying to be Johnny Depp, I mean they both have the same sort of looks. (www.whimsicalstrawberries.set/archives/00000049.html)

(6.56)

“The days in the palace of art were over, and before the busy harassed idealist there now stretched an unsuspected wilderness of inner desolation and pain.” This kind of hyperbolic writing does Morris no favours. (CW-Times)

(6.57)

“One day, somebody will try to get his skull for a black magic ritual,” she said: “He seems to attract that kind of person.” (CWTimes)

(6.58)

The poster has mug shots of the likes of Mr and Mrs Le Bon and Bryan Adams, all wearing T-shirts bearing the slogan “Nuclear Free Pacific.” This sort of empty posturing was deemed acceptable behaviour when people wore Katharine Hamnett T-shirts instead of voting Labour. (CW-Times)

296Chapter 6: TN-constructions (6.59)

“The only fight going on here,” he said, “is between the Ministry of Tourism, which wants to build another ski lift, and the Ministry of Conservation, which won't cut down the trees to let them do it.” While this sort of impasse continues, Bansko's slow development will protect it as one of the most beautiful and interesting small resorts in Europe. (CW-Times)

(6.60)

She becomes a heroine by cutting off her own finger and presenting an attacker with her ring. This leaves the blacks awestruck, and saves Arnaud. This sort of behaviour wears thin, when Arnaud again stuns the blacks by rubbing a fire brand over her body. (CW-Times)

(6.61)

“He puts a lot of effort into the area. He was born in a council flat here and has never forgotten that. That sort of attitude is typical of many in Liverpool.” (CW-Times)

(6.62)

“When we came under pressure the lads literally threw themselves in front of the ball,” Wilkins said. “If we continue with that sort of attitude we can stay up.” (CW-Times)

(6.63)

Both Dannies have had their moments specifically on the 606 football phone-in show but how the BBC believes that two ageing, plump cockneys cutting up newspapers for one another could entertain anyone but their families is beyond me. Even the families can't stand this type of behaviour any more. (CW-Times)

(6.64)

Falling down and collapsing on the floor with hysterics. Yeah I don't know why Catherine finds that sort of thing funny. (COLT)

(6.65)

Branson was dropped off by his mother in a field in Devon to find his own way back home. Pop psychologists who are paid to have views on this sort of thing have suggested that this early initiative test caused Branson to develop a life-long compulsion to repeat the feeling of being in danger. (CW-Times)

In (6.55), the same sort of looks expresses reference to a generalization over the looks of two individuals mentioned in the previous clause, i.e. Orlando Bloom and Johnny Depp. This example is very comparable to (6.54), which has the same shape and strength, but in (6.55) sort of additionally nuances the claim somewhat. The presence of the postdetermining adjective same here contradicts Denison’s (2002: 3) claim that the postdeterminer use of TNs is incompatible with other postdeterminers. In this kind

6.3 Classification of NP-internal TN-uses and extended uses

297

of hyperbolic writing (6.56) a concrete instance of writing is generalized over anaphorically into a contextual type. In (6.57) that kind of person again does not refer to a subtype of ‘people’ in any generic way. Instead it generalizes, starting from reference to a very specific person. The same mechanism applies to examples (6.58) and (6.65), where in each case a very concrete and specific situation, undertaking or way of behaving is generalized over and conjured up as a type. The determiner complex construction occurs mostly with abstract N2s, or human nouns as collocates, but apart from a set of N2s that refers to ‘attitude/behaviour’, there is no significant collocational clustering, except for the determiner complex uses that have thing as N2, illustrated here by examples that sort of thing (6.64) and this sort of thing (6.65). Examples (6.66) to (6.69) construe similar meanings as examples (6.55)-(6.65) but feature plural determiners followed by a singular TNexpression and plural N2, i.e. the only structure Denison singled out as instantiating the postdeterminer use of TNs. (6.66)

“’What the f are you doing riding that thing,’” Powell recalled with amusement. “I get used to those kind of comments. I just say, “Well, somebody's got to ride it.” (CW-Times)

(6.67)

One survey suggests that about 90% of chief executives are of above-average height, while another found senior civil servants tend to be taller than junior counterparts. Perhaps most concerning was a study showing recruitment consultants selected the taller of two evenly matched candidates 72% of the time. Of course, we all know talent has many dimensions but let us hope these sort of studies are not cause for ill-founded bias in appointments. (CWTimes)

(6.68)

A pet owner could bury a wire carrying a radio signal which activated the collar device when the animal approached the boundary, he said. Dr. Blackshaw said barking inhibitors were usually only effective in 25 percent of cases. Using these sort of devices means you are treating the symptom rather than the cause she said. (CWOZnews)

(6.69)

We were only able to respond that we were unaware of any evidence linking plastic milk bottles and cancer, but your investigative report puts the whole issue into perspective. Unfortunately these sort of scare tactics do a lot of harm. (CW-OZnews)

298Chapter 6: TN-constructions (6.70)

It went bankrupt in the early 20th century, attracting Cuban refugees and American and European artists and writers and people who just wanted to hang out. It was once home to Ernest Hemingway, Tennessee Williams and John Audubon. These sorts of people are still here, and you can stay in their small ramshackle boarding houses. (CW-Times)

In examples (6.66) and (6.67), for instance, an anaphoric relation to the concrete instances mentioned in the previous discourse is construed, as well as a generalization of reference to higher-order concepts named by comments and studies respectively. In (6.68) to (6.70) the same anaphorically generalizing mechanism is construed referring to higher-order concepts devices, scare tactics and people respectively. Plural NPs with these sort/kind of imply a broadening of reference. For instance, in these sort of scare tactics, the speaker includes not only the examples previously mentioned but also any other instances covered by that generalization. In these sorts of people (6.70) the demonstrative, TN-expression and N2 are plural, but the construction as such realizes the same sort of generalized and broadening reference as examples (6.66) to (6.69). In many of the determiner complex constructions, the referential meaning of the determiner complex can be substituted by a NP with predeterminer such. In (6.57): He seems to attract such a person, in (6.58) such empty posturing, in (6.60) such behaviour, etc. (For further discussion of the affinities between demonstrative and TN + of and such, see Denison 2002: 6 and Mackenzie 1997: 89). Substitutability by such indicates that the postdeterminer use often involves what Ward and Birner (1995: 732) have characterized as dual reference, i.e. the mechanism by which reference to a concrete instance is implied by a NP whose overt reference is generalized. Actual pragmatic reference then is to one instance and not to the generalized concept as such. In Dutch such uses can be translated by predeterminer zo/zulk (een) (‘such a’), but also by a postdeterminer construction een dergelijke95 (‘a similar’), which also anaphorically generalizes over instances. Whereas in examples (6.55) to (6.70) generalized anaphoric reference invokes mainly the qualitative specifications of the concepts in question, examples (6.71) to (6.77) activate size implications of the N2 at hand.96 This postdeterminer use will be called the size-related postdeterminer use. (6.71)

She said: “I think the girls are amazing. There’s no way I could muster up that kind of energy during a Texas gig. I’d be knackered for a week.” (CW-Sunnow)

6.3 Classification of NP-internal TN-uses and extended uses

299

(6.72)

I could never afford a boat, nor an oceanside house (Marianne is editor of the Nantucket Inquirer and Mirror) because we don't have that kind of money. (CW-Times)

(6.73)

I was a bit worried that we came racing out of the starting blocks at such a pace, because I doubted sometimes whether we would be able to keep up that kind of effort for a full 80 minutes. (CWTimes)

(6.74)

The work of the director and his actors consists in asking questions about the play. This should turn all but the most routine rehearsal into a self-examination. What am I attempting to do here? Is Hamlet meant to shock or mystify the audience? Am I telling a story? Communicating Feelings: Am I giving a lecture on morality or am I directing a whodunit with bits of poetry stuck on? Most audiences are not aware that this kind of self-analysis has gone into a production. (CW-Times)

(6.75)

Conran walked away from a chance of a lease at the top of the Oxo building overlooking the Thames, since snapped up by Harvey Nichols. “We didn't want to pay that sort of money.” (CWTimes)

(6.76)

These are people who may be facing an armed robber one minute, or trying to quell a riot, then having to knock on someone's door to tell them their six-year-old child has been knocked over. No other job requires that sort of range of skills. (CW-Times)

(6.77)

Killik said that the new low-cost dealing services would “squeeze” some of the providers in the market. “Some may not wish to compete at these sort of prices. (CW-Times)

In (6.71) for example that kind of energy construes an anaphoric relation to the instance of energy associated with the girls’ performance in the preceding discourse, as well as reference to a more general concept of energy, by activating the size implications of the noun energy. At the same time, there also seems to be some intensification involved in this use of that kind of, conveying ‘a lot of’, or ‘more than some assumed standard’ of energy that a person has. In (6.74) too this kind of self-analysis translates as ‘a lot of selfanalysis’ and ‘more than people would expect’. The size-related postdeterminer has two easily recognizable sets of favoured collocates, i.e. one to do with energy/effort e.g. that kind of energy (6.71), that kind of effort (6.73), this kind of self-analysis (6.74) and that

300Chapter 6: TN-constructions sort of range of skills (6.76); and one to do with money, as in that kind of money (6.72), that sort of money (6.75) and these sort of prices (6.77). In the frequently attested uses with money, the intensification of the quantitative concept is even stronger, as it conveys ‘too much’. In (6.72) that kind of money refers to ‘the amount of money needed to buy a boat and oceanside house’, which is assumed to be more than the speaker actually has. It is reference to this assumed norm that aids identification of the referent involved. Example (6.77), with these sort of prices, exemplifies a size-related postdeterminer use with supposed concord mismatch, i.e. plural determiner + singular TN-expression + plural N2. In addition to anaphoric relations, the determiner complex may also point forward in the NP and realize cataphoric relations, as in examples (6.78) to (6.88): (6.78)

If the war reignited, it could spread and spark the kind of conflict that has drawn Americans into two larger wars this century. (CWTimes)

(6.79)

It might as well go to Richard Holbroke. It was he who pitched together the Dayton accord, bringing to Bosnia the kind of peace Stalin and Hitler brought to Poland. (CW-Times)

(6.80)

The two-mile handicap will be worth a minimum of 100,000 pound, second only to the Champion Hurdle in prize-money, and should help to attract the kind of big fields which graced the race in the 1970s and 1980s when it was sponsored by Schweppes. (CW-Times)

(6.81)

Both were adept at mouth-to-mouth resuscitation when their daughter stopped breathing on the bedroom floor. Precisely the kind of parents who, like any of us, thought it could never happen, at home, to their A-level student daughter. (CW-Times)

(6.82)

In America her latest movie, The Scarlet Letter, received the kind of reviews that you could distil into pure sulphuric acid. (CWTimes)

(6.83)

In other words, it is just the sort of movie nobody with any money is supposed to want to make. (CW-Times)

(6.84)

The illustrations are suitably funky cartoons, and it's the sort of book which could convert kids who think that reading is stuffy. (CW-Times)

6.3 Classification of NP-internal TN-uses and extended uses

301

(6.85)

Elizabeth Taylor (1912-1975) is the sort of minor English author consistent accomplished, controlled who inspires devotion in practising writers. (CW-Times)

(6.86)

Not only is Captain Ramsey the only one who doesn't break sweat, once, in the entire film, but he also chomps on the sort of cigar that looks as if it's primed to torpedo the Russians at the first available opportunity. (CW-Times)

(6.87)

“I was not forced out. I was not nobbled. I am not the sort of person who gets nobbled.” (CW-Times)

(6.88)

Starstruck is the type of programme Radio 4 does particularly well these days. (CW-Times)

In these examples, the NPs with postdetermining TN refer cataphorically to concrete instances described by the following restrictive relative clause, and generalize from these, e.g. the kind of peace Stalin and Hitler brought to Poland (6.79). They may also point forward to a more general description which as such defines the generalization expressed by the NP, e.g. the kind of reviews that you could distil into pure sulphuric acid (6.82). Both uses occur with all three singular variants of the TN-expressions. For the identification of the discourse’s referent, signalled to be retrievable by definite determiner the, the determiner complex in the kind of big fields (6.80) for instance points forward to N2 + relative clause, which contains the description of the very specific outlook of fields that graced a specific race in the 1970s and 1980s when it was sponsored by Schweppes. In addition, the determiner complex generalizes this one specific instance into a contextual type. The restrictive relative clauses often contain very detailed, expressive and particular descriptions of events or people, in a superficial way reminiscent of the longish nonce expressions found in front of the semi-suffix use in examples such as a blink thrice or you’ll miss it kind of town (6.39), not just a towel for a cape or underwear on the outside kinda costume (6.50) and a non-‘worship and serve me forever’ typa way (6.51), however, without the phoricity and identifiability marking criterial to the determiner complex use. As with the anaphoric examples discussed above, some of these can be paraphrased with predeterminer such, e.g. (6.79) such a peace as Stalin and Hitler brought to Poland. In examples (6.78)-(6.88) the cataphoric determiner complex with TNexpression interacts with qualitative specifications. Just like the anaphoric examples they can also interact with size implications of the concepts expressed by N2, as in examples (6.89) to (6.94):

302Chapter 6: TN-constructions (6.89)

We have spent a great deal of money trying to keep our players out of the reach of predators, but there was no way we could match the kind of offer made to Craig from our own resources. (CW-Times)

(6.90)

He went on to say that after looking deep into his own soul he had concluded that to offer himself as a candidate would require “a passion to run the race and win the quest; the kind of passion and commitment that I felt every day of my 35 years as a soldier.” (CW-Times)

(6.91)

So, while never supposing to pay the kind of fees that English clubs do for foreign talents, Ajax import them young and teach them a second language to their native tongue. (CW-Times)

(6.92)

No one at the PIA, of course, expects the total of complaints to be even near this figure, but it is the sort of number that concentrates the mind. (CW-Times)

(6.93)

It begins a week today, and it has attracted the sort of prepublicity usually reserved for the more bloodthirsty films. (CWTimes)

(6.94)

The war had been a major stimulus to computer technology in other parts of the world and while American and British scientists were receiving proper funding for their own computers ENIAC and Colossos respectively Zuse was not given the sort of money he required to develop an electronic version of his calculator. (CW-Times)

As with the anaphoric examples of size-related postdeterminer uses, frequent N2-collocates relate to money, as in the kind of offer (6.89), the kind of fees (6.91) and the sort of money (6.94); or an ‘effort’, as in the kind of passion (6.90), in addition to collocates such as the quantity word number in (6.92) or pre-publicity in (6.93). Again, the quantitative semantics often convey a meaning of surpassing an assumed norm of something or can even signify ‘too much’. In addition to determiner complexes which construe generalizing identifying relations, either anaphoric or cataphoric ones, TNs can also feature in non-phoric determiner complexes. Primary determiners in these non-phoric complexes seem to be restricted to what, any and some. The latter two determiners often have quantitative implicatures, in addition to identifier function, which is probably why many non-phoric constructions seem to be

6.3 Classification of NP-internal TN-uses and extended uses

303

size-related determiner complexes. The other main set of non-phorics are emphasizing postdeterminers with what. Examples (6.95) to (6.99) illustrate the size-related uses: (6.95)

Before you get involved in this kind of adventure you need to check very carefully just what kind of load your car can take. (CW-Times)

(6.96)

But anyone who watches the programme for any kind of insight would be misguided. (CW-Times)

(6.97)

If he were any sort of stylist he would not use the identical phrase (“a singularly unattractive individual") to describe his subject in the course of only seven pages. (CW-Times)

(6.98)

If she was any sort of psychic, she'd have made a blanket announcement at the start. (CW-Times)

(6.99)

Exasperated Bath know that, without entering some kind of contractual arrangement, the talent that exists in their second team will not remain if clubs with financial muscle come calling. (CWTimes)

In these examples the TN-expression functions as a postdeterminer, further aiding the primary determiner with which it is closely related. Any and some are grounding predications that combine identification semantics with quantifying implications. Most of these non-phoric determiner complexes hence fit the interpretation of size-related postdeterminers discussed earlier; they simply lack anaphoric or cataphoric generalizing reference. Examples (6.95) and (6.96) identify the load and insight respectively in terms of quantitative features, but no phoric relation pointing forward or backward in the discourse is expressed. Any is polarity sensitive, which is demonstrated by the frequent interrogative or conditional contexts in which it features, and the presence of negation markers. The downtoning semantics of any and sort of in such contexts give rise to emphatic pragmatic meaning. In (6.96) the speaker makes clear that no insight whatsoever can be gained from a particular programme. (6.97) and (6.98) similarly indicate that the speaker does not think the person being talked about to be anything near a stylist or psychic respectively. In (6.99) some kind of indicates a small, but asserted quantity. Examples (6.100) to (6.103) then are examples of what I call the emphasizing postdeterminer use.

304Chapter 6: TN-constructions (6.100) “Why doesn’t your child live with you? What sort of mother are you anyway?” (www.questia.com/PM.qst?a=refresh&docId= 0899319&type=book) (6.101) He wrote and starred in the successful musicals Stop the World - I Want to Get Off, The Roar of the Greasepaint - the Smell of the Crowd, and a number of hit singles, including What Kind of Fool Am I Who Can I Turn To? and The Candy Man. (CW-Times) (6.102) What kinda jerk would I be if I simply turned my back on her? (www.bigblack.blogspot.com/2005_01_01_bigback_archive.html) (6.103) What sorta question is that?! I assume You’re a beginner… (www.menshealth.co.uk/talk/ thread.phtml/post740523/-63K) Collocates in this use typically refer to humans or human behaviour. In (6.100), for instance, what sort of mother does not activate the size implications of N2, but rather degree of category membership, in that the speaker questions how much of a prototypical member of a certain class the person talked about is, in this case how much of a mother. The emphasizing postdeterminer use here occurs in the context of a rhetorical question, with strong exclamative value, which enhances the emphatic effect. The implicature then is that the person can hardly be called a mother anymore and hence is a bad one. In (6.101) the song title What Kind of Fool Am I is similarly featured within such a rhetorical question, with the I-persona knowing he is a big fool. Following Traugott (2006: 14), one might say that whereas size-related postdeterminers construe identifiability relations by positioning a referent on a scale of measurement, emphasizing ones such as (6.100) to (6.103) assess the scale itself. Bolinger (1972: 32) offers a sharp interpretation of the semantic shifts from describing to evaluating and intensifying involved in the establishment of this emphasizing use. In the first stage what sort of is used in descriptive generic statements, asking about a subtype of some superordinate category, as in what sort of fish in (6.11). This example was classified as a head noun use. In the second stage, the speaker decries something by pretending that it does not deserve its name (Bolinger 1972: 32). In what sort of mother are you anyway (6.100), for instance, it is implied that the addressee is a bad or unworthy instance of the category ‘mother’. Via metaphorization, this use develops further into an intensifying use (Bolinger 1972: 32), as in What Kind of Fool (6.101) and what kinda jerk (6.102), where kind of and kinda further emphasize the negative features that are part of the semantics of fool and jerk. In uses with neutral nouns such as question in what sorta question is that?! (6.103), finally,

6.3 Classification of NP-internal TN-uses and extended uses

305

what sort/kind of seems to have acquired a certain negative value of its own. Examples such as (6.11), (6.100) and (6.102) show that one and the same surface structure containing a TN-expression, i.e. what + TN + of + N2, can realize different constructional meanings, displaying progressive degrees of grammaticalizatity and subjectifivity. In Section 6.5 the chronologies of the various subtypes of postdeterminer uses will be discussed in detail.

6.3.4.3 Conclusion: A functional approach to the complex determiner construction with TNs Having looked at the postdeterminer uses of TNs illustrated by (6.55)– (6.103), we are now in a position to compare the category proposed here with Denison’s (2002/2005) approach to the postdeterminer construction. As we saw Denison (2002/2005) tentatively posits a postdeterminer use for TNs, the main formal evidence of which is the number incongruence between plural determiner and singular TN, as illustrated in (6.66)–(6.69). Little functional description is provided for the category. He moreover wonders whether this construction is not just a variant of the binominal construction and hence is unsure whether it has independent status as a truly distinct construction, formally and functionally. The approach to the postdeterminer uses of TNs presented here, on the other hand, is essentially functional, in that it notes the parallelisms with postdeterminer uses of adjectives. In a first step, this led to a discussion of the referential values expressed by determiner complexes involving TNs, i.e. generalizing reference, size-related and emphasizing uses. I will now further argue why the postdeterminer use of TN-expressions has to be viewed as a distinct construction, different from the head use of TNs, as well as the modifier uses, both in semantic and grammatical terms. As noted in Section 6.3.1, NPs with TNs as head are concerned with discussing subclasses and hyponymic relations between classes in a generic way. Head TN-constructions operate in what Langacker (1991) has referred to as the type universe of interpretation, making statements such as ‘There are n subclasses T of class Y’, ‘t is an instance of subclass T which is a hyperonym of class Y’, etc. TNs used as head followed by of + N2 all have generic reference in some way. This reference is realized by the interaction between the determination and categorization zones of the NP, in which TN+of+N2 realizes the categorization function, as reflected in the parsing of 5 types of animals in (6.6), as [determination: 5] [categorization: types of animals] or the new kinds of pay flexibilities in (6.10) as [determination: the] [modification: new] [categorization: kinds of pay flexibilities]. By contrast, TNs used as postdeterminers construe, together with primary deter-

306Chapter 6: TN-constructions miners, referential relations to concrete instances and generalize from them. These can be realized anaphorically, cataphorically (cf. Denison 2002: 3), or non-phorically. An example such as the same sort of looks in (6.55) thus has to be parsed as [determination: the same sort of] [categorization: looks]. The formal realization of these referential meanings is not restricted to NPs with these/those sort/kind of, as shown in the semantic discussion of examples (6.55)-(6.65), (6.70)-(6.76) and (6.78)-(6.100). However, number incongruence does typically characterize plural NPs with demonstrative determiner and postdetermining TN (but see example 6.70). This number incongruence reflects the fact that the number of such NPs is determined by N2, not by the TN, as the latter has coalesced with the determiner to express a referential and phoric modifier meaning comparable to such (a). This coalescence is indicative of the fact that postdeterminers tend to derive from lexically full, more autonomous units such as attributive adjectives – or head nouns in the case of TNs – via a grammaticalization process. In this process they shift towards general non-referential meanings and are bound to the primary determiner. NPs with singular N2, or with determiners that are not overtly marked for plural, cannot manifest number incongruence, and hence do not have an overt recognition mark of the postdeterminer status of the TN. However, all TNs used as postdeterminer are phonologically reduced (cf. Denison 2002: 3), which also reflects that they do not have autonomous head status. Table 6.8 represents relative token frequencies of postdeterminer meanings for each TN-expression. Table 6.9 represents relative frequencies of anaphoric, cataphoric and non-phoric postdeterminer uses. In both tables the raw frequencies for the Times data are reduced to normalized frequencies per half a million words: Table 6.8. Frequencies of postdeterminer TN-uses

Times p0.5mw COLT

Sort

Kind

Type

Sorts

Kinds

Types

Total

408

347

45

11

1

1

813

35.39

30.1

3.9

0.95

0.09

0.09

70.53

9

3

1

0

0

0

13

307

6.3 Classification of NP-internal TN-uses and extended uses Table 6.9. Frequencies of anaphoric, cataphoric and non-phoric postdeterminer uses

Times p0.5mw COLT

Anaphoric

Cataphoric

Non-phoric

Total

#

%

#

%

#

%

#

243

29.89

478

58.79

92

11.32

813

21.08

29.89

41.47

58.79

7.98

11.32

70.53

6

46.15

4

30.77

3

23.08

13

It turns out that these three types of phoricity relate both to the semantics TN postdeterminers can construe and to the determiners that typically precede the TN. Anaphoric, cataphoric and non-phoric postdeterminers can all have size-related uses, whereas emphasizing uses are most common with non-phorics with what. In terms of sets of determiners, it can be noted that anaphoric postdeterminers are typically preceded by demonstratives, while cataphorics mostly have the definite article the preceding the TN-string, and are typically followed by a restrictive that/wh-clause. Non-phoric determiner complexes typically have any, some or what as primary determiner. This phenomenon corresponds to the Firthian notion of colligation, defined as the process or result of grouping a set of words on the basis of their similarity in entering into syntagmatic grammatical relations. For example, a set of verbs which take a certain kind of complement construction would be said to be ‘in colligation with’ that construction, e.g. agree, choose, decline, manage, etc. colligate with to-infinitive constructions, as opposed to -ing forms. (Crystal 1991: 61-62)

Colligations can be seen as collocations in “syntagmatic syntactic relations” (Bublitz 1996: 3), whereas collocations “refer to groups of words considered as individual lexical items irrespective of their grammatical classes or relations” (Robins 1971: 63). The idea of colligational patterns also fits in with the notion of constraints applying to constructions in construction grammar, and collocationally constrained constructions. In conclusion, in this section I have argued that the postdeterminer use of TNs constitutes a construction in its own right, distinct from the head use, with its own functional meanings and with its own formal features and constructional-grammatical behaviour. 6.3.5 Qualifying use

308Chapter 6: TN-constructions The qualifying uses of TNs described in this section have already been insightfully discussed in the literature. Denison (2002), for instance, offers a plausible scenario for the shifts leading to these uses, and Aijmer (2002) gives a very fine-grained, data-based description of their various pragmatic values. In their qualifying use, TN-strings basically hedge the type specifications incorporated by the linguistic predications which they hold in their scope. In Quirk et al.’s (1972: 4552) terms, they function as approximators, hedgers or downtoners of these categorizations. Examples of qualifying TN-expressions are (6.104) to (6.106), with their scope ranging from scope over a noun (6.104), to an adjectival predicate (6.105) and finally an entire sentence (6.106): (6.104) But I suppose it’s as a that’s as a sort of holiday, kind of doing you know nothing but sitting around. (quoted in Denison 2002: 2) (6.105) Overall, Reign of Fire is a good popcorn film but not a spectacular film. I wont give anything away but it feels sort of idiotic and a waste of Matthew McConaughey’s character. (www.bigscreen. excepc.com/ReaderReview) (6.106) Now Stan Collymore had moved up to the A-Listers, having his way with Hollywood siren Sharon Stone in the back of a car. For 30 seconds. Well, sort of. (www.sky.com/showbiz/article) Sort of in (6.104) indicates that the nominal category holiday is used ironically. In (6.105) it downtones the adjectival predicate idiotic, and in (6.106) it modalizes the whole preceding proposition, taking back some of the strength of that assertion. According to the basic type of grammatical unit being qualified, I will make a distinction between nominal, adverbial and sentential qualifiers. The terms nominal qualifying and sentential qualifying use refer to the nature of the element they hold in their scope, whereas adverbial qualifying use refers to its adverbial function with regard to adjectival, verbal and adverbial predicates. In the nominal qualifier use, Denison (2002: 11) points out, the TN has undergone a shift from head to modifier status. With reference to Langacker (1991), he (2002: 11) notes that the TN is thus defocused and moves “towards a less holistic construal of [a] category, recognising its internal structure (degrees of centrality)”. The qualifying construction is “in effect about the nature of membership of the class of N2” (ibid.), that is, it can be used to refer to a possible, arguable, or a peripheral member of that class. Denison (2002: 11) also notes that type does not take part in this development.

6.3 Classification of NP-internal TN-uses and extended uses

309

According to Denison (2002: 12), one of the scenarios by which the adverbial construction might have arisen is by extending the syntactic range of sort of or kind of as a hedging device from modifying a noun, as in a sort of holiday (6.104), to modifying other categories as in (6.107), in which kind of is used first with a verb, and then, after a pause to reframe the VP, before a noun. (6.107) and they kind of group – put people into kind of categories (quoted in Denison 2002: 12) Sections 6.3.5.1 to 6.3.5.3 discuss the nominal qualifier use, adverbial qualifier use and sentential qualifier use respectively.

6.3.5.1 Nominal qualifying use Relative token frequencies of the nominal qualifying use for TNexpressions are represented in Table 6.10. For the Times data normalized frequencies per half a million words are added in a second row. Plural sorts, kinds and types do not appear as nominal qualifier in the corpus data. Table 6.10. Frequencies of the nominal qualifying use

Times p0.5mw COLT

Sort

Kind

Type

Sorts

Kinds

Types

Total

157

141

4

0

0

0

302

13.62

12.23

0.35

0

0

0

26.2

7

2

0

0

0

0

9

In general, the nominal qualifying construction with sort of and kind of is, as pointed out by Denison (2002), concerned with the relation of its referent to the categorization used in N2-position. Some examples of nominal qualifying constructions are (6.108)-(6.114): (6.108) They had allowed the economy to ride on a tide of credit and debt and called it an economic miracle. --- Well, it was a kind of miracle. After all, they made almost £100 billion on North Sea Oil revenue vanish without trace. (CW-Today) (6.109) Are they a kind of Shakespeare Sister then? (COLT)

310Chapter 6: TN-constructions (6.110) Pillars of Hercules, By Paul Theroux. This smashing book is a sort of anthology, or epitome, of Therouxness. (CW-Times) (6.111) Turn the bird skin-side up and flatten out firmly with the heel of your hand to form a sort of butterfly. (CW-Times) (6.112) She started off as a kind of supergroupie, but then he couldn't be without her. (CW-Times) (6.113) He’s a sorta Leftist (with whom I often find myself in agreement) whose blog is always worth a visit. (www.blimpish.typepad .com/blog/2005/04) (6.114) Virtually every Republican candidate this year has endorsed some type of flatter or flat tax structure along with radical simplification of the system. (CW-Times) The structural analysis proposed by Denison (2002: 2 and 3) for this construction sees N2 as the semantic head and the sort/kind-string, without the article, as a unit which qualifies N2, i.e. [det phrase[det a][NP[qual sort of] [N holiday]]] (example cited in Denison 2002: 2). Kruisinga (1932: 396) similarly argues that “in a sort of pleasure the article qualifies the whole group with pleasure as its leading element.” However, there are a number of arguments against this analysis. As noted by Margerie (2010), if the determiner in this construction is some, omission of the TN-string (mostly with kind of) leads to awkward results and change in meaning, e.g. Is it really some kind of sex test? (CW-Times) vs. ?Is it really some sex test? and They wanted to enshrine the World Cup triumph as some kind of national treasure (CW-Times) vs. ?They wanted to enshrine the World Cup triumph as some national treasure. In (6.114), a rare example with type, some flatter or flat tax structure equally sounds awkward and the meaning is changed. In addition to Margerie’s (2010) observation, it can be noted that if the determiner is a and N2 is uncount, omission of the TN-string leads to results that are simply ungrammatical, e.g. It’s a kind of magic (CW-Times) vs. *It’s a magic. N2 may moreover have its own determiner, e.g. Some kind of an artist (CW-Times). Hence, the present study proposes a structural analysis that sees the (indefinite) determiner preceding the TN-string as part of one complex qualifying unit,97 i.e. [qualifying unit: determiner + TN + of] + (determiner) + [head: N2]. This qualifying unit then functions as a grounding predication, similar to a bit of. An advantage of this structural analysis is that constructional-semantic overlap between the qualifying uses of a bit of and a/some sort/kind/type of is captured. In addition, it also provides diachronic conti-

6.3 Classification of NP-internal TN-uses and extended uses

311

nuity with the structural analysis of the postdeterminer use (see Section 6.3.4 and 6.5 on the diachrony of TN-constructions). The general meaning of the nominal qualifier is one of hedging and approximation, vis-à-vis categorization statements, but this can be done for various reasons and further meanings can be added to this core meaning. In a kind of miracle (6.108), for instance, the speaker seems ironic in that the concept of a ‘miracle’ and that of ‘making £ 100 billion vanish without a trace’ is incompatible. In examples (6.109) to (6.114) the speakers’ use of the TN-expression in front of N2 shows uncertainty about the category label used. This can be because N2 is too strong or possibly an offensive label, as in a kind of supergroupie (6.112) or a sorta Leftist (6.113); or in (6.109)-(6.111) the speaker uses a more familiar category label as a reference point for describing something that only approximates the prototypical member of that category label, e.g. a sort of anthology (6.110). Some type of flatter or flat tax structure (6.114), referred to earlier, shows that, against Denison (2002), a type of can also construe hedging semantics, though it rarely does (see Table 6.10), and always functions as approximator in the Times data. In addition to approximators, qualifying sort/kind of are also often used as more explicit metalinguistic operators (Aijmer 2002: 178). As pointed out by Aijmer (2002: 195-196), they can function as marker of lexical imprecision, as a warning of a style shift, or as a self-conscious marker of creative idiom or metaphor, which “enables the speaker to be creative, to use words in an innovative and humorous way, to borrow phrases belonging to literary or more formal style and to use a slangy turn of speech” (Aijmer 2002: 195). For instance in (6.115) to (6.120), (6.115) Well I don’t like these er … what’s … for want of a better word these sort of organised paths like the Pennine way. (CWUKspoken) (6.116) Essentially, Richard Julin, the curator for the museum, was curious about the linkages between architecture, history, and how film can be seen a [sic] kind of hyper-textual archaeology. (www.djspooky.com/art.html) (6.117) Offering the viewer a kind of a la cart theme park ride through the future, “The Fifth Element” reminded me an awful lot of the Paul Verhoeven megamovie “Total Recall”. (www.dcox.customer. netspace.net.au/fifth.html/) (6.118) More than any other nation, the Scots regard the passing of the year as some kind of baptism in booze. (CW-Times)

312Chapter 6: TN-constructions (6.119) Full breathing is also an important tool for encouraging waste elimination; a kind of spring-cleaning process that can go on all year round. (CW-Times) (6.120) His wit earned him the friendship of Virgil and, through Virgil's good offices, of the millionaire Maecenas, who was a sort of oneman Arts Council (or intelligent lottery). (CW-Times) In examples (6.115) to (6.120), N2 is a creative coinage of the speaker, rather than a more familiar category label, as in examples (6.109)-(6.111). In these sort of organized paths (6.115), the speaker explicitly shows uncertainty about the precision of the category label by for want of a better word. Such metalinguistic uses typically occur in contexts where a metaphorical comparison is signalled to be a creative approximation of what the speaker wants to communicate, for instance the comparison of full breathing with a daily spring-cleaning in (6.119). In addition to hedging approximator semantics and metalinguistic meaning, qualifying TN-uses can have intensification semantics in some ironic contexts, similar to the pragmatic reversal mechanism noted for SSN bit of in Chapter 5. Approximating hedging uses, such as a kind of Shakespeare Sister (6.109) and a kind of supergroupie (6.112) are pragmatically mostly intersubjective in meaning, in that they attempt to clarify the type specification in the NP for the hearer or try to protect his face. In contrast, the intensifying booster meaning displays attitudinally subjective meaning, in that it shows a speaker’s attitude about the proposition. It is evident that context and prosodic patterns are essential in triggering off this subjective value, and booster semantics remain implicational in nature. Examples are (6.121) to (6.123): (6.121) He might not have had a memorable hit in years, but fancies himself as some kind of an artist. (CW-Times) (6.122) Well, you know, I don't want to rain on your parade literally, but won't it be kind of a bust if Joe isn't there? (CW-NPR) (6.123) To find himself chasing after teenage flesh like some farcical sugar-daddy was a kind of nightmare. (CW-Times) In these examples, instead of hedging the appropriateness of the category label, categorization judgments are reinforced, by means of a mechanism similar to that of understatement expressions. In (6.121) the speaker implies that the person talked about considers himself to be a big artist, and this is frowned upon by the speaker. Kind of a bust in (6.122) similarly translates

6.3 Classification of NP-internal TN-uses and extended uses

313

as ‘very much so a bust’, when read as a rhetorical question. In a kind of nightmare (6.123), rather than saying that chasing after teenage flesh was more or less a nightmare, the speaker apparently wants to convey that it really was a nightmare. In some kind of an artist (6.121) and kind of a bust (6.122), N2 is moreover premodified by the indefinite article, which was also typically the case in the subjective and intersubjective uses of a bit of. The semantic prosody of N2 in nominal qualifying uses is, unsurprisingly, typically negative for the hedging uses. In the intensifying uses N2 is also typically negative, but need not be so, e.g. some kind of an artist (6.121), even though the broader context is negatively coloured. N2 may also be positive, as in a kind of miracle (6.108), a kind of spring-cleaning process (6.119), a sort of one-man Arts Council (6.120). N2 is neutral in the more metalinguistic qualifying uses, e.g. these sort of organized paths (6.115).

6.3.5.2 Adverbial qualifying use Relative token frequencies for the adverbial qualifying use are represented in Table 6.11. The second row of the Times frequencies are frequencies per half a million words: Table 6.11. Frequencies of adverbial qualifying uses

Times p0.5mw COLT

Sort

Kind98

Type

Sorts

Kinds

Types

Total

18

12

0

0

0

0

30

1.56

1.04

0

0

0

0

2.6

53

7

0

0

0

0

60

In their adverbial use, sort of and kind of can similarly be used as approximators to qualify the categorizations in their scope as approximate or imperfect in relation to the instances being depicted (Quirk et al. 1972: 452; Huddleston and Pullum 2002: 623-624; Bolinger 1972: 223; Aijmer 2002: 49; Margerie 2010). Type of does not appear in this pattern in my corpus data, nor do plural sorts, kinds and types. The structural analysis of the qualifying unit is the same as the one provided for the nominal qualifying use, except for the determiner in front of the TN-expression, which is absent from the construction because of the non-nominal nature of the element being qualified.

314Chapter 6: TN-constructions In some cases, adverbial qualifying constructions are used as downtoners, “suggesting an incomplete or low degree” of, for instance, the quality referred to (Quirk et al. 1972: 452). The types of grammatical unit that can be adverbially qualified in this way include adjectives, numbers, verbs and whole predications: (6.124) Then later in the night we took a walk in our underwear around the campus. That was sorta weird. (www.yaledailynews. com/article) (6.125) “I don't think my paintings are ‘universal’ or ‘pure’” she wrote, “they're kind of phoney abstract.” (CW-Times) (6.126) You know them hopper things you bounce on, the round balloon thing? She's got one, it's flat, it's not flat but it's kind of deflated. (COLT) (6.127) No it's not that bad the game actually it's alright but, it is a bit, sort of boring when you play it every day. (COLT) (6.128) It's an industrial process done by hand, which is kind of ironic. (CW-Times) When used as an adverbial qualifier of an adjective as in (6.124)-(6.128), the TN-string indicates that the description is only approximate, or that the quality described is not present to a very high degree in the instance being considered. In a bit, sort of boring (6.127), the TN-string functions as an alternative for downtoning adverbial bit. Adverbial qualifying uses also seem to allow for intensifying implicatures, as in (6.129) to (6.131). (6.129) It’s kind of stupid but it’s kind of funny of you know what I mean cos the baby don’t talk and as soon as the Mum and Dad’s out of the room its’ let’s go and find the chocolate. It’s so funny. (COLT) (6.130) Oh I like that song. That’s wicked! (…) There’s gonna be DJ and all that. With all hard core music. It’s safe. Kinda wicked! Well I’m buying my ticket today. (COLT) (6.131) “This show is a spectacular flop.” Unsurprisingly, only the first of James Thurber's two-sentence review “It had only one fault. It was kind of lousy.” (CW-Times)

6.3 Classification of NP-internal TN-uses and extended uses

315

In (6.129) kind of funny is later on rephrased as so funny. In (6.130) the speaker is very much looking forward to the upcoming DJ event and in this emphatic context his kinda wicked conveys intensifying semantics. Example (6.131) cites a very negative review and rather ironically decides that it was very lousy, literally coded as kind of lousy. Just like in the discussion of bit of in Chapter 5, it has to be noted here that intersubjective hedging meaning and subjective intensification meaning intertwine and cannot always be distinguished in a straightforward manner, not only due to lack of contextual and prosodic markers, but also because some instances are (intentionally) mixed in contexts where face work and emphatic meaning combine. As illustrated by (6.132), the qualifying TN-string can also serve as an indicator of number approximation (Aijmer 2002: 196): (6.132) It was the late ‘70s and everyone was getting into punk rock, and I was kind of sixteen and going for something else. (www.npr.org/ programs/thistle/features reader_int2.html) Likewise, with verbs, adverbial qualifiers can function either as approximators or as downtoners (Quirk et al. 1972: 276-454): (6.133) He sort of smiled at us. (Quirk et al. 1972: 454) (6.134) Until then, I had felt numb and sort of sleep-walked through each day. (CW-Times) (6.135) Yeah I mean I’m not, I’m not saying you’re a bad influence on her now, I think she’s ... just sort of jumping in at the chance for someone to victimise personally […] (COLT) (6.136) I am never happy unless I am winning; I really thought I would run a lot faster: I kind of let myself down. (CW-Times) In sort of smiled (6.133) and sort of sleep-walked through each day (6.134), the TN-string can be rephrased as you could almost say, which implies a denial of the truth-value of what is denoted by the verb. In sort of jumping in at the chance (6.135) and kind of let myself down (6.136), the TN-string invokes an assumed norm with reference to which the force of the verb is reduced. When the TN-string frames whole predications, it can occur in front of them or in end position. The latter is illustrated by (6.137):

316Chapter 6: TN-constructions (6.137) This plugin is designed to solve that problem, sort of. (www.nbcs.rutgers.edu/hedrick) As noted by Quirk et al. (1972: 456-457), front position has to be distinguished further into pre-operator position, where the qualifier lies “within the scope of clause interrogation and negation”, as in (6.138), and postoperator position, where it merely frames the type specifications of the predication, as in (6.139). (6.138) He kinda sorta maybe loves him. (www.alykat.hispeed.com /unfrozen/fanfic/mooks) (6.139) I’ve sort of become part of the mountain bike world in a way without actually having a mountain bike; it’s quite strange. (CWUKspoken)

6.3.5.3 Sentential qualifying use Relative token frequencies for the sentential qualifying use are represented in Table 6.12. Table 6.12. Frequencies of sentential qualifying use

Times p0.5mw COLT

Sort

Kind

Type

Sorts

Kinds

Types

Total

3

1

0

0

0

0

4

0.26

0.09

0

0

0

0

0.35

2

3

0

0

0

0

5

In its (as yet rare) use as sentential qualifier, sort of/kind of have rid themselves even further of the structural constraints of the NP and VP or other predicates they qualify. In these uses, they have scope over an entire proposition, and function like adverbs such as perhaps, largely, etc., qualifying the truth or accuracy of that proposition. Because of their propositional scope, sentential TN-uses may involve conversational moves by distinct speakers, as in (6.140) and (6.141). (6.140) CMG: Sort of like In The Fishtank [LB: two bands have to record an album together within a certain time span]? Beam: Kinda.

6.3 Classification of NP-internal TN-uses and extended uses

317

Definitely like that but not with all the parameters. Not like you have to go and perform it all in one day. (www.cokemachineglow. com/feature/interview) (6.141) A: Do our eyes deceive us or are we witnessing a return to the brash, rich, greed-is-good, yuppie 1980s. B: Well, kind of. (CWTimes) In these examples, the sentential qualifying use is also similar to yes/no as answers to a polar question. The TN-string, then, is a hedged variant of these. Type of was not attested in the corpus data in this qualifying use, nor were plural sorts, kinds and types.

6.3.6 Discourse marker use As discourse markers, TN-strings are not tied to grammatical class boundaries anymore, and lack clear notions of scopal domain. They apply more diffusely to the discourse, through which they are scattered. According to Denison (2002: 4, 14), the discourse marker use developed from the approximator value of the adverbial use through processes of subjectification and semantic bleaching (see Section 6.5). Relative token frequencies for discourse marker uses are represented in Table 6.13. Type does not appear as discourse marker in the corpus data, nor do plural sorts, kinds and types. Table 6.13. Frequencies of discourse marker uses

Times p0.5mw COLT

Sort

Kind

Type

Sorts

Kinds

Types

Total

5

0

0

0

0

0

5

0.43

0

0

0

0

0

0.43

28

4

0

0

0

0

32

As discourse markers, TN-strings are used as indicators of tentativity, fillers and hesitation markers. They may convey general tentativity about what is being said, comparable to the present-day discourse marker use of like (Meehan 1991), which is often used together with the TN-string in this sense, as in (6.142), (6.144) and (6.145).

318Chapter 6: TN-constructions (6.142) As I remember it used to be sort of like fairly common for a Tuesday, that I’d pretend to be sick. (quoted in Denison 2002: 4) (6.143) I don't want to, you know, sort of, well, effing badmouth him. (CW-Times) (6.144) And do you know those advert features that they do in the States, they're like sort of like ... half hour adverts. (COLT) (6.145) I've never really been, like, single, cos I'm always sort of like, not actually going out with somebody but I'm always, I've always got someone who sort of fancies me. (COLT) It can be noted that tentative sort of also signals social non-dominance by the speaker, to the extent that it may be over-used by a speaker in an inferior position, as a marker of deference to the other speaker in superior position, as in the following exchange between lecturer (L) and student (S), quoted by Martin (1980), which includes discourse marker uses as well as qualifying uses. (6.146) L: Do you find the system, er, makes sense? S: Yes, I think I’m er beginning to understand it better now and I must say, I’m sort of impressed… I quite like the idea of, er, sort of, er, sort of, flexibility I think is the keyword, isn’t it? … in the sort of Prospectus. L: So they tell us. S: And yeah, and it seems to me as well that the, er, the exam system is, er, a much better idea if you sort of, er, take it, the whole, the whole, um, six terms and, um, you know, work it out on assessment like that rather than sort of three hours, pass or fail, sort of. L: How are you getting on with Don Juan? S: Er, um, I quite like it, really. It’s, er, part of it I find, er, sort of, er, a bit contrived but I suppose that’s sort of Byron’s style that, er, um. L: What’s Byron’s style?

6.3 Classification of NP-internal TN-uses and extended uses

319

S: One I think generally he’s, er, it’s, um, somewhat satirical style in that L: Uh huh. S: Uh, he’s particularly in Don Juan he’s sort of bringing out the, er, bitterness of sort of family life, I mean his wife, er, left him on, er… Sort of and kind of may also be strewn across the discourse as fillers or hesitation markers like er. Whereas some of these seem to be used largely unintentionally (6.147), other uses are more strategic (6.148), and serve the function of floor-holding and buying time “while the speaker plans what to say next” (Aijmer 2002: 188). (6.147) …and I sort of opened the door, and looked out, and I sort of saw Richard... (COLT, quoted in Denison 2002: 4) (6.148) ^well I !don’t think .^it’s ^((sort of a)) . a com:plete con:cl\usion= you’re sort of ^left with the - - you ^sort of [∂:m] – it’s ^sort [∂?] an :end to a :story in a :w\/ay= . you can ^just im’agine_these_things_going \on# it ^sort of !winds \up# it’s [∂:m].^rather an _arti!f\icial .{^[d\u:nei’mal]#}# ^rather ‘like [?] ‘one of [∂:m] ‘Moli!\ere’s ‘plays# (COLT, quoted in Aijmer 2002: 189)

6.3.7 Marker of onomatopoeia and quoted speech and thought Finally, TN-strings can frame onomatopoeic expressions as in (6.149), and, like innovative quotative markers such as go and be like, can represent speech as in (6.150) (cf. Aijmer 2002). (6.149) I’ve ^never s/\een a ‘sortof# ^bottle ‘after :b\ottle# . sort of ^pop ‘pop p/opping# âll the t/\ime# - - and ^everybody got :awfully dr\unk I rem/ember# (COLT, quoted in Aijmer 2002: 186) (6.150) “I just got a visual, Sharon standing in front of the class going, (SCREAM), while these little kids kinda ‘Señorita Flynn? Hee hee hee hee hee.” (Santa Barbara Corpus, quoted in De Smedt 2005: 112)

320Chapter 6: TN-constructions In these uses, kinda/kind of, sorta/sort of are often used together with be and go, as in (6.151) to (6.153). It is not always easy to decide to what extent the quotative semantics are copied from these verbs. (6.151) im just being kinda hey i can hear murkin (www.livejournal.com/ users/andyhello) (6.152) Then it homes in on one of the men right and it goes sort of “whee” one hour later “whee ... whee bom bom bom” and the other one goes one hour later. (COLT) (6.153) He kinda went, ‘Yeah, I think so!’ (www.rockconfidential.com/ Testament.html) The onomatopoeic and quotative use is still very rare, with one attestation in the Times subcorpus for sort of and six in COLT, all of which equally for sort of. 6.3.8 Ambivalent contexts99 In addition to the seven functional categories of TN-uses discussed above, the Times and COLT data sets also yield examples that are ambivalent between two constructional uses. Such synchronic ambivalence indicates semantic and syntactic similarities between constructional uses and may, but need not, point towards diachronic bridging contexts between constructional meanings (cf. Traugott & Trousdale 2010). Since the ambivalent contexts discussed here are attested by synchronic data, they can equally point out non-chronological interference between constructional meanings or constructional blends (cf. Bolinger 1961 and Aarts 2007) (see the diachronic corpus study in Section 6.5). Aarts (2007: 187-192) distinguishes between different types of constructional gradience “in terms of the extent to which they [i.e. mixed structures] are separable into different constructions”. Blends fuse “particular properties” of two constructions into “novel” constructions. Rather than interpreting it as a performance error, Aarts (2007: 187) considers blending to be “a process” that “probably” plays “an extremely important role in language change”. Following Bolinger (1961), blends are characterized as tightly integrating characteristics of two constructions, to the point that these are no longer “independently recognizable” (Aarts 2007: 189). As opposed to fusions for instance, blended constructions are not perceived as unusual, and may gain general currency. In Aarts (2007: 188) and Bolinger (1961), blending primarily pertains to the structural properties of two constructions,

6.3 Classification of NP-internal TN-uses and extended uses

321

but also includes the semantics, whereby the meaning of the resulting blend is implicational in nature only. Aarts (2007: 188) also notes that a blend allows the speaker to “imply both [meanings] at the same time.” I would argue that the type of blended meaning of the TN-expressions discussed in this section surpasses implicational meaning, though the saliency of both meanings can shift depending on the context. Blends with TN-expressions are structures in which the constructional properties of two TN-uses are combined. These blended examples can hence be related to the notion of vagueness described for SN-expressions, but in contrast to the latter, blends incorporating TN-expressions do not typically involve metaphorization. The ambivalent contexts discussed in this section are either ambiguous in nature or constitute blends. Only a restricted set of examples can be called vague. In the Times data set the following ambivalent contexts were attested: head/postdeterminer (H/PD), head/modifier (H/M), attributive/semi-suffix (Attr/Ss), head/quantifier (H/Q), postdeterminer/qualifying (PD/Qual.) and attributive modifier use/nominal qualifier (Attr/Nqual.). For the COLT data set the mixed categories are head/postdeterminer and qualifying/discourse marker (Qual./DM). As the figures in Table 6.14 show, the frequency of these ambivalent categories is altogether relatively low. Table 6.14. Frequencies of dual reading contexts H/PD

H/M Attr/Ss H/Q PD/Qual

Attr/ Nqual

Qual/DM

Total

Times

45

8

2

23

2

3

0

83

p0.5mw

3.9

0.69

0.17

1.99

0.17

0.26

0

7.2

5

0

0

0

0

0

16

21

COLT

I will now discuss examples of each, starting with the head/postdeter-miner category: (6.154) Lee himself, making one of his cameos, is to be seen in the crowd of onlookers that gather around the corpse, as if curious to see what sort of plot will sprout up around it. The seeds for a thriller, perhaps? (CW-Times) (6.155) This [study] found that in 1993, for instance, of the total of 3,201 crimes committed by persons on bail, 60 were larcenies and rob-

322Chapter 6: TN-constructions beries, the sort of crimes most typically committed by drug addicts. (CW-Times) (6.156) What sort of people will be there? Oh the mums will be there. (COLT) In these examples the main ambivalence is triggered by the question of whether meaning is construed in type space or not. More specifically, in (6.154) for instance what sort of plot can be interpreted as a head noun construction, cf. what sort of fish in (6.11), referring to a true subtype of the superordinate category of ‘plots/genres’, e.g. thriller, science-fiction. One can also read the ambiguous sentence as a non-phoric postdeterminer, as in what kind of load (6.95), with the TN demoted and generic reference faded into instantial reference. In (6.155) the sort of crimes on the one hand refers back to real types of crime, i.e. larceny and robbery. On the other hand, it also sets up generalizing reference, conjuring up a type that extends from the 60 instances of larcenies and robberies mentioned in the previous discourse, as often happens in the postdeterminer use. The sort of crimes then anaphorically relates to these, but at the same time cataphorically creates the generalization to drug addicts as typically associated with these crimes. Both constructional meanings seem to be blended here. The ambiguity in (6.156) hinges on whether the mums can be considered a true subtype of people here or whether it is just an answer to the question what sort of people in terms of instances, in which case it would be a non-phoric postdeterminer. It was observed earlier that definite determiner + TN + of and what + TN + of are structures that can realize either head noun or postdeterminer uses. It is hence unsurprising that they give rise to ambiguity and blends. Example (6.157) exemplifies ambiguity between a head and attributive modifier reading: (6.157) Leiths at Fyvie inherited a dreadful curse on the castle, now in National Trust hands, which says the eldest son never inherits (a pretty standard sort of curse in the 16th century). (CW-Times) The ambiguity here depends on the phonetic realization of the TNexpression and the scope of the adjective preceding it. In (6.157) either a real subtype of all types of curses is referred to, or reference is made to an instance. Pretty standard can ascribe a quality to the phonetically salient type specification [sort of curse]. In that case the TN has head noun status and refers to an actual subtype of the superordinate category ‘curses’, i.e. ‘curses on castles’. The TN-expression can also be phonetically reduced, with stress on pretty standard, in which case the construction has instantial

6.3 Classification of NP-internal TN-uses and extended uses

323

meaning and comes to mean ‘a standard curse’. The TN-expression is then part of an attributive modifier construction and designates a qualitative variant rather than a type. The next ambivalent context is that of head/quantifier, exemplified by (6.158) to (6.160) (6.158) They are just three of dozens of companies wooing investors with tempting offers to buy their shares. All kinds of discounts and freebies are being touted to investors by quoted companies. (CWTimes) (6.159) Performers need to arrange specialist insurance against all kinds of eventuality to ensure that the band plays on. (CW-Times) (6.160) I’ve met all sorts of people from social strata I would not otherwise have encountered. (CW-Times) As discussed in Section 6.3.2, there are fuzzy boundaries between the head and quantifier use. In (6.159) all kinds of eventuality can refer either to all the different subtypes of eventuality an insurance policy can ensure, or to ‘a lot of’ them without invoking types. In examples (6.158) to (6.160) all kinds/sorts of seem to have lost its universal relative quantifier reading, but may still invoke the subtype meaning. In example (6.158) all kinds of discounts and freebies in one reading refers to different types of discounts and freebies, in which case kinds is head noun and is preceded by a quantifier, cf. many kinds of courage in (6.12). Alternatively, it can be interpreted as just referring to ‘a lot of discounts and freebies’, i.e. a quantifier reading. The same goes for (6.160) in which the ambiguity is triggered off by the variable scope of from social strata. On the one hand, the sentence can be read as referring to different types of people, in the sense of ‘coming from different social strata’. On the other hand, the sentence can mean that the speaker encountered many people from different strata. An example from the postdeterminer/qualifying category is: (6.161) Government plans to tear up the 25-year licences it issued just before privatisation in 1989. Instead, it will promote the sort of “new era” of competition that has already done so much to multiply complaints against British Gas. (CW-Times) In this example structural properties of the postdeterminer and qualifying use are blended. The sort of “new era” invokes both a cata-phoric postdeterminer meaning to the following relative clause and a qualifying reading.

324Chapter 6: TN-constructions The latter is indicated by the quotation marks around new era, which frame it and set it off as hedged. The attributive modifier/nominal qualifier category is exemplified by (6.162) and (6.163): (6.162) Climbing the stairs from the entrance, I was shown into a weird sort of Gothic library-cum-bar complete with a hidden door, behind which waiters lurked, forgetting to take my order. (CWTimes) (6.163) Five months ago, King had already mapped out most of his ‘serial’ novel. Funny sort of high-wire act. (CW-Times) Both instances contain a qualitative adjective in front of the TN-expression and have classifying expressions in front of N2. Both examples seem to be blends of hedging semantics, typical of the qualifying construc-tion, and of attributive modifier semantics. (6.162) describes a weird variant of a Gothic library-cum-bar, but also qualifies the category label Gothic library-cumbar. The same goes for (6.163). Phonetic salience of the TN-expression versus the adjective premodifying the TN may make either the qualifying or attributive modifier interpretation more salient. Finally, there is the qualifying/discourse marker category, attested in the COLT corpus and exemplified by (6.164) to (6.169): (6.164) It’s kind of like, baggy. Yeah they're, MC Hammery like. (COLT) (6.165) You can hear everyone talking like sort of really clearly. (COLT) (6.166) Apparently she's always walking down the streets and like taking her top off and showing her tits to everyone and sort of like pulling her trousers down. (COLT) (6.167) I can't work out whether they actually think that or they're just sort of saying that because he doesn't talk to them as much or something (COLT) (6.168) Yeah but she might sort of like expose herself but is she vulgar? (COLT) (6.169) I, I won't like ... when I'm with other people like and I talk about other people like when I talk about Jenny and then I feel sort of like really two faced (COLT)

6.3 Classification of NP-internal TN-uses and extended uses

325

These examples show that there is no clear-cut boundary between the hedging semantics expressed by qualifying uses, with scope over various types of predicate, and the broader scope and further detachment from NP or clause structure typical of the discourse marker use. The presence of other hesitation or hedging markers, such as like, as well as the semantic prosody of the nouns, adjectives and verbs included in the discourse may make such examples gravitate to either of both categories. In a general way all of these ambivalent contexts point out that the constructional semantics and structural features of TN-uses are quite closely related, suggesting constructional gradience (cf. Aarts 2007). Table 6.14 showed that in the Times corpus head/quantifier and head/postdeterminer uses are most frequently attested and in COLT the qualifying/discourse marker uses occur the most. This may be an indication of ongoing change since both the qualifying and discourse marker use are rather recent uses. In Section 6.5 we will address the chronologies of the various TN-uses in detail and answer the question of whether the attested synchronic ambivalence can be linked to diachronic bridging contexts. It seems that factors in synchronic ambivalence such as the phonetic realization of the TN-expression, collocational values of N2, including semantic prosody and colligational relations between determiners and postdeterminer uses of adjectives can diachronically serve as triggers as well. As hinted at earlier, the constructional blends discussed here differ from the vagueness attested with SN-constructions. The vague instances of SNconstructions were cases in which the original lexical meaning of the SN was rekindled, e.g. striding past the compost heap of nostalgia. In a constructional blend such as the sort of “new era” of competition that has already done so much to multiply complaints against British Gas (6.161) two readings are activated at the same time, because of the presence of structural properties typical of two TN-uses. Vagueness of SNs and TN blends hence compare to the extent that the two readings involved in them cannot and should not be disambiguated. The difference is that blends do not involve relexicalized cases, but integrate two constructional TN-uses that do not cancel each other out. In example (6.161), then, qualifying and postdeterminer meaning combine. The synchronically attested uses of TNconstructions, including the ambivalent contexts, also constitute a semistable system, similar in certain respects to that argued for (S)SNconstructions. Section 6.3.9 will explore in greater detail how the coexisting TN-uses can be distinguished synchronically and will also explore functional overlap between (S)SN- and TN-constructions.

326Chapter 6: TN-constructions 6.3.9 Conclusion: Synchronic layering of TN-constructions and comparison with S(S)N-constructions 6.3.9.1 Synchronically coexisting TN-constructions With reference to close analysis of corpus data, it was argued that seven functionally different categories of TN-uses have to be distinguished, each with their own distinct constructional semantics. In support of these synchronically coexisting constructions, semantic and formal criteria were adduced, some of which had been pointed out in descriptive grammars and specialized studies of TNs. In addition, several ambivalent contexts were distinguished for examples that could not be unambiguously classified in these seven categories or that proved to be constructional blends. Every TN-use has been described as constituting a construction and semantic, structural, phonetic and pragmatic features have been discussed for each one (cf. Goldberg 2006). In Section 6.5 diachronic evidence will be provided for the fact that the various TN-uses are the result of processes of grammaticalization, subjectification and decategorialization. In another way than SN-constructions, the different synchronic layers of TN-uses are also shaped by collocational and colligational clusterings, primarily in the prenominal string. The distinct TN-uses with their constructional similarities and differences fit the idea of categories with prototype structure (see Goldberg 2006) and constructional gradience (see Aarts 2007). Structurally, the main fault line lies between the head use and all other uses. In the head noun construction the TN is head noun and the of-phrase acts as postmodifier to that head, whereas in the other six uses the TN and of form a TN-string that has modifier status.100 Partial tests for (non-)head status include concord patterns between TN/N2 and finite verbs or phoric determiners. Verbal concord is difficult to check systematically, since the NP with TN has to occur as subject and the TN and N2 have to differ in number to see which of the two determines concord. Moreover, verb concord can follow other principles than strict grammatical concord and corpus examples can always contain mistakes. Concord patterns between phoric determiners and TNs are similarly inconclusive in that for the postdeterminer category for instance concord mismatch was not criterial, but did make up a subset (cf. Denison 2002). Another test for non-head status of the TN-expression is substitutability by monomorphemic modifier expressions of the same function. In the quantifier use all sorts/kinds/types of are functionally equivalent to quantifiers like many/much or various. Predeterminer such (a) can replace some postdeterminer uses. In the qualifying uses the TN-string can be substituted by more established hedges or adverbials, such as rather. Phonetic non-

6.3 Classification of NP-internal TN-uses and extended uses

327

salience, sometimes reflected by hyphenation or coalesced spelling, also in part corroborates non-head status (but see note 92). Preposing of the of-phrase, a test suggested for instance in Traugott (2006) and the transformational literature, seems only a partial test for head status. In a head noun construction the of-phrase can be preposed, as in of stone it is a type; whereas in the attributive modifier construction or other modifier constructions, such as the qualifying construction, it typically cannot: *of person a scattered sort and *of spring cleaning a sort, or it would change the constructional meaning. Alternation between (determiner) + TN + of+ N2 and (determiner) + TN + of + a + N2 also seems to distinguish only partially between head and modifier status in binominal TN-constructions, in that head noun uses typically do not allow an article in front of N2 and modifier uses often do: *A type of a stone is not possible whereas A scattered sort of a person and It is sort of a miracle are acceptable. As discussed in Section 6.3.3, this was interpreted as a formal reflex of the fact that modifier uses refer to instances and not types. However, this test does not apply if N2 is uncount or plural count. Generalizing from this, we can say that the main constructionalsemantic divide between head noun and modifier uses is that the head noun use construes its meaning within type space, whereas the modifier and postdeterminer uses essentially refer to instances. The specific semantics of certain modifier uses to a large extent come about through interaction with premodifiers, e.g. qualitative adjectives in the attributive modifier use, classifying nonce expressions in the semi-suffix use, determiners in the postdeterminer use. Qualifying modifier uses are about assessment of the categorial status of instances and not about setting up types. In these qualifying uses several (inter)subjective pragmatic values to do with politeness and intensification can add to the general meaning. Collocationally, qualifying uses often co-occur with negative collocates, which they then (inter)subjectively modify. The specific semantics of all these modifier uses are reflected by the collocational and colligational clusterings described for each. Adverbial and sentential qualifying uses, discourse marker and quotative uses, then, function outside of the binominal NP and take on yet other meanings that do not refer to nominal instantial meanings anymore.

6.3.9.2 Constructional semantic overlap between SN- and TN-constructions I am now in a position to chart specific synchronic functional overlap between SNs and TNs at the micro-constructional level. In addition, more general parallels can be drawn at the meso-constructional level. Firstly, the quantifier use of TN-expressions is comparable to the quantifier semantics that SN-expressions develop in their grammaticalization

328Chapter 6: TN-constructions process, albeit by different means. The scalar implications in TNexpressions are less apparent than in SN-expressions and need the help of other, more manifestly quantitative elements of structure to reinforce them. In the quantifier TN-use this role is taken on by quantifier all and in some rare cases by no, which in the process shift from relative to absolute quantifier semantics. The size-related postdeterminer use exploits the scalarity of the N2 being quantified in determining it, e.g. that kind of energy (6.71). In the diachronic case study discussed in Section 6.5 it will become clear that a (good/great) sort of diachronically had a quantifier use very similar to the quantifier use of SN-expressions such as a lot of which disappeared in Late Modern English. This SN-use of a TN-expression is exclusive to singular sort and is typically premodified by good or great. TN-uses that involve attitudinally subjective values and the emergence of intersubjective values, such as the modifier uses and qualifying uses, are functionally equivalent to certain uses of the SSN-expression bit of (a).101 Both can express hedging semantics as well as pragmatic intensification. In example (6.170) the TN-expression can be substituted by bit of to convey roughly the same intensifying meaning, reformulated in (6.170’): (6.170) He might not have had a memorable hit in years, but fancies himself as some kind of an artist. (CW-Times) (6.170’) He might not have had a memorable hit in years, but fancies himself as a bit of an artist. (6.170’) is very similar to attested SSN-constructions with bit of a, such as (6.171): (6.171) She is, after all, the woman often dismissed as cranky or opportunist who for nearly two decades has talked ceaselessly about her conversion to vegetarianism. Frankly, she was a bit of a bore. (CW-OZnews) Examples (6.172) and (6.173), in which the qualifying use of the TN has an intensifying value, in appropriate contexts similarly seem to allow substitution by a bit of: (6.172) So they come in with these Band-Aids, and they say, “Look, we're reforming, aren't we a bunch of great guys?” And it really is kind of a joke. (CW-NPR)

6.3 Classification of NP-internal TN-uses and extended uses

329

(6.173) “It implicated the Premier in the matter,” Kovacs said. “It was sort of an unpleasant matter. She should not have been involved.” (CW-OZnews) Example (6.174) is an attested equivalent of (6.172): (6.174) I know that's a bit of a joke. (CW-UKspoken) Corpus data show that SSN- and TN-constructions typically convey these intensifying semantics in structures in which an indefinite article precedes N2. The SSN-expression itself is also typically preceded by a. The parallel between a bit of and TN-expressions with regard to intensifying semantics furthermore extends to adverbial uses of both expressions, as in (6.175) and (6.176), compared to (6.177) with a TN-expression, cited earlier as (6.131):102 (6.175) Still eager to race around stage like a decapitated chicken, percussionist enticed the black queen into some dirty dancing. In the end it all became a bit tiresome and predictable. (CW-OZnews) (6.176) It is a bit late for social services to talk about reassessing the case now. The time for action was long ago. (CW-Today) (6.177) “This show is a spectacular flop.” Unsurprisingly, only the first of James Thurber's two-sentence review “It had only one fault. It was kind of lousy.” (CW-Times) Again it is contextual clues that support the intensifying meaning, such as the derogatory ironic tone in (6.175) and (6.177) and long ago in (6.176). Bit continues to be preceded by a in this construction, but appears without of, whereas the TN-expressions are not preceded by a but retain of. In addition to intensifying semantics, both bit of a and TN-expressions can express hedging or approximator semantics in binominal constructions, as in (6.178) and (6.179): (6.178) I am a bit of a gypsy. For the last four years, I haven't lived in one place longer than three months." (CW-OZnews) (6.179) I hear he's sort of a mystic. I hear he's sort of crazy. I hear he's unpredictable. (CW-USbooks) Both meanings can also be expressed in adverbial constructions, as in We were a bit slow) and it is a bit, sort of boring (6.127), in which a bit is re-

330Chapter 6: TN-constructions phrased by sort of. A bit and sort/kind of can also both be used as independent answers to bipolar questions, where they have mitigating or hedging values, as in (5.25) Q: Did you like the movie? A: A bit and (6.141) A: Do our eyes deceive us or are we witnessing a return to the brash, rich, greedis-good, yuppie 1980s. B: Well, kind of. In both the intensifying and hedging binominal and adverbial constructions N2 or the other predicates are typically negative, but need not be so; however, positive predicates are mostly construed ironically. The structure what + (a) + SN/TN + of + N2 is equally shared as a construction for expressing emphatic meaning by both types of nouns. One can think here of valuing (quantifier) uses of bunch of and load of, as in (6.180) and (6.181), cited earlier as (4.220) and (4.170). These can be compared to example (6.182) with a TN-expression: (6.180) What a bunch of gobbledygook. (CW-OZnews) (6.181) What a load of rubbish. (CW-Today) (6.182) What kind of discipline is it that gives actors and dancers stage presence? (CW-Times) The emphasizing postdeterminer use of TN-expressions is mostly attested without a in front of N2 in my corpus data, but additional extractions on what + sort|kind + of also yield variants with a, as in (6.183) and (6.184).103 (6.183) “Shit. What kind of a question is that?” (USbooks) (6.184) What kind of a monster was this! (USbooks) Naturally, one could argue that it is the emphatically (interrogative) semantics of what combined with the equally emphatic rhetorical contexts and mostly negative N2s that coerce SN- and TN-expressions into this intensifying meaning. This is certainly partly the case, though examples of the valuing SN-use have shown that SN- and TN-expressions can express emphatically evaluative meaning outside of this specific construction too, as illustrated by (6.185) for instance: (6.185) The selectors, basically, are a load of crap. They couldn't pick a good team to save their bloody lives-both cricket and football. (CW-UKmags) The SSN-expression a bit of seems to act as a nexus within this set of specific functional overlap between SN- and TN-expressions. This may be

6.4 Register analysis of the functions of TN-constructions

331

explained by the fact that it combines size implications with implicatures of categorization in its ability to express exemplar meaning (cf. Traugott 2010b: 44), i.e. bit implies a rather abstract scale of small quantitative values, with further implicatures of small being inadequate or insufficient, but still exemplar. The exemplar meaning in its turn is conducive to various (inter)subjective values and subjectification. In addition to more specific instances of functional overlap, (S)SN- and TN-expressions also show at least two more general parallels at the mesoconstructional level. Firstly, they share the synchronic alternation between head and modifier status, which Section 6.5 will argue to be the result of syntactic reanalyses in processes of grammaticalization. Secondly, (S)SNand TN-expressions display parallels in their subjectification processes,104 which are characterized by an overall leftward movement within the NP towards the determiner zone (cf. Adamson 2000). All of these features can be accounted for in a dynamic model of the NP, as proposed in Chapter 2, which allows interaction and shifts between the functional zones of determination, quantification, modification and categorization through grammaticalization, delexicalization, subjectification and the development of (inter)subjective values. There hence seem to be deeply-rooted mesoconstructional similarities between the three sets of nominal constructions discussed in this study. Before I turn to the diachronic corpus study of TN-expressions, Section 6.4 presents a synchronic register study that compares TN-uses across the Times and COLT corpus.

6.4 Register analysis of the functions of TN-constructions In this section I will report on a register study, based on the two sets of British English data used to set up the descriptive framework in the previous sections. Their main features will be briefly reintroduced here. The first data set consists of exhaustive extractions on the TNs sort/kind/type + of from The Times, a subcorpus of the Collins Wordbanks corpus. The second data set comprises exhaustive extractions on sort/kind/type + of (and sorta/typa/kinda) from COLT. These two data sets (De Smedt 2005) represent the two ends of various register clines. The language in The Times is in the formal, written mode, and is, because of its normative function in British culture, not thought of as representing very progressive usage. By the same token, it is very professional and stylistically carefully wrought. The COLT corpus, by contrast, contains the highly informal, spontaneously spoken language of teenagers. According to Halliday (1978), Chafe (2003) and Du Bois (2003), casual conversation between peers constitutes the

332Chapter 6: TN-constructions most important locus of language change and innovation, without being subject to overly conscious forms of monitoring or engineering. Table 6.15 represents the relative percentages of the seven main functions of TNs in The Times and COLT data sets, as well as the mixed categories. Table 6.15. Corpus analysis of TN-functions in The Times and COLT The Times

Head use Quantifier use Modifier use Attributive Semi-suffix use Postdeterminer use Qualifying use Nominal qualifier Adverbial qualifier Sentential qualifier Discourse marker use Quotative use Head/postdeterminer Head/modifier Attributive/semi suffix Head/quantifier Postdeterminer/qualifying use Attributive modifier/nom. qualifier Qualifying use/discourse marker Unclear Total

COLT

#

%

#

%

348 60 69 42 27 813 336 302 30 4 5 1 45 8 2 23 2

20.28 3.20 4.02 2.45 1.57 47.38 19.58 17.60 1.75 0.23 0.29 0.06 2.62 0.47 0.12 1.34 0.12

4 1 10 4 6 13 74 9 60 5 32 6 5 0 0 0 0

2.38 0.6 5.95 2.38 3.57 7.74 44.05 5.36 35.71 2.98 19.05 3.57 2.98 0 0 0 0

3

0.17

0

0

0

0

16

9.52

1 1716

0.06 100

7 168

4.17 100

When we compare the quantified data from The Times and COLT, it is immediately clear that the relative frequencies peak for different clusters of functions in the two data sets. In The Times corpus the NP-internal TN-uses predominate, particularly the postdeterminer and head constructions, as well as the nominal qualifier use. In COLT, on the other hand, it is precisely uses that are not tied to NP structure that are most frequent, i.e. the adverbial qualifier and discourse marker use.

6.4 Register analysis of the functions of TN-constructions

333

The large number of postdeterminer, head and nominal qualifier uses of TNs in The Times and the prevalence of adverbial qualifiers and discourse markers in COLT can be explained by a variety of factors. Obviously, these differences correlate with the different stylistic properties and communicative aims of the two registers. The differences in the use of TNs also reflect the different tendencies of the two registers with regard to degree of innovativeness. Both these points will be discussed in more detail in the following paragraphs. The typical uses of TNs in The Times data are directly or indirectly concerned with the classification of entities. In head noun function, which accounts for 20.28% of all occurrences, the TNs straightforwardly describe subclasses as being related to superordinate classes. This form of taxonomization and subclassification is clearly rather important in newspaper texts. The nominal qualifying function of TNs is only slightly less frequent in The Times corpus (17.60%). The most common use made of this construction in The Times has the writer signalling to the reader not to take the expression in question literally, but to interpret it as an illustrative approximation or metaphoric reformulation of a discourse referent, as in (6.186) and (6.187), cited earlier as (6.119) and (1.49): (6.186) Full breathing is also an important tool for encouraging waste elimination a kind of spring-cleaning process that can go on all year around, every day of your life. (CW-Times) (6.187) The principle of earth sheltering is the same, whether applied to homes above or below the ground. An 18in covering of earth acts as a kind of blanket. (CW-Times) By far the most frequent function of TNs in The Times data, however, is the postdeterminer use, which constitutes half of all occurrences (47.38%). We have to keep in mind that the mixed category head/postdeterminer holds another 45 instances. As we saw in Section 6.3.4, NPs with postdeterminer uses of TNs realize anaphorically or cataphorically motivated reference to generalized concepts. These generalizations often express a property of the instance(s) ana- or cataphorically referred to, which is then applied to (an)other instance(s) as well. For instance, in (6.69) these sort of scare tactics refers back to ‘linking plastic milk bottles and cancer’ and at the same time refers to any other instances falling under that common denominator. In (6.79) the kind of peace Stalin and Hitler brought to Poland compares the peace installed by the Dayton agreement with the peace brought to Poland by Stalin and Hitler. These and other examples quoted above illustrate the peculiar discourse effects which these NPs have, such as expressing a

334Chapter 6: TN-constructions personal assessment of specific instances as flowing from generally accepted properties of other instances. The functional category of phorically motivated generalized reference masks, as it were, the fact that highly personal evaluations are being expressed and sanctioned. De Smedt (2005: 119) also suggests that this form of reference with its abstract, objectified aura serves to draw the reader into this evaluation, and create an evaluative consensus between speaker and hearer, as illustrated by (6.188) (6.188) And when Brosnan swings his gun around a corner, it comes to a rock-steady halt. It’s the sort of skill that comes in handy amid the swift, crisply defined action sequences of the film by far the best since The Spy Who Loved Me. (CW-Times) The high frequency of uses concerned with classification of entities no doubt generally correlates with the high incidence of heavy NPs in The Times corpus. As noted by Halliday (1985b), NPs tend to do a lot of the meaning-making work in written texts such as quality journalism. Moreover, the way in which these NP-internal TN-constructions are used plays a role in the relation which is set up between writer and readers in this kind of journalism. They help the writer establish an authoritative tone, in which s/he projects expertise and sure judgment. In the COLT data, by contrast, we are dealing with a register of language use in which there are fewer heavy NPs, and in which verb phrases do a lot more work: spontaneously spoken language tends to juxtapose and combine shorter clauses with simpler NPs. The largest sets of TN-uses are formed by adverbial qualifiers (35.71%) and discourse marker uses (19.05%) and the qualifying use/discourse marker category contains another 16 instances. The adverbial qualifiers mostly qualify adjectival and verbal phrases, that is, predications in utterances. Inspection of the data shows that in most cases the qualifiers play down the force of the utterance. The discourse markers are scattered throughout the discourse and convey tentativeness and solidarity. The high incidence of adverbial qualifiers and discourse markers in the teenage language recorded thus has the general effect of conveying non-dominance and of establishing solidarity. As for the innovative character of TN-uses, it is striking that the strong semantic shift and detachment from NP structure are very much associated with the COLT data, and barely represented in The Times. The relative frequencies of the new uses in COLT may to a certain extent reflect the progression of changes manifested by TN-constructions. Adverbial qualifiers, in which the TN has shifted away from the qualification of entitycategorizations, are by far the most common in the COLT data (35.71%). The further step in which the TN-string qualifies a whole proposition accounts for a much smaller portion (2.98%). The discourse marker use, hy-

6.5 Diachronic case study of sort(s), kind(s) and type(s)

335

pothesized by Denison (2002) to be a further development of the adverbial qualifier, is well represented too (19.05%). By contrast, the newer quotative uses appear in only 3.57% of the attestations. The only structurally innovative use attested with some frequency in The Times data is the semi-suffix use. It seems fairly uncontroversial to state that, as far as frequent and creative semi-suffix uses are concerned, Internet uses are playing a more important pioneering role than the examples found in The Times. In Section 6.5 the diachronic dimensions of the synchronic classification presented in Section 6.3, and the discourse features discussed in Section 6.4 will be studied in great detail. It will be argued that the synchronic variation in TN-uses is the result of complex and interlocking processes of grammaticalization and (inter)subjectification. 6.5 Diachronic case study of sort(s), kind(s) and type(s)105 6.5.1 Introduction Section 6.3 presented a classification of synchronic TN-patterns distinguishing seven main constructional uses. A starting point of this classification was Denison (2002) and (2005) who distinguishes three main NP constructions with TNs in present-day English, namely the head, postdeterminer and qualifier constructions. In addition, Denison attempted to reconstruct the chronology and the paths of change linking the main constructions. In his account, the postdeterminer and qualifier construction developed from the binominal construction in which lexically full sort/kind/type is the head followed by a second noun designating a superordinate class. In the chronology he posits the postdeterminer construction as an early reanalysis of the binominal construction (c1390 for all kind of and c1550 for kind and sort of), whereas the qualifying constructions developed later from it (c1580 for kind of and c1710 for sort of), via the mediation of the postdeterminer construction. I have distinguished two additional NP-constructions with TNs, namely quantifier, and descriptive modifier (i.e. attributive modifier and semisuffix), on the basis of syntactic, semantic and collocational features. In this section I will consider the diachronic import of these newly distinguished constructions and argue that they are key pivots in the developmental paths that have led from the head construction to constructions in which the TN is not the head. By thus refining Denison’s proposed chronology, I argue that new constructions emerge as the result of complex interlocking paths in which the quantifier and descriptive modifier constructions pre-dated, and helped facilitate and entrench, the postdeterminer and qualifying constructions. In the diachronic corpus study the focus will be on the NP-

336Chapter 6: TN-constructions internal TN-constructions, i.e. head, postdeterminer, quantifier, descriptive modifier and nominal qualifier. The discussion will also mainly be restricted to sort(s) and kind(s). The structure of this Section is as follows. In Section 6.5.2 the diachronic data sets and the methodology used in the analysis will be discussed. In Section 6.5.3 the TN-uses will be presented from a diachronic point of view. This section will show that the five nominal construction types with TNs that are part of the extended synchronic classification proposed in Section 6.3 are also attested in the diachronic data sets. Section 6.5.4 will reconstruct the diachronic paths of change between the various uses. Section 6.5.4.1 will first summarize Denison's chronology of TN-uses and Section 6.5.4.2 will then revise this proposal in view of the quantifier and attributive modifier construction, which up till now have not been included in chronologies. It will be argued that these constructions are important missing links in the complex paths of development of sort and kind that started around 1500 and led to five different nominal constructions in Late Modern English. Section 6.5.5 rounds off with a more general conclusion of the diachronic analysis.

6.5.2 Data sets and methodology The aim of this study is to reconstruct, on the basis of qualitative and quantitative corpus study, the chronology and development of the nominal constructions with sort(s) and kind(s) that came into existence by reanalysis within the binominal construction. As exhaustive extractions of sort(s) and kind(s) from the Penn-Helsinki Parsed Corpus of Middle English, Second Edition (PPCME) (1150– 1500) revealed, they hardly occurred in the syntagm NP + of + NP in that early period. However, from 1500 onwards sort(s) and kind(s) started to be followed commonly by of + NP. From that time period onward, therefore, extractions were made on the keywords sort(s) of/kind(s) of, as these queries netted all the uses I was concerned with and whose development relative to each other I wanted to chart. The successive diachronic slices were compiled as follows. Diachronic data were extracted from the Penn-Helsinki Parsed Corpus of Early Modern English (PPCEME) (1500–1640) and the extended version of the Corpus of Late Modern English Texts (CLMETEV) (1640–1920). The extractions from PPCME and PPCEME were analysed exhaustively. They consist of 31 tokens for kind(s) of and 23 for sort(s) of for the period 1500–70, and 41 tokens for kind(s) of and 44 for sort(s) of for the period of 1570–1640. The data sets from CLMETEV consist of 239 tokens for kind(s) of and 73 tokens for sort(s) of for the period 1640–1710. Data for the periods 1710–80, 1780–1850 and 1850–1920 consist of random samples of 250 instances for

6.5 Diachronic case study of sort(s), kind(s) and type(s)

337

sort(s) of and kind(s) of. For present-day English, all occurrences of sort of/sorta/ kind of/kinda in the Times subcorpus of the Collins Wordbanks corpus were extracted and analysed. All the quantitative findings discussed in this section result from the systematic classification and quantification of these samples in terms of the constructions introduced in Section 6.3. For illustrative purposes I also quote examples from the MED, the OED, and other subcorpora of the Collins Wordbanks corpus, indicating the source. I will approach the diachronic analysis not only in terms of first occurrences of new constructions, but will also consider the proportions of all constructions in each time slice. This study will, therefore, be based on the qualitative and quantitative analysis of successive diachronic samples of sort(s) and kind(s). Type, a third important member of the TN set in present-day English, will be left out of the discussion. As a later addition to that set it developed at a different rate than sort and kind; qualifying uses with type, for instance, are only just starting to be attested.

6.5.3 Nominal constructions with type nouns 6.5.3.1 Type nouns as head nouns of the binominal construction The binominal construction in which the TN functions as the head noun is the source construction of the reanalyzed uses that will be discussed. In it, we find the lexically full use of sort(s) and kind(s) meaning ‘(sub)kind’. As observed by Denison (2002: 7), sort and kind had more specific lexical senses earlier on. For instance, in Old English kind meant ‘birth, descent, nature‘. The NP + of + NP-construction appears in English from the Middle English period onwards. The binominal construction in which sort(s) and kind(s) feature their abstract subtype sense first occurs in my data in the period 1500–70. The TNs are in bold: (6.189) These two sortes of the chyldren of Israel (OED, Pilgr. Perf. 1526) (6.190) . . . of whiche course sorte of kerseyes . . . (PPCEME, Statutes 1540–70) (6.191) . . . the third sort of Leemynster wolle . . . (PPCEME, Statutes 1540–70) These NPs have generic reference: they refer to (whole) subclasses of the superordinate classes expressed by N2. The TNs in this construction may be singular or plural and can be premodified by quantifiers, indefinite and

338Chapter 6: TN-constructions definite determiners as well as other modifiers. This is because they are used in contexts building up generic and taxonomic interpretations of the world, often quantifying subtypes (6.189 and 6.191) and ascribing properties to them (6.190). Syntactically, this construction is parsed as [determination: whiche] + [adjective: course] + [head: sorte (postmodifier: of + NP: kerseyes)].

6.5.3.2 Type nouns as postdeterminers In the postdeterminer construction use, as described in Section 6.3.4, the TN together with the primary determiner forms a complex determiner that can express more intricate identifying and phoric relations than a simple determiner can. TNs functioning within a complex determiner no longer refer to a generic subclass, but help express more abstract, textual relations such as anaphora. This can be illustrated by comparing a diachronic postdeterminer use in (6.193) with a diachronic example of the head use preceded by an anaphorical determiner in (6.193): (6.192) "Our very pride, methinks, should be a sufficient guard, and turn whatever favourable thoughts we might have of such a one, unknowing his design, into aversion, when once convinced he presumed upon our weakness". In these kind of reasonings did she continue some time (CLMETEV, Haywood 1744) (6.193) [about peppermint] The Bed, where it grows, must . . . be pricked up . . . with a small Fork, or the Earth made fine with a Trowel; because the Runners of this sort of Mint shoot along upon the Surface of the Ground, (. . .), which is contrary to other Sorts of Mint (CLMETEV, Bradley 1732) In (6.193) this sort of Mint realizes generic reference to the whole subclass of peppermint. The determiner this simply points back into the text to identify the species 'peppermint' as the antecedent. The structural parsing of this example is [determiner: this] + [head: sort (postmodifier: of + mint)]. In (6.192), on the other hand, these kind of functions as the complex determiner of reasonings and as a whole points back into the discourse to the specific instances of reasoning quoted; it abstracts properties such as 'proud' from them, adding these to the determination of reasonings. The pointing relation expressed by these kind of thus involves a local generalization of instances into an ad hoc type, which aids identification of the specific instances referred to. The structural parsing of these kind of reasonings is [determiner: these kind of] [head: reasonings], with these sort of conveying

6.5 Diachronic case study of sort(s), kind(s) and type(s)

339

the meaning of such. Similar to what we have seen in Section 6.3, the expression of abstract textual relations such as anaphora diachronically is not restricted to plural NPs. This kind of discourse in (6.194) is an example analogous to (6.192) except that its head noun is non-plural discourse. This kind of points back to the questions asked by Faithful, and includes their property of 'unexpected seriousness' into the reference to the discourse in question. (6.194) [following a series of questions from Faithful] Then Talkative at first began to blush; but, recovering himself, thus he replied: "You come now to experience, to conscience, and God; and to appeal to him for justification of what is spoken. This kind of discourse I did not expect." (CLMETEV, Bunyan 1678) The first postdeterminer uses with sort(s) in my datasets date from c1553 and those with kind(s) from c1563. Fairly soon, by c1672 for sort and c1605 for kind, the postdeterminer constructions had established different subtypes relating to the type of phoricity expressed, namely anaphoric, cataphoric or non-phoric. Examples (6.192) and (6.194) illustrate anaphoric uses referring back to specific antecedents for which the head noun provides an appropriate categorization. The primary determiner in such complexes is typically a demonstrative. Cataphoric complex determiners are mostly introduced by the; they point forward to a postcedent described in the following relative clause, which forms the basis of the local generalization construed by the + TN of, as in (6.195): (6.195) It is impossible for a poet to read this without being filled with that sort of enthusiasm which is peculiar to the inspired tribe, and which Dryden largely felt when he composed it. (CLMETEV, Cibber 1753) Non-phoric uses form the smallest portion and are mostly introduced by some or what. In these complex determiners too the addition of the TN string invokes a generalising relation, even though no antecedent or postcedent is referred to, as in (6.196). (6.196) You see what a kind of Shuffling there has been to stifle the Truth. (PPCEME, 1685) Diachronically, complex determiners with TNs are rather commonly intensifying, as illustrated by (6.195)–(6.197). This is similar to what was observed for the synchronic data. If N2 is a gradable noun (Paradis 2000: 239243), it may be intensified qualitatively (Bolinger 1972: 58), as in (6.197),

340Chapter 6: TN-constructions in which implied qualities of heretics such as 'blasphemous' are heightened by suche a sorte of. (6.197) Suche a sorte of herytykes ho I ever sawe, that wyll nother reverence the croose of Chryste (PPCEME, Mowntayne c.1553) If N2 has size implications, it may be intensified quantitatively, as in (6.195), in which the amount of enthusiasm is enlarged by cataphoric that sort of. The restrictions to specific sets of primary determiners, dependent on the type of phoricity, can be interpreted as a sign of decategorialization. Complex determiners incorporating a TN-expression typically do not have any adjectival premodifiers other than postdeterminer adjectives. There still is the possibility of either a singular or plural form of the TN.

6.5.3.3 Type nouns as part of descriptive modifiers In the diachronic data sets too two subtypes of the descriptive modifier category are attested, the attributive modifier and the semi-suffix use. In both uses the TN is preceded by lexical material: qualitative adjectives in the attributive modifier use (6.198)–(6.200), and classifying adjectives or nouns, or longer expressions in the semi-suffix use (6.201)–(6.203). The lexical material preceding the TN-expression is underlined. (6.198) This is but a scandalous sort of an Office. (PPCEME, Farquhar 1707) (6.199) He let his mind have its full flight, and shewed by a generous kind of negligence, that he did not write for praise. (CLMETEV, Cibber 1753) (6.200) He’s a good honest kind of a Fellow, and one of us. (CLMETEV, Gay 1728) (6.201) "A Good Knock-about kind of a Wife". (CLMETEV, Baker 1845–53) (6.202) David knew nothing about this mincing, half-and-half, milk-andwater Sort of religion. (CLMETEV, Booth 1880)

6.5 Diachronic case study of sort(s), kind(s) and type(s)

341

(6.203) But that, of course, for himself—well, he preferred, as a general rule, the Pension Frensham sort of Thing. (CLMETEV, Bennett 1908) In these diachronic examples the qualitative adjective in attributive modifier uses does not apply to sort or kind but to N2 and the NPs in (6.198)– (6.200) do not refer to subtypes of Office, negligence or Fellow, but to instances of them. As with the synchronic attributive modifier use, an indefinite article may appear in front of N2, as in (6.198) and (6.200). Examples of the attributive modifier use appear in my datasets from 1500 onwards, both for sort and kind of. A diachronically stable feature of this use is that the prenominal adjectives are carefully chosen and often rather infrequent adjectives. I also attested similar motifs in the diachronic data, such as the attribution of qualities to persons as in (6.200). The reanalysis as attributive modifier is for a large part enabled by interaction between semantic bleaching of the TN-expression and the position of attributive adjectives in the prenominal string in the English NP. A comparison between English and French TN-expressions by Willemse, Davidse and Brems (2007) confirmed the influence of such language internal factors on the frequency of occurrence of certain TN-uses. In the French data hardly any attributive modifier uses were attested, because in French qualitative adjectives typically follow the head. In the semi-suffix uses, the element preceding the TN is of a more classifying nature and can be of varying length, e.g. classifying adjectives, proper names (6.203), fixed or nonce expressions (6.201) and (6.202). The semi-suffix use, whose earliest attestations in my datasets are c1850, pushes certain features of the attributive modifier use further. The often considerable length of the lexical modifier stimulates further demotion and phonetic attrition of the TN, which functions more or less as a clitic or suffix to the preceding lexical material. This is why Denison (2002: 4) proposed the term semi-suffix use for this use. It has a metalinguistic value, signalling that the lexical items preceding it have to be interpreted as an ad hoc, often very creative, classifier. In addition, it also often has hedging meaning, indicating that the description is only approximative. Again the restrictions to a specific set of premodifiers and the typical association with singular forms of the TN can be interpreted as a reflex of decategorialization.

6.5.3.4 Type nouns as part of quantifiers Diachronically, the most common subtype of quantifier use, which is still used in present-day English, is the pattern in which plural kinds and sorts

342Chapter 6: TN-constructions combine with all. This quantifier use is an early reanalysis of the binominal construction, appearing c1380 for kind and c1550 for sort. The argumentation for this reanalysis hinges on the shift from the universal quantifier sense of all in the binominal construction to the many sense of the quantifier construction. A diachronic example such as (6.204) illustrates how such a reanalysis may have come about. (6.204) I Answered, That Religion being a design to recover and save Mankind, was to be so opened as to awaken and work upon all sorts of people, and generally men of a simplicity of Mind, were those that were the fittest Objects. (PPCEME, Burnet 1680) In (6.204) all can be read as quantifying over all subsets of people. The example contains a number of contextual clues supporting this reading, as it stresses that religion should be so opened as to save (all) Mankind. The explicit description of one subset, men of a simplicity of Mind, supports the generic subclass-reading of sorts. This example also illustrates how a NP with totality quantifier all often invites an inference of a great quantity in an absolute sense: work upon all sorts of people implies 'on very many people'. In examples (6.205)–(6.207), then, reference is no longer to the exhaustive subtypes of exception, passion or motes, but to very many instances or a large amount of them. In other words, what was an invited implicature of 'large quantity' in the head use (6.204) has semanticized (Traugott 1989) and has become the conventionalized meaning of the uses in (6.205)–(6.207). It can be noted that particularly in the earlier stages the TN could also be singular as in (6.205) and (6.206). The diachronic reanalysis and pragmatic enrichment is hence nicely reflected by the internally graded nature of the synchronic quantifier use discussed in Section 6.3.2. (6.205) The present condition of humane nature doth not admit of any constitution of things, whether in religion or civil matters which is free from all kind of exception or inconvenience. (PPCEME, Tillotson c1679) (6.206) … to extirpate all sort of passion or concern for anything. (PPCEME, Burnet 1680) (6.207) Then wash the curd till it be as white and cleane from all sorts of motes as is possible. (PPCEME, Markham 1615) The clearest instances of quantifier uses diachronically as well as synchronically occur in negative contexts, in the sense that the absence of a great

6.5 Diachronic case study of sort(s), kind(s) and type(s)

343

quantity is referred to: free from (6.205), extirpate (6.206) and cleane from (6.207). However, the boundaries between a head noun use in which the TN is modified by a quantifier, and the quantifier use are both synchronically and diachronically rather fuzzy and, as we have seen in Section 6.3.2, the quantifier use itself remains internally graded. In examples such as (6.208), for instance, both the categorization and the quantity meaning are present. All sorts of pepull refers to people of different ranks and backgrounds as well as to many people as such. (6.208) And at ys bereing was the gretest mone mad for hym of ys deth as ever was hard or sene, boyth of all sorts of pepull. (PPCEME, Machyn 1553–9) The diachronic data contain many such ambivalent examples, e.g. all sorts of fruits, all kinde of pastimes, whose contexts activate both the variety and large quantity meaning. In present-day English too, such examples still occur. In addition to quantifying uses with all, the diachronic data also contain quantifiers with negative totality quantifiers no or (not) any, as in (6.209)– (6.211). (6.209) He had no sort of virtue. (PPCEME, Burnet 1683) (6.210) ... and that there was much wisdom, and no kind of shame, in making use of bad people on certain occasions. (CLMETEV, Fielding 1751) (6.211) That was saied without any kind of compulsion. (PPCEME, Edward 1550–2) As with the rare synchronic example with no kind of, reference is not to the zero-attestation of certain subtypes of 'virtue', 'shame' or 'compulsion', but to the absence of any degree or amount of virtue, shame or compulsion as such. No sort/kind of was common as a quantifier in Modern English—and as such got a separate entry in the (1933) OED—but has virtually disappeared now. In addition to the lexicalizations all/no sort(s)/kind(s) of, another quantifying use found only with sort, namely a (good/great, etc.) sort of, was very common in the period 1500–1640, and is already attested in the Middle English data, making this in fact the earliest reanalysed use of sort. It disappeared in Late Modern English:

344Chapter 6: TN-constructions (6.212) We shall cum everychon Mo then a good sorte. (PPCEME, Mankind c1475) (6.213) ... he cute downe a greate sorte of brakes [i.e. fern]. (PPCEME, Mowntayne c1553) As suggested by the gloss provided by the OED, i.e. "a (great, good, etc.) number or lot of persons or things; a considerable body or quantity; a multitude", this expression features a quantifying use of sort, analogous to the quantifier use of SN-expressions such as a (whole) lot of. Less commonly a sort of was also used without intensifier, as in: (6.214) If the world shal turn, A sort of you shal burn. (OED, Strype 1548) The fact that a (good/great) sort of was used with a variety of animate and inanimate collocates shows that it had also delexicalized to a great extent. We thus find at a very early stage, preceding all other reanalyses, two different quantifying uses resulting both via reanalysis from head uses of TNs. Strikingly, with both, their meaning of large absolute quantity is mostly intensified. The restriction to quantifier all, and more rarely no, and the lack of premodifiers of the TN-expressions can be seen as reflexes of decategorialization.

6.5.3.5 Type nouns as nominal qualifiers The last reanalysis of the binominal construction that stayed within the boundaries of the NP106 is the nominal qualifying construction, which appeared in the period 1570–1640 for kind and in 1640–1710 for sort. It is commonly accepted that this reanalysis was enabled by bleaching of the lexical subtype meaning into the pragmatic sense of peripheral membership (see Bolinger 1972, Aijmer 2002, Denison 2002 and Margerie 2010). In this use the nominal description offered by N2 is qualified, that is hedged, toned down or otherwise nuanced, for example: (6.215) we chafed very well one of the blunt edges of it upon a kind of large Pin-cushion cover’d with a course and black woollen stuff. (PPCEME, Boyle 1675–6) (6.216) The Trees we saw were a small kind of Cabbage Palms. (CLMETEV, Cook 1768–71)

6.5 Diachronic case study of sort(s), kind(s) and type(s)

345

(6.217) . . . which is the Palace of the King’s Women, a sort of seraglio (PPCEME, Behn 1688) Nominal qualifiers are speaker-related, that is subjective, when they signal that the speaker is unsure, humorous or ironic about the categorization of some referent (e.g. 6.215). Qualifier uses can also be intersubjective in that their approximator semantics may show the speaker’s concern for the face of the hearer or his or her effort to explain a less familiar item by classifying it as a non-prototypical member of some category (e.g. 6.216 and 6.217). The latter subtype is rather prominent diachronically, for instance in travel reports such as the one by British explorer Captain James Cook. The emergence of nominal qualifying uses of TN-expressions also involved decategorialization in that the use became restricted to the singular variant of the TN. This restriction also applies to the further development of extended NP-external TN-uses such as adverbial and sentential qualifiers, discourse markers and quotatives. In addition, the TN is typically only preceded by the determiners a or some and displays more restrictions with regard to other premodifiers than head noun uses of TNs.

6.5.4 Diachronic paths of change and chronology Section 6.5.4.1 will survey Denison’s chronology proposals for the TNuses he distinguishes within the NP, namely head, postdeterminer and qualifier. Section 6.5.4.2 then examines the validity of this chronology in view of the additional intra-nominal TN-uses discussed in Section 6.5.3 and proposes alternative pathways of change. It will assess in detail the specific roles that the various quantifier uses and descriptive modifier uses have played in the establishment of the constructional network of TNconstructions

6.5.4.1 Denison’s chronology of head, postdeterminer and nominal qualifier uses Denison (2002/2005) posits a twofold path of change, allowing for multiple inheritance links between the various TN-constructions. One path leads from what he calls the binominal, or head noun use, to the postdeterminer construction and the other from the binominal to the nominal qualifying construction via the mediation of the postdeterminer construction. This is visualized in Figure 6.1.

346Chapter 6: TN-constructions

Figure 6.1. Denison’s chronology

Figure 6.1 suggests that the postdeterminer use, restricted to patterns with incongruous concord as in these kind of reasonings (see Section 6.3.4), is an early reanalysis of the binominal construction and "is at least as old as the qualifying use, probably much older" (Denison 2002: 9). The reanalysis would then have started in head use patterns with a singular TN and plural N2 in which the TN was not premodified except for a determiner not marked for number, as in (6.218): (6.218) as giving him some kind of hopes that . . . (1627; quoted in Denison 2002: 11) Qualifying uses such as a kind of large Pin-cushion (6.215) arise in contexts where the TN has low information status and instead the NP following of receives focus. In such contexts a conversational implicature could arise from "a normal member of the class of N2" to "a possible member" (Denison 2002: 11). The semanticization of this inference affected the structural status of the TN and gave rise to the qualifying construction as a new distinct use. As the postdeterminer use of TNs is very early, preceding the qualifying uses, Denison (2002: 11) "provisionally assume[s]" that it also contributed to the formation of the qualifying use, but he does not expand on this.

6.5.4.2 Chronology and pathways of change revisited In this section I assess how the quantifier and descriptive modifier construction fit into the semantico-syntactic pathways proposed by Denison. I argue that particularly these two non-considered constructions provide crucial missing links in the interlocking paths of development starting around 1500: they have played key roles in the emergence and entrenchment of the other reanalysed uses. The role of the descriptive modifier construction will, for chronological reasons, be restricted mainly to the earlier attributive use, even though some reference will also be made to the semi-suffix use. Table 6.16 charts the relative frequencies of the five nominal constructions for sort(s) of (S) and kind(s) of (K) from 1150 to the present: head

347

6.5 Diachronic case study of sort(s), kind(s) and type(s)

(H)107, attributive modifier (AM), quantifier (Q), postdeterminer (PDM) and nominal qualifier (NQ). It thus represents the relative quantitative importance and development (rise or decline) of these constructions vis-à-vis each other over time. Table 6.16. Relative Frequencies of sort(s) of (S)/kind(s) of (K) from 1150-PDE 11501500

15001570

15701640

16401710

17101780

17801850

18501920

1993-

S

S

S

S

K

S

K

S

S

H

75 99.7 44 74 62 44 56 55 42 24

6

14 15 24

8 20

AM

0

0

Q

25

0.3

PDM

0

NQ

0

S

K

4

K

3

K

K

K

K

2

5

3

5

7

5

8

7

10

7

5

3

48 13 34

1

22

8

9

14

3

8

3

8

7

1

0

4

10

2

12

7

19 24 19 32 18 24 17 52 51

0

0

0

0

5

12 13 19 34 46 49 44 37 21 21

Figure 6.2 visualizes the relative frequencies of quantifier uses with sort(s) of/kind(s) of from 1500 to present-day English.108

Figure 6.2. Relative Frequencies of Quantifier TN-uses from 1500 to PDE

Overall, quantifier uses of TNs have been present in a fairly stable proportion from 1500 onwards, hovering around a relative proportion of 10% of all the uses with sort/kind. The peak of quantifier uses (34%) manifested by sort in 1500–1640 is due to the then common SN-expressions a (good/great) sort of. As noted in Section 6.5.3.4, the quantifier uses were the earliest reanalyses of TNs within the binominal construction to emerge

348Chapter 6: TN-constructions at the end of the fourteenth century for kind and at the end of the fifteenth century for sort. These grammaticalized quantifier uses can be traced back to source head uses in Middle English in which the TN was preceded by quantifiers, for example: (6.219) Þe moder in þe femelle fongeþ semen, and þis is general in all kynde of bestes in þe whiche is male and female. (MED, Trevisa c.1398) In the diachronic data, head noun uses of TNs are preceded by quantifiers other than all and no, for example There be divers sorts of loves (PPCEME, Deloney 1619), several sortes of witches (PPCEME, Gifford 1593), and so forth. However, it is only the totality quantifiers no and especially all that triggered an ambiguity between generic taxonomic meaning and quantifying meaning. Non-totality quantifiers such as many, several and diverse did not invite pragmatic inferences leading to a quantifying reinterpretation. The quantifying expression a sort of, which is typically intensified by scalar adjectives good/great also resulted from reanalysis of head uses, in which sort was used in the sense of 'band, company', for example Paris now with his unmanly sorte (OED, Surrey 1547). The emphatic and often negative discourse contexts in which these uses emerge suggest that the early semanticization of scalar implicatures in them was motivated by the language users’ needs to use more creative and emphatic quantifying expressions (cf. Haspelmath 1998 and Hopper & Traugott 2003). Intensifying and emphatic expressions are well-known to engage in renewal processes of grammaticalization (Lorenz 2002). Interestingly, the emphatic quantifier uses with kind/sort seem to have enabled the entrenchment of the more neutral, purely identifying, postdeterminer uses. In Early Modern English postdeterminer uses were much less frequent than quantifier uses, but they overtook the latter in Late Modern English. Diachronic corpus analysis of sort shows clearly that the quantifying uses and, interestingly, also the intensifying postdeterminer uses pre-dated the general system of identifying and phoric postdeterminers. In the periods 1500–70 and 1570–1640, the various quantifier uses (all/no sort(s) of, a (good/great) sort of) account for a half and a third of the datasets respectively, while just one postdeterminer use is attested in each of these periods. Moreover, the very first postdeterminer use with sort is intensifying, namely (6.220), cited earlier. In fact, the emergence of cataphoric (6.220) and anaphoric (6.221) postdeterminer uses is associated with intensifying and evaluative contexts, such as illustrated by (6.220) and (6.221).

6.5 Diachronic case study of sort(s), kind(s) and type(s)

349

(6.220) Suche a sorte of herytykes ho I ever sawe, that wyll nother reverence the croose of Chryste (PPCEME, Mowntayne c1553) (6.221) The Rhadars here had apprehended a sturdy Thief, who had set upon a poor Merchant, and drove away his Ass loaded with Indian Cloth. These sort of Vermin are rare . . . (PPCEME, Fryer 1672– 81) The grafting of the new postdeterminer meanings on the already established intensifying-quantifying uses seems to have helped the postdeterminer use to gain a foothold as a separate constructional use. It can also be observed that intensifying postdeterminer uses of TNs — particularly ones intensifying quantity — have continued to account for a sizeable proportion of the whole construction — at least one third in present-day English. The intensifying-quantifying contexts that facilitated the emergence of the general postdeterminer system continue to form a substrate of it. As pointed out by Bolinger (1972: 60) determiners — in the broad sense — can often convey both identifying and quantifying-intensifying meanings, as in Such a person always frightens me versus Such a blunderer always frightens me. However, the path from intensifying to non-intensifying found here is still a surprising one, as subjectification clines (Bolinger 1972, Traugott 1989) generally predict a unidirectional change from non-intensifying to intensifying (but see Ghesquière 2010). Having paved the way for the general system of postdeterminer uses that came to form the majority, the quantifying uses receded to forming only a minor but stable proportion. No sort(s)/kind(s) of and a (good/great) sort of died off, and only all sorts/kinds of has persisted into present-day English with its mix of quantity and diversity meaning. In other words, what might have been interpreted from a purely synchronic perspective as a fairly recent new use, still showing signs of transition and not having become very productive yet, is really the oldest reanalysis of the binominal construction. Postdeterminer uses of TNs with determiner some have, in turn, probably facilitated the emergence of nominal qualifier uses with some via a transfer from quantifying meaning to qualitative approximation, illustrated by (6.222) and (6.223) respectively. (6.222) Taking shelter in a miserable cottage, we remained shivering and shaking till the carriage was in some sort of order (CLMETEV, Beckford 1783) (6.223) Before men are put forward into the great trusts of the State, they ought by their conduct to have obtained such a degree of estima-

350Chapter 6: TN-constructions tion in their country as may be some sort of pledge and security to the public that they will not abuse those trusts. (CLMETEV, Burke 1770) In (6.222) some asserts, against negative contextual expectations (cf. McGregor 1997: 281-282), that at least a certain degree of order was established in the carriage. In (6.223) by contrast, some, as Margerie (2010: 336) puts it, "reinforces the idea of approximation or uncertainty, indicating that the referent of the head noun is qualitatively vague". Margerie also observes that determiner some forms a prominent collocation with qualifying kind in present-day English. In my historical data, some started collocating with both kind and sort in the period 1640–1710, the same period in which some started to be used with some frequency in the head noun and postdeterminer constructions as well. Not surprisingly, one also finds examples that blend properties of the postdeterminer and qualifying constructions, and that illustrate specific pragmatic mechanisms involved in the shift from quantification to qualification, for example (6.224). (6.224) But a system unfavourable to freedom may be so formed as considerably to exalt the grandeur of the State, and men may find in the pride and splendour of that prosperity some sort of consolation for the loss of their solid privileges. (CLMETEV, Burke 1770) Some in (6.224) asserts a minimal quantity of consolation and also hedges consolation qualitatively as perhaps not the most appropriate notion in the circumstances. Both meanings pragmatically involve an element of surprise, as both quantitatively and qualitatively a positive assertion is made that is incongruous with the negative context of loss. The observable interaction between postdeterminer and qualifying uses with some puts flesh on Denison’s (2002: 11) assumption that the former must have played some mediating role in the derivation of the qualifying from the head noun construction. In this context it can be noted that another subtype of qualifying uses seems to derive directly from the head noun construction via the description of a peripheral or unknown hyponymic relation. As noted earlier, this use appears for instance in travelogues where explorers seek to describe exotic fauna, flora or rituals in terms of categories or concepts that people in the old world are familiar with. By implication these descriptions are approximations only, as in examples (6.216) above and (6.225) below from Captain James Cook’s Journal:

6.5 Diachronic case study of sort(s), kind(s) and type(s)

351

(6.225) These she caused to be carried from her Canoes up to the Fort in a kind of Procession. (CLMETEV, Cook 1768–71) In general, a historical shift can be observed within the nominal qualifier uses, which first predominantly involve approximate categorization but come to be used increasingly with other pragmatic values such as irony, hedging and the metalinguistic marking of stylistic changes, and so forth. Thus, in the period in which the qualifying construction emerges, its approximate categorization use predominates, namely in 12 out of 13 cases with kind and in 16 out of 18 cases with sort. This proportion had already gone down to 14 out of 47 qualifying uses with sort in 1710–80 and to 12 out of 37 with kind in 1850–1920. The attributive modifier use can be traced back to a Middle English head use of the TN, namely one in which a singular TN is modified by a qualitative adjective: (6.226) . . . þe gentyl kende (PPCME, Arundel 57 1340) In the change from premodified TNs used as heads to attributive modifier uses, the reference shifts from generic to instantial. As noted earlier, this can be shown by the insertion of the indefinite article before N2, as in a good honest kind of a Fellow (6.200). In head uses such as (6.226) the adjective pertains to the description of (sub)types. By contrast, in attributive modifier uses the adjective together with the TN string applies to concrete instances, as in (6.227) and (6.228): (6.227) The sensation produced by the properties of this shrub is a wild, dreamy kind of happiness. (CLMETEV, Baker 1845–53) (6.228) He consulted his note-book in a dazed, flustered kind of way. (CLMETEV Blackwood 1915) Chronologically, the attributive modifier use is also an early reanalysis, appearing at the beginning of the sixteenth century: (6.229) all the worlde woulde saye that wee wer a wyse sorte of counsaylers (PPCEME, More c1513) These uses often function as complements of copular verbs, classifying the instances referred to by the subject, for example (6.198) and (6.200) and (6.227). Figure 6.3 represents the relative frequencies of attributive uses from 1500 to present-day English.

352Chapter 6: TN-constructions

Figure 6.3. Relative frequencies of attributive modifier uses from 1500 to presentday English

Quantitatively, the attributive modifier use has not increased greatly in frequency, even though it manifested a minor peak towards the end of Late Modern English. In the early uses a great variety of N2s is found: for example counsaylers (6.229), falsehood, fever, meat, compliment, satisfaction, negligence (6.199), and so forth. A discourse context associated with this use from 1500 onwards right until the present-day is the attribution of a character trait to a person, for example a surly kind of fellow, wicked kind of people, a goody-goody sort of young woman, paternalistic sort of chap and a good honest kind of a Fellow (6.200). From the end of Late Modern English onwards the attributive modifier use is increasingly restricted to a few discourse topoi such as character trait attribution and uses with the semi-lexicalized string kind/sort of way, which is attested from 1850 onwards, for example a dazed, flustered kind of way (6.228). The attributive modifier use with its syntagmatic structure determiner + adjective + TN + of + N2 (e.g. a generous kind of negligence) also seems to have played a facilitating role in the emergence of qualifying uses with the structure determiner + TN + of + adjective + N2 (e.g. a sort of irregular vows). Qualifiers with adjectives are an important subtype of the nominal qualifier construction, which accounts for varying portions of more than half to about a third of the different diachronic slices. Structurally, the only change involved is transfer of the adjective from pre-TN to post-TN position. Quite a few examples were found in which it was possible to put the adjective in the alternative position without it causing great semantic or pragmatic changes. Example (6.230) is a clear illustration as it contains a qualifying NP with one adjective that precedes and one that follows the TN sort. (6.230) Being asked many years after if he had any remembrance of the Queen, he had a confused but somehow a sort of solemn recollection of a lady in diamonds and a long black hood. (CLMETEV, Carey 1792)

6.5 Diachronic case study of sort(s), kind(s) and type(s)

353

Moreover, there are also structurally blended examples, such as (6.231) a finical sort of a tawdry Fellow (PPCEME, Vanbrugh 1696) The possibilities of word order variation offered by the syntagm noun + of + noun invite a sort of verbal play that is part of language change, and that can be observed in the transfer of the adjective in an example such as a strong cup of tea from a cup of strong tea (Halliday 1994: 195). Chronologically, the attributive modifier use was the first to detach qualitative adjectives from reference to subtypes. In the qualifying use that emerged later, qualitative adjectives were also put to use for the characterization of instances. Syntactically, both uses have a strong disposition towards occurring as predicative complements and are almost exclusively introduced by indefinite determiners. Apart from all these grammatical similarities, there is also overlap between some of their pragmatic effects. As noted earlier, descriptive modifier uses veer towards metalinguistic, hedging and approximative meanings, particularly with their more elaborate semi-suffix uses. These are precisely core pragmatic values associated with nominal qualifying constructions. In view of all these similarities, it seems justified to propose that the descriptive modifier construction facilitated the emergence of the qualifier construction, and continued to impact particularly on the subtype containing adjectives.

6.5.5 Summary I have shown that Denison’s chronology of TN-constructions overlooks two early reanalyses of the head noun construction within binominal syntagms, namely the quantifier and descriptive modifier constructions. The combination of TNs and totality quantifiers all and no triggered the potential of scalar meaning in TNs, as did adjectives with scalar meaning like great and good. Both are cases of deictic and attitudinal subjectification, as they involve the expression of speaker-stance in the form of assessment of size as well as allow for the expression of hyperbolic attitude. In the development of attributive modifier uses the combination of TNs and adjectives such as scandalous, dazed and dreamy elicited inferences of evaluation from an original categorization meaning. These inferences draw on the type/token implication of TNs, which involves a meronymic shift. The semanticization of evaluative attributive meaning in the attributive modifier construction is a case of grammaticalization and attitudinal subjectification, as it involves an increase of the expression of the speaker's subjective attitude with regard to a referent.

354Chapter 6: TN-constructions Diachronic corpus analysis furthermore showed that they are also important links in the paths of development of other TN-constructions. If we add them to Denison’s twofold pathways from head to postdeterminer and from head to nominal qualifier, we end up with more complex paths of development, which can be schematized as in Figure 6.4. R indicates relationships involving reanalysis and E indicates entrenchment relationships.

Figure 6.4. Main paths of development and multiple inheritance links in the type noun constructional network

The emphatic quantifier uses are early reanalyses (attested for kind from c1380 onwards) of head noun uses on the basis of textual and attitudinal subjectification (a). The purely textual postdeterminer uses appear from 1550 onwards as a reanalysis of the head noun use (b) involving textual subjectification. In Early Modern English the older quantifier construction is more frequent than the new postdeterminer construction, but by the Late Modern English period the latter has overtaken the former. I have proposed that the intensifying-quantifying uses, with their expressive functionality, helped facilitate and entrench the postdeterminer use. The descriptive modifier use is an early reanalysis of the premodified head noun use, emerging around 1500 (c). The nominal qualifying construction is also a reanalysis of the binominal head noun construction, but arises later and is attested from c1600 for kind of and c1700 for sort of (g). It emerges in contexts that describe a peripheral or unknown hyponymic relation and this leads to a different syntactic parsing in which the TN + of are reanalyzed as a qualifier. The qualifying construction is also facilitated by earlier reanalyses of the head noun use, namely the postdeterminer and

6.6 Conclusion

355

descriptive modifier constructions. Postdeterminer uses probably facilitated the emergence of qualifier uses via a transfer from identifying-quantifying meaning to the approximation of qualitative meaning. The combination of some + kind/sort + of +N2 is a concrete bridging context in this respect. The descriptive modifier construction influenced the nominal qualifier both constructionally and in terms of shared subjectified meaning. Both constructions are associated with characterizing individuals and have approximative semantics. The difference is that attributive modifiers are primarily attitudinally subjective in nature in that a speaker selects precise and often uncommon adjectives to characterize the instance in question. However, the semi-suffix use does involve some intersubjective, namely metalinguistic and hedging, meaning. The qualifying use arguably veers more strongly towards intersubjective meaning in its attempt to characterize an instance which is unfamiliar to the hearer as a peripheral member of a better-known category, or as a face-saving device (cf. Brown & Levinson 1987). With regard to the NP-external TN-uses, it can be noted that the discourse marker use involves the development of attitudinally and textually subjective values. It may function as a text-structuring device which can serve discourse-strategic purposes, as well as acquire more attitudinal values. In addition, it may display intersubjective hedging meaning. The quotative use seems to involve textual subjectification as it displays a semantic shift from indicating peripheral membership within a category to marking the upcoming expression or stretch of discourse as different in level from the surrounding discourse. It has thus become clear that non-head uses of TNs have emerged as the result of complex and interlocking paths in which the quantifier and descriptive modifier constructions play pivotal roles. These paths and developments are remarkably parallel for kind of and sort of, but with the former generally leading the way.

6.6 Conclusion In this chapter I have presented a functional classification of TN-uses, i.e. seven distinct constructions defined in terms of the main structural and semantic differences manifested in synchronic and diachronic data. In comparison with the classifications proposed in the existing literature, I have advocated further distinctions for the NP-internal uses, distinguishing modifier and postdeterminer uses from head uses. In accordance with the literature on the subjectified uses of TNs, nominal, adverbial and sentential qualifiers were distinguished from discourse and quotative markers. Finally, a number of ambivalent contexts were described which include am-

356Chapter 6: TN-constructions biguous examples and constructional blends, both of which can point up transitional contexts. This descriptive framework proposed in Section 6.3 was applied to two British English data sets from opposing registers: written texts from the quality newspaper The Times (Collins Wordbanks corpus) and spontaneously spoken conversation between teenagers (COLT). Quantification of these analyses revealed strong asymmetries in the relative frequencies of the various TN-uses in the two data sets. TN-constructions in The Times were predominantly NP-internal and were concerned with the classification of entities in different forms. These uses, it was suggested, contribute to the authoritative voice of the writer. In the COLT data, adverbial qualifiers and discourse markers predominated; these convey tentativeness and solidarity. The figures thus overwhelmingly confirm the hypothesis that spontaneous conversation between peers is the prime locus of language change and innovation (cf. Halliday 1978, Chafe 2003, Du Bois 2003). In a second step the synchronically coexisting TN-uses were argued to be layered uses resulting from processes of grammaticalization and subjectification. Each layer was then argued to link up with distinct collocational patterns synchronically as well as diachronically. The diachronic corpus study also proposed a network of TN-constructions based on the way the various TN-uses impacted upon one another diachronically, enabling the entrenchment of constructional uses or their reanalysis from other uses. TNs were furthermore noted to share several general properties with SN-constructions at the meso-constructional level on the basis of the processes of change they went through. The quantifying and valuing SN-uses, just like the quantifier, postdeterminer and qualifying TN-uses, all display textual and more attitudinal subjectification and a general leftward movement in the NP towards the determiner zone. In addition to these core similarities, TN-constructions also display several differences compared to SN-constructions. In (S)SN-constructions collocational shifts and semantic prosody mark head versus modifier uses primarily in the postnominal string, i.e. they pertain to collocations between the (S)SN and N2 in the of-phrase. In TN-constructions, however, the different layers of TN-uses typically relate to distinct collocational patterns in the prenominal string, i.e. the modifiers and determining elements preceding the TN. Mostly only general tendencies can be noted with regard to the N2-collocates. Semantic prosody is involved in several qualifying or hedging uses. Delexicalization is less saliently visible in the developmental paths displayed by TN-constructions because the source semantics of each of the three TN-nouns are already more general than those of the SNexpressions discussed in Chapters 4 and 5.

Chapter 7 Descriptive-theoretical consequences and prospects for further research In this study I have argued for fine-grained classifications of English size noun (SN) and type noun (TN) expressions in terms of synchronically layered constructions, which revolve around head versus modifier status. It was argued that the current ability of SNs to function as head, quantifier and valuing quantifier and of TNs to perform functions such as head, postdeterminer, quantifier, attributive modifier, semi-suffix, hedge and quotative is due to diachronic processes of grammaticalization and (inter)subjectification involving a basic reanalysis from head status to various modifier functions. It was also argued that SN- and TN-constructions show zones of functional overlap as well as share basic processes of textual or deictic and attitudinal (inter)subjectification. It was thus shown that the joint treatment of SNs and TNs, which was unmotivated in the traditional literature, is warranted in terms of micro- and meso-constructional parallels. Descriptively, I have elaborated my arguments by means of extensive analyses of synchronic and diachronic corpus data. One of the most important findings was that each layer of SN- and TN-use involves specific preand/or post-nominal collocational patterns in the broad sense (i.e. extension or reduction of collocational range, semantic prosody and colligation). The observations made in these descriptive analyses fed back into a more dynamic functional-cognitive approach to the English NP, as well as a construction-based approach to grammaticalization and (inter)subjectification. The descriptive-theoretical issues addressed in this book have repercussions which stretch far beyond the specific case studies of (S)SN- and TNconstructions. These pertain to how the English NP, as a locus of synchronic variation and change, is modelled, the development of a constructional approach to grammaticalization and the rethinking of certain grammaticalization factors. Firstly, functional-cognitive models of the English NP as proposed by Halliday (1994) and Langacker (1991) generally assume that all positions in the NP are variable and rely on free choices, whereas many structures in English and other languages, such as SN- and TN-syntagms, incorporate a great amount of determination in terms of collocational and colligational patterning (also see Hilpert 2008 on collocations). In this study these patterns surfaced through close analysis of premodification patterns in SNand TN-constructions. This type of analysis is fairly new in grammaticalization research but, as the case studies in this book have shown, it yields previously unacknowledged uses as well as crucial insights into the micro-

358Chapter 7: Descriptive-theoretical consequences developments and chronology of these uses and their interaction within the constructional network. It was argued that such partially filled structures can be accounted for by incorporating a constructional component into my dependency-based functional-cognitive model of the NP. Collocations are then viewed as a lexical syntagmatic relation with the potential of defining a construction’s individuality, i.e. collocationally constrained constructions (also see Brems 2010). The latter, as form-meaning pairs, can be described multidimensionally with reference to collocational patterns, general semantics, semantic prosody, polarity sensitivity and formal, prosodic, register and dialectal features. With regard to register it was furthermore noted that formal and colloquial language each have their own type of creativity which may promote certain types of grammaticalization. Collocational and colligational patterns not only distinguished between head noun, descriptive modifier and postdeterminer uses of TNs, but, strikingly, also link up systematically with the subtypes of postdeterminer uses. Anaphoric postdeterminers have a demonstrative as primary determiner, e.g. these sort of scare tactics, while cataphoric ones typically have the definite article in that position, e.g. the kind of peace Stalin and Hitler brought to Poland. Non-phoric postdeterminers have what, any or some as primary determiner, e.g. what sort of jerk would do that! Secondly, this constructional approach also entailed a specific approach to grammaticalization which integrates it with diachronic construction grammar and constructionalization (cf. Fried & Östman 2004, Traugott 2007 & 2008a-b, Noël 2007, Trousdale 2008a-b, Bisang 2010, Sag 2010). As a consequence, factors or parameters of grammaticalization described in the literature have received a distinct interpretation. I have argued that grammaticalization primarily works on and results in substantive constructions or constructs, but that generalizations can be made over these concrete changes at micro-, meso- and macro-constructional levels of schematicity. This led to a proposal for a construction grammar-inspired trajectory of grammaticalization as moving from a more lexical source-construction to a more grammatical target-construction, in terms of gradual functional approximation or realignment to the latter as the result of analogical reanalysis. Intermediate stages can be formed by very specific constructions, which model themselves on (several) more or less distant, complex or substantive analogies. The concepts of reanalysis and analogy hence become highly integrated (cf. Fischer 2007) as I argued for an interpretation of reanalysis that recognizes the pull of an analogical model construction. Specifically with regard to infrequent SSN-expressions it was pointed out that this analogical pull may come from several models at different levels of schematicity. Analogy was thus interpreted as more than the extension of a new analysis to other syntactic environments, but as itself effecting (reconstruction of) change. Such a constructional approach to grammaticaliza-

359 tion enables to compare constructions at various levels of schematicity within and across networks, e.g. SNs and TNs as binominal structures and (monomorphemic) quantifiers, postdeterminers, etc. Shifts in semantico-pragmatics (i.e. delexicalization) were operationalized in terms of collocational extension and reclustering, while also recognizing persistence effects. Delexicalization and shifts in categorial status (i.e. decategorialization) were considered within a loss-and-gain model. Grammaticalization paths leading from source to target constructions are essentially a story of decategorialization and recategorialization. This nuanced approach allows for the extensive description of ambivalent constructions, which were further subclassified in terms of ambiguity, vagueness and blends, with reference to Bolinger (1961), Aarts (2007) and Willemse (2007). As with SN- and TN-constructions, lexical, grammaticalized and ambivalent uses of constructions can be argued to constitute semistable systems that form subparadigms with their own functionality within already existing paradigms of quantifiers, postdeterminers, hedges, quotatives, etc. It is in this context that the attested partial decategorialization of (SN- and TN-)constructions can be interpreted as a form of paradigmatic enrichment (also see Brems & Davidse 2010b). I have argued extensively that the restricted or particular premodification patterns allowed in the grammaticalized uses of SNs and TNs do not detract from the grammaticalization argument. With regard to constructional networks constituting semi-stable systems, it was noted that these may persist over long periods of time. In other words, grammaticalization processes do not have to go on to completion, in the sense of all SNs developing (valuing) quantifier uses for instance. This seems unlikely at this point since these expressions thrive on the tension between lexical and grammatical layers of meaning for their functionality within existing paradigms. This observation also ties in with the important point that previous analyses of a grammaticalizing construction need not have disappeared in order to argue for reanalysis (pace Langacker 1991 and 2009). With regard to the concepts of (inter)subjectivity and (inter)subjectification it was suggested that these may have to be broken down into subtypes, e.g. attitudinal versus textual/deictic (cf. Brems & Ghesquière 2009). It was moreover observed that these values may co-occur. In keeping with Ghesquière (2010), postdeterminers were argued to be textually intersubjective in nature, since a speaker uses them to negotiate the tracking of discourse referents by the hearer. They are textual because they are typically phoric in doing so. Qualifying uses of TNs were said to be in between subjective and intersubjective values depending on the subtype. Examples intended to save the hearer's face by hedging a possibly offensive nominal label, as in she's a kind of super groupie, can be argued to be inter-

360Chapter 7: Descriptive-theoretical consequences subjective in nature as well as a reflection of the speaker's attitude towards the she talked about, i.e. attitudinally intersubjective. The same goes for qualifying uses in which an unfamiliar referent is described in terms of a more familiar one, e.g. it is a sort of anthology, which are intended to guide the hearer to a better understanding. One might wonder whether there may also be a textual dimension to these uses. Other examples of qualifying TN uses seem to be first and foremost a reflection of the speaker's creativity in typifying a referent and may be considered primarily (attitudinally) subjective in nature, e.g. Offering the viewer a kind of a la carte theme park ride through the future or They regard the passing of the year as some kind of baptism in booze. Discourse markers that function as floor-holders may be labelled textual because they help structure a text strategically in terms of turn-taking and holding the floor, but at the same time they show deference towards a hearer, i.e. they are intersubjective. Quotative markers serve to set a stretch of discourse off from preceding discourse, signalling to the hearer that it is reported in nature. This would mean they have an intersubjective function which seems mainly textual in nature. On the other hand, the reported material is to some extent only the speaker's reconstruction of a speech event and hence also subjective in that respect. It is clear from this short discussion that this is an area of research that needs much more work (see Ghesquière 2010, Brems, Ghesquière & Van de Velde forthcoming for possible new venues in the demarcation and intersection of (inter)subjectivity). The study also made clear that more grammaticalization factors than the established ones have to be brought into the picture. In order to account for differences in degree of grammaticality and grammaticalization within the set of SNs, it was pointed out that prosodic factors and syllabic structure may play a larger role than commonly recognized. Bunches was the only plural SN with two syllables and did not grammaticalize. By contrast, valuing quantifier uses of SNs were largely restricted to the singular versions of SNs which seem to attract emotively coloured collocates more easily, since they are able to generalize over a group of entities or people and evaluate them all in one go (typically negatively). However, it is only by means of more research that the nature and importance of such prosodic factors can be fathomed. In the previous paragraphs I have summarized the main descriptive findings of this study of the synchronic layering of SN- and TN-expressions, as well as the most important theoretical contributions to grammaticalization research, cognitive-functional models of the NP and construction grammar. With these case studies of originally binominal structures, I have shown that the NP, as somewhat uncharted territory, is an interesting and important locus of synchronic variation and diachronic change that deserves multidimensional study within a dynamic theoretical framework. Since this

361 kind of study was largely unprecedented and the literature did not put forward a ready-made theoretical framework, established concepts had to be modified or reinterpreted and occasionally new concepts had to be added in order to accommodate the data. The more general applicability of this constructional-functional framework beyond the analyses presented here will of course need further research. Future research will also have to determine whether the wide-ranging collocational changes noted for SN- and TNconstructions merely mark such shifts, or whether they are a conditioning or causal factor of grammaticalization. Future studies will also have to look further into the nature of semantic prosody in order to confirm that it can be a facilitating factor that can promote emergent readings, particularly in processes of grammaticalization that involve attitudinal subjectification. In any case, collocational patterning has to be recognized as a fundamental research topic in grammaticalization studies.

Corpora The Bergen Corpus of London Teenage Language (COLT) The COLT corpus was collected in 1993 and consists of the spoken language of 13 to 17-year-old teenagers from different parts of London. The complete corpus, half a million words, has been orthographically transcribed and word-class tagged, and is a constituent of the British National Corpus. The CD-ROM version of COLT includes sound files and a prosodically annotated version in addition to the orthographically transcribed text. The prosodically annotated version, however, does not comprise the whole corpus. (http://www.hd.uib.no/ colt/) Collins WordbanksOnline (CW) A 57 million word synchronic corpus containing spoken and written English mainly from the 1990s onwards. It is stratified in terms of register, medium and region. The combination of these parameters defines 12 subcorpora. The subcorpora are indicated in the examples following CW: BBC= BBC World Service radio broadcasts, London; NPR= National Public Radio broadcasts, Washington (USA); OZnews= Australian newspapers, Brisbane (Australia); Sunnow= Sun Newspaper, London; Times= Times and Sunday Times, London; Today= Today newspaper, London; UKbooks= fiction and non-fiction books published in Britain; UKephem= ephemera produced in Britain (leaflets, adverts, junk mail, personal letters, etc.); UKmags= British periodicals (both thematically general and specialist); UKspoken= transcribed informal speech from all parts of Britain; USbooks=fiction and non-fiction books published in the USA; USephem= ephemera produced in the USA (leaflets, adverts, junk mail, personal letters, etc.). (http://www.collinslanguage.com/ wordbanks/default.aspx) Corpus of Early Modern English texts (CEMET) CEMET has been compiled by Hendrik Desmet and consists of texts drawn from Project Gutenberg and the Oxford Text Archive. It spans the period 16401717 and contains about 2 million words. Corpus of English Dialogues (CED) The CED contains Early Modern English speech-related texts such as constructed dialogues and authentic speech situations, and amounts to almost 1.2 million words. (http://www.engelska.uu.se/Research/English_Language/Research_Areas/Elect ronic_Resource_Projects/A_Corpus_of_English_Dialogues/)

Corpora

363

Corpus of Late Modern English Texts, extended version (CLMETEV) CLMETEV has been compiled by Hendrik De Smet and consists of texts drawn from Project Gutenberg and the Oxford Text Archive. It contains about 10 million words and spans the period 1710-1920 (see De Smet 2005). (https://perswww.kuleuven.be/~u0044428/clmetev.htm) Innsbruck Middle English Prose Corpus Sampler (IMPECS) The IMEPCS contains about 4 million words of Middle English prose. Many of its texts overlap with the PPCME (see below), but unlike the latter it contains complete versions of texts. Overlap has always been checked. (http://www.uibk.ac.at/anglistik/projects/icamet/) Instituut voor Nederlandse Lexicologie (‘Institute for Dutch Lexicology’) (INL-corpus) 38 million word corpus. The synchronic part of the INL corpus consists of a miscellaneous subcorpus (1970-1995) with 5 million words and a newspaper subcorpus (Meppeler Courant 1992-1995) comprising 27 million words. (http://www.inl.nl/) Old Bailey Corpus (OBC) The OBC consists of the proceedings from London’s central criminal court from 1674-1834. It contains circa 52 million words and profiles itself as reflecting more colloquial, spoken language. (http://www.uni-giessen.de/ oldbaileycorpus/) Penn-Helsinki corpora The Penn-Helsinki Parsed Corpus of Middle English 2 (PPCME2) contains 2,311,930 words and is based largely on the Middle English section of the Diachronic Part of the Helsinki Corpus of English Texts. It spans the period 11501500 The Penn-Helsinki Parsed Corpus Early Modern English (PPCEME) contains over 1.7 million words and spans the period 1500-1710 (http://www.ling.upenn. edu/hist-corpora/) The York-Toronto-Helsinki Parsed Corpus of Old English Prose (YCOE) YCOE is a 1.5 million word syntactically annotated corpus and the Old English sister corpus to the Penn-Helsinki Parsed Corpus of Middle English. (http:// www-users.york.ac.uk/~lang22/YcoeHome1.htm)



Notes 1. 2.

3.

4.

5.

6.

7.

8.

For the explicit argumentation in favour of treating these structures as constructions see Chapter 2, Section 2.2. In Brems (2001), (2002), (2003a) and (2004a) SN-constructions were called measure noun constructions, following the more narrow definitions and descriptions offered in traditional grammars, such as Quirk et al. (1972) and (1985), as well as Halliday’s (1994) account of “measure expressions” (see Section 2.1.1), and Dodge and Wright (2002). The present term size noun was opted for from Brems (2006) onwards primarily because it ties in with Langacker’s (1991) approach to the semantic function of quantification in the NP, which serves to indicate “the size of the profiled instance” referred to (Langacker 1991: 81), and is hence most in keeping with the theoretical framework used in this study (see Section 2.1.2). TN-constructions were previously referred to as SKT-constructions, e.g. in Brems (2003c), following Denison’s (2002) and Keizer’s (2001) terminology. Brems (2004b), Brems and Davidse (2005) and (2010) and De Smedt, Brems and Davidse (2007) use type noun constructions, because it is a more general term signaling the similarity in constructional semantics between uses of not only sort of, kind of and type of, but also of other nouns such as species, class, etc. Reference to type also ties in with Langacker’s semantic function of “type specification” within the NP (see Section 2.1.2), which makes this term analogous to and theoretically more cohesive with the term size noun. See Section 2.1.2 on Langacker’s grouping together of uncount nouns and plural count nouns as (two kinds of) mass nouns in terms of both semantic and formal similarities. In the study itself either the SN/TN will be in bold, or the SN/TN together with of, reflecting my structural analysis of each example. In the present section normally no such indications are made so that the reader is not yet influenced by the analysis that will be proposed later on. CW stands for Collins WordbanksOnline corpus, which is the main synchronic corpus used in this study. Examples which derive from other sources, such as the Internet or corpora other than CW, will be marked as such. Examples with no reference to the source are constructed examples. The original spelling and diacritics of the corpus extractions have largely been retained; hesitation markers, typos and tagging symbols hindering the reading of extractions have been edited out. Information on the composition of CW and the other corpora used in this study can be found in the corpus references. See Aarts (1998) for instance for a list of “binominal noun phrases in English”, or Traugott (2008a) and Trousdale (2010b) in the wake of Brems (2003). In the literature survey in Section 1.2 it will also become clear that certain authors generalize over several types of binominal phrases in their analysis of the specific ones looked at here (e.g. transformational accounts such as Akmajian and Lehrer 1976, Jackendoff 1981 and Lehrer 1986). See the reference section on corpora for more information.

Notes

365

9. See the reference section on corpora for more information. 10. The citation forms of the “quantifying nouns” are taken from Biber et al. (1999), who offer no explanation as to when of and the indefinite determiner a belong to the quantifying expression or not. If you take the cited forms to reflect an (implicit) structural analysis, this makes for a confused picture. As the expressions show, the difference cannot just be about grammaticalized versus head noun uses of quantifying nouns, e.g. a load of is said to act most like a quantifier, but dozens is also said to have quantifier-like features, yet is listed without of (Biber et al. 1999: 253 and 257). In other cases where SNs can have both head and quantifier uses only one cited form is offered. 11. Dodge and Wright (2002: 3-11), within the larger approach of Embodied Construction Grammar, presented a “typology” of “English Measure Phrases” (or “EMPs”) based primarily on their conceptual source semantics, “image schematic structures”, and the scene and frame of measurement they exploit. They do not make many explicit syntactic claims about the internal structure of such patterns in terms of head versus quantifier uses (but they do distinguish between two constructions at the end of the paper, which seems to coincide with such a distinction (Dodge and Wright 2002: 10). They (ibid.: 2-3) also distinguish EMPs that go with plural nouns from EMPs that team up with uncount nouns, which in its turn depends on the conceptual semantics of the measure phrases (cf. Biber et al. 1999). They too note fluctuating agreement patterns and unusual modification patterns with (mostly collective) EMPs, which are in part explained by the traditional notional versus grammatical concord opposition, reformulated here within a conceptual constructional framework (ibid.: 10). The main function of EMPs is described in terms of individuation of an instance in the case of uncount nouns (ibid.: 4; cf. Quirk et al. 1985, Jespersen 1932). In the case of plural count nouns, EMPs add information of shape or configuration to a by definition individuated count noun, much in the same way as classifiers in classifier languages do (ibid.: 45; cf. Lehrer 1986, Lucy 2000, and Joosten 2003 and Joosten et al. (2007) who discuss the conceptual semantics of Dutch collective and aggregate nouns.) 12. In transformational grammar rule operations, such as PP extraposition, interact with proposed levels of constituency or “NP cycles”. A prepositional phrase for instance can only be extraposed one cycle or constituency level higher than the one it belongs to itself (i.e. the NP cycle hypothesis in Akmajian and Lehrer 1976: 396). Proposed constituency analyses hence tend to tailor to such generalized rules. 13. This analysis explains why in the sentence A photograph of a man with three arms was published last year (=cycle 3) only of a man with three arms as a whole (=cycle 2) can be extraposed, but not with three arms by itself, because this PP belongs to cycle 1 and cannot climb two cycles (Akmajian and Lehrer 1976: 396). However, Akmajian and Lehrer (1986) noted that applying Jackendoff’s (1968) analysis to NPs containing NP-like quantifiers “seemed to lead to counterexamples to the NP cycle hypothesis”, in that they allow PP extrapositions beyond the cycle hypothesis and disallow extraposition that is predicted by the hypothesis (Akmajian and Lehrer 1976: 397). Hence a dif-

366Notes

14.

15.

16.

17.

18.

19.

20.

ferent constituency analysis is proposed in order to ensure the validity of the established rule of extraposition. This new proposal basically reverts to the pre-article analysis in which of is disconnected from the second noun, and, additionally, also from the NP-like quantifier (i.e. it is “transformationally adjoined to it” Akmajian and Lehrer 1976: 399). This broken up constituent analysis is set up to prevent unwarranted PP extraposition. Denison (2002) and (2005) do not differ in terms of descriptive analyses, but (2005) explicitly states that some of the diachronic shifts are related to grammaticalization, whereas this is more implicit in (2002). We will see in the corpus studies that type of also participates in my definition of the complex determiner construction, which detracts from Denison’s argument here. This specific postdeterminer use with type, however, might be a recent innovation and need not conflict with the development of the postdeterminer construction as such within the set of TN-constructions. This ties in with Brown and Levinson’s (1978/1987) face theory, which is a theory of politeness that tries to explain how and why social relations are encoded in language in the way they are. It argues that all people in society have two kinds of face, which both have specific needs. The negative face is associated with the right to personal sovereignty and freedom from imposition. The positive face typically seeks approval from other people and is associated with the human need to be liked and respected. Halliday uses the term (nominal) group instead of the more common noun phrase or NP, because it is essentially a “group of words”. “A phrase is different from a group in that, whereas a group is an expansion of a word, a phrase is a contraction of a clause.” (Halliday 1985a: 159). In Halliday and Matthiessen (2004) TN-expressions are added to the list of nominal groups that display a dissociation of Head and Thing, as well as an agnate structure that is “construed in the same way, with the Head dissociated from the Thing and the two linked by of, as in a cup of coffee.” They also add that “in measure or type expressions […] the Head word has become partially grammaticalized”, without any further comments (Halliday and Matthiessen 2004: 334). The inclusion of TNs as well as the reference to grammaticalization as such are probably due to the growing interest in both topics around the same time, and the emergence of grammaticalization studies about the NP in specific (e.g. Brems 2002 and 2003a, Denison 2002, Aijmer 2002, Traugott 2007). Halliday’s idea of two simultaneously present layers might describe intrinsically vague cases or certain metaphorical phenomena, as in The British have forged a fine tradition of gardening and cannot afford to sit on their wellclipped laurels. Striding past the compost heap of nostalgia, come Christopher Lloyd. (CW-Times), where the SN-expression is an irreducible blend of head and quantifier semantics (cf. Bolinger 1961). Later on, Langacker also reinterprets singular count nouns as a type of mass nouns, in the sense that they designate a “degenerate mass”, i.e. a mass that consists of only one component entity (Langacker 1991: 84). Hence all nouns can be conceptualized as mass nouns to some degree, even though the extension towards singular nouns is rather controversial.

Notes

367

21. Milsark (1976) makes a similar conceptual distinction between “cardinal” and “universal/non-universal” and adds further recognition criteria. Even though Milsark predates Langacker, he proposes a classification of quantificational potential that is more flexible than Langacker (1991), in that it attributes both absolute and relative uses to indefinite quantifiers such as some, any, no and much, little, few, many. Langacker on the other hand sees the first three as inherently relative and the second set as inherently absolute. Brems and Davidse (2003) argue for rethinking Langacker and Milsark’s claims in terms of prototypical quantifier uses of the two sets, and beyond Milsark argue that even cardinal numbers can express relative quantification in some contexts, as first suggested by Laffut (1998). 22. Even though he stresses the importance of dependency relations, McGregor offers few criteria for head (versus dependent) status, other than distributional equivalence. In binominal patterns like the SN and TN-syntagm, however, the main problem precisely is assessing this distributional status. Without further criteria for distributional equivalence itself, such an assessment is hard to make. Langacker’s profile determinant leads to similar difficulties (Langacker 1987: 289). 23. Construction Grammar is used here as a cover term for a set of “constructionist approaches” (Goldberg 2006), and not as a term for the specific version proposed by Kay and Fillmore (1999). Differences between the various strands are described in Section 2.2.3. 24. Francis (1993: 140) similarly states that the implication of the view given in traditional grammars is that “so long as you choose a word from the right class to fill the slot, then the result will be a grammatical utterance. But, while it may indeed be grammatical in the narrowest sense of the term, it is unlikely that it will be natural or acceptable unless lexical factors are taken into account.” 25. Hunston and Francis (2000) seems interesting in this context. Their Sinclairian Pattern Grammar: A Corpus-driven Approach to the Lexical Grammar of English offers a description of lexical items in terms of the phraseological patterns they occur in, both general patterns, such as V n (i.e. a verb followed by a noun group) and very specific ones. The interaction between lexicon and grammar is stressed, by noting that knowing the meaning of a word means knowing the grammatical distribution of that word, and its collocational behaviour. However, Hunston and Francis (2000), like much of the Birmingham work, shy away from theoretical presuppositions and generalizations about grammatical structure. The absence of such a unified framework generalizing over the individual observations decreases the value of the book. This is why CxG is taken as the starting point for a constructional approach here and not Pattern Grammar. The latter can, however, add interesting specifics to this CxG framework. 26. This introduction is based on Croft (2001), Croft and Cruse (2004: 225-247) and Goldberg (2006). 27. Fillmore et al. (1988) use the term formal idiom. Croft and Cruse (2004: 234) use Langacker’s conceptually similar term schematic construction instead.

368Notes

28. 29.

30.

31.

32.

33.

34.

35.

Goldberg (2006) mainly talks about filled or partially filled idioms (e.g. Goldberg 2006: 5). I will use the terms constructionist and constructional interchangeably throughout this study. Croft and Cruse (2004: 265-290) discuss differences between Fillmore et al.’s “Construction Grammar”, Lakoff’s/Goldberg’s “construction grammar”, Langacker’s “Cognitive Grammar” and Croft’s “Radical Construction Grammar” by means of four general questions: “(i) What is the status of the categories of the syntactic elements in construction grammar, given the existence of constructions? (ii) What sorts of syntactic relations are posited? (iii) What sorts of relations are found between constructions? (iv) How is information stored in the construction taxonomy?” The main difference between Goldberg (1995) and (2006) is the focus on a usage-based model. This has led, for instance, to an extension of the definition of a construction to include fully predictable patterns, in addition to the traditional (partially) unpredictable ones, provided certain frequency criteria are fulfilled (Goldberg 2006: 12-13). In fact the lexical openness in schematic constructions that Croft and Cruse refer to is not merely "lexical", in that it can, somewhat contradictorily, also be the case that the construction is fixed except for "grammatical inflectional categories", as they themselves observe (Croft and Cruse 2004 : 248). The idiom to kick the bucket for instance is nearly fully substantive, except for the inflectional categories and some variation in the subject, e.g. He kicked the bucket/Her daughter has kicked the bucket. The difference between (more or less) lexical and grammatical meaning, essential to the concept of grammaticalization, likewise is in a way a nonissue in CxG (cf. Langacker 2003). This flexibility warrants interaction between CxG and grammaticalization theory, but makes it problematic as well (cf. Noël 2007). In Chapter 3 it will be interpreted positively. This semantic model has much in common with Langacker’s cognitive model of profile/base. Langacker says that a word form symbolizes a concept profile and this profile has to be understood against a certain base, i.e. "that knowledge or conceptual structure that is presupposed by the profiled concept." (Langacker 1987 discussed in Croft and Cruse 2004: 15). An alternative name for the base is domain. The base/domain can be largely equated with Fillmore’s notion of a frame. According to Langacker (2005: 160) Construction Grammar neglects certain dimensions of construal that figure prominently in his Cognitive Grammar, such as profiling, metaphor, metonymy, mental spaces and blending, all of which are essential to meaning and essential to Cognitive Grammar. Fillmorean CxG has strong affinities with Head-driven Phrase Structure Grammar (HPSG) but uses notions of modifier and head in keeping with this specific theory. RCxG in principle does not accept the existence of syntactic categories as NP, Verb, etc. and only takes constructions as primitives. Other CxGs, and case studies within them, do use these traditional labels rather straightforwardly.

Notes

369

36. Fillmorean Unification CxG and other strands have an elaborate formalized apparatus supplying this phonological, morphological, syntactic, semantic, pragmatic, prosodic and discourse information about each construction, in terms of Attribute-Value Matrices, which are represented in (nested) box diagrams or alternatively in square-bracket notation. 37. I will use the term constraints in way that essentially makes it coincide with such notions as properties or characteristics. 38. Unification CxG is “not uniformly” usage-based according to Goldberg (2006: 215-216), since individual authors within this framework differ in how they address the matter. Unification CxG mainly aims at accounting for generalizations in language. Predictable patterns, however frequent, are presumed not to be part of a speaker’s knowledge of a language. They are concreter instantiations of more schematic constructions. 39. Goldberg (2006) aims at psychological plausibility in proposing her views on the nature of constructions. The present study assumes that observations arising from this specific concern translate rather straightforwardly into tools for the kind of descriptive analysis we are interested in here. Still, the psychological concern might not always coincide with descriptive aims, and this should be kept in mind. 40. In Croft and Cruse (2004: 233) schematic constructions refer to all constructions in which at least one part of the construction is lexically open. They hence also subsume fully schematic constructions, in which all slots are lexically variable. In order to avoid confusion between schematic and fully schematic, the term partially filled in constructions can be used to refer to patterns in which certain items are filled-in and others are not (cf. Goldberg 2006: 5). Schematic will mainly be used for constructions which are non-substantive. 41. It has to be observed here that Goldberg does not refer to formal or semantic aspects in the historical development of these adjectives that might explain why mere and aghast are restricted to attributive and predicative usage respectively. As such diachrony is not an important factor in Goldberg (2006), even though in theory it could be part of the CxG framework. 42. The literature on semantic prosody is often vague about the directionality of the phenomenon and its origins: is it the emotive colouring of the set of collocates that primes the node for either positive or negative semantic prosody, or is it something in the node itself that radiates it? Stubbs (1995) suggests that it is the emotive colouring of the set of collocates that affects the (semantically neutral) node and may eventually restructure the semantics of the node as a result of it (see Brems 2010 for a more extensive critical survey of the various interpretations of semantic prosody in the literature). 43. Grammaticalization is also referred to in the literature as grammaticization or grammatization, but I systematically use the term grammaticalization in this study, except when citing an author who uses one of the alternative terms. 44. This holds, provided one accepts that the class of discourse markers in fact constitutes a grammatical category and hence can serve as a grammaticalization output (see Hopper and Traugott 2003: 129, Traugott 1995b and 2003, and Hopper 1991: 19 on what is potentially “‘in’ the grammar”, as well as Himmelmann 2004).

370Notes 45. To classify Lehmann (1985) within this formal approach to grammaticalization amounts to an oversimplification. Lehmann (1985) subsumed a semantic parameter and also introduced interesting distinctions such as a synchronic versus a diachronic interpretation of grammaticalization. In its diachronic interpretation, grammaticalization refers to the specific process of change. In its synchronic interpretation it can account for variation in degree of grammatical versus lexical character of members of one grammatical category, which is explained as the synchronic result of such diachronic processes (e.g. the range of devices for marking case relations in Latin, Lehmann 1985: 303). This “synchronic variation” is similar to Hopper’s (1991) concept of layering, which is discussed in greater detail later in Section 3.2.2. Lehmann (1985: 311) also discusses the concept of renewal, which will be discussed in Section 3.4.8. His parameters are moreover still widely used in many studies today and by no means superseded. Criticism of Lehmann’s parameters is discussed in the following sections. 46. Traugott (2003: 643) argues that different components in the grammar may serve quite different functions and more pragmatically-functioning components may not all be subject to the same kinds of syntactic scope reduction as case and tense categories for instance. Linguists who do not consider discourse markers to be a grammatical category and hence not a product of grammaticalization, sometimes refer to their development as “pragmaticalization” (e.g. Erman and Kotsinas 1993) or “postgrammaticalization” (Vincent et al. 1993) to make this clear. 47. Traugott’s original cline referred to Halliday’s ideational, textual and interpersonal metafunctions (see Section 2.1.1). In Traugott (1982) interpersonal was rephrased as expressive and, following Benveniste (1966) for instance, Traugott and Dasher (2002) referred to it as subjective and intersubjective meaning. 48. Hopper (1991), as well as Heine (2003), Hopper and Traugott (2003) and Bisang et al. (2004) note explicitly that none of these principles are exclusive to grammaticalization, but common to linguistic change in general. The significant degree to which these related principles co-occur warrants them being singled out in this fashion. Heine (2003: 583-584) argues that his four principles (see below) “and the way they are interrelated are part of one and the same explanatory framework”. Heine (2003) does argue for at least a certain amount of predictive and explanatory power for grammaticalization theory, against Campbell (2001) (see Section 3.2.4). 49. Bisang et al. (2004: 4) describe the field of grammaticalization theory in terms of the interaction between three “mutually independent components which form the core of grammaticalization”, i.e. “morphosyntactic change”, “semantic or functional change” and “constructions”, and several linguistic fields “that create its fringes”, i.e. “pragmatics, phonology and the lexicon”. Interaction with constructions, the lexicon and pragmatics all belong to recent trends in grammaticalization research, which include an active interest in the difference between grammaticalization and lexicalization (see Section 3.4.9). 50. In Hopper and Traugott (2003: 1 and 40) process is rephrased as change, because the former is potentially misleading. Referring to grammaticalization

Notes

51.

52.

53.

54.

55.

56.

57.

371

as a process may seem to imply that grammaticalization is a deterministic force with an impetus of its own, independent of language users. In Bisang et al. (2004: 9) emergence of new constructions is not excluded from grammaticalization studies. They (2004: 13) ask if such “types of structural change should be allowed in [a definition of grammaticalization] that can hardly be described by (or even contradict) prevalent parameters of grammaticalization, first of all those established by Lehmann.” Traugott (2003: 631) argues that grammaticalization works on extant constructions, whereas Traugott (2006) seems to consider the possibility of grammaticalization directly working on very schematic constructions. Trousdale (2005) similarly sees schematic constructions as a direct input for grammaticalization. Goldberg (2006) considers generalizations or constructions as having a prototype structure, but does not interpret this in terms of a generalization serving as a model for new instances of the schematic category. It should be noted here that Bisang et al. (2004), who argue explicitly for a constructional approach to grammaticalization, take a very broad definition of grammaticalization that not only includes the description of "all the processes involved in the diachronic change and emergence of such systems" [i.e. "more or less stable, regular and productive form-meaning mappings"]. It also includes "the emergence of new grammatical structures which either replace older structures to express the same concept or which express new concepts for which there was no structure in earlier times." Following Noël (2007), I have argued in Section 3.2.3.2 that symbolization belongs to the job description of diachronic Construction Grammar, as distinguished from a constructional approach to grammaticalization. The case studies in Bisang et al. (2004), however, often invoke a more narrow definition of grammaticalization. Himmelmann (2004: 34 and 39) for instance argues that it is narrowly defined grammaticalization processes that share a significant subset of changes, which warrants their theoretical significance. Himmelmann (2004: 34) also argues that grammaticalization works on extant substantive constructions, when he says that “grammaticization applies only to the context expansion of constructions which include at least one grammaticizing element. Context expansion may also occur with other types of constructions, for example a certain word order pattern, a compounding pattern or a reduplication pattern. These are not considered instances of grammaticization here”. In addition, work has been done on non- or minimally lexical sources of grammaticalization, e.g. Lehmann (2008) on information structure, Diessel (2006) on demonstratives and Traugott (2010a) on pseudo-clefts. In a way, all grammaticalization studies take place with hindsight. Still, the distinction between parameters, mechanisms, motivations and epiphenomena presented here seems justified and warrants descriptive accuracy. Cf. Traugott (2003: 644 and 647), who does say that new categories may arise, but only through accretion of many individual changes, not as the direct outcome of one change. Hopper (1991:17) notes: “what is distinctively

372Notes

58.

59.

60.

61.

62.

63. 64.

65. 66.

67.

68.

grammatical in a language is not its broader structures but the individual grammatical forms which comprise this structure”. Campbell (2001) does not consider semantic bleaching from lexical to grammatical or less to more grammatical meaning as an independent empirical test for grammaticalization because this shift is already part of the standard narrow definition of grammaticalization incorporating unidirectionality. Hopper (1991: 20) does note that “[t]he occurrence of certain lexical items in frequent collocations (for example, when the word ‘foot’ repeatedly occurs in phrases like ‘at the foot of the hill’, etc.) may be prima facie evidence of incipient grammaticization.” Certain textual markers such as discourse markers or topicalizers also serve to express "speaker’s attitude to the text under production” (Traugott 2010b: 31). In Brems (2003a) and (2007) only attitudinal subjectification is explicitly discussed, whereas textual subjectification is more implicitly referred to. In De Smedt, Brems and Davidse (2007) reference to textual subjectification likewise is rather implicit. In addition, the distinction between pragmatic and semantic (inter)subjective meaning was glossed over in these articles. Breban (2010) makes use of Langacker’s concept of deictic subjectification for the non-attitudinal subjectification developments displayed by the grammaticalized uses of adjectives of comparison. Target is used here in the sense of Campbell (2001: 32) as “an outcome of grammaticalization”. Obviously, within a theory of grammaticalization, classes or categories, just like language itself, can never be truly closed (cf. Bybee 2003a). It would hence be more appropriate to refer to the latter as ‘less open’, but closed-class is probably too entrenched to be jettisoned (e.g. Quirk et al. 1985). On the other hand, other people might interpret this as degrammaticalization (see Fitzmaurice 2000). Hoffmann (2004: 189), following Bybee and Thompson (2000: 380), notes that extremely high frequency, rather than facilitating grammaticalization and change in general, can have a “conserving effect”. A very high level of entrenchment may cause an expression to resist change; one can think here of highly frequent verbs such as be, which retains its irregular past tense morphology despite pressure from innovations in this domain in the form of the productive and regular –ed pattern or its allomorphs. In Section 2.2 it was observed that in Construction Grammar grammaticalization is, to a certain extent, a non-issue. The same goes for lexicalization, but again I interpret this positively in the sense that lexicalization is not disallowed. As Himmelmann (2004) notes, confusion between both types of change is mainly brought on by the simplistic scales used in the morpheme-based view on grammaticalization: lexical item > grammatical item in grammaticalization and grammatical item > lexical item for lexicalization. These scales do not incorporate criteria to decide on what is in the grammar and what is in the lexicon, which is a notorious domain of dispute and depends strongly on theoretical affiliation.

Notes

373

69. Within speaker-hearer interaction, roles shift between interlocutors. Hearer and speaker are not used in a strict way here, but first and foremost make the point that social considerations as well as egotist ones shape the discourse between people and determine the arsenal of communicative strategies and means of expression in most languages. Haspelmath’s (1999) concept of extravagance similarly links a speaker’s expressivity to impressing some hearer. As such, subjective and intersubjective values are not always easy to tease apart. 70. Chapter 6 will show in greater detail how the analysis proposed there differs from that in De Smedt, Brems and Davidse (2007) and Brems and Davidse (2010). 71. UK data= 42,099,593 words. US data= 9,980,368 words. Australian data= 5,337,528 words. 72. In an interview on commercial Flemish television station VTM a few years ago host Marlène de Wouters started her interview with actor Kevin Costner by saying that she had heaps of questions for him, to which Costner replied “ ‘Heaps’?, that’s Australian!”. 73. Whole can also occur in the head noun construction with SNs, where it means ‘as a whole’, as in Internal haemorrhoids can bleed, blood with the bowel motion. They can also pass down through the anal sphincter where they protrude, feeling like a whole bunch of grapes. (CB-OZnews) 74. AV is a computer programme that combines tools for classifying data and performing statistical analyses on them. It has been developed by Dirk Speelman at the University of Leuven. For more information see http://wwwling.arts.kuleuven.ac.be/genling/abundant. 75. The Ormulum is an exegetic text most probably written between 1150 and 1200 by a monk named Orm. It is known especially for its idiosyncratic orthographical system and for introducing several metrical innovations that would inspire future English poets. Together with Ancrene Wisse and the Ayenbite of Inwyt it is one of the crucial texts to attest to the transition from Old English to Middle English (see Parkes 1983 and http://www2.english.su.se/ nlj/ormproj/ormulum.htm). 76. The observations made in this section are based on the case studies presented by Brems and Kimps (2005), but the data analysis offered here was carried out by myself and differs from the one presented in 2005. It also includes more detailed collocational information. 77. The English translations of the Dutch examples try to stay close to the Dutch originals in the choice of syntactic structures and types of expressions. English SN-expressions+N2 are underlined, as well as its premodifiers and contextual markers that emphasize the meaning of the SN. 78. Neither Traugott (2005) and (2010b), nor Israel (1996) and (2004) uses the term SSN-expression. In Traugott’s terminology (S)SNs are said to shift from a partitive construction to a degree modifier construction, in which the SN has modifier status. 79. Traugott (2005) uses the term polarity sensitive items, whereas Israel (2004) and Traugott (2010b) mostly refer to polarity items.

374Notes 80. In his study, Hoffmann uses a list of 132 P-N-P-sequences drawn from the written component of the British National Corpus and a Newspaper corpus (i.e. editions of 7 British and American newspapers from the year 1999) with 175,465,397 words in it. The Newspaper corpus and the part of the BNC together make up the synchronic corpus for his study. The size of the BNC corpus is not mentioned in his article, but according to the official website the written component of the BNC amounts to 89,740,544 words. For the diachronic part of the corpus study The Gutenberg texts were used (i.e. texts by British authors written in the 17th, 18th and 19th centuries), which amount to 23.5 million words, as well as the OED (Oxford English Dictionary) database of illustrative examples on CD-ROM. 81. Hoffmann (2004) systematically seems to equate "being grammaticalized" with having "grammatical status" and being "units of language structure" (ibid.: 171), "indivisible units which are retrieved from memory as one single chunk" (ibid.: 181), or "units of storage in the mental representation of grammar" (ibid.: 184), reminiscent of Goldbergian Construction Grammar (Goldberg 2006). Equating grammatical status and grammaticalized status is, in my view, problematic. 82. By dint of occurs 66 times in Hoffmann’s synchronic corpus, in contradistinction to occurs 15 times. I think at least in contravention of and in furtherance of can also be seen as incorporating infrequent lexical items. The former P-NP string occurs 90 times in Hoffmann’s study, the latter occurs 24 times (Hoffmann 2004: 183). In Collins Wordbanks furtherance has 11 attestations, only 2 of which are outside of a P-N-P string, 4 appear in in furtherance of, 1 in for the furtherance of, 3 in in the furtherance of and 1 in at the furtherance of. Contradistinction only occurs 1 time in Collins Wordbanks and this is within in contradistinction to. Contravention occurs 30 times, 8 attestations of which occur outside of a P-N-P string, 21 within in contravention of and 1 in for contravention of. The version of the Collins Wordbanks corpus used for the present study is of course smaller than the corpora used by Hoffmann, but the proportions and normalized frequencies give an idea of the relative frequencies. 83. Polarity sensitivity is of course dealt with by many authors, but Israel (2004) is chosen here because of his clarity and focus on the pragmatics and semantics of polarity sensitivity. Israel (2004: 716) does not discuss SSNs as such, but includes what I call (S)SNs in his list of PIs. 84. Adverbial qualifying uses of type nouns, by contrast, do retain the of, as in He sort of smiled at us (Quirk et al. 1972: 454). It might be the case that a bit of is more likely to lose the of than sort of because only the former invokes a partitive meaning, which does not square well with extended uses. 85. This ties in with Kel Richards’ account. Kel Richards is a writer and presenter of the daily wordwatch feature on Australian ABC news radio who offers the following etymology for skerrick on the ABC website (http//: www.abc.gov.au/classic/breakfast/index2002-12-full.htm): Skerrick is one of those words that began life as a British dialect word, came to Australia with the early settlers, and survived here in colloquial Australian English while fading out of existence in the land of its birth. It’s recorded in

Notes

86. 87.

88.

89. 90. 91.

92.

93. 94.

375

Australia as early as 1854 (in a book called Gallops and Gossips) in the statement: “I have plenty of tobacco, but not a skerrick of tea or sugar” (which is, clearly, the modern sense of the word). So, where did the word come from? The 1823 edition of Grose’s Dictionary of the Vulgar Tongue records the word “scurrick” which is said to be thieves’ cant for a half-penny (it’s recorded in the same sense, in the same year, in a Dictionary of Turf by “Jon Bee”). And this word “scurrick” is sometimes recorded as “scuttick” and sometimes as “skiddick” so it is probably the origin of “skerrick” – especially as the meaning seems to match: a half-penny being “a small amount”. The following example with skerrick from the NSOED might be a (relatively recent) illustration of this obsolete monetary meaning: These wadless blokes of the Never-Never have to pay road, car, petrol, State, Federal and Unemployment Relief taxes, and never get a skerrick in exchange (F. CLUNE: Roaming round Darling 1936) Kind of occurs 14334 times in the entire corpus, kinds of 1601 times, type of 2628 times, types of 1471 times, sort of 23431 times and sorts of 1483 times. This section presents my classification of TN-uses, based on the data sets used in De Smedt, Brems and Davidse (2007), but diverging from it on several points. Additional categories have been set up to accommodate ambivalent contexts, e.g. head/postdeterminer, qualifying use/discourse marker. Subcategories have also been added within the larger categories, e.g. size-related postdeterminer, emphasizing postdeterminer, within the postdeterminer category. In addition, the quantifier use is added, as well as systematic information about collocational patterns. Denison (2002/2005) calls the head noun use the binominal use. I prefer to use the term binominal to refer more generally to NPs that contain two nouns and use the label head noun use for this specific binominal use here. It should be noted that the subtype can also be indicated by a classifying element preceding the TN, as in The Danish sort of reform in (6.17). The Collins Wordbanks data are not coded for intonation, so I am relying on my own reading of the examples here. Frequencies for kind also include those attested for the coalesced spelling kinda. One attributive modifier use was attested with kinda in the COLT corpus. These non-standard spelling variants are not systematic or consistent in informal registers in the case of premodified modifier uses. What’s more, the sorta/kinda/typa-spelling occasionally gets extended to head uses in informal contexts, as in What paper should I use? What kinda pencils and pens are best to draw with? Is it okay to copy pictures? (www.organicmetal.co.uk/ pages/tutorials_qa.htm/). Coalesced spelling as such is hence not a straightforward sign of modifier status, nor of grammaticalization, but typical of many routinization processes (cf. Joseph 2007). These metalinguistic values to some extent link up these modifier uses with the qualifying uses that will be discussed in Section 6.3.5. According to the OED (1933, vol. 9: 74-75), same, which derives from IndoGermanic *some-, meaning ‘level’, came into English as a postdeterminer ad-

376Notes

95.

96. 97.

98. 99.

100.

101. 102. 103.

104. 105. 106.

107.

108.

jective. Throughout the history of English, it has always formed a fixed unit with definite determiners. In een dergelijke, een means ‘one’ or ‘a(n)’. The first part of dergelijke derives from obsolete case marking on Dutch definite article de, while gelijk means ‘equal’. One could of course argue that quantity implications are simply the qualities involved in concepts dealing with size. Margerie (2010) calls this complex unit a “complex determiner”, even though elsewhere she denies the existence of a functional category of determiner complexes incorporating TNs. Frequencies for kind also include attestations with kinda. Two adverbial qualifying uses with kinda are attested in the COLT corpus. Examples in this section have the TN in bold when ambivalence is between head and modifier status. If ambivalence is between two subtypes of modifier uses, the entire TN-expression is in bold. Contrary to S(S)N-expressions, the TN-expressions generally retain of in their extended uses, or at least represent it in a reduced way in coalesced spellings, such as kinda/sorta/typa. For this comparison additional, exhaustive, extractions were made on a|Ø+sort|kind|bit+of+a|an+NOUN|JJ|VERB. JJ stands for the adjectives. Additional extractions were made on adverbial uses of bit. Remember that alternation of a TN-construction with a in front of N2 was argued to be a test for distinguishing generally between head and modifier uses. Of all postdeterminer uses, emphasizing postdeterminers allow this most easily, probably because of their increased subjective value. Type of did not participate in all of the shifts that sort and kind participated in. The qualitative and quantified diachronic analysis presented here is based on the data sets of Brems & Davidse (2010a). It is in fact with the qualifying use that TN-constructions broke free from the boundaries of the NP, as TN + of also began to qualify other syntagms than nominal ones (see Section 6.3.5). Because non-nominal constructions with TNs are much more recent phenomena, typically associated with informal spoken language, for which there are no diachronic sources, I have limited myself in this section to nominal TN-constructions. The Middle English data contain head uses, but not within binominal constructions. In the Modern English data, by contrast, binominal constructions are common and from this period on extractions were restricted to the NP + of + NP syntagm, which defines the locus of reanalysis. Middle English data with sort/kind followed by of + N2, which are extremely rare, have not been included in this overview as they would deform the overall picture.

References Aarts, Bas 1998 2007

“Binominal Noun Phrases in English”. Transactions of the Philological Society 96: 117-158. Syntactic gradience: the nature of grammatical indeterminacy. Oxford: Oxford University Press.

Adamson, Sylvia 2000 “A lovely little example: word order options and category shift in the premodifying string”. In Olga Fischer, Anette Rosenbach and Dieter Stein, eds. Pathways of Change: Grammaticalization in English. Amsterdam/Philadelphia: John Benjamins. 39-66. Aijmer, Karen 2002

English Discourse Particles: Evidence from a Corpus. Amsterdam/Philadelphia: John Benjamins.

Akmajian, Adrian and Adrienne Lehrer 1976 “NP-like quantifiers and the problem of determining the head of an NP”. Linguistic Analysis 2: 395-413. Allan, Keith 1977

“Classifiers”. Language 53: 285-311.

Andersen, Henning 2001 “Actualization and the (uni)directionality”. In Henning Andersen, ed. Actualization: Linguistic Change in Progress (Current Issues in Linguistic Theory, 219). Amsterdam/Philadelphia: Benjamins. 222-248. Athanasiadou, Angeliki, Costas Canakis & Bert Cornillie (eds.) 2006 Subjectification: Various paths to Subjectivity. Berlin/New York: Mouton de Gruyter. Bache, Carl 2000

Essentials of mastering English: A Concise Grammar. Berlin/New York: Mouton de Gruyter.

Bache, Carl & Niels Davidsen-Nielsen 1997 Mastering English: an advanced grammar for non-native and native speakers. Berlin: Mouton de Gruyter.

378References Bakema, Peter 1998

Het verkleinwoord verklaard: een morfosemantische studie over dimunitieven in het Nederlands. Unpublished PhD dissertation. University of Leuven.

Benveniste, Émile 1966 “De la subjectivité dans le langage”. In Émile Benveniste, ed. Problèmes de Linguistique Générale. Paris: Gallimard. 258-266. Biber, Douglas, Stig Johansson and Geoffrey Leech 1999 Longman Grammar of Spoken and Written English. London: Longman. Bisang, Walter 2010

“Grammaticalization in Chinese: A construction-based account”. In Gradience, Gradualness and Grammaticalization, Traugott, Elizabeth Closs and Graeme Trousdale, eds., 245–277.

Bisang, Walter and Björn Wiemer 2004 “What makes grammaticalization? An appraisal of its components and its fringes”. In Walter Bisang, Nikolaus P. Himmelmann and Björn Wiemer, eds. What makes Grammaticalization? A Look from its Fringes and its Components. Berlin/New York: Mouton de Gruyter. 3-20. Bisang, Walter, Nicholaus P. Himmelmann and Björn Wiemer 2004, eds. What makes Grammaticalization? A Look from its Fringes and its Components. Berlin/New York: Mouton de Gruyter. Blank, Andreas 2001 “Pathways of Lexicalization”. In Martin Haspelmath, Ekkehard König, Wulf Oesterreicher & Wolfgang Raible. Language Typology and Language Universals (HSK 20). Berlin: Mouton de Gruyter. 1596-1608. Bolinger, Dwight 1961 “Syntactic blends and other matters”. Language 37: 366-381. 1967 “Adjectives in English: attribution and predication”. Lingua 18: 1-34. 1972 Degree words. The Hague: Mouton de Gruyter. Breban, Tine 2002

Adjectives of comparison: Postdeterminer, Epithet and Classifier Uses. Unpublished MA Thesis. University of Leuven.

References 2006a

2010

379

“Grammaticalisation and subjectification of the English adjectives of general comparison”. In Angeliki Athanasiadou, Costas Canakis and Bert Cornillie, eds. Subjectification: Various Paths to Subjectivity. Berlin/New York: Mouton de Gruyter. 241-278. English adjectives of comparison: Lexical and grammaticalized uses. (Topics in English Linguistics). Berlin: Mouton de Gruyter.

Breban, Tine and Kristin Davidse 2003 “Adjectives of comparison: the grammaticalisation of their attribute uses into postdeterminer and classifier uses”. Folia Linguistica 37 (3/4): 269-317. Brems, Lieselotte 2001 From Head to Quantifier: Grammaticalization and delexicalization in Measure Noun constructions. Unpublished MA thesis. University of Leuven. 2002 “Measure Noun constructions: an instance of semantically-driven grammaticalization”. Preprints of the Department of Linguistics nr. 193. University of Leuven. 2003a “Measure Noun constructions: an instance of semantically-driven grammaticalization”. International Journal of Corpus Linguistics 8 (2): 283-312. 2003b “The SKT-construction: sort of, kind of and type of”. Paper presented at the third International Workshop of the Systemic Functional Research Community on Interpersonal and Ideational Grammar, Ghent University, 27-29 November 2003. 2004a “Measure Noun constructions: degrees of delexicalization and grammaticalization”. In Karin Aijmer and Bengt Altenberg, eds. Advances in Corpus Linguistics: Papers from the 23rd International Conference on English Language Research on Computerized Corpora (ICAME 23). Amsterdam: Rodopi. 249-265. 2004b “Type noun constructions”. Paper presented at ICAME 25 (The 25th International Conference on English Language Research on Computerized Corpora of Modern and Medieval English) in Verona, 19-23 May 2004. 2006 “The grammaticalization of small size nouns: reconsidering frequency and analogy”. Preprints of the Department of Linguistics nr. 254. University of Leuven. 2007 “The grammaticalization of small size nouns: reconsidering frequency and analogy”. Journal of English Linguistics 34 (4): 293-324. 2010 “Size noun constructions as collocationally constrained constructions: lexical and grammaticalized uses”. English Language and Linguistics 14 (1): 83-109.

380References Brems, Lieselotte and Kristin Davidse 2003 “Absolute and relative quantification: Beyond mutually exclusive word classes”. Belgian Journal of English Language and Literatures. New Series 1: 49-60. 2005 “Type noun constructions: Paths of grammaticalization”. Paper presented at NRG 3 (New Reflections on Grammaticalization 3), Santiago de Compostela, 17-20 July 2005. 2010a “The grammaticalization of nominal type noun constructions with kind/sort of: chronology and paths of change”. English Studies 91: 180-202. 2010b “Complex subordinators derived from noun complement clauses”. Acta Linguistica Hafniensia 42 (2): 101-116. Brems, Lieselotte and Ditte Kimps 2005 “Ideational and interpersonal driving factors of grammaticalization and collocational patterns: Dutch and English quantity nouns”. Paper presented at FITIGRA (From Ideational to Interpersonal: perspectives from grammaticalization), University of Leuven, 10-12 February 2005. Brems, Lieselotte & Katrien Verveckken 2008 “English and Spanish Size Noun-constructions, a locus of ongoing grammaticalization”. SLE 41 (41th Annual meeting of SLE: Languages in Contrast, Grammar,Translation). 17-20 september 2008, Forli (Italië). Brems, Lieselotte & Lobke Ghesquière 2009 “Implications for directionality and (inter)subjectification hypotheses: Evidence from case studies on type nouns and prenominal adjectives". Paper presented at Third Annual IAP meeting, 11 December 2009, Brussels. Brems, Lieselotte, Lobke Ghesquière & Freek van de Velde, eds. Forthcoming Intersections of intersubjectivity. Special issue of English Text Construction. Amsterdam/Philadelphia: John Benjamins. Brinton, Laurel and Elizabeth Closs Traugott 2005 Lexicalization and language change. Cambridge: Cambridge University Press. Brown, Lesley 1993, ed.

The New Shorter Oxford English Dictionary on Historical Principles. Oxford: Clarendon Press.

References

381

Brown, Penelope and Stephen Levinson 1978 “Universals in language usage: Politeness phenomena”. In E. N. Goody, ed. Questions and politeness: Strategies in social interaction. Cambridge: Cambridge University Press. 56-289. 1987 Politeness: Some Universals in Language Usage. Cambridge: Cambridge University Press. Bublitz, Wolfram 1996 “Semantic prosody and cohesive company: ‘somewhat predictable’”. Leuvense Bijdragen 85: 1-32. 2003 “Emotive prosody: how attitudinal frames help construct context”. In Ewald Mengel, Hans-Jörg Schmid and Michael Steppat, eds. Proceedings of Anglistentag 2002 Bayreuth. Trier: WVT.381-391. Bybee, Joan 2003a “Cognitive processes in grammaticalization”. In Michael Tomasello, ed. The New Psychology of Language, volume 2. New Jersey: Lawrence Erlbaum. 145-167. 2003b “Mechanisms of change in grammaticization: The role of frequency”. In Brian D. Joseph and Richard D. Janda, eds. The Handbook of Historical Linguistics. Oxford: Blackwell. 602-623 Bybee, Joan, Revere D. Perkins and William Pagliuca 1994 The evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. Chicago/London: University of Chicago Press. Bybee, Joan and Sandra Thompson 2000 “Three frequency effects in syntax”. Berkeley Linguistic Society 23: 378-388. Campbell, Lyle 2001

“What’s wrong with grammaticalization?”. Language Sciences 23: 113-161. (online access via Science Direct)

Campbell, Lyle and Richard Janda 2001 “Introduction: Conceptions of grammaticalization and their problems”. Language Sciences 23: 93- 112 (online access via Science Direct) Cappelle, Bert 2005

Particle patterns in English: a comprehensive coverage. Unpublished PhD Thesis. University of Leuven.

382References Chafe, Wallace 2003

“Language and the flow of thought”. In Michael Tomasello, ed. The New Psychology of Language: Cognitive and Functional approaches to Language Structure. New Jersey: Lawrence Erlbaum. 93-111.

Chomsky, Noam 1970 “Remarks on Nominalization”. In Roderick A. Jacobs and Peter S. Rosenbaum, eds. Readings in English Transformational Grammar. Waltham, Mass.: Ginn. 184-221. Church, Kenneth and Patrick Hanks 1990 “Word association norms, mutual information and lexicography”. Computational Linguistics 16 (1): 22-29. Church, Kenneth, William Gale, Patrick Hanks and Donald Hindle 1991 “Using statistics in lexical analysis”. In Uri Zernik, ed. Lexical Acquisition. New Jersey: Lawrence Erlbaum. 115-164. Clear, Jeremy 1993

Croft, William 2000 2001

2006

“From Firth principles: Computational tools for the study of collocation”. In Mona Baker, Gill Francis and Elena TogniniBonelli, eds. Text and Technology. In Honour of John Sinclair. Philadelphia/Amsterdam: John Benjamins. 271-292.

Explaining Language Change: An Evolutionary Approach. Harlow: Longman. Radical Construction Grammar: Syntactic Theory in Typological Perspective. Oxford: Oxford University Press. “A Radical Construction Grammar analysis of noun phrase structure”. Plenary paper presented at HDLS 7 (The High Desert Linguistics Society), University of Mexico, 9-11 November 2006.

Croft, William and D. Alan Cruse 2004 Cognitive Linguistics. Cambridge: Cambridge University Press. Crystal, David 1991

A Dictionary of Linguistics and Phonetics. 3rd edition. Oxford: Blackwell Paperback

References

383

Davidse, Kristin 1997 Introduction to English Grammar. Course notes. University of Leuven 2004 “The interaction of identification and quantification in English determiners”. In Michel Achard and Suzanne Kemmer, eds. Language, Culture and Mind. Stanford: CSLI Publications. 507-533. De Beaugrande, Robert 1997 “Linguistics–systemic and functional. Renewing the ‘warrant’”. In Anne-Marie Simon-Vandenbergen, Kristin Davidse and Dirk Noël, eds. Reconnecting Language: morphology and syntax in functional perspectives. Amsterdam: John Benjamins. 49-70. Den Boon, Ton, Dirk Geeraerts et al., eds. 2008 Van Dale Groot Woordenboek van de Nederlandse Taal. 14th edition. Utrecht: Van Dale Lexicografie. Denison, David 1998 “Syntax”. In Suzanne Romaine, ed. The Cambridge history of the English language, vol. 4, 1776-1997. Cambridge: Cambridge University Press. 92-329. 2002 “History of the sort of construction family”. Paper presented at the Second International Conference on Construction Grammar, University of Helsinki, 7 September 2002. [Online draft version available at http://lings.In.man.ac.uk/staff/dd/papers/sortof_ iccg2.pdf.] 2005 “The grammaticalizations of sort of, kind of and type of in English”. Paper presented at NRG 3 (New Reflections on Grammaticalization 3), Santiago de Compostela, 17-20 July 2005. 2006 “Category change and gradience in the determiner system.” In Ans Van Kemenade & Bettelou Los, eds. The Handbook of the History of English. 279-304. De Smedt, Liesbeth 2005 Functions of the T-nouns kind, sort and type: a comprehensive, data-based description. Unpublished MA Thesis. University of Leuven. De Smedt, Liesbeth, Lieselotte Brems and Kristin Davidse 2007 “NP-internal functions and extended uses of the 'type' nouns kind, sort, and type: Towards a comprehensive, corpus-based description”. In Roberta Facchinetti, ed. Corpus linguistics 25 years on. Amsterdam: Rodopi. 225-255.

384References De Smet, Hendrik 2005 “A corpus of Late Modern English texts”. ICAME Journal 29: 69-82. 2007

2009

“In search of syntactic reanalysis: a usage-based approach”. Preprints of the Department of Linguistics nr. 261. University of Leuven. “Analyzing reanalysis”. Lingua 119: 1728-1755.

De Smet, Hendrik and Jean-Christophe Verstraete 2006 “Coming to terms with subjectivity”. Cognitive Linguistics 17: 365-392. De Smet, Hendrik and Lobke Ghesquière 2010 “When grammaticalization hesitates”. Paper presented at GRAMIS, 11-13 November 2010, Brussels. Diessel, Holger 2006 “Demonstratives, joint attention, and the emergence of grammar”. Cognitive Linguistics 17: 463-489 Diewald, Gabriele 2010 “On some problem areas in grammaticalization theory”. In Katerina Stathi, Elke Gehweiler & Ekkehard König, eds. Grammaticalization: Current Views and Issues. Amstedam/Philadelphia: John Benjamins. (Studies in Language Companion Series 119). 17-50 Dodge, Ellen and Abby Wright 2002 “Herds of wildebeests, flasks of vodka, heaps of trouble: an embodied construction grammar approach to English measure phrases”. Paper presented at the 28th Annual Meeting of the Berkeley Linguistics Society, 15-18 February 2002. Du Bois, John W. 2003 “Discourse and grammar”. In Michael Tomasello, ed. The New Psychology of Language: Cognitive and Functional approaches to Language Structure. New Jersey: Lawrence Erlbaum. 47-87 Erman, Britt and Ulla-Britt Kotsinas 1993 “Pragmaticalization: the case of ba' and you know”. Studier i modern sprakvetenskap 10: 76-92.

References

385

Evans, Nicholas and David Wilkins 2000 “In the mind’s ear: the semantic extensions of perception verbs in Australian languages”. Language 76: 546-592. Fillmore, Charles J. 1968 “The case for case”. In E. Bach and R.T. Harms, eds. Universals in linguistic theory. New York: Holt, Reinhart and Winston. 1-88. 1975 “An alternative to checklist theories of meaning”. In Cathy Cogen et al., eds. Proceedings of the First Annual Meeting of the Berkeley Linguistics Society. Berkeley: Berkeley Linguistics Society. 123-131. 1977 “Topics in lexical semantics.” In R. Cole, ed. Current Issues in Linguistic Theory. Bloomington: Indiana University Press. 76-138. 1985 “Frames and the Semantics of Understanding”. Quaderni di Semantica VI.2 Fillmore, Charles J., Paul Kay and Mary Catherine O’Connor 1988 “Regularity and idiomaticity in grammatical constructions: The case of let alone”. Language 64 (3): 501-538. Fischer, Olga 1997

1999 2000

2007

“The grammaticalisation of infinitival to in English compared with German and Dutch”. In Raymond Hickey and Staninslaw Puppel, eds. Language History and Linguistic Modelling: a Festschrift for Jacek Fisiak. Berlin: Mouton de Gruyter. 265-280. “Grammaticalisation: Unidirectional, nonreversible? The case of to before the infinitive in English”. Views 7: 5-24. “Grammaticalisation: Unidirectional, non-reversable? The case of to before the infinitive in English”. In Olga Fischer, Anette Rosenbach and Dieter Stein, eds. Pathways of Change: Grammaticalization in English. Amsterdam: John Benjamins. 149-169. Morphosyntactic Change: Functional and Formal Perspectives. Oxford: Oxford University Press.

Fitzmaurice, Susan 2000 “Remarks on the degrammaticalization of infinitival to in present-day American English”. In Olga Fischer, Anette Rosenbach and Dieter Stein, eds. Pathways of Change: Grammaticalization in English. Amsterdam: Benjamins. 171-186.

386References Foolen, Ad 2004

Francis, Gill 1993

1995

“Expressive binominal NPs in Germanic and Romance languages”. In Günter Radden & KlausǦUwe Panther, eds. Studies in Linguistic Motivation. Berlin: Mouton de Gruyter. 75Ǧ100.

“A Corpus-Driven Approach to Grammar: Principles, Methods and Examples”. In Mona Baker, Gill Francis and Elena TogniniBonelli, eds. Text and Technology. In Honour of John Sinclair. Philadelphia/Amsterdam: John Benjamins. 137-156. “Corpus-driven grammar and its relevance to the learning of English in a cross-cultural situation”. In A. Pakin, ed. English in Education: Multicultural perspectives. Singapore: Unipress.

Francis, Elaine J. & Etsuyo Yuasa 2008 “A multi–modular approach to gradual change in grammaticalization”. Journal of Linguistics 44 (1), 45–86.

Fried, Mirjam and Jan-Ola Östman, eds. 2004 Construction Grammar in a Cross-Language Perspective. Amsterdam/Philadelphia: John Benjamins. Geeraerts, Dirk 1993

“Vagueness’s puzzles, polysemy’s vagaries”. Cognitive Linguistics 4-3: 223-272.

Ghesquière, Lobke 2010 On the subjectification and intersubjectification paths followed by the adjectives of completeness. In Kristin Davidse, Lieven Vandelanotte and Hubert Cuyckens (eds.). Subjectification, intersubjectification and grammaticalization. (Topics in English Linguistics 66) Berlin: Mouton. 277–314. Givón, Talmy 1995

Functionalism and Grammar. Amsterdam: John Benjamins

Goldberg, Adele 1989 “A unified account of the semantics of the ditransitive”. In Proceedings of the Berkeley Linguistic Society volume 15. 1991 “On the problems with lexical rule accounts of argument structure”. In Cognitive Science Conference Proceedings.

References 1992

1995

2003

2006

387

“The inherent semantics of argument structure: The case of the English ditransitive construction”. Cognitive Linguistics 3-1: 37-74. Constructions: a construction grammar approach to argument structure. Chicago: University of Chicago Press. “Words by Default: inheritance and the Persian complex predicate construction”. In E. Francis and L. Michaelis, eds. Mismatch: Form-Function Incongruity and the Architecture of Grammar. Stanford: CSLI Publications. 84-112. Constructions at Work: The Nature of Generalization in Language. New York: Oxford University Press.

Goossens, Louis 1999 “Metonymic bridges in modal shifts”. In Klaus-Uwe Panther and Günter Radden, eds. Metonymy in Language and Thought. Amsterdam: John Benjamins. 193-210. 2000 “Patterns of meaning extension, ‘parallel chaining’, subjectification, and modal shifts”. In Antonio Barcelona, ed. Metaphor and metonymy at the crossroads: A cognitive perspective. Berlin: Mouton de Gruyter. 149-169. Gries, Stefan Th. and Caroline David 2006 “This is kind of/sort of interesting: Variations in hedging in English”. Paper presented at ICAME 27 (The 27th International Computer Archive of Medieval and Modern English) in Helsinki, 24-28 May 2006. Grose, Francis 1823

Dictionary of the Vulgar Tongue. London: S. Hooper.

Haiman, John, ed. 1985 Iconicity in Syntax. Proceedings of a Symposium on Iconicity in Syntax (June 24-6-1983). Amsterdam: John Benjamins. Haiman, John 1994

“Ritualization and the development of language”. In William Pagliuca, ed. Perspectives on Grammaticalization. Amsterdam: John Benjamins. 3-28.

388References Halliday, Michael A.K. 1978 Language as social semiotic. The social interpretation of language and meaning. London: Arnold. 1985a An introduction to Functional Grammar. London: Arnold. 1985b Spoken and written language. Deakin: Deakin University Press. 1994 An Introduction to Functional Grammar. 2nd edition. London: Arnold. Halliday, Michael A.K. & John R. Martin 1993 Writing Science: Literacy and Discursive Power. London: Routledge. Halliday, Michael A.K. and Zoe James 1993 “A quantitative study of polarity and primary tense in the English finite clause”. In John Sinclair, Michael Hoey and Gwyneth Fox, eds. Techniques of Description. Spoken and written discourse. London: Routledge. 32-66. Halliday, Michael A.K and Christian M.I.M. Matthiessen 2004 An Introduction to Functional Grammar. London: Arnold. Harris, Alice C. and Lyle Campbell 1995 Historical syntax in cross-linguistic perspective. Cambridge: Cambridge University Press. Haspelmath, Martin 1989 “From purposive to infinitive- A universal path of grammaticalization”. Folia Linguistica Historica 10: 287-310. 1998 “Does grammaticalization need reanalysis?” Studies in Language 22 (2): 315-351. 2004 “On directionality in language change with particular reference to grammaticalization”. In Olga Fischer, Muriel Norde and Harry Perridon, eds. Up and down the cline: The nature of grammaticalization. Amsterdam: John Benjamins. 17-44 Heine, Bernd 1993 2003

Auxiliaries: Cognitive Forces and Grammaticalization. New York: Oxford University Press. “Grammaticalization”. In Brian D. Joseph and Richard D. Janda, eds. The Handbook of Historical Linguistics. Oxford: Blackwell. 575-601.

References

389

Heine, Bernd and Mechthild Reh 1984 Grammaticalization and Reanalysis in African Languages. Hamburg: Helmut Buske. Heine, Bernd, Ulrike Claudi and Friederike Hünnemeyer 1991 Grammaticalization: a Conceptual Framework. Chicago: University of Chicago. Hilpert, Martin 2008

Germanic Future Constructions: A Usage-based Approach to grammaticalization. Constructional Approaches to Language. Amsterdam: John Benjamins.

Himmelmann, Nikolaus P. 2004 “Lexicalization and grammaticization: Opposite or orthogonal?” In Walter Bisang, Nikolaus P. Himmelmann and Björn Wiemer, eds. What makes Grammaticalization? A Look from its Fringes and its Components. Berlin/New York: Mouton de Gruyter. 21-42. Hoey, Michael 1998

“Introducing Applied Linguistics: 25 Years on”. Paper presented at The 31st BAAL Annual Meeting: Language and Literacies. The University of Manchester.

Hoffmann, Sebastian 2004 “Are low-frequency complex prepositions grammaticalized?” In Hans Lindquist and Christian Mair, eds. Corpus Approaches to Grammaticalization in English. Amsterdam/Philadelphia: John Benjamins. 171-210. Hopper, Paul J. 1987

1991

“Emergent Grammar”. In Jon Aske, Natasha Berry, Laura Michaelis and Hana Filip, eds. Berkeley Linguistics Society 13: General Session and Parasession on Grammar and Cognition. Berkeley: Berkely Linguistics Society. 139-157. “On Some Principles of Grammaticization”. In Elizabeth Closs Traugott and Bernd Heine, eds. Approaches to Grammaticalization, Vol I. Amsterdam: John Benjamins. 17-36.

Hopper, Paul J. and Sandra A. Thompson 1984 “The discourse basis for lexical categories in Universal Grammar”. Language 60-4: 703-752.

390References Hopper, Paul J. and Elizabeth Closs Traugott 2003 [1993] Grammaticalization. Second Edition. [First Edition]. Cambridge: Cambridge University Press. Huddleston, Rodney and Geoffrey K. Pullum 2002 The Cambridge Grammar of the English Language. Cambridge: Cambridge University Press. Hudson, Richard A. 1971 English Complex Sentences: An Introduction to Systemic Grammar. Amsterdam/London: North-Holland Publishing Company. 1987 “Zwicky on heads”. Journal of Linguistics 23: 109-132. Hunston, Susan and Gill Francis 2000 Pattern Grammar:. A Corpus-driven Approach to the Lexical Grammar of English Amsterdam: John Benjamins. Hunston, Susan and Geoff Thompson, eds. 2001 Evaluation in text: authorial stance and the construction of discourse. Oxford: Oxford University Press. Israel, Michael 1996 2004

“Polarity sensitivity as lexical semantics”. Linguistics and Philosophy 19: 619-666. “The pragmatics of polarity”. In Laurence Robert Horn and Gregory Ward, eds. The Handbook of Pragmatics. Oxford/Malden, MA: Blackwell. 701-723

Jackendoff, Ray 1968 “Speculations on presentences and determiners”. Bloomington: Indiana University. Linguistics club. 1981 X Syntax : A Study of Phrase Structure. Cambridge: MIT press. Jespersen, Otto 1932 1970

Essentials of English Grammar. London: Allen and Unwin. A Modern English Grammar on Historical Principles. London: Allen and Unwin.

References

391

Johnson, Christopher 2000 “Review of Pattern grammar: a corpus-driven approach to the lexical grammar of English by Susan Hunston and Gill Francis. John Benjamins 2000”. Computational Linguistics 27 (2): 318320. Joosten, Frank 2003

Collectiva en Aggregaatsnamen in het Nederlands: Begripsbepaling en Typologie. Unpublished PhD thesis. University of Leuven.

Joosten, Frank, Gert De Sutter, Denis Drieghe, Stef Grondelaers, Robert J. Hartsuiker and Dirk Speelman 2007 “Dutch collective nouns and conceptual profiling”. Linguistics 45-1: 85-132. Joseph, Brian D. 1992 “Diachronic Explanation: Putting Speakers Back into the Picture”. In Garry W. Davis and Gregory K. Iverson, eds. Explanation in Historical Linguistics (Current Issues in Linguistic Theory, 84). Amsterdam: John Benjamins. 123-144. 2001 “Is there such a thing as ‘grammaticalization’?”. Language Sciences 23: 163-186. 2007 “Why I don’t believe in grammaticalization (as it is usually portrayed) or: Why I only kinda/kind of believe in it”. Plenary lecture at the LOT Summer School, Leuven, 10 June 2007. Kay, Paul 1997

“Notes on Argument Structure Constructions”. Unpublished manuscript, University of California, Berkeley.

Kay, Paul and Charles J. Fillmore 1999 “Grammatical constructions and linguistic generalizations: The What’s X doing Y? construction”. Language 75 (1): 1-33. Keizer, Evelien 2001 “A classification of sort/kind/type-constructions”. Ms. University College London.

392References Keller, Rudi 1990 1994

Kroch, A. 1989

Sprachwandel: Von der unsichtbaren Hand in der Sprache. Tübingen: Francke. Language change: The Invisible Hand in Language (translation of Keller 1990)

“Reflexes of Grammar in Patterns of Language Change”. Language Variation and Change 1:199-244.

Krug, Manfred G. 2000 Emerging English Modals: A Corpus-Based Study of Grammaticalization. Berlin/New York: Mouton de Gruyter. Kruisinga, Etsko 1925 A Handbook of Present-day English. 4th edition. Utrecht: Kemink en Zoon. 1932 A Handbook of Present-day English. 5th edition. Groningen: Noordhoff. Kurtböke, Petek 2001 “YAP- and OL- as delexical nominalising devices in diaspora Turkish”. Paper presented at the University of Leuven, 15 March 2001. Kuteva, Tania 2000

Laffut, An 1998

Lakoff, George 1977

Auxiliation: an Enquiry into the Nature of Grammaticalization. Oxford: Oxford University Press.

“A quantitative approach to the locative alternation.” Belgian Essays on Language and Literature. 83-94.

“Linguistic gestalts”. In Beach W.A., Fox S.E., Philosoph S., eds. Papers from the Thirteenth Regional Meeting, Chicago Linguistic Society, University of Chicago, Chicago, Illinois 1977. 236-287.

Lambrecht, Knud 1994 Information Structure and Sentence Form: Topic, Focus, and the Mental Representations of Discourse Referents. Cambridge: Cambridge University Press. 2004 “Focus-structure construction and language typology: the case of spoken French”. Paper presented at the Third International Conference on Construction Grammar, Marseilles, July 2004.

References

393

Langacker, Ronald W. 1977 “Syntactic reanalysis”. In Charles N. Li, ed. Mechanisms of Syntactic Change. Austin: University of Texas Press. 59-139. 1985

1987

1990 1991

1998

1999

2002

2003

2005

2009

to appear Lasersohn, P. 1999

“Observations and speculations on subjectivity”. In John Haiman, ed. Iconicity in Syntax. Amsterdam: John Benjamins. 109-150. Foundations of Cognitive Grammar. Volume 1: Theoretical Prerequisites. Stanford: Stanford University Press. “Subjectification”. Cognitive Linguistics 1-1: 5-38. Foundations of Cognitive Grammar. Volume 2: Descriptive Application. Stanford: Stanford University Press. “On subjectification and grammaticization”. In Jean-Pierre Koenig, ed. Discourse and Cognition: Bridging the Gap. Stanford: CSLI Publications. 71-89. “Losing control: grammaticization, subjectification and transparency”. In Andreas Blank and Peter Koch, eds. Historical Semantics and Cognition. Berlin/New York: Mouton de Gruyter. 147-175. “Deixis and subjectivity”. In Frank Brisard, ed. Grounding: The Epistemic Footing of Deixis and Reference. Berlin/New York: Mouton de Gruyter. 1-28 “Extreme subjectification: English tense and modals”. In Hubert Cuyckens, Thomas Berg, René Dirven and Klaus-Uwe Panther, eds. Motivations in Language: Studies in Honor of Günter Radden. Amsterdam/Philadelphia: John Benjamins. 3-26. “Integration, Grammaticization, and Constructional Meaning”. In Mirjam Fried and Hans C. Boas, eds. Grammatical Constructions: Back to the Roots. Constructional Approaches to Language 4. Amsterdam/Philidelphia: John Benjamins. 157-189. Investigations in Cognitive Grammar. Cognitive linguistics research 42. Berlin/New York: Mouton de Gruyter. “A lot of quantifiers”.

“Pragmatic Halos”. Language 75 (3): 522-551.

394References Lavrysen, Joke, Eva Nys and Evy Vreven 2005 “Type noun constructions”. Term paper. Linguistics Department. University of Leuven. Lehmann, Christian 1985 “Grammaticalization: Synchronic variation and diachronic change”. Lingua e Stile 20: 303-318. 1995 [1982] Thoughts on Grammaticalization. Munich: Lincom Europa. [First published as Thoughts on grammaticalization: a Programmatic Sketch, no. 48 in the series Arbeiten des Kölner Universalien-Projektes, University of Cologne, Institut für Sparchwissenschaft, 1982] 2004 Theory and method in grammaticalization. Zeitschrift für germanistische Linguistik. Volume 32 (2): 152–187 2005 “Theory and method in grammaticalization”. In Gabriele Diewald, ed. Grammatikalisierung. Berlin: Mouton de Gruyter. 152-187. 2008 “Information structure and grammaticalization.” In Elena Seoane, & Maria José Lopez Couso, eds., Theoretical and empirical issues in grammaticalization. (Typological Studies in Language, 77). Amsterdam: John Benjamins. 207-229. Lehrer, Adrienne 1986 “English classifier constructions”. Lingua 66: 109-148. Lessau, Donald A. 1994 A Dictionary of Grammaticalization. Bochum: Brockmeyer. Lindquist, Hans and Christian Mair 2004, eds. Corpus Approaches to Grammaticalization in English. Amsterdam/Philadelphia: John Benjamins. Lorenz, Gunter 2002

Los, Bettelou 1999

“Really worthwhile or not really significant?: a corpus-based approach to delexicalization and grammaticalization of intensifiers in Modern English”. In Ilse Wischer and Gabrielle Diewald, eds. New Reflections on Grammaticalization. Amsterdam/Philadelphia: John Benjamins. 143-161.

Infinitival Complementation in Old and Middle English. The Hague: Thesus, Holland Academic Graphics.

References Louw, Bill 1993

Lucy, John A. 2000

395

“Irony in the Text or Insincerity in the Writer? The Diagnostic Potential of Semantic Prosodies”. In Mona Baker, Gill Francis and Elena Tognini-Bonelli, eds. Text and Technology. In Honour of John Sinclair. Philadelphia/Amsterdam: John Benjamins. 157-176.

“Systems of Nominal Classification: a concluding discussion”. In Gunter Senft, ed. Systems of Nominal Classification. Cambridge: Cambridge University Press. 326-341.

Mackenzie, Lachlan J. 1997 “Grammar, discourse and knowledge: the use of such”. In J. Aarts, I. de Mönnink and H. Wekker, eds. Studies in English language and teaching. In honour of Flor Aarts. Amsterdam: Rodopi. 85-105. Mair, Christian 1997

“The spread of the going-to future in written English: a corpusbased investigation into language change in progress”. In Raymond Hickey and Staninslaw Puppel, eds. Language History and Linguistic Modelling: a Festschrift for Jacek Fisiak. Berlin: Mouton de Gruyter. 1537-43.

Margerie, Hélène 2010 ‘‘On the Rise of Subjective Meaning in the Grammaticalisation of kind of/kinda.’’ In Kristin Davidse, Lieven Vandelanotte & Hubert Cuyckens, eds. Subjectification, Intersubjectification and Grammaticalization. Berlin: Mouton de Gruyter. 315-346. Martin, James R. 1980 Reference as semantic choice. Mimeo, University of Sydney. 1992 English text: systems and structures. Amsterdam/Philadelphia: John Benjamins. Martin, Willy, ed. 1991 Van Dale Groot Woordenboek Nederlands-Engels. Second edition. Utrecht/Antwerpen: Van Dale Lexicografie. Matthews, Peter H. 1981 Syntax. Cambridge: Cambridge University Press.

396References McGregor, William B. 1997 Semiotic Grammar. Oxford: Clarendon. Meehan, Teresa 1991 “It’s like, ‘What’s happening in the evolution of like?’: a theory of grammaticalization”. Kansas working papers in linguistics 16: 37-51. Meillet, Antoine 1912 L’évolution des formes grammaticales. Scienta (Rivista di Scienza) 12 no. 26: 6. [Reprinted in Meillet 1958: 130-48] 1915-1916 “Le renouvellement des conjunctions.” Annuaire de l’École Pratique des Hautes Etudes. [Reprinted in Meillet, Antoine. 1958. Linguistique Historique et Linguistique Générale. Paris: Champion. 159-174] Michaelis, Laura A. 2004 “Type shifting in construction grammar: An integrated approach to aspectual coercion”. Cognitive Linguistics 15: 1-67. 2005

“Entity and Event Coercion in a Symbolic Theory of Syntax”. In Jan-Olla. Östman and Mirjam Fried, eds. Construction Grammar(s): Cognitive Grounding and Theoretical Extensions. Constructional Approaches to Language, Volume 3. Amsterdam: John Benjamins. 45-88.

Milsark, Garry Lee 1976 Existential Sentences in English. Bloomington: Indiana University. Linguistics Club. Narrog, Heiko 2010

“Modality and speaker orientation”. Plenary at GRAMIS: Grammaticalization and (Inter)subjectification, 11-13 November 2010, Brussels.

Newmeyer, Frederick J. 1998 Language Form and Language Function. Cambridge, Mass.: MIT Press.

References Noël, Dirk 2005

2007

Norde, Muriel 2009

Paradis, Carita 2000

397

“The productivity of a ‘source of information’ construction: or, where grammaticalization theory and construction grammar meet”. Paper presented at FITIGRA (From Ideational to Interpersonal: perspectives from grammaticalization), University of Leuven, 10-12 February 2005. “Diachronic construction grammar and grammaticalization theory”. Functions of Language 14: 177-202.

Degrammaticalization. Oxford: Oxford University Press. OED (Oxford English Dictionary, 3rd edition) online: http://www.oed.com

“Reinforcing adjectives: a cognitive sematic perspective on grammaticalization”. In Ricardo Bermúdez-Otero, David Denison, Richard M. Hogg and C.B. McCully, eds. Generative Theory and Corpus Linguistics: A Dialogue from 10 ICEHL. Berlin/New York: Mouton de Gruyter. 233-258.

Parkes, M. B. “On the Presumed Date and Possible Origin of the Manuscript of the Orrmulum”, in Five Hundred Years of Words and Sounds: A Festschrift for Eric Dobson. E. G. Stanley and Douglas Gray, eds. Cambridge: D.S. Brewer, 1983. 15–27. Peters, Pam, Peter Collins and Adam Smith, eds. 2009 Comparative studies in Australian and New Zealand English: grammar and beyond. Amsterdam: John Benjamins. Peters, Stanley and Dag Westerståhl 2006 Quantifiers in language and logic. Oxford: Clarendon. Polinsky, Maria 1998 “A non-syntactic account of some asymmetries in the double object construction”. In Jean-Pierre Koenig, ed. Conceptual Structure and Language: Bridging the Gap. Stanford: CSLI Publications. 401-420.

398References Poutsma, Hendrik 1928 A Grammar of Late Modern English: 1. The sentence - First half: The elements of the sentence. 2nd edition. Groningen: Noordhoff. Quirk, Randolph, Sidney Greenbaum, Geoffrey Leech and Jan Svartvik 1972 A Grammar of Contemporary English. London/ New York: Longman. 1985 A Comprehensive Grammar of the English Language. London/ New York: Longman. Ramat, Paolo 1982

“Ein Beispiel von ‘reanalysis’ typologisch betrachtet’. Functions of Language 16: 365—383.

Robins, Robert Henry 1971 General Linguistics. An Introductory Survey. London: Longman. Sag, Ivan 2010

“Sign-based construction grammar: An informal synopsis”. In Hans C. Boas & Ivan A. Sag, eds. Sign-based Construction Grammar. Centre for the Study of Language and Information. 36-170. (available online: http://lingo.stanford.edu/sag/papers/ theo-syno.pdf)

Sankoff, Gillian 1990 “The grammaticalization of tense and aspect in Tok Pisin and Sranan”. Language Variation and Change 2: 295-312. Schank, Roger C. and Robert P. Abelson 1977 Scripts, plans, goals and understanding. New Jersey: Lawrence Erlbaum. Schwenter, Scott A. and Elizabeth Closs Traugott 1995 “The semantic and pragmatic development of Substitutive Complex Prepositions in English”. In Andreas Jucker, ed. Historical Pragmatics. Amsterdam: John Benjamins. 243-273. Selkirk, Elisabeth 1977 “Some remarks on noun phrase structure”. In Peter S. Culicover, Thomas Wasow and Adrian Akmajian, eds. Formal Syntax. New York: Academic Press. 285-316.

References

399

Sinclair, John 1991 [1992] Corpus, concordance, collocation. [Second edition]. Oxford: Oxford University Press. 1997 Collins Cobuild English Grammar. London: HarperCollins Publishers.

Sinclair, John and Antoinette Renouf 1991 “Collocational frameworks in English”. In Karin Aijmer and Bengt Altenberg, eds. English corpus linguistics : studies in honour of Jan Svartvik. London: Longman. 128-144. Sinclair, John, Founding Editor-in-Chief 2001 Collins Cobuild English Dictionary for Advanced Learners. Third edition. Glasgow: HarperCollins Publishers. Sperber, Dan and Deirdre Wilson 1991 “Irony and the use-mention distinction”. In Steven Davis, ed. Pragmatics: A reader. Oxford: Oxford University Press. Stubbs, Michael 1995 “Collocations and semantic profiles: On the cause of the trouble with quantitative studies”. Functions of Language (2) 1: 23-55. Sweetser, Eve 1988

“Grammaticalization and semantic bleaching”. In Shelley Axmaker, Annie Jaiser and Helen Singmaster, eds. Proceedings of the 14th Annual Meeting of the Berkeley Linguistics Society. General Session and Parasession on Grammaticalization. Berkeley: Berkeley Linguistics Society. 389-405.

Tabor, Whitney 1993 “The gradual development of degree modifier sort of and kind of: a corpus proximity model”. In Katherine Beals, Gina Cooke, David Kathman, Karl-Erik McCullough, Sotaro Kita and David Testen, eds. Papers from the 29th Regional Meeting of the Chicago Linguistic Society. Chicago: Chicago Linguistic Society. 451-465. Thompson, Sandra A. 1990 “Information flow and the dative shift in English discourse”. In J. Edmondson, K. Feagin and P. Mühlhäusler, eds. Development and Diversity: Linguistic Variation across Time and Space. Summer Institute of Linguistics. 239-253.

400References Thomspon, Della, Henry Watson Fowler and Francis Georges Fowler 1995 The Concise Oxford Dictionary of Current English. Ninth edition. Oxford: Clarendon. Traugott, Elizabeth Closs 1982 “From propositional to textual and expressive meaning: some semantic-pragmatic aspects of grammaticalization”. In Winfred P. Lehmann and Yakov Malkiel, eds. Perspectives on Historical Linguistics. Amsterdam/Philadelphia: John Benjamins. 245-271. 1988 “Pragmatic strengthening and grammaticalization”. In Shelley Axmaker, Annie Jaisser and Helen Singmaster, eds. Proceedings of the Fourteenth Annual Meeting of the Berkeley Linguistics Society. Berkeley: Berkeley Linguistics Society. 406-416 1989 “On the rise of epistemic meanings in English: An example of subjectification in semantic change”. Language 65: 31-55. 1990 “From less to more situated in language: the unidirectionality of semantic change”. In Sylvia Adamson, Vivien Law, Nigel Vincent and Susan Wright, eds. Papers from the Fifth International Conference on English Historical Linguistics. Amsterdam/Philadelphia: John Benjamins. 497-517. 1994 “Grammaticalization and lexicalization”. In R.E. Asher et al., eds. The Encyclopedia of Language and Linguistics. Volume 3. Oxford/New York/Seoul/Tokyo: Pergamon Press. 1995a “Subjectification in grammaticalization”. In Dieter Stein and Susan Wright, eds. Subjectivity and Subjectivisation: Linguistic Perspectives. Cambridge: Cambridge University Press. 31-54. 1995b “The role of the development of discourse markers in a theory of grammaticalization”. Paper presented at ICHL XII, Manchester 1995. 1996 “Grammaticalization and lexicalization”. In Keith Brown and Jim Miller, eds. Concise Encyclopedia of Syntactic Theories. Oxford: Pergamon. 181-7. 2003 “Constructions in grammaticalization”. In Brian D. Joseph and Richard D. Janda, eds. The Handbook of Historical Linguistics. Oxford: Blackwell. 624-647. 2006 “Constructions and language change revisited: the concepts of constructional emergence and coercion from the perspective of grammaticalization”. Paper presented at CGN 3, Düsseldorf, 1-2 April 2006.

References

401

2007

“The concepts of constructional mismatch and type-shifting from the perspective of grammaticalization”, Cognitive Linguistics 18: 523-557. 2008a “Grammaticalization, constructions and the incremental development of language: Suggestions from the development of degree modifiers in English". In Regine Eckardt, Gerhard Jäger & Tonjes Veenstra, eds. Variation, Selection, Development-Probing the Evolutionary Model of Language Change. Berlin/New York: Mouton de Gruyter. 219-250. 2008b “The grammaticalization of NP of NP constructions”. In Alexander Bergs & Gabriele Diewald, eds. Constructions and Language Change . Berlin/New York: Mouton de Gruyter. 21-43. 2008c Workshop on gradience, gradualness and emergence in grammaticalization. Comments by Elizabeth Closs Traugott. Presented at NRG 4 (New Reflections on Grammaticalization 4), University of Leuven, 16-19 July 2008. 2010a “Dialogic contexts as Motivations for Syntactic Change”. In Robert A. Cloutier, Anne Marie Hamilton-Brehm, and William A Jr.,Kretzschmar, eds. Studies in the History of the English Language V: Variation and Change in English Grammar and Lexicon: Contemporary Approaches. Berlin/New York: Mouton de Gruyter. 11-27. 2010b “From ideational to interpersonal: a reassessment”. In Hubert Cuyckens, Kristin Davidse and Lieven Vandelanotte, eds. Subjectification, Intersubjectification and Grammaticalization. Berlin/New York: Mouton de Gruyter. 29-74. Forthcoming “Intersubjectification and clause periphery”. In Lieselotte Brems, Lobke Ghesquière & Freek Van de Velde, eds. Intersections of intersubjectivity: special issue of English Text Construction. Traugott, Elizabeth Closs and Richard Dasher 2002 Regularity in Semantic Change. Cambridge: Cambridge University Press. Traugott, Elizabeth Closs and Graeme Trousdale, eds. 2010 Gradience, gradualness and grammaticalization. Typological Studies in Language series. Amsterdam: John Benjamins.

402References Trousdale, Graeme 2008a “Constructions in grammaticalization and lexicalization: evidence from the history of a composite predicate construction in English”. In Graeme Trousdale & Nikolas Gisborne, eds. Constructional Approaches to English Grammar [Topics in English Linguistics 57]. Berlin: Mouton de Gruyter. 33-68 2008b “Words and constructions in grammaticalization: the end of the English impersonal construction”. In Susan Fitzmaurice & Donka Minkova, eds. Studies in the History of the English Language IV: Empirical and Analytical Advances in the Study of English Language Change [Topics in English Linguistics 61]. Berlin: Mouton de Gruyter. 310-326. 2008c “A constructional approach to lexicalization processes in the history of English: evidence from possessive constructions”. Word Structure 1: 156-177. 2010a Issues in constructional approaches to grammaticalization in English. In Katerina Stathi, Elke Gehweiler & Ekkehard König (eds), Grammaticalization: Current Views and Issues [Studies in Language Companion Series 119]. Amsterdam: Benjamins. 2010b “Binominal constructions in nonǦstandard dialects”. Plenary paper presented in the workshop Binominal syntagms as a neglected locus of synchronic variation and diachronic change: Towards a unified approach at SLE 43, Vilnius, 2-5 September 2010. Forthcoming “Grammaticalization, constructions and the grammaticalization of constructions”. In Tine Breban, Lieselotte Brems, Kristin Davidse & Tanja Mortelmans, eds. proceedings of NRG 4 (New Reflections on Grammaticalization). Studies in Language Companion Series. John Benjamins. Vandelanotte, Lieven 2005 Types of speech and thought representation in English: Syntagmatic structure, deixis and expressivity, semantics. Unpublished PhD thesis. University of Leuven. Vanden Eynde, Martine 2003 Grammaticalization and delexicalization of edge-noun expressions. Unpublished MA thesis. University of Leuven. Van Langendonck, Willy 2007 Theory and typology of proper names. Berlin: Mouton de Gruyter.

References

403

Vincent, Diane, Sebastião Votre and Marty LaForest 1993 “Grammaticalisation et postgrammaticalisation”. Languages et Linguistique 19: 71-103. Vos, Hinderika Margaretha 1999 A grammar of partitive constructions. Published PhD thesis, Tilburg University. Ward, Gregory and Betty Birner 1995 “Definiteness and the English existential”. Language 71.4: 722-742. Willemse, Peter 2005 Nominal reference point constructions: possessive and esphoric NPs in English. Unpublished PhD Thesis. University of Leuven. 2007 “Indefinite possessive NPs and the distinction between determining and non-determining genitives in English”. English Language and Linguistics 11(3). Willemse, Peter, Kristin Davidse and Lieselotte Brems 2007 “The development of extended type noun uses: a comparison between English sort/kind and French sorte/espèce. Paper presented at ICAME 28 (The 28th International Computer Archive of Medieval and Modern English), Stratford-upon-Avon, 23-27 May 2007. Wischer, Ilse 2000

“Grammaticalization versus lexicalization – ‘methinks’ there is some confusion”. In Olga Fischer, Anette Rosenbach and Dieter Stein, eds. Pathways of Change: Grammaticalization in English. Amsterdam: John Benjamins. 355-370.

Index

absolute quantification, 209 adverbial qualifying, 308, 313 alotta, 99, 111, 115, 159 ambiguity ambiguous, 99, 100, 147, 152, 167, 173, 212, 322, 348, 359 ambivalence ambivalent, 6, 8, 12, 37, 46, 84, 192, 320, 322, 325 American English, 7, 28, 77, 116, 126, 145, 150, 169, 180, 254, 373 analogy analog, 8, 13, 16, 81, 96, 100, 111, 113, 117, 121, 125, 236, 240, 244, 265, 268, 358, 367 anaphoric, 6, 30, 76, 89, 193, 293, 295, 298, 306, 339, 348 animate nouns, 128, 131, 133, 138, 143, 151, 163, 167, 169, 172, 179, 181, 186, 212, 214, 222, 226 attitudinal subjectification attitudinal subjectivity, 11, 107, 126, 174, 176, 183, 189, 228, 231, 235, 250, 271, 353, 356, 361 attraction, 21, 23, 101 attributive use, 37, 38, 346 attrition, 82 Australian English, 126, 145, 150, 169, 239 binominal, 2, 9, 12, 20, 30, 37, 57, 90, 98, 123, 133, 136, 150, 162, 179, 188, 208, 217, 229, 234, 270, 280, 292, 305, 327, 329, 330, 335, 342, 344, 349, 353, 359, 374, 400 bleaching, 83, 89, 97, 317, 341, 344, 387

blend blends, 200, 321, 325, 350 bonding, 34, 185 booster, 312 boundedness, 105, 133, 164, 165 bridging contexts, 95, 99, 100, 212, 274, 320, 325 British English, 7, 23, 28, 77, 105, 126, 145, 150, 169, 172, 272, 331, 356 canonical codings, 41, 47, 49, 59 cataphoric, 295, 300, 306, 339, 340, 348, 358 categorization, 2, 4, 16, 22, 41, 55, 59, 98, 250, 271, 280, 286, 292, 305, 309, 311, 331, 339, 343, 345, 351 classification, 2, 12, 16, 18, 23, 27, 29, 56, 246, 271, 274, 284, 333, 355, 379 classifying adjective classifying adjectives, 55 clines, 65, 82, 205, 208, 215, 331, 349 coalescence coalesced spelling, 82, 85, 94, 116, 119, 191, 240, 245, 306 coercion, 69, 384, 388 coerce, 69 coextensiveness, 128 cognitive-functional model, 8 collective nouns, 23, 209, 379 colligation, 62, 73, 307, 357 collocation collocates, 2, 8, 74, 115, 118, 166, 178, 274, 278, 350, 370, 387

Index collocationally constrained constructions, 71, 191, 194, 201, 229, 307, 358 completion, 86, 120, 232, 266, 359 complex determiner, 25, 30, 33, 110, 305, 338 compositionality, 74 concord concord patterns, 6, 10, 19, 27, 46, 76, 129, 135, 162, 210, 212, 214, 229, 246, 286, 293, 300, 326, 346 condensation, 82 conjugational relations, 56, 58 constellation, 5, 10, 42, 71, 76, 108, 128, 133, 143, 145, 155, 161, 178, 183, 208, 214 constituency constituency tests, 26, 43, 46, 52, 55, 67, 71, 98, 129, 229 constraints, 25, 38, 61, 65, 68, 70, 75, 105, 120, 237, 307, 316 construct, 65, 71, 369 constructs, 70, 90, 102, 120, 136, 138, 143, 169, 211, 246, 358 construction, 58, 60, 61,70, 89, 93 construction grammar, 358, 385 constructional family, 52 constructional gradience, 320, 325, 326 constructional grammaticalization, 89, 270 context expansion, 88, 113, 120, 230 contiguity, 12, 142, 145, 180, 181, 183, 213 conversational implicature, 33, 346 corpus-based grammars, 19 CxG framework, 60, 86 decategorialization, 24, 85, 91, 94, 95, 103, 111, 115, 119, 124, 143, 159, 172, 191, 194, 200, 230, 232, 241, 246, 326, 340, 341, 344, 359 partial decategorialization, 84 degree of grammaticality

405

degree of grammaticalization, 126, 132, 155, 159, 203, 232, 245, 360 delexicalization, 7, 9, 12, 37, 40, 76, 79, 91, 95, 103, 116, 119, 130, 138, 146, 155, 164, 167, 176, 181, 189, 203, 229, 236, 238, 243, 245, 248, 265, 331, 359, 367, 382, 390 dependency, 41, 44, 55, 67, 70, 78, 98, 358 descriptive modifier construction, 346, 355 desemanticization, 85 determination, 4, 7, 22, 38, 47, 56, 59, 61, 71, 75, 271, 274, 280, 286, 305, 331, 338, 357 diagnostics, 80, 82, 96, 113, 191 diminutive, 124, 126, 217, 223, 225, 227, 228 disambiguation, 6, 100, 130 discourse discourse markers, 14, 33, 68, 77, 82, 265, 272, 287, 293, 298, 303, 325, 332, 338, 345, 348, 352, 355, 359, 376, 383, 387 divergence, 23, 64, 84 downtoning, 13, 34, 237, 248, 266, 303, 314 emergent grammar, 79, 92 emotive prosody. See semantic prosody enrich target-paradigms, 112 enrichment pragmatic enrichment, 34, 84, 90, 98, 108, 113, 121, 231, 342, 359 erosion, 85, 89, 94, 103, 111, 115, 117, 212 evaluation, 9, 111, 146, 149, 175, 190, 198, 201, 230, 237, 250, 267, 334, 353 evidential marker, 34 experiential structure, 42 expressive, 9, 12, 80, 84, 94, 106, 111, 117, 125, 139, 141, 144, 147,

406Index 169, 171, 183, 196, 225, 230, 237, 266, 291, 301, 354, 388 extension, 2, 6, 12, 27, 33, 53, 59, 61, 89, 98, 100, 114, 120, 127, 131, 138, 143, 159, 168, 188, 205, 218, 227, 235, 246, 251, 262, 270, 336, 345, 364, 371, 391 extravagance. See expressive factors, 4, 12, 39, 60, 76, 82, 91, 103, 117, 124, 129, 132, 156, 191, 203, 233, 241, 260, 268, 325, 333, 341, 357, 360, 368 filler, 18, 82 fixation, 82, 83, 119 Frame Semantics, 66 framing, 58, 59, 70, 292 frequency, 8, 13, 24, 75, 115, 125, 127, 130, 136, 145, 155, 160, 168, 186, 236, 240, 248, 260, 265, 273, 321, 334, 341, 350, 367, 377 functional overlap, 2, 7, 16, 19, 325, 327, 330, 357 generalization. See generalized reference generalized reference generalizing reference, 334 generative grammar, 26, 62 generic, 16, 47, 276, 277, 280, 286, 290, 297, 304, 322, 337, 342, 348, 351 gob, 9 grammaticalization, 2, 7, 79-94 ongoing grammaticalization, 24 grounding, 47, 49, 52, 55, 108, 110, 112, 303, 310 headness, 57 heapsa, 140, 150 hedge hedging, 5, 32, 34, 110, 225, 250, 308, 357 hella, 9 hesitant grammaticalization, 120 hesitation marker, 18

iconicity, 77, 199 ideational, 42, 84, 131, 155, 160, 249, 389 identification, 6, 18, 38, 43, 271, 294, 300, 338, 371 idiomatization, 104, 118, 125, 266 idioms, 61, 71 implicature implicatures, 77, 187, 212, 281, 304, 342 incongruence, 30, 32, 36, 305 individualization, 20, 21 infrequency, 13, 15, 116, 124, 157, 216, 236, 240, 247, 251, 255, 263, 268, 274, 283, 288, 341, 358 inheritance network, 65 instantiation, 27, 42, 47, 48, 49, 50 intensifying, 13, 112, 115, 117, 120, 141, 143, 147, 154, 185, 190, 199, 304, 312, 328, 339, 348, 354 interpersonal, 34, 42, 57, 131, 160, 288, 291, 368, 389 intersubjectification intersubjective, 105, 107, 121, 249, 374 irony, 30, 148, 171, 182, 247, 288, 311, 329, 345 kinda, 17, 34, 39, 77, 99, 114, 270, 274, 286, 290, 301, 304, 315, 319, 331, 337, 379, 383 landmark, 26, 50, 51 layering, 2, 81, 84, 91, 94, 96, 103, 105, 126, 159, 228, 270, 326, 360 leftward movement, 331, 356 levels of constructional schematicity, 16, 90 lexicalization lexicalized, 88, 92, 96, 118, 284, 388, 390 lexicon, 62, 65, 90, 118, 236, 248 linking relations, 56, 61 logical structure, 44, 45 loss-and-gain, 85, 89, 94, 103, 111, 120, 359

Index

407

low-frequency See infrequency, 116, 236, 240, 241

259, 271, 281, 290, 302, 310, 315, 326, 333, 342, 355

manner, 32, 36, 205, 220, 241, 315 mass noun, 69 measure nominals, 42, 45 measure nouns, 21, 26, 48, 53 mechanisms, 13, 16, 41, 70, 80, 91, 94, 106, 118, 125, 147, 152, 156, 250, 270, 350 meso-construction, 118, 216, 232, 266 metafunctions, 42, 57 metaphorical, 23, 103, 130, 142, 147, 155, 163, 186, 200, 208, 210, 214, 312 metaphorization, 99, 146, 154, 214, 229, 237, 247, 265, 304, 321 metonymy, 84, 147, 151, 156, 166, 228, 261, 375 MI-score, 274, 278, 281 modification, 2, 4, 7, 22, 24, 38, 44, 55, 101, 200, 271, 280, 286, 305, 331 morphology-based approach, 82 multi-dimensional model, 68 multi-modular analysis, 46 multiple inheritance. See multiple parents multiple models, 120 multiple parents, 33, 66

obligatorification, 82 of-phrase, 5, 10, 23, 45, 54, 128, 133, 149, 155, 162, 194, 208, 214, 229, 276, 280, 326, 356 Old English, 14, 32, 216, 258, 261, 337, 364 on the brink, 4 on the edge of, 4 on the verge, 4, 95 onomatopoeia, 82, 319 Ormulum, 211 paradigmaticization, 82, 113, 119, 230 parameters, 41, 59, 65, 80, 92, 94, 111, 114, 119, 192, 243, 317, 358, 363 partially filled constructions, 60, 71, 73 particle particle uses, 26, 34, 38, 52, 120, 159, 217 partition, 22, 28 partitive, 20, 25, 52, 126, 160, 205, 209, 236, 246, 391 pathways of change, 29, 32, 38, 215, 345, 346 Pattern Grammar, 74, 378 persistence persistent, 76, 84, 94, 103, 104, 125, 132, 174, 176, 203, 243, 247, 359 phonological reduction, 97 phonology, 68 phraseologies, 75 pluralization, 203 polarity polarity sensitive, 13, 41, 77, 124, 236, 243, 246, 249, 251, 257, 259, 261, 303, 358, 376, 378 postdeterminer, 30, 35, 37, 275, 278, 284, 292, 302, 311, 321, 325, 330, 332, 335, 338, 345, 354, 367 postmodifier, 30, 33, 45, 128, 229, 276, 280, 326, 338

N of N-scheme, 27 networks, 40, 64, 65, 359 nominal qualifying use, 284, 308, 313, 333, 344, 345, 353 nominal syntax, 60, 69, 81 non-canonical coding, 9, 47 non-directionality, 119 non-replicate mass, 48, 50 NP-external, 3, 6, 12, 33, 82, 271, 345, 355 NP-internal, 3, 16, 27, 35, 81, 271, 284, 294, 332, 355, 371 number, 3, 19, 24, 30, 36, 48, 58, 66, 90, 124, 133, 138, 152, 167, 176, 180, 191, 201, 212, 227, 239, 247,

408Index pragmatic inferencing, 8, 76, 91, 98, 100, 108, 117 pragmatic strengthening, 84, 388 pragmaticalization, 372 predeterminer, 298, 301 premodification, 2, 4, 24, 30, 55, 61, 76, 85, 112, 126, 135, 143, 155, 160, 171, 179, 181, 185, 191, 198, 217, 224, 230, 239, 246, 357, 359 primary stress, 35, 279, 287, 290 propositional, 11, 42, 84, 106, 121, 231, 316, 388 prosodic features, 41, 190, 203 prosody, 4, 41, 62, 72, 95, 104, 111, 122, 131, 140, 148, 160, 164, 170, 176, 182, 186, 190, 216, 219, 222, 224, 231, 238, 247, 254, 261, 313, 325, 356, 361, 369 proximity concord, 247 qualifying construction, 30 qualitative adjective, 196, 197, 324, 341, 351 quantification, 7, 9, 19, 38, 41, 47, 51, 121, 157, 165, 175, 195, 201, 223, 226, 231, 237, 267, 271, 274, 331, 337, 350, 368, 371, 400 relative quantification, 47, 49 quantifier use, 11, 71, 102, 125, 141, 143, 150, 154, 163, 168, 181, 190, 202, 210, 213, 220, 232, 236, 245, 256, 264, 281, 323, 341 quantity, 5, 9, 22, 28, 43, 47, 57, 138, 145, 156, 164, 168, 176, 181, 190, 198, 201, 212, 232, 248, 256, 265, 266, 283, 302, 342, 349, 368 quotative marker, 18 quotative markers, 272, 319, 355 reanalysis, 8, 28, 46, 54, 73, 79, 85, 92, 100, 111, 117, 119, 212, 229, 235, 240, 245, 286, 335, 336, 341, 346, 351, 354, 356, 372, 376, 381, 386 recategorialization, 94, 112, 120, 359 referential, 10, 58, 84, 136, 162, 204, 298, 305

regional regional differences, 7, 105, 180, 189 regionally restricted, 9 register register variation, 66 registers, 7, 23, 99, 139, 157, 189, 230, 238, 272, 287, 291, 333, 356 regular SNs, 8, 12, 76, 123, 125, 234, 246, 249 relative quantifiers, 48, 52 renewal, 9, 117, 122, 125, 154, 157, 230, 348 replicate mass, 48, 50 routinization, 103, 115, 122, 230, 236, 240, 248 scale of grammaticalization, 203 scale of magnitude, 50, 52, 53, 205, 210 schematicity, 8, 61, 65, 69, 78, 90, 95, 102, 120, 125, 232, 237, 240, 266, 268, 358 scopal domain scope, 38, 317 scoping, 58, 59, 70 selectional restrictions, 26, 73, 132 semantic prosody, 74 semantic prosodies, 4, 72, 77, 104, 125, 143, 170, 173, 183, 227, 231, 247 semanticization, 108, 231, 235, 346, 348, 353 semi-lexicalized, 119, 146, 286, 290, 352 semiotic, 7, 26, 56, 57, 60, 63, 94, 103, 376 Semiotic Grammar, 56, 384 semi-stable semi-stable systems, 9, 245, 325 semi-suffix semi-suffix use, 31, 37, 272, 284, 288, 294, 301, 321, 327, 335, 340, 346, 353, 355, 357 sentential qualifying, 308, 316, 327

Index singular/plural contrast, 24, 112, 159, 232 size, 2, 9, 12, 19, 39, 48, 58, 71, 94, 103, 116, 138, 148, 152, 167, 176, 191, 194, 205, 209, 213, 231, 235, 239, 246, 295, 298, 301, 307, 328, 331, 340, 353, 357, 367 SKT-patterns SKT-constructions, 31 SKT-nouns, 28, 32 small SNs, 8, 123, 125, 234 sorta, 17, 34, 77, 99, 114, 274, 290, 304, 310, 314, 320, 331, 337 source semantics, 7, 12, 21, 85, 183, 189, 204, 217, 356 source-construction, 112, 120, 218, 358 spatio-temporally bounded, 164, 209 specialization, 84, 228 spelling, 116, 140, 167, 189, 206, 212, 239, 254, 255, 291, 327 SSNs, 9, 13, 38, 76, 95, 102, 117, 123, 234, 254, 263, 271 stress stress pattern, 5, 35 stressed, 35, 93, 243, 287 subjectification subjective, 11, 58, 71, 76, 84, 104, 111, 124, 131, 160, 176, 205, 210, 215, 246, 249, 264, 291, 312, 327, 331, 345, 353, 355, 359 subject-verb concord, 129 substitutability, 129, 229, 233, 245, 65, 326 syllable structure, 76, 190 symbolization, 89 systemic-functional approach

409

systemic-functional grammar, 22 target-construction, 102, 120, 128, 218, 358 taxonomy, 65 textual, 42, 84, 106, 311, 338, 354, 359, 388 textual subjectification, 108, 354, 355 transformational-generative grammar, 19, 25 transformationalist, 18 T-score, 274, 281 typa, 3, 17, 99, 115, 274, 287, 290, 301, 331 type specification, 5, 18, 39, 47, 55, 114, 280, 312, 322 unidirectionality unidirectional, 84, 94, 349 vague, 12, 92, 99, 130, 133, 146, 152, 160, 167, 173, 186, 197, 199, 231, 247, 253, 262, 321, 325, 350 vagueness, 100, 325, 374 vagueness marker, 29 valuing quantifier, 11, 111, 125, 131, 136, 154, 160, 169, 174, 191, 195, 203, 217, 225, 228, 357, 360 valuing quantifying uses, 145, 183 verb agreement, 212 wodge, 9 word class grammars, 19 zones of constructional overlap, 38, 271