Late Modern English Syntax
The Late Modern period is the first in the history of English for which an unprecedented wealth of textual material exists. Using increasingly sophisticated databases, the contributions in this volume explore grammatical usage from the period, specifically morphological and syntactic change, in a broad context. Some chapters explore the socio-historical background of the period while others provide information on prescriptivism, newspaper language, language contact and regional variation in British and American English. Internal processes of change are discussed against grammaticalisation theory and construction grammar and the rich body of textual evidence is used to draw inferences on the precise nature of historical change. Exposing readers to a wealth of data that informs the description of a broad range of syntactic phenomena, this book is ideal for graduate students and researchers interested in historical linguistics, corpus linguistics and language development. is Full Professor of English Linguistics in the English Department at the University of Zurich, Switzerland. M a r ianne H u n d t
S tu d ies i n E n g lish L a n g ua g e
General editor Merja Kytö (Uppsala University) Editorial Board Bas Aarts (University College London) John Algeo (University of Georgia) Susan Fitzmaurice (University of Sheffield) Christian Mair (University of Freiburg) Charles F. Meyer (University of Massachusetts) The aim of this series is to provide a framework for original studies of English, both present day and past. All books are based securely on empirical research, and represent theoretical and descriptive contributions to our knowledge of national and international varieties of English, both written and spoken. The series covers a broad range of topics and approaches, including syntax, phonology, g rammar, vocabulary, discourse, pragmatics and sociolinguistics, and is aimed at an international readership. Already published in this series: Hilde Hasselgård: Adjunct Adverbials in English Raymond Hickey: Eighteenth-Century English: Ideology and Change Charles Boberg: The English Language in Canada: Status, History and Comparative Analysis Thomas Hoffmann: Preposition Placement in English: A Usage-based Approach Claudia Claridge: Hyperbole in English: A Corpus-based Study of Exaggeration Päivi Pahta and Andreas H. Jucker (eds.): Communicating Early English Manuscripts Irma Taavitsainen and Päivi Pahta (eds.): Medical Writing in Early Modern English Colette Moore: Quoting Speech in Early English David Denison, Ricardo Bermúdez-Otero, Chris McCully and Emma Moore (eds.): Analysing Older English Jim Feist: Premodifiers in English: Their Structure and Significance Steven Jones, M. Lynne Murphy, Carita Paradis and Caroline Willners: Antonyms in English: Construals, Constructions and Canonicity Christiane Meierkord: Interactions across Englishes: Linguistic Choices in Local and International Contact Situations Haruko Momma: From Philology to English Studies: Language and Culture in the Nineteenth Century Raymond Hickey (ed.): Standards of English: Codified Varieties Around the World Benedikt Szmrecsanyi: Grammatical Variation in British English Dialects: A Study in Corpus-Based Dialectometry Daniel Schreier and Marianne Hundt (eds.): English as a Contact Language Bas Aarts, Joanne Close, Geoffrey Leech and Sean Wallis (eds.): The Verb Phrase in English: Investigating Recent Language Change with Corpora Martin Hilpert: Constructional Change in English: Developments in Allomorphy, Word Formation, and Syntax Jakob R. E. Leimgruber: Singapore English: Structure, Variation and Usage Christoph Rühlemann: Narrative in English Conversation Dagmar Deuber: English in the Caribbean: Variation, Style and Standards in Jamaica and Trinidad Jock O. Wong: The Culture of Singapore English Eva Berlage: Noun Phrase Complexity in English Nicole Dehé: Parentheticals in Spoken English: The Syntax–Prosody Relation Anita Auer, Daniel Schreier and Richard J. Watts (eds.): Letter Writing and Language Change Marianne Hundt (ed.): Late Modern English Syntax Earlier titles not listed are also available
Late Modern English Syntax
Edited by Marianne Hundt University of Zurich
University Printing House, Cambridge C B 2 8B S , United Kingdom Cambridge University Press is part of the University of Cambridge. It furthers the University’s mission by disseminating knowledge in the pursuit of education, learning and research at the highest international levels of excellence. www.cambridge.org Information on this title: www.cambridge.org/9781107032798 © Cambridge University Press 2014 This publication is in copyright. Subject to statutory exception and to the provisions of relevant collective licensing agreements, no reproduction of any part may take place without the written permission of Cambridge University Press. First published 2014 Printed in the United Kingdom by Clays, St Ives plc A catalogue record for this publication is available from the British Library Library of Congress Cataloguing in Publication data Late modern English syntax / edited by Marianne Hundt. pages cm – (Studies in English language) Includes bibliographical references and index. I S B N 978-1-107-03279-8 (hardback) 1. English language–Syntax. 2. English language–Grammar, Historical. 3. English language–20th century–History. I. Hundt, Marianne, editor. P E 1361.L 38 2014 425–dc 3 2014020708 ISBN
978-1-107-03279-8 Hardback
Cambridge University Press has no responsibility for the persistence or accuracy of U R L s for external or third-party internet websites referred to in this publication, and does not guarantee that any content on such websites is, or will remain, accurate or appropriate.
This book is dedicated to Hans-Joachim Kölbel (17.10.1931–25.3.2014)
v
Contents
List of figures List of tables List of data sources (corpora) List of contributors Acknowledgements 1 Introduction: Late Modern English syntax in its linguistic and socio-historical context
page x xiii xvii xix xxi 1
M ar ia nne Hu ndt
Part I Changes in the VP
11
2 The decline of the be -perfect, linguistic relativity, and grammar writing in the nineteenth century
13
Lieselotte A nde rwald
3 Let’s not, let’s don’t and don’t let’s in British and American English
38
An ita Aue r
4 Do we got a difference? Divergent developments of semi-auxiliary (have) got (to) in British and American English
56
Chr istia n Mai r
5 From contraction to construction? The recent life of ’ll
77
Na d ja Nesselhauf
6 Books that sell – mediopassives and the modification ‘constraint’
90
M ar ia nne Hu ndt
vii
viii Contents
Part II Changes in the NP
111
7 Beyond mere syntactic change: a micro-analytical study of various and numerous
113
T i ne B reba n
8 Culturally conditioned language change? A multivariate analysis of genitive constructions in ARCHER
133
B e ne dikt S zmrecsany i , A n ette Rose n bach, J oa n B res na n a nd C h ristoph Wolk
Part III Changes in complementation patterns
153
9 On the changing status of that-clauses
155
G ünte r Roh de nburg
10 Variability in verb complementation in Late Modern English: finite vs. non-finite patterns
182
Hube rt C u ycke ns, Frauke D ’ hoe d t a n d B e ne dikt S zmrecsany i
Part IV Category change 11 Opposite developments in composite predicate constructions: the case of take advantage of and make use of
205
207
E va B e rlage
12 Constrained confusion: the gerund/participle distinction in Late Modern English
224
He ndrik De S met
13 ‘You are a bit of a sneak’: exploring a degree modifier in the Old Bailey Corpus
239
C lau dia C lari dge a nd M e r ja K y tö
Part V The syntax–pragmatics interface 14 If you choose/like/prefer/want/wish: the origin of metalinguistic and politeness functions L au rel J. B ri nton
269 271
Contents ix
15 Epistemic parentheticals with seem: Late Modern English in focus
291
M ar í a J os é L ó pez -C ouso a nd Bel é n M én dez -Naya
Part VI Text-type related change 16 Syntactic stability and change in nineteenthcentury newspaper language
309 311
Er ik S mitterbe rg
Part VII Language contact and syntactic change 17 ‘[W]ell are you not got over thinking about going to Ireland yet’: the be -perfect in eighteenth- and nineteenth-century Irish English
331
333
Kevi n Mc C affert y
References Author index Subject index
352 380 382
Figures
The decline of the be-perfect with mutative intransitive verbs, after Kytö (1997) page 15 2.2 The decline of the be-perfect in COHA 16 2.3 Number of grammar books included per decade in the CNG 19 2.4 Stages in the change from active – passive – neute r to t ra nsitive – i nt r a n sitive in the CNG 21 2.5 Illustration of active – passive – n eute r verbs (and acts) 22 2.6 The shift to t ra nsitive – i nt r a n sitive in British grammars (CNG) 23 2.7 The shift to t ra nsitive – i nt r a n sitive in American grammars (CNG) 24 2.8 Defining the be-perfect in American grammars (CNG) 26 2.9 Defining the be-perfect in British grammars (CNG) 26 2.10 Evaluating the be-perfect (CNG) 28 2.11 Terminology (verb types) employed and analysis of the be-perfect (British grammars) 29 2.12 Terminology (verb types) employed and analysis of the be-perfect (American grammars) 30 2.13 Analysing the be-perfect as be + adjective 32 3.1 Distribution and historical development of negated let and let’s constructions in COHA 50 5.1 Pronouns with ’ll in the eighteenth and twentieth centuries (ARCHER-3.1) 81 5.2 Use of ’ll in different registers in the eighteenth and twentieth centuries (ARCHER-3.1) 81 5.3 Semantic development of ’ll in ARCHER-3.1 82 5.4 Proportion of pure future uses (in relation to intention uses) in the OED 83 5.5 Use of ’ll and will to express a spontaneous decision (ARCHER-3.1 1950–99) 86 2.1
x
List of figures xi
6.1 Mediopassives and alternative constructions in four Sears & Roebuck catalogues 6.2 Modification of mediopassives in four Sears & Roebuck catalogues 6.3 Modification of mediopassives in four Sears & Roebuck catalogues – skewing verbs removed 6.4 Constructional taxonomy for English mediopassive constructions 8.1 Proportion of s-genitive (y-axis) against real time (x-axis) 8.2 S-genitive rates (y-axis) as a function of possessum (top) and possessor (bottom) length and variety 8.3 S-genitive rates (y-axis), as a function of ARCHER time slice (x-axis) and several animacy categories 8.4 Animacy classification for general noun phrase sample by ARCHER period 11.1 Historical development of determiners and/or modifiers preceding advantage in the construction take advantage of 11.2 Historical development of determiners, modifiers and determiner+modifier preceding advantage in the construction take advantage of 11.3 Historical development of determiners and/or modifiers preceding use in the construction make use of 11.4 Historical development of determiners, modifiers and determiner+modifier preceding use in the construction make use of 13.1 The development of three uses of bit (normalised) 13.2 Distribution of the three NP of NP constructions (normalised) 13.3 Semantic categories of nouns occurring in all three NP of NP constructions 13.4 Items modified by adverbial a bit (normalised) 13.5 Functions of bit-constructions (all variants, raw frequencies within the columns) Mosaic plot 13.1 The development of three uses of bit Mosaic plot 13.2 Distribution of the three NP of NP constructions Mosaic plot 13.3 Semantic categories of nouns occurring in all three NP of NP constructions Mosaic plot 13.4 Items modified by adverbial a bit Mosaic plot 13.5 Functions of bit-constructions Mosaic plot 13.6 Items modifying bit
94 99 99 108 136 143 145 152 216 217 218 219 246 247 248 251 256 261 262 263 264 265 266
xii List of figures
14.1 Typology of clauses: if you choose, like, prefer, want, and wish 14.2 Percentage of second-person if-ECs with choose, want, wish, like, and prefer, and percentage of these that are indirect or metalinguistic in three corpora of PDE 15.1 Adverbial and bare seem-parentheticals across time 16.1 Progressives and phrasal verbs per 1,000 words by genre in CONCE and in CNNE 16.2 The proportion of super-phrasal co-ordination by paper and period in CNNE 17.1 B e / have variation with intransitives for subperiods 1350–1990 17.2 B e / have -perfect with intransitive verbs in IrE (CORIECOR) 17.3 B e / have with intransitives in letters only, ARCHER and CORIECOR compared 17.4 Percentage use of be -perfect in IrE, 1761–1890 17.5 Percentage use of be -perfect in IrE, 1761–1890 17.6 Percentage use of be -perfect in IrE, 1761–1890
274 279 301 318 324 335 337 337 349 350 350
Tables
2.1 The decline of the be -perfect in COHA page 36 2.2 Defining verb types (British grammars) 37 2.3 Defining verb types (American grammars) 37 3.1 Historical trajectory of adhortative let constructions 42 3.2 Negation patterns of the let’s hortative in ARCHER 45 3.3 Negation patterns of the let’s hortative in CEN 48 3.4 Verbs occurring with let’s not, don’t let’s, and let’s don’t in CEN 49 3.5 Distribution of negated let’s constructions in COHA 51 3.6 Comparison of let’s do and let’s don’t occurrences in COHA 52 3.7 Verbs occurring with don’t let’s and let’s don’t in COHA 53 4.1 Standard, traditional vernacular and modernvernacular negations of (have) got (to) in the COCA corpus 65 4.2 What-questions in the standard and in the modern vernacular in the spoken COCA material 71 5.1 Number of words in the relevant subperiods of ARCHER–3.1 80 7.1 Semantico-functional analysis of various 117 7.2 Formal and contextual analysis of various 118 7.3 Head nouns in NPs with various 121 7.4 Semantico-functional analysis of numerous 123 7.5 Formal features of numerous 124 7.6 Formal and semantic analysis of numerous 125 7.7 Nouns modified by numerous 126 7.8 Types of nouns modified by numerous 126 8.1 Interchangeable genitive frequencies (raw hits in ARCHER, and variant frequencies) 136 8.2 Fixed effects in the minimal adequate mixed-effects logistic regression model for genitive variation in ARCHER 142 xiii
xiv List of tables
8.3 9.1 9.2 9.3 9.4 9.5 9.6 9.7 9.8 9.9 9.10 9.11 9.12 9.13 9.14 9.15 9.16 9.17 9.18 9.19
Comparing different explanations for the decreasing effect of animacy in genitive variation, dative variation, and in progressives The evolution of finite and non-finite complements after flatter + [+/- reflexive] object in the quotations of the OED between 1650 and the present The evolution of finite and gerundial complements after congratulate + object and the corresponding passive in a series of historical databases Finite and non-finite complements after pride Finite and gerundial complements after upbraid + object Finite and gerundial complements after reproach + object and the corresponding passive in a series of historical databases including the BNC Finite and gerundial complements after blame + object and the corresponding passive as well as the type X is to blame in a series of historical databases Long-term detransitivisation trends with reproach Long-term detransitivisation trends with blame The distribution of that-clauses and other complements after acquaint + object The evolution of that-clauses after answer + object and the corresponding passive in the quotations of the OED between 1550 and the present The evolution of that-clauses after answer + object and the corresponding passive in a series of historical databases including the BNC The evolution of that-clauses after certify + object and the corresponding passive in the quotations of the OED The evolution of that-clauses after certify + object and the corresponding passive in a series of historical databases including the BNC The evolution of that-clauses after confirm + object The evolution of the that-clause after satisfy + object The distribution of that-clauses and raising structures (AcI constructions) in the passive after directive and related verbs The evolution of that-clauses and raising structures involving be-predicates after (active) prefer The evolution of that-clauses and raising structures in the passive after (active uses of) require The evolution of that-clauses and raising structures in the passive after (active uses of) request
151 157 158 159 160 161 161 162 162 164 164 165 166 166 167 168 170 171 172 173
List of tables xv
9.20 Finite, gerundial and infinitival clauses after for fear 9.21 Finite and gerundial clauses after for fear in British and American newspapers 9.22 The use of the emergent conjunction in (the) event (that) ‘if ’ and rival gerundial options in a series of historical databases 9.23 The use of the emergent conjunction in (the) event (that) ‘if ’ and rival gerundial options in British and American newspapers 9.24 Finite, gerundial and infinitival clauses after on/upon (the) condition in a series of historical databases 9.25 Finite and gerundial clauses after on/upon (the) condition in British and American newspapers 9.26 Overview of corpora 10.1 Distribution of finite vs. non-finite CCs with the CTPs remember, regret and deny across the two LModE subperiods 10.2 The distribution of finite vs. non-finite CCs for the factor in te rveni ng mate rial i n wo r d s 10.3 The distribution of finite vs. non-finite CCs for the factor meani ng cc 10.4 The distribution of finite vs. non-finite CCs for the factor tempo ral relatio n 10.5 Fixed effects (main effects only) in the minimal adequate mixed-effects logistic regression model 10.6 Interactions between language-internal constraints and real time pe riod in the minimal adequate mixedeffects logistic regression model 10.7 The distribution of non-finites under the condition denotatio n = ‘same’ 11.1 Choice of databases and corpora used for the analyses presented in Section 11.4 11.2 Inside and outside passives with take advantage of and make use of in the ECF and the BNC (wri1) 11.3 Inside and outside passives with take advantage of and make use of in the BNC 13.1 OBC – periods, word counts, and data 13.2 The development of three uses of bit 13.3 Distribution of the three ‘NP of NP’ constructions 13.4 Semantic categories of nouns occurring in all three ‘NP of NP’ constructions 13.5 Items modified by adverbial a bit 13.6 Functions of bit-constructions 13.7 Items modifying bit
174 175 175 176 177 177 179 192 193 193 194 196 197 201 212 213 213 243 260 261 262 263 264 257
xvi List of tables
14.1 Dating of if-ECs 15.1 Distribution of parenthetical types with seem in the complete ARCHER 15.2 Distribution of verb forms according to parenthetical type in the complete ARCHER 15.3 Medial vs. final position of seem-parentheticals in the complete ARCHER corpus 16.1 Newspapers included in CNNE as of December 2012 16.2 Descriptions of the genres in CONCE 16.3 Progressives and phrasal verbs by period in CNNE 16.4 Progressives by period and paper in CNNE 16.5 Phrasal verbs by period and paper in CNNE 16.6 Conjoins of and in CNNE by period 16.7 Conjoins of and by paper and period in CNNE 17.1 Verbs included in CORIECOR searches for beperfects 17.2 Use of be -perfect in IrE, 1701–1890 (n = 1,110)
282 302 303 305 314 315 317 319 319 323 329 346 348
Data sources (corpora)
ARCHER BNC Brown BYU-BNC CB CED CEMET CEN CLMET3.0 CLMETEV CNG CNNE COCA COHA CONCE CORIECOR d91–95 D92–95 EAF ECCO ECF ED EEPF EPD ETC FRED g90–95 GloWbE
A Representative Corpus of Historical English Registers Versions 3.1, 3.2 British National Corpus Versions 1.0 and BNCweb CPQ-edition. Version 4.3 Brown Corpus (Written American English, 1961) Based on the British National Corpus from Oxford University Press Collins Cobuild Corpus A Corpus of English Dialogues 1560–1760 Corpus of Early Modern English Texts Corpus of English Novels Corpus of Late Modern English Texts (version 3.0) Corpus of Late Modern English Texts, Extended Version (1710–1920) Collection of Nineteenth-Century Grammars The Corpus of Nineteenth-Century Newspaper English Corpus of Contemporary American English Corpus of Historical American English A Corpus of Nineteenth-Century English The Corpus of Irish English Correspondence Daily Telegraph and Sunday Telegraph Detroit Free Press Early American Fiction Eighteenth-Century Collections Online Eighteenth-Century Fiction English Drama Early English Prose Fiction English Prose Drama Early Twentieth-Century Corpus Freiburg English Dialect Corpus Guardian Corpus of Global Web-Based English xvii
xviii List of data sources
HC i93–94 ICE L92–95 LC LION LNC M93–95 MED MNC N01 NCF OBC OBP OED OED online PCEEC PPCEME PPCMBE Sears & Roebuck corpus Strathy t90–95 TIME UofV W90–92 wridom1
The Helsinki Corpus of English Texts The Independent and the Independent on Sunday International Corpus of English Los Angeles Times The Lampeter Corpus of Early Modern English Tracts Literature Online Late Nineteenth-Century Corpus Daily Mail and Mail on Sunday Middle English Dictionary Mid Nineteenth-Century Corpus The New York Times Nineteenth-Century Fiction Old Bailey Corpus (www.uni-giessen.de/oldbaileycorpus/ index.html; 1720 to 1913) The Proceedings of the Old Bailey – London’s Central Criminal Court, 1674 to 1913 Oxford English Dictionary Oxford English Dictionary Online Parsed Corpus of Early English Correspondence, text version Penn-Helsinki Parsed Corpus of Early Modern English Penn Parsed Corpus of Modern British English non-digitised corpus of 1897, 1927, 1957 and 1987 mailorder catalogues The Strathy Corpus of Canadian English The Times and The Sunday Times Time Magazine Corpus The Modern English Collection, University of Virginia Electronic Text Center The Washington Times Fictional / Narrative Component of the BNC WordSmith Tools, version 5.0. 2010. PC software by Mike Scott. Published by Lexical Analysis Software and Oxford University Press
Contributors
Lieselotte A nde rwald ,
of Kiel
Anita Aue r ,
English Department, University
English Department, Université de Lausanne
Eva Ber lage ,
Department of English and American Studies, University of Hamburg T in e Breba n ,
School of Arts, Languages and Cultures, University of Manchester Joan Br es na n ,
Center for the Study of Language and Information, Stanford University Lau r el J. Bri nton , English Department, University of British
Columbia, Vancouver
Clau dia C lari dge ,
Department of Anglophone Studies, University of Duisburg-Essen Hubert C u ycke ns , Hend r ik De S met , Fr auke D ’ hoe dt ,
Department of Linguistics, KU Leuven
Department of Linguistics, KU Leuven
M ar ian ne Hu ndt ,
Zurich
M er ja Ky tö ,
Department of Linguistics, KU Leuven
Department of English, University of
Department of English, Uppsala University
M ar í a J os é L ó pez -C ouso ,
Department of English and German, University of Santiago de Compostela Chr istia n Mai r ,
English Department, University of Freiburg
Kevin Mc C affe rt y ,
University of Bergen
Department of Foreign Languages,
xix
xx List of contributors B el é n Mé ndez-Naya ,
Department of English and German, University of Santiago de Compostela Nadja Nesselhauf ,
Heidelberg
Department of English, University of
G ünte r Roh de nburg ,
Department of English and American Studies, University of Paderborn A nette Rose nbach , Tanagra Wines, E rik
S mitterbe rg ,
University
Department of English, Uppsala
B e ne dikt S zmrecsany i ,
Leuven
C h ristoph Wolk ,
South Africa
Department of Linguistics, KU
Freiburg Institute for Advanced Studies, University of Freiburg
Acknowledgements
Many people have been involved in the preparation of this volume. First of all, I would like to thank the contributors, not only for the work they produced for this book but also for meeting the various deadlines set to complete the project. This is not to be taken for granted, especially in view of at times more-than-busy schedules and the changes (both professional and private) some of them went through while this volume was taking shape. I would also like to express my thanks to all the colleagues who kindly agreed to make the time and review the chapters submitted for this volume (in alphabetical order): Bas Aarts, Douglas Biber, Marion Elenbaas, Udo Fries, Lobke Ghesquière, Martin Hilpert, Andreas Jucker, Juhani Klemola, Manfred Krug, Maarten Lemmens, Juhani Rudanko, Daniel Schreier, Barbora Skarabela, Aaron K. Smith, Sali Tagliamonte, Gunnel Tottie and Elizabeth Traugott. Nine of the seventeen chapters were presented and discussed at an invitational workshop held immediately before the International Conference on English Historical Linguistics in Zurich in 2012. I would like to thank the University of Zurich for the generous funding that made the organisation of the workshop possible. As always, it has been a joy to work with Helen Barton (Cambridge University Press) and the series’ editor, Merja Kytö: I am grateful for their enthusiasm, encouragement and flexibility as well as the valuable input they provided at various stages of the project. Jennifer Keller and Georgina Wood, the editorial assistants at the English department, did a great job in helping to prepare the manuscript and the index, respectively. Georgina Wood also helped at the proof-reading stage, Helena Dowson (Cambridge University Press) and Jenny Slater (Out of House Publishing) saw the book through production, and Sue Browning did the copy-editing. Last but not least, I am indebted to my family: editing a volume like this meant that there were times when they saw less of me than they deserved; but it is their love and support that keeps me going.
xxi
1
Introduction: Late Modern English syntax in its linguistic and socio-historical context M arianne Hundt
1.1 Late Modern English syntax: setting the scene At the end of the twentieth century, the Late Modern period still had to be described as ‘the Cinderella of English historical linguistic study’ (Jones 1989: 279). Fortunately, this scenario has changed over the last fifteen years or so. Among other things, the change of emphasis within historical linguistics to socio-historical and corpus-based approaches led to a surge of interest in Late Modern English (roughly the period between 1700 and 1900). In August 2001, the University of Edinburgh hosted the first international conference on Late Modern English (LModE). These are now organised on a triennial basis. Research on aspects of LModE has since been published in conference proceedings and more focused thematic volumes such as the one on the nineteenth century edited by Kytö et al. (2006). Introductory textbooks (Beal 2004 and Tieken-Boon van Ostade 2009) have also been added to the growing body of literature, and new (editions of) language histories of English now routinely add a chapter or section on the Late Modern period (e.g. Algeo and Pyles 2005, Barber et al. 2009 or Brinton and Arnovick 2011).1 Despite the recent progress in the historiography of the English language between 1700 and 1900, morphological and syntactic change in LModE is still the least researched aspect of this period. (According to the eighteenthcentury grammarian Lowth, English grammar could be described in only ten lines;2 this is obviously a contemporary misjudgement on the basis of the importance that people attributed to morphology.) Late Modern syntax has received excellent and fairly comprehensive coverage in a book-length chapter of the Cambridge History of the English Language, volume i v (Denison 1998). But while this chapter provides a 1
The chapter in Algeo and Pyles is essentially a short comparative history of British and American English from 1800 up until the twenty-first century. 2 See the extract from the facsimile of A Short Introduction to English Grammar (1762) on the cover of this book. In full, this extract from the preface reads “The English Grammar which hath been last presented to the public, and by the person best qualified to have given us a perfect one, comprises the whole Syntax in ten lines: For this reason; ‘because our language has so little inflexion, that its construction neither requires not admits many rules.’”
1
2 Hundt
wealth of examples to illustrate syntactic developments (some of them from corpora), it does not attempt to exploit the resources available at the time in a systematic corpus-linguistic approach. In fact, the rich body of primary material has only recently begun to be studied in more depth and employing corpus-linguistic methodology (e.g. the contributions on syntactic change in Kytö et al. 2006 for the nineteenth century). This kind of research is facilitated by an ever growing number of available corpora:3 in addition to general reference corpora like ARCHER (A Representative Corpus of Historical English Registers) and CONCE (Corpus of Nineteenth-Century English), there are more specialised resources like the Corpus of Early English Correspondence Extension Corpus, the Old Bailey Corpus (OBC), the Network of EighteenthCentury Texts (NEET) corpus or the collections of letters from immigrants (e.g. CORIECOR). It is also possible to make use of large databases of fiction writing as a source of evidence (e.g. Eighteenth-Century Collections Online or Literature Online), even though these were not compiled with linguistic research on LModE in mind. Increasingly, corpus-linguistic resources afford rich annotation: the Corpus of Historical American English (COHA) and the Corpus of Late Modern English Texts (CLMET 3.0) are both partof-speech-tagged (albeit with different tagsets) and the Penn Parsed Corpus of Modern British English (PPCMBE) is also syntactically annotated. An extended version of the original ARCHER corpus has recently also been tagged and parsed, with different POS-tagging that allows direct comparison with COHA, on the one hand, and the PPCMBE, on the other.4 Rydén (1984: 512) refers to the eighteenth century as ‘the first century in the development of English where the extant texts allow us to view and evaluate syntactic usage in a reasonably full contextual light (stylistic, social or otherwise)’. In fact, a wealth of contextual information is available for the whole of the Late Modern period, enabling new kinds of research, e.g. on the sociolinguistic details concerning the spread of innovative patterns or the decline of recessive constructions.5 In addition, scholars have started to systematically relate corpus data on syntactic change to contemporary grammatical descriptions of the phenomena in order to gauge the possible effects that grammars from the period may have had on language change (see e.g. Auer 2006 or Anderwald submitted and this volume). The present volume brings together state-of-the art research into LModE syntax by prominent scholars in the field in the first book exclusively dedicated to the topic. The contributions all make systematic use of existing 3
Some of these resources became available as contributors to this volume were working on their chapters. 4 The accuracy of the tagging and parsing of ARCHER 3.2 with the Treebank Tagger and dependency parser could be improved significant by normalising the spelling before the corpus was annotated (see Schneider 2012). 5 The great majority of syntactic studies are concerned with the emergence of new patterns or distinctions. Research on recessive constructions or syntactic demise is much rarer (see e.g. Hundt, Forthcoming a and Anderwald or McCafferty, this volume).
Introduction: LModE syntax in context 3
corpora. Corpus evidence is not merely used to arrive at more detailed descriptions of syntactic patterns in the LModE period. As the subtitle of this chapter indicates, the aim is to provide both linguistic and socio-historical context for the case studies. Accordingly, a number of contributions in this volume discuss their findings against the background of internal processes of change and within a theoretical framework, such as grammaticalisation theory and/or construction grammar (e.g. the chapters by Brinton, Claridge and Kytö or Nesselhauf). The rich body of textual evidence available is used in these chapters to draw inferences on the precise nature of historical change. This, in turn, is used to feed into grammaticalisation theory (e.g. Breban or Cuyckens et al.), including the discussion of sociolinguistic foundations for grammaticalisation discussed in the chapter by Mair. True to the aim of this volume, some contributions contextualise their findings within the socio-historical background of the period (e.g. Anderwald, Smitterberg or Szmrecsanyi et al.); others provide information on regional variation in British (BrE) and American English (AmE) (e.g. Anderwald, Auer, Mair, Rohdenburg) and language contact (McCafferty). In fact, language contact in the Late Modern period merits a volume in its own right, so the single contribution selected for this volume obviously only covers a tiny aspect of syntactic change in varieties other than (standard) BrE and AmE. However, the compilation of suitable corpora that will allow detailed studies on language contact and the emergence of New Englishes is still one of the lacunae in English historical linguistics (see Hundt Forthcoming b). Yet other contributions in this volume take variation across text types into account as one factor in the ongoing change they study (e.g. Nesselhauf or Szmrecsanyi et al.). Smitterberg takes a complementary approach in that he puts text types centre stage, using different syntactic patterns as a diagnostic for stylistic change in LModE syntax. As regards methodology, two complementary approaches can be observed in the chapters of this volume: some studies make use of the corpus data for fine-grained qualitative analyses and micro-analyses of change (e.g. the contributions by Breban or López-Couso and Méndez-Naya); others apply sophisticated statistical modelling to variation and change (i.e. Szmrecsanyi et al. and Cuyckens et al.). Some studies focus exclusively on data from the Late Modern period (e.g. Breban; Anderwald even restricts her study to change and related grammatical descriptions in the nineteenth century); others go beyond 1945, contrasting and comparing evidence from the eighteenth and/or nineteenth century with more recent corpus data (e.g. the contributions by Hundt or Mair). While there are various connections in terms of outlook and methodology between the chapters of this volume, they are grouped together according to their thematic focus: case studies on change in the verb phrase, the noun phrase, complementation patterns and category change; in addition, there are two chapters on the syntax–pragmatics interface, and one each on texttype related change and contact-induced change.
4 Hundt
1.2 Studies on change in the verb phrase Lieselotte Anderwald (Chapter 2) relates corpus-linguistic evidence on the final decline of the b e -perfect in nineteenth-century BrE and AmE (a change that was led by AmE) to evidence from contemporary grammars. Her investigation of a large corpus of nineteenth-century British and American grammars shows that grammarians on both sides of the Atlantic initially simply lacked adequate grammatical terminology to describe the recessive syntactic pattern, resulting in a misanalysis of the construction. Moreover, she is able to make a link between prescriptive comments, on the one hand, and the lack of suitable grammatical concepts available to the grammarians. When adequate terminology did become available, the ‘faulty’ description lingered longer in American grammars where the b e -passive had declined more quickly. In Britain, where the b e -passive was still a low-frequency phenomenon in language use for a little longer, grammarians switched to a more ‘adequate’ description earlier. The following two chapters look at variation in the use of do-support with two grammaticalising (modal) (auxiliary) constructions. Anita Auer (Chapter 3) investigates the diachronic development of negation patterns with let us/let’s, taking both regional and social variation into account: she finds that do-support in negation of let’s is more frequent in British than in American English and that men are ahead (by about eighty years) of women in adopting do-support. The results are slightly different for the variant in which do-periphrasis follows the verb: it is exclusively found in AmE, with evidence suggesting regional predominance in southern dialects and women as early adopters. Christian Mair (Chapter 4) uses evidence from LModE and corpora of present-day AmE for a detailed study on the social embedding of ongoing grammaticalisation of semi-auxiliary (have) got (to). He traces the origin of main-verb syntax with the semi-auxiliary to regional and ethnic varieties of AmE in the second half of the nineteenth century. Evidence from the Corpus of Contemporary American English (COCA) is used to critically discuss the problems that both main-verb syntax and social variation pose for the study of grammaticalisation. His study demonstrates the usefulness, in particular, of large corpora for the study of ongoing grammatical change that has its origin in non-standard usage of previous centuries: it is only these vast historical collections of text that provide evidence on the ultimate roots of this particular change, precisely because it originates in non-standard usage and is therefore rarely attested in writing. The last two chapters in this section use a construction grammar approach to variation. Nadja Nesselhauf (Chapter 5) employs data from the British part of ARCHER in her case study on ’ll and its development in comparison with both will and shall. Fine-grained semantic analyses of the forms in the second half of the eighteenth against the second half of the twentieth century show that the increase in the contracted form is not simply an example
Introduction: LModE syntax in context 5
of the frequently discussed development towards more colloquial written styles. She argues that, in addition to its old uses, a new form–meaning pattern (i.e. a ‘construction’ in the construction grammar sense of the term) emerged, that of expressing a spontaneous decision. Constructionalisation of this additional meaning would find support in the frequent co-occurrence of enclitic ’ll with certain hosts (i.e. pronouns, and in particular the firstperson singular I), which she assumes to be stored as a whole rather than as two separate items. Marianne Hundt (Chapter 6) also takes a construction grammar approach in her investigation of mediopassive constructions, i.e. patterns such as … books that (won’t) sell. She uses evidence from a corpus of advertising copy, COHA and COCA to verify whether these constructions are more likely to require some kind of modification in earlier decades of the Late Modern period or whether this constraint is stable across time. Five different case studies show that the constraint affects different verbs to different degrees and that variability is closely related to the interaction between constructional meaning and verb meaning. While bare mediopassives, on the whole, are typical of twentieth-century usage and facilitated by a prior strengthening of the modal semantics of the construction, weakening of the modification constraint may affect individual verbs differently, making instantaneous use of bare mediopassives possible for a verb like clamp on without a prior increase in the modified mediopassive use. 1.3 Studies on change in the noun phrase The two chapters on changes in the noun phrase illustrate the two complementary methodological approaches mentioned above: fine-grained qualitative analyses of individual lexico-grammatical patterns, on the one hand, and rigorous statistical modelling of a binary syntactic choice, on the other. While quantifier uses of adjectives such as various and numerous are first attested in the Early Modern period, it is during LModE that the change is consolidated. Tine Breban (Chapter 7) uses data from the PPCMBE and CLMET in a micro-analytical study of the two items to detail aspects of actualisation and paradigmaticisation of this syntactic change. Her study goes ‘beyond mere syntactic change’ in that collocational expansion and contextual embedding are taken to be key steps in the process. Furthermore, the case studies are used for a critical discussion of the role that semantic reanalysis plays in grammaticalisation. The case studies that Breban presents suggest that this role has to be reconsidered in two respects: (a) instead of semantic change, changes in the collocational behaviour of the grammaticalising item can be the starting point for change, and (b) instead of ambiguity the emergence of new meanings in underspecified contexts triggers change. Benedikt Szmrecsanyi and his co-authors (Chapter 8) provide a detailed quantitative study of the genitive alternation in two text types that are
6 Hundt
relatively open to innovation, i.e. newspaper writing and personal letters. The evidence from ARCHER shows that, after an initial decline, the s- genitive gained ground again in the twentieth century. More detailed analyses of the nominal heads shows that the decline in the Late Modern period has to be attributed to cultural change (as reflected in text frequencies) rather than grammatical reasons, namely a decline in the frequency of possessors that have always had a preference for the s-genitive (animate nouns). The recent popularity of the s-genitive, by contrast, can be attributed to genuine grammatical change, i.e. the weakening of the animacy constraint. The latter type of change is seen as a case of grammaticalisation as it involves an expansion of the host class. In their discussion of possible explanations the authors offer a subtle argument that shows how cultural and technological change, change in discourse frequency of certain possessor types, changes in conceptualisation of animacy and/or possession are all connected to ‘real’ grammatical change, and how this can be modelled statistically. 1.4 Changes in complementation patterns The two chapters in this section are concerned with what has been termed ‘the great complement shift’, i.e. the change from finite to non-finite clausal complements (CCs) (e.g. Vosberg 2006). Günther Rohdenburg (Chapter 9) contributes to the sizeable body of studies in this area by focusing on thatclauses following expressive, representative and directive verbs as well as a set of complex conjunctions. His data are drawn from various corpora and historical databases, as well as the OED quotation database. The detailed analyses confirm that complementation patterns of expressive and representative verbs show similar diachronic trends: that-clauses as complements are either lost completely, become restricted to certain patterns or, in a minority of cases, remain at the low level of usage found in the Late Modern period. Rohdenburg places these findings in the larger context of changes in the transitivity of these verbs. After directive verbs, that-clauses are a less recessive complementation type, partly because of the revival of the mandative subjunctive in AmE, as Rohdenburg argues. Complex conjunctions like in (the) event or on/upon (the) condition provide another stronghold for finite complementation. He is thus able to provide some diachronic evidence against the general trend of a shift from finite towards non-finite complement clauses. Hubert Cuyckens and his co-authors (Chapter 10) apply regression analysis to the alternation between finite that-clauses (including the variant with a zero complementiser) and non-finite clauses (both gerundial and to-infinitival) as complements of three factual verbs, namely remember, regret and deny. Empirical evidence comes from the OBC and CLMETEV. The data are coded for various factors (semantic, structural, medium, etc.) that may influence the choice between the two complement clauses. Their
Introduction: LModE syntax in context 7
case studies contribute a new perspective on Rohdenburg’s cognitive complexity principle, both by adding additional complexity factors that have a disfavouring effect on non-finite CCs, and by showing that other complexity factors in fact do not have such a strong effect. In terms of diachronic change, the evidence confirms previous research on the shift towards nonfinite CCs. The authors conclude their study with a discussion of whether changes in the preference for certain complementation patterns can be considered instances of grammaticalisation or not. 1.5 Category change Category change can be observed for individual lexical items in specific constructions or with respect to constructional types, as the first two chapters in this part of the book show. The third chapter in this section studies category change as part of items undergoing grammaticalisation. Eva Berlage (Chapter 11) discusses subtle changes in two apparently similar composite predicates, namely take advantage of and make use of, i.e. combinations of a light verb (take/make) with a noun and preposition. Significantly, only make use of alternates with a simplex verb (use). Category change in this chapter is part of the changes in the cohesiveness of the two constructions: differences in the degree of noun-phrasiness of advantage and use in these two composite predicates is taken to account for differences in the degree of cohesiveness. Against a set of syntactic and semantic criteria, Berlage shows that advantage becomes less noun-phrasy whereas use becomes more noun-phrasy in the course of the Late Modern period: these opposite developments mean that take advantage of becomes more cohesive across time (i.e. it undergoes lexicalisation) whereas make use of becomes less cohesive (i.e. it undergoes delexicalisation). Future studies on composite predicates will have to show whether the two case studies can be generalised to constructions with different kinds of nominal (noun vs. deverbal noun) elements. On the basis of CLMETEV, Hendrik de Smet (Chapter 12) shows that the trend for participial and gerund ing-clauses to become less distinct across time is continued in the Late Modern period, to the extent that the two are often superficially identical. However, the changes evidenced in CLMETEV do not only provide evidence of a further blurring of the distinction between verbal and nominal ing-clauses (e.g. in what De Smet calls ‘blended complement construction’). The data also indicate that language users continue to be aware of the difference between participle and gerund, particularly in certain lexico-grammatical contexts. The chapter contextualises the partial collapse of the participle/gerund distinction within a more general discussion of phrasal and clausal syntax and the wider framework of construction grammar (i.e. with changes starting locally and working their way up on the constructional hierarchy to the macro-level).
8 Hundt
Claudia Claridge and Merja Kytö (Chapter 13) provide a comprehensive investigation of a bit (of) (a) in the OBC during a period when the degree adverb grammaticalises and increases in discourse frequency. The study encompasses the whole range of constructional variants (lexical uses, quantifier, degree adverb, hedge, others). The authors’ aim is to provide broad empirical coverage in a speech-related corpus and thus test hypotheses advanced in previous research. In the OBC data from the 1720s to 1913, non-lexical uses dominate. Their findings corroborate that the end of the eighteenth century is the crucial period for the grammaticalisation of the binominal quantifier a bit of (a) NP as it expands and is used with nonconcrete nouns. Moreover, the quantitative evidence shows that this process happens faster in the case of the variant with the article. As far as the development of the degree adverb is concerned, the OBC data add important detail to previous research, especially with respect to co-occurrence patterns with verbs: according to Claridge and Kytö, these are likely to have spearheaded the grammaticalisation of the degree adverb. While adding valuable quantitative analyses to the existing body of research, the chapter nicely illustrates how even a very large, speech-related corpus such as OBC may still be too small for the analysis of some low-frequency items undergoing grammaticalisation. 1.6 The syntax–pragmatics interface The two contributions in this section go beyond ‘mere syntactic change’, too, in that they study phenomena at the interface of syntax and pragmatics. The focus of Laurel Brinton’s contribution (Chapter 14) is on the metalinguistic and politeness functions of a set of if-clauses, namely those with second-person subjects and verbs of choice (such as choose, like, wish); more precisely, she studies those with ellipsis of the complement, which she refers to as if-elliptical clauses (if-ECs). These make their first appearance with the pragmatic meanings in the Late Modern period. The chapter takes a detailed corpus-based description of the form and function of these clauses in Present-Day English (PDE) as its starting point before tracing their development on the basis of a broad range of LModE corpora and text collections. Corpus data show that it is ultimately impossible to find a single source for the development of the metalinguistic function of if-ECs; this leads Brinton to suggest that it developed by a process of constructionalisation from a set of similar constructs involving verbs of choice which became increasingly abstract and less compositional while at the same time increasing in productivity. Building on their earlier work on the history of epistemic parentheticals with seem, María José López-Couso and Belén Méndez-Naya (Chapter 15) use evidence from ARCHER to trace the history of this construction in
Introduction: LModE syntax in context 9
LModE. The focus of the chapter is on a detailed description of construction types with epistemic/evidential seem, in terms of both their grammatical properties and their pragmatic functions. Parenthetical clauses are attested in two subtypes in ARCHER, namely as bare parentheticals and as an (older) adverbial variant preceded by as/so. Bare parenthetical it seems first occurs at the end of the Early Modern period; in LModE, it is more frequent than the adverbial variant. The authors show that it is also less variable and thus the more grammaticalised of the two seem-parentheticals. While the pragmatic function of seem-parentheticals is to express the ‘speaker’s commitment towards the truth of the proposition expressed in their anchor clause’, first-person experiencer NPs are rare in corpus data. But the authors also point out that the evidence provided in a relatively small corpus such as ARCHER can only be the starting point for a fuller description of epistemic parentheticals in LModE; in particular, the evidence is not conclusive with respect to the minutiae of change, e.g. in scientific texts. 1.7 Text-type and contact-induced change One of the advantages of studying syntactic change in the Late Modern period is that the rich body of textual material allows us to trace how changes spread through different kinds of text because genre is typically an important factor in language variation and change. In research on ongoing grammatical change, newspapers were found to be a text type particularly open to syntactic changes related to changing stylistic norms, i.e. a trend towards more colloquial language use in writing (see e.g. Hundt and Mair 1999). Smitterberg (Chapter 16) uses a corpus of nineteenth-century newspapers and four case studies (progressives, phrasal verbs, not-contraction and conjoins of and) to verify whether this tendency can already be observed in the Late Modern period. It turns out that this is only the case for two of the four syntactic patterns, namely the progressive and conjoins of and. The results are contextualised within ongoing socio-historical change during the 1800s and contemporary corpus data from other text types. He is thus able to show how similar market forces may result in different stylistic preferences and patterns of change. The volume opens with a chapter on the b e -perfect in the Late Modern period and it closes with the same topic, but from a different perspective: while Lieselotte Anderwald looked at the relation between grammatical description and change in her contribution, Kevin McCafferty (Chapter 17) uses evidence from a corpus of personal letters to investigate language contact as a factor in the use of b e -perfects in eighteenth- and nineteenthcentury Irish English (IrE), i.e. the period when the majority of speakers in Ireland shifted from Irish to English as their first (and often only) language. Even though be -perfects are currently also recessive in this regional variety, they are used not only with (intransitive) mutative and motion verbs
10 Hundt
(e.g. change and go) but also with transitives (e.g. finish). McCafferty’s corpus data show that, while IrE follows the general trend towards a greater use of the h av e -perfect, it is lagging somewhat behind BrE and AmE in the Late Modern period. b e -perfects in IrE are still used with a larger variety of verbs than in the reference varieties and even extended to transitives during this period. Pending more conclusive evidence on regional distribution, i.e. particularly in areas with prolonged language contact between Irish and English, McCafferty puts forward arguments for Irish influence on the emergence of b e -perfects with transitives during LModE. The contributions in this volume all make use of the rich body of textual evidence available to draw inferences on the precise nature of historical change, including both external and internal factors. In this respect, the present volume also provides evidence that can be made fruitful to the study of earlier periods and language change in general.
Part I Changes in the VP
2
The decline of the be -perfect, linguistic relativity, and grammar writing in the nineteenth century Lies elotte Ande rwald
2.1 Introduction In this chapter, I expand in a quantitative way on the valuable qualitative work that has so far been done on the topic of historical grammar writing, and link my findings to facts about language change that historical corpuslinguistic studies have produced (for qualitative work on grammar writing, or work concentrating on individual authors, see (among others) TiekenBoon van Ostade 2011; Fens-de Zeeuw 2011; Navest 2011; Straaijer 2009; as well as contributions in Tieken-Boon van Ostade 2008; Leitner 1986, 1991). Combining these two fields of investigation gives us a stronger base for claims on prescriptive vs. descriptive stances in the nineteenth century, allows us to compare British and American grammar writing in quantitative terms, and to evaluate the potential success (or not) of prescriptivism. In this particular case study of the treatment of the b e -perfect in British and American grammar writing, we will see a phenomenon that is consistently described incorrectly, due to the inadequate terminology that was available at the time to the majority of grammar writers, and this incorrect description then leads to condemnation of a linguistic phenomenon that, as a traditional and conservative feature of the language, we might have expected to be cherished and perhaps venerated by the community of rather conservative normative grammarians. 2.2 Grammar in flux in the nineteenth century It has been demonstrated sufficiently by now that the nineteenth century was by no means the century of stagnation that some twentieth-century historical linguists still claimed it was (see Beal 1999 for treatments of the Late Modern period more generally; Aarts et al. 2012 for an overview of syntactic change; Denison 1998 for the most detailed description of change in Late Modern syntax available to date). In particular, changes in the verb phrase have been investigated in much detail (Arnaud 1983, 1998; Hundt 2001, 2004a, b; Smitterberg 2005; Kytö et al. 2006). However, only rarely have prescriptive grammar books been correlated with phenomena of language 13
14 Anderwald
change, at least not more than anecdotally, and references are typically based on the collection of quotes in Visser (1963–73).1 In this chapter I will concentrate on a change that came to near completion in the nineteenth century, the decline of the b e -perfect. Forms of the perfect with both b e and have are quite usual in the Indo-European languages, and the two auxiliaries typically have a neat division of labour (see McFadden 2007b for a useful overview). Thus, in German (intransitive) verbs of motion typically require a perfect with b e , whereas the default in all other cases is hav e (thus, ich bin gelaufen ‘I have run’ with b e , vs. ich habe gesessen ‘I have sat’ with hav e ); in French the choice is similarly constrained by properties at the predicate level (McFadden 2007b), and intransitive verbs of motion form one (although not the only) group of verbs that form the perfect with ê t r e rather than 2 avo i r . In English, we can observe a rather long-term development with concomitant competition between the two auxiliaries. Forms of the perfect with both hav e and b e can be traced back to Old English times. The standard account is that the hav e -perfect evolved from a stative (possessive) resultative construction (of the kind Ic haefde hine gebundene), where the participle was inflected and showed concord with the direct object (‘I had him in a bound state’). From this origin with transitive constructions, hav e then expanded to the use in intransitive constructions too, and thus grammaticalised (see the discussion in Denison 1993: ch. 12). According to Kytö’s figures, stative intransitive verbs seem to have favoured hav e at all times (Kytö 1997: 34). Perfect forms with b e are found in particular with mutative intransitive verbs like c o m e , a rriv e , or ru n , a subclass of intransitives, but over the course of the eighteenth and nineteenth centuries the perfect with be comes to be more and more restricted, and finally is attested only with a small subset of lexical items. As Kytö (1997) has shown, in the group of mutative verbs, process verbs like b e c o m e or g row persist with the b e perfect for longer, whereas action verbs like a r r i v e or ru n come to be used with the hav e -perfect earlier (see Figure 2.1). In the nineteenth century, we see the rapid decline in the use of mutative verbs with b e , and the restriction to the lexemes c o m e and g o .3 G o in 1
One exception I am aware of is Rydén and Brorström (1987), who, like me, try to link language change with the comments in grammars at the time. However, since they use only twenty-seven grammar books from the nineteenth century, their approach is not quantitative. For a similar, but temporally more extended approach cf. Auer (2006, 2009). Visser himself, of course, has to be mentioned as the first linguist who took normative grammar evaluations seriously and correlated them with linguistic phenomena (Visser 1963–73). 2 According to general reference grammars, reflexive verbs and intransitive verbs of motion (alle r, v e nir, arriver, partir, entrer, sortir , descendr e, monter , naîtr e, mo u ri r, d écéder, retourner, rentrer, rester , tomb er , devenir ) take êtr e rather than avoir as the perfect marker. 3 It is striking that hardly any remnants of the be -perfect are attested in traditional dialects of English – despite the fact that it is still encountered in the standard in the nineteenth century. This is quite in contrast to other pervasive features of non-standard grammar, which can typically be traced back historically much further (e.g. adverbs without -ly,
The BE -perfect and 19c grammar writing 15 100 90
% occurrences
80 70 60 50 40 30 20 10 0 1385
1460
1535
1605
1675
process verbs HC action verbs HC
1725 1775 Period
1825
1875
1925
1975
process verbs ARCHER action verbs ARCHER
Figure 2.1 The decline of the b e -perfect with mutative intransitive verbs, after Kytö (1997) [Note: the periods of the corpora have been converted to points in time based on the middle of the period in each case. In this way, differences in periodisation can be converted to a coherent temporal scale, facilitating comparison across different materials.]
particular is the only form that is still encountered today in the combination of a past participle with a form of b e ( he is gone) in substantial numbers, although, of course, have gone exists next to it (see Figure 2.2 and Table 2.1 for figures). However, b e gone is synchronically no longer seen as a form of the perfect, and even some of the best grammarians have difficulties analysing it. (See the odd classification in Quirk et al. 1985: 170, Note a, where it is called a ‘pseudo-passive’.) As this chapter will show, the problems that the b e -perfect presents are by no means new, but are quite a characteristic feature of nineteenth-century grammar writing. 2.3 Corpus studies of the b e -perfect Corpus studies that have investigated the competition between the hav e perfect and the be -perfect are faced with the methodological question of multiple negation, or the lack of -s in measurement nouns; for a list see Kortmann 2008a). The only varieties where the be -perfect seems to have survived, and is perhaps expanding, are the Shetland Islands (Melchers 2008; Melchers and Sundkvist 2010), perhaps mainland Scottish English, and Irish English (see Filppula 1999: 116–22, 2008; Hickey 2007a: 196–97; McCafferty, this volume).
16 Anderwald 100 90
% occurrences
80 70 60 50 40 30 20 10 90 00 20
80
19
19
60 70 19
50
19
40
19
30
19
20
19
10
19
00
19
90
19
80
18
70
18
60
18
50
18
40
18
30
18
20
18
18
18
10
0 Period % be gone % be come % be fled
% be arrived % be become % be risen
% be fallen % be departed
Figure 2.2 The decline of the b e -perfect in COHA
what to include in their count, since – depending on the time frame of investigation – the envelope of variation varies considerably, and with the envelope of variation text frequencies also differ significantly between studies. Rydén and Brorström (1987; see also Rydén 1991) investigate 120 verbs which were still variable in the eighteenth and nineteenth centuries on the basis of their own collection of dramas and letters. Kytö (1997) replicates the investigation across the Helsinki Corpus and a pre-final version of ARCHER 1 (for details on ARCHER 1 see Biber et al. 1994) on the basis of 194 verbs (although many occur only very marginally), as Figure. 2.1 above illustrates. I have extended Kytö’s investigation to cover the nineteenth and twentieth centuries in written American English and turned to the Corpus of Historical American English (COHA) to do so (see Davies 2012 for an introduction). Initially I searched all lexemes that were still variable in LModE in Kytö’s data. It turned out that at least in COHA, the b e -perfect was much less frequently attested than in the British data investigated by the earlier studies.4 As Figure 2.2 shows, even for the eight lexemes where b e 4
Note that the search string employed ([be] *.[v?n*]) is relatively restrictive, as it only returns occurrences of the be -perfect where the participle directly follows the auxiliary. However, considering the size of COHA, and the absolute frequency of the phenomenon, it was deemed acceptable as a compromise to exclude negation (he is not arrived), inserted
The BE -perfect and 19c grammar writing 17
still occurs with the past participles more than sporadically, the relative frequency of the be -perfect declines noticeably in almost all cases, and we can see that the be -perfect becomes a decidedly marked choice for all lexemes after 1840 at the latest; for many lexemes, frequencies of the b e -perfect (as opposed to the hav e -perfect) are never more than 10 per cent (d e part, fall , r i s e ). Although Figure 2.2 shows that b e is still attested during the nineteenth century for these lexemes, it also makes it very clear that the alternatives with have are much more frequent in written American English of the time. Some of the attestations of the (by then) minority pattern with b e are given in (1).5 (1)
a. Farewell, dreamer – waking or sleeping! Your brightest dreams
are fled (1835, FIC)
b. I yesterday heard that a French vessel was arrived at Portsmouth
(1840, MAG)
c. the evil influence of this house is departed (1875, FIC) d. it [sc. copying paintings] was fallen from being a master art
(1882, NF)
All these verbs are today formed with hav e as the perfect marker in the majority of cases. The only real exception is the verb g o , which, as mentioned before, today can be analysed as a construction involving a participial adjective: he is gone ~ he is absent. This use seems to be the dominant, indeed the only possible use today, but since b e gone mostly occurs without further complementation, occurrences like he is gone are ambiguous for the nineteenth century between actual perfect and adjective uses, i.e. between meaning ‘he has gone somewhere’ and ‘he is absent’. Of course, one reason for this ambiguity, and probably a facilitating factor in the ensuing meaning shift towards PDE, is the fact that the former activity reading ‘he has gone somewhere’ implies the PDE adjective reading ‘he is (now) absent’. As Figure 2.2 shows, the frequency of b e g o ne in written American English declines (if nowhere near the rate of the other lexemes) until the 1880s, and then slowly increases again. At least part of this increase may be due to the rise of the adjective reading. Kytö similarly finds that occurrences of b e g o n e in the later periods in ARCHER are all stative (Kytö 1997: 54). In COHA, we find clear cases of adjectival uses in co-ordination with other adjectives from the 1860s (our boys that are gone and dead 1864 FIC, Teresa was gone and safe adverbials (he is lately arrived) or inverted constructions (is he arrived?), in order to arrive at maximally accurate returns, which could then be directly compared to the respective constructions with hav e . 5 The abbreviations after the years relate to the text type: FIC=fiction, MAG=magazines, NF=non-fiction books.
18 Anderwald
from pursuit, 1883 FIC). Instances where no literal movement is possible, and a literal b e -perfect is thus not a plausible reading, occur even earlier, in fact as early as the 1820s (e.g. my hat was gone, 1823 FIC; our money is gone, 1839 NF).6 The be -perfect is thus a clear example of a linguistic feature undergoing rapid change over the course of the nineteenth century, so it is a valid question to ask whether this change was perceived as such in the grammar books of the time, and if so, whether this engendered any positive or negative evaluations. In theory, we would expect prescriptive grammarians to be conservative and backward-oriented, and therefore critical of new developments. More general criticism of new forms would also entail an attempt at preserving obsolescent constructions, or even attempts to revive dead ones. From the obsolescent character of the b e -perfect in the nineteenth century we might therefore expect a rather positive treatment in grammar books of the time. As we will see, however, this is not the case. As I will argue, the unexpected negative evaluation of this one phenomenon can be directly linked to the faulty description of this linguistic feature, itself linked to the inadequate linguistic terminology that was available to grammar writers at the time. 2.4 The be -perfect in nineteenth-century grammars 2.4.1 The Collection of Nineteenth-Century Grammars (CNG) In order to answer questions on nineteenth-century grammar writing in a principled way, I have collected a number of British and American grammar books over the past few years. Based on the Google books project, between 2007 and 2011, I downloaded all treatments of English grammar, as they became available as full texts, written by native speakers of English for native speakers of English (see Figure 2.3). This excludes usage guides or collections of aphorisms and also books for very young children that simply did not contain much information. Today, the CNG is a collection of 258 grammars covering all ten decades of the nineteenth century.7 Publication dates and editions were checked against Görlach (1998) and only the earliest editions were included (although later editions were kept for reference). First editions are not always available, and have often not been preserved. However, for over 70 per cent of all grammars, the first edition is in fact included in the CNG. 6
This is not to say that the hav e -perfect cannot have a stative reading as well. The tree has gone is possible in the meaning ‘has disappeared’ (although not ‘has moved somewhere’). Indeed, as I have said above, the have -perfect has always been preferred in stative contexts. 7 A complete bibliographic list can be found on www.anglistik.uni-kiel.de/index.php/cng. html. In all diagrams, the date of first publication has been taken as decisive where known.
The BE -perfect and 19c grammar writing 19 25
20
15
10
5
0
1800
1810
1820
1830
1840
US grammars
1850
1860
1870
1880
1890
British grammars
Figure 2.3 Number of grammar books included per decade in the CNG
2.4.2 Describing the be-perfect In the CNG, 122 grammars mention the b e -perfect explicitly or use it in their pages of exercises. This means that roughly every other grammar deals with the be -perfect in some way. However, as we will see, b e -perfect is an anachronistic term that was never employed by nineteenth-century grammarians themselves. One representative attempt at a description of the b e -perfect can be found in Fisk (1822). Although Fisk was elsewhere unashamedly based on Murray’s influential grammar (Murray 1795), this passage is not copied from the ‘master’: Active intransitive verbs are sometimes construed in the passive form; as, ‘I am come; He is arrived; She was gone; It is grown’; &c. but the regularity and propriety of such constructions are questionable. ‘I have come; He has arrived; She has gone; It has grown’; in the examples just cited, would be preferable to the forms of expressions there used. The regular passive verb can be formed only from the active transitive verb, because it denotes action received from one object and endured by another; this is, it denotes the passing of action from one object to another; which is directly opposed to the nature of intransitive verbs. (Fisk 1822: 69)
The reasoning behind this passage is spelled out in (2). (2)
(Premise 1) com e is an intransitive verb (Premise 2) be + participle = passive → I am come is the passive of an intransitive verb (Premise 3) Passives can only be formed from transitive verbs → I am come is incorrect
20 Anderwald
Since all three premises are largely correct, it is perhaps puzzling to see where the wrong conclusions come from. However, from our present-day perspective we would say that Premise 2 (although partly correct) is incomplete. In the syllogism underlying Fisk’s reasoning, a second part to his second premise seems to be missing: (3) Premise 2b: b e + participle of intransitive verb = perfect Adding this premise shows that Fisk’s conclusion is in fact based on faulty reasoning. Adding Premise 2b could change the first conclusion (I am come cannot be the passive, since it is intransitive), and therefore the second conclusion could be: I am come must be the b e -perfect. We can only understand Fisk’s ‘mistake’ if we consider it in the context of terminology and rewriting that this passage must have undergone. Consider the original in Murray, itself copied almost verbatim from Lowth (1762) some thirty years earlier: The neuter verb is conjugated like the active; but as it partakes somewhat of the nature of the passive, it admits, in many instances, of the passive form, retaining still the neuter signification; as, ‘I am arrived’; ‘I was gone’; ‘I am grown.’ The auxiliary verb am, was, in this case, precisely defines the time of the action or event, but does not change the nature of it; the passive form still expressing, not properly a passion, but only a state or condition of being. (Murray 1795: 67)
Although the hedging (somewhat, in many instances, retaining the neuter signification, expressing not properly a passion) already indicates that Murray (or rather Lowth) himself was rather uneasy with his own analysis of I am arrived as a passive, the terminology employed is telling. From Murray onwards, ‘neuter verbs’ that ‘partake somewhat of the nature of passive verbs’ were to become an enduring topos of English grammar writing.8 In order to be able to understand this definition, we have to digress briefly into the history of grammar terms. 2.4.3 Excursus: division of verbs English inherited the division of verbs into act i v e – pas s i v e – n e u t e r from mediaeval Latin grammars. In fact, the tripartite distinction is usually traced back to Donatus’ influential ars grammatica from the fourth century.9 8
For the continuing influence of Murray (and through him Lowth) see Figure 2.13, where his formulation can be traced verbatim or part verbatim up to the 1840s. Available online at http://htl2.linguist.jussieu.fr:8080/CGL/text.jsp?id=T28. It seems that the short version of the ars minor was particularly influential, because here we have the explicit tripartite division of verbs: uerbum quid est? pars orationis cum tempore et persona sine casu aut agere aliquid | aut pati aut neutrum significans [‘a verb is what? A part of speech with tense and person without case signifying either acting something or suffering or neuter’], whereas the ars maior has five divisions of verbs into activa passiva neutra deponentia communia [active, passive, neuter, deponent, common]. Schönberger translates deponentia by ‘middle’, and communia by ‘mediopassive’ (Schönberger 2009: 85), but this translation seems to be idiosyncratic. See also the critical edition of Donatus (Holtz 1981).
9
The BE -perfect and 19c grammar writing 21 ACTIVE PASSIVE NEUTER
ACTIVE (action, passion) NEUTER
ACTIVE (transitive intransitive) PASSIVE NEUTER
ACTIVE (transitive intransitive=neuter) PASSIVE
ACTIVE=TRANSITIVE NEUTER= INTRANSITIVE PASSIVE
ACTIVE (transitive intransitive) PASSIVE ACTIVE=TRANSITIVE NEUTER=INTRANSITIVE TRANSITIVE (active, passive) INTRANSITIVE TRANSITIVE INTRANSITIVE=neuter TRANSITIVE INTRANSITIVE
Figure 2.4 Stages in the change from ac t i v e – pas s iv e – ne ut e r to t r a n s i t i v e – i n t r a n s i t i v e in the CNG
Early English grammar writing took over this distinction wholesale, and only slowly over the course of the nineteenth century replaced it with the more syntactic distinction of t ra ns it iv e vs. int ra n s i t i v e verbs that we are familiar with from modern grammar writing.10 As Figure 2.4 shows, this process of terminological changeover proceeded via several intermediate steps, many of which used both ac t iv e – pas s iv e – n e u t e r and t r a n s i t iv e – i nt r ans i t i v e terminologies, foregrounding the one or the other. Figure 2.4 depicts which terms were used as primary terms (indicated by capitals) and which were meant to be subordinate, by individual grammar writers, without indicating quantitatively which systems were used by how many grammarians. As the lines show, the different systems that emerge from the grammars themselves can be ordered logically according to the terms employed. As I have shown elsewhere at greater length (Anderwald submitted: ch. 2.2), the choice of terminology was often commented on explicitly, and the merits (and demerits) of each system were discussed at length, especially during the first decades of the nineteenth century, the phase of the most intense change. The change is not just one in terminology, but is also part of a more widespread change from semantic to syntactic definitions. The essentially semantic distinction of ac t iv e verbs (designating activities), pas s i v e verbs (designating the ‘undergoing’ of an activity, or ‘receiving’ an activity), and 10
The definition of i n t r ansitive vs. transitive was itself initially semantic, and only slowly came to be phrased in purely syntactic terms. This is not relevant for my argument here, but it shows how persistent the tradition of semanticising (often over-semanticising) the terminology was.
22 Anderwald
Figure 2.5 Illustration of ac t i v e – pas s ive – ne ut e r verbs (and acts) (Cobbin 1864: 37, digitised by Google books)
verbs (as being neither active nor passive, or alternatively as verbs of being, or states of being) was sometimes even illustrated: In the transition phase, the definition of neuter verbs was often augmented by syntactic criteria, such that neuter verbs were now also verbs whose ‘action was limited to the subject’ (i.e. intransitive verbs), and in this shift away from semantics the confusion becomes palpable, as the passage from Ingersoll from the 1820s, written in the popular catechism format, exemplifies. (Dramatis personae are Caroline, a model student, and George, a bit slower than his sister. The Tutor is portrayed as knowing it all, and is clearly the mouthpiece of the author, making the following passage doubly ironic.) n eu t er
T u to r .
verbs.
I believe, Caroline, that you understand how to form the passive
But there are many active verbs that are intransitive; such as go, fly, arrive, &c. Suppose that I should put the perfect participle of an active intransitive verb after the neuter verb be, and say, ‘He is gone’; ‘He is arrived’; ‘The bird is flown’, &c. what kind of a verb shall I have then? T u to r . A neuter verb, in a passive form. (Ingersoll 1824: 149; italics in the original) Ge o rge .
The BE -perfect and 19c grammar writing 23 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 1800
1810
1820
1830
1840
1850
1860
1870
1880
1890
Period 1: active – passive – neuter 3: active=transitive, neuter=intransitive 5: transitive – intransitive
Figure 2.6 The shift to grammars (CNG)
2: active – transitive vs. active – intransitive 4: transitive – intransitive, + neuter
transitive
–
i nt r a n s it ive
in British
The justified and pertinent question why verbs like g o , fly, or ru n should not be considered to be active verbs, if they clearly designate activities, was never satisfactorily answered, and must have contributed to the decline of the mediaeval system of terminology. Clearly, the syntactic criterion (‘action limited to subject’) cross-cuts the semantic classification of ‘active’ vs. ‘neuter’, which must have led to persistent confusion in the classroom. When we investigate the shift in terminology in British and American grammars, striking differences become apparent. As Figure 2.6 (figures in Table 2.2) shows, the terminology ac t iv e – pas s i v e – n e u t e r persists until the 1860s in Britain, and is only slowly superseded by the more modern terminology of t ra ns it iv e – int ra ns it iv e . It is only in the 1860s that this more recent terminology becomes the majority position in British grammars in the CNG. A comparison with American grammars shows a very different picture, as Figure 2.7 illustrates (figures in Table 2.3). In American grammar writing, the position of ac t i v e – pas s i v e – n e u t e r is never as prominent as in British grammar writing. The alternative terminology t r ans i t iv e – int ra ns it iv e is available much earlier, and
24 Anderwald 100%
80%
60%
40%
20%
0% 1800
1810
1820
1830
1840
1: active – passive – neuter 3: active=transitive, neuter=intransitive 5: transitive – intransitive
Figure 2.7 The shift to grammars (CNG)
1850
1860
1870
1880
1890
1900
2: active – transitive vs. active – intransitive 4: transitive – intransitive, + neuter
t r a n s i t i ve – in t r ans it iv e
in American
becomes the accepted majority position after the 1840s, i.e. at least twenty years earlier than in Britain. Intermediate stages between the two systems are also not as pronounced as in British grammar writing. 2.4.4 The be-perfect as the passive of neuter verbs After the look at the traditional terminology, we can now reconstruct why the be-perfect was analysed as a ‘passive’ for several decades. Recall from the introductory description of the variable distribution of the be-perfect that it could still occur with (some) intransitive verbs of motion and change of state, but was variable throughout, and decreased rapidly in frequency at the expense of have. In the system of active, passive and neuter verbs, the situation of periphrastic verb forms presented itself as at least partly straightforward: active verbs could be used with have to form the perfect (thus have was seen as the ‘sign’ of the perfect), active verbs combined with be turned into passive verbs (thus be was the ‘sign’ of the passive), and neuter verbs could also be combined with have to form the perfect. It followed quite logically that a combination of a neuter verb with be then had to be a form of the passive. This also illustrates how neuter verbs were perceived to be situated ‘between’ active and passive verbs: they formed the perfect like
The BE -perfect and 19c grammar writing 25
active verbs, but the ‘passive’ like passive verbs. A ‘correct’ interpretation of be + past participle as a form of the perfect would have destroyed this symmetry, and would also have gone against one of the general principles underlying much of nineteenth-century grammar writing, the principle of one form – one meaning. In fact, a correct analysis of the be-perfect would have violated this principle in both directions: On the one hand, analysing be + past participle of intransitive verbs as a perfect would have entailed acknowledging that the perfect could be signalled by two different auxiliaries (have and be) with virtually no difference in meaning. On the other hand, it would have entailed admitting that be + past participle could indicate either the passive (i.e. of transitive verbs), or the perfect (i.e. of intransitive verbs). In this second formulation, it also becomes quite clear that having adequate terminology at hand plays a central a role in the correct description of this phenomenon. Only the terminology of transitive – intransitive would allow the majority of later grammarians to make the required distinctions in order to correctly identify the be-perfect as a perfect. As Figure 2.8 and Figure 2.9 show, an adequate definition is only achieved very late in American grammars, where the hedged version (the b e -perfect is a passive form but does not have passive meaning) persists as a majority position until the 1890s. Only in grammar books published from 1900 onwards do we see an analysis that explicitly calls this construction a perfect. In British grammars, on the other hand, we see the switch-over much earlier, and in the 1860s a majority of grammars define forms of b e with the past participle of an intransitive verb as a perfect. 2.4.5 Attitudes to the be-perfect In contrast to claims by Rydén that in LModE grammar writing the b e -perfect was ‘only marginally … encompassed within the “doctrine of correctness”’ (Rydén 1991: 344), in the CNG the b e -perfect is frequently criticised as being an anomaly, unnecessary, or simply incorrect. In comparison with other phenomena, this criticism is quite strong, and this strong criticism is clearly linked to the essentially faulty analysis as a passive, and to the unclear status of the b e -perfect in the system of ac t i v e – pa s s i v e – n e ut e r verbs, as the representative quotation from Parker and Fox illustrates: 50. Neuter verbs of motion and change are, by some grammarians, used in the passive form; but such usage must be considered a departure from grammatical correctness; as, ‘I am come’, ‘I was gone’, ‘I am grown’, ‘I was fallen’.* *Although this form of expression is sanctioned by Murray, Lowth, and other good authorities, yet reason and analogy will not justify us in assenting to their decision; for, besides the awkwardness of the expression, it is objectionable as being an unnecessary anomaly. (Parker and Fox 1835)
100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 1800 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 Period 4=perfect 3=passive form, perfect meaning 2=passive form, other meaning 1=passive
Figure 2.8 Defining the b e -perfect in American grammars (CNG) 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% b1800 1800 1810 1820 1830 1840 1850 1860 1870 1880 1890 Period 4=perfect 3=passive form, perfect meaning 2=passive form, other meaning 1=passive
Figure 2.9 Defining the b e -perfect in British grammars (CNG)
The BE -perfect and 19c grammar writing 27
It is particularly interesting to see that the arguments used to defend other obsolescent constructions (like the passival, e.g. the bridge is building), e.g. referring to the use of established authorities, especially other grammarians, are refuted here (viz. their footnote). The phrase ‘unnecessary anomaly’ hints at the variability that still existed, and that is criticised, implicitly appealing to the principle of one meaning – one form mentioned earlier. While the be -perfect was criticised both in American and British grammars, only in American grammars do we find an emancipation of criticism of the be -perfect from an actual description of the phenomenon in the grammars themselves. Thus, in Harvey, examples of the b e -perfect are only encountered in a chapter on false syntax under the heading ‘improper forms’. FALSE SYNTAX./ II. IMPROPER FORMS. / […] /Caution XIV. – Avoid the use of improper passive forms. / Ex. – 1. He was retired from active service. 2. Evening was come when we reached the summit of the mountain. 3. He is possessed of a large number of farms. 4. The disputants were agreed on that. 5. The hour for adjournment is arrived. (Harvey 1868: 203)
It is worth noting here that the author mixes his examples (only examples 1, 2 and 5 are examples of the b e -perfect), and that he neither provides a definition (beyond the heading ‘improper passive forms’), nor does he give any indications of how to correct the forms. By the 1860s in America, the criticism of the b e -perfect must have been common enough to make a passage like this one self-evident. Quite clearly, Harvey expected his readers (teachers, and perhaps even the students) to be instantly aware of the ‘error’ he was speaking of, and to be able to provide alternative correct forms themselves. Indeed, if we look at the critical stances in America and Britain towards the b e -perfect, we can see that criticism was very widespread throughout the century, peaking towards the middle, as Figure 2.10 shows. In fact, in many grammar books the b e -perfect was criticised because it was regarded as a ‘passive’, especially from the 1820s to the 1860s, as marked in Figure 2.10. Only three grammars over all of the nineteenth century prefer the b e -perfect wholesale. These are Smith (1841) and Kenny (1858), who recommend changing the hav e -perfect to the b e -perfect without discussing this phenomenon further, and Allen and Cornwell (1841), who seem to feel obliged to defend this construction as ‘more idiomatic’ (Allen and Cornwell 1841: 152) at a time when it was criticised the most. Other positive evaluations are rather tentative, and by many the b e -perfect is only permitted in some contexts, often with reference to the very frequent lexemes it still occurred with, and sometimes with reference to a restriction to stative meanings. Thus, writing at roughly the same time as Allen and Cornwell, Wells is generally critical, but admits the b e -perfect
28 Anderwald 10 9 8 7 6 5 4 3 2 1 0
1800
1810
1820
1830
1840 1850 Period
1860
1870
1880
1890
0=no preference 4=BE preferred 3=BE preferred in some contexts 2=criticised 1=criticised because passive
Figure 2.10 Evaluating the b e -perfect (CNG; only included are grammars that mention the b e -perfect)
with c o m e and g o , and supports this with the use by roughly contemporaneous authors: Most intransitive verbs do not admit of the passive form. Thus, instead of saying, ‘Very great abuses are crept into this entertainment,’ it would be better to say, ‘Very great abuses have crept into this entertainment.’ But the verbs come and go, and perhaps a few others, may, in some cases, properly assume the passive form; as, ‘The time is come.’ – Channing. ‘The sharp touches of the chisel are gone from the rich tracery of the arches.’ – Irving. (Wells 1847: 80)
The overall climate of the times, however, is highly disapproving, and the criticism of the b e -perfect is explicitly linked in many cases to the analysis of this form as an improper passive. 2.4.6 Linguistic relativity Summing up, we have seen that the b e -perfect was described (inadequately, compared to linguistic reality) as the passive of a neuter verb because of the
The BE -perfect and 19c grammar writing 29 10 9 8 7 6 5 4 3 2 1 0
b1800 1800 1810 1820 1830 1840 1850 1860 1870 1880 1890 Period 0=terms missing
3=neuter+perfect
4=intransitive+perfect
1=neuter+passive
2=intransitive+passive
Figure 2.11 Terminology (verb types) employed and analysis of the b e -perfect (British grammars)
inadequate mediaeval terminology available of ac t i v e , pas s i v e and n e u t e r verb types that English grammar writing had inherited from mediaeval Latin grammar writing, especially Donatus’ ars grammatica. We have also seen that in the 1860s this mediaeval terminology was ousted, and the more adequate distinction of t ra ns it iv e vs. int ra ns i t i v e verbs became the accepted view of the majority of grammar writers in Britain. After a short period of transition when the b e -perfect is described as the passive of an intransitive verb, we find accurate descriptions of the b e -perfect as a perfect, which quickly becomes the majority view as well, as Figure 2.11 shows. Almost as soon as the terminology became available in Britain, then, we find an adequate description of the phenomenon of the b e -perfect. This temporal order suggests that the majority of British grammar writers literally could not ‘see’ the be -perfect as a perfect before they had the terminology of t r an s i t i ve – i n t ra ns it iv e at their disposal. Once this terminology was in place, the transition in description happened very quickly. This process can be interpreted as a specialised kind of linguistic relativity, where the available technical terminology determined to a large degree what could actually be perceived and described accurately by grammarians.
30 Anderwald 10 9 8 7 6 5 4 3 2 1 0
1800 1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 Period 0=terms missing
3=neuter+perfect
4=intransitive+perfect
1=neuter+passive
2=intransitive+passive
Figure 2.12 Terminology (verb types) employed and analysis of the b e -perfect (American grammars)
In American grammar writing, the temporal sequence is quite different, as Figure 2.12 illustrates. On the one hand, as we have seen, the more adequate terminology of t r a n s i t i v e – i n t r a n s i t i v e was available much earlier, and became a majority position more than two decades before it did so in British grammar writing. However, a faulty description of the b e -perfect persists in American grammars until the 1890s. As Figure 2.12 shows clearly, the faulty analysis of the b e -perfect as a passive was simply transferred to the new terminology, and the combination of using the new term i n t r a n s i t i v e , but continuing to analyse the b e -perfect as a passive, is in fact the majority opinion of American grammar writers from the 1840s to the 1890s. In other words, despite the transition in terminology from ac t i v e – pa s s i v e – n e u t e r to t r a n s i t i v e – i n t r a n s i t i v e , the faulty description of the b e -perfect as a ‘passive’ is transferred from neuter verbs to intransitive verbs, perhaps compounding the impression of it being a highly irregular form, and therefore worth criticising. The first quotation in section 2.4.2 from Fisk from the 1820s already shows such a mixed system.
The BE -perfect and 19c grammar writing 31
Perhaps we can even link the persistent incorrect stance in American grammars to the fact that by the 1840s, the b e -perfect was only very rarely encountered in contemporaneous writing, as the corpus evidence in section 2.3 has shown. With the obsolescence of the b e -perfect construction, grammar writers’ intuitions about this form must have declined as well, adding to the discomfort that the faulty description quite clearly caused. In this way, criticising ‘improper passive forms’ could emancipate itself from actual linguistic reality and become a self-perpetuating topos in American grammar writing. As we have seen, just mentioning the key terms of ‘incorrect passive’ could invoke censure and trigger an automatic substitution by hav e , like a knee-jerk reaction, as clearly intended by Harvey in section 2.4.5. 2.4.7 Ways out of the quagmire We have seen two solutions to the descriptive problem of the b e -perfect. In British grammar writing, adequate terminology results in an adequate description (after a short transition period). I have argued that this can be regarded as an instance of linguistic relativity. In American grammar writing, a faulty description becomes independent of linguistic relativity, and turns into a self-perpetuating topos of grammar writing that was clearly copied from earlier sources, as near-identical formulations and lists of examples indicate. In fact, however, some grammar writers opted for a third option, and reinterpreted the remaining instances of the b e -perfect as an adjectival construction. As mentioned above, this analysis is clearly correct for the only remaining form, be + gone, today, and a representative summary of this stance is provided by Harriet Mathews at the end of the nineteenth century. A few intransitive verbs have the composition of the passive form while the subject represents the actor; as, ‘He is not here; for he is risen’ – B i bl e … ‘The melancholy days are come’ – B ryan t . / ‘All else is gone’ – Wh i t tie r … / This idiom is most frequently used in the verb to go. It expresses a delicate shade of meaning, making reference to the state which is the result of the action rather than to the act expressed by the verb. It should never be mistaken for the passive form, which belongs to the transitive verb only and the subject of which always represents the object of the action or relation expressed by the verb. In parsing, this composition of the intransitive verb is sometimes called the idiomatic form. (Mathews 1892)
However, we also see that, despite the essentially correct description, and the s tat e vs. ac t i o n distinction, Mathews’ passage is still formulated around the old passive terminology. Note that she speaks of a ‘passive form’, an ‘idiom’, or an ‘idiomatic form’. Nowhere does she use the term
32 Anderwald 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% b1800
1810
1830
1850 Period
1870
0=not adj
1=ADJ
2=only gone is ADJ
1M=Murray
1890
Figure 2.13 Analysing the b e -perfect as be + adjective
p erfect, which indicates how pervasive the American discourse of treating this phenomenon as a kind of faulty passive had become by the 1890s. As Figure 2.13 shows, trying to see, or to construct, a difference between the hav e -perfect and the b e - perfect is a persistent theme throughout the nineteenth century. Until the 1840s we see repetitions of Murray’s definition and description, stressing that the b e -perfect is not a passive, but ‘expressing … a state or condition of being’. From the 1840s, the separate status of b e g o n e is occasionally commented on explicitly, or g o is singled out as the only lexeme that could still occur with b e in this construction. As our corpus overview in Figure 2.2 has shown, this is quite an accurate reflection of linguistic facts, based on the figures from COHA for written American English, and corroborated by earlier studies for other varieties. On the other hand, the need to see semantic differences where we know none existed yet at the beginning of the century can again be traced back to a belief in the principle of o n e f o r m – o n e m e a n i n g : if we observe two forms, there must be a difference in meaning.
The BE -perfect and 19c grammar writing 33
2.4.8 A minority position: the be-perfect as a perfect As the preceding discussion has made clear, seeing the b e -perfect as a perfect was a minority position in grammar writing throughout much of the nineteenth century. This is particularly striking for American grammar writing, where the first correct description of the b e -perfect in terms of the perfect is not encountered until Quackenbos’ grammar from the 1860s. Quackenbos joins the current debate and explicitly rejects the prevalent definition of the be -perfect in terms of a passive. C aut i ons . –
In old writers we sometimes find the perfect of certain intransitive verbs formed with am in stead of have, and the pluperfect with was in stead of had. Thus: – Winter is [has] come; they are [have] arrived; when they were [had] gone; happiness was [had] flown. These forms are now rarely used, and should be avoided. Do not take them for passive tenses, which they resemble, but parse thus: Is come is an intransitive verb, used for has come; in the indicative mood, perfect tense, &c. (Quackenbos 1868: 123)11
It is also quite striking that knowledge of modern languages, like French or German, which we know was a typical accomplishment of educated men (and indeed women) at the time, does not seem to have helped in the analysis of the b e -perfect; indeed, only a dozen grammars mention modern languages in the context of the b e -perfect at all, and these all fall squarely in the second half of the century, a time when the faulty analysis in terms of a passive was already firmly established (at least in American grammar writing). This oddity is occasionally remarked upon in the few correct descriptions of the b e -perfect that we find. Thus Mason engages with the prevalent discussion of the b e -perfect in a footnote (after defining the form correctly as a form of the perfect), employing quite sarcastic language: * Some grammarians are pleased to order us to alter these forms [sc. with b e ] into ‘I have come’, ‘He has arrived’, &c. They had better at the same time mend the French and German languages, which at present still tolerate the forms, Je suis venu, Ich bin gekommen. (Mason 1858: 31)
Mulligan is another grammarian who refers to other modern languages: ‘In French, most neuter verbs are conjugated to this day with être. The same remark applies to other languages’ (Mulligan 1868: 132). In British grammar writing, on the other hand, correct descriptions of the be -perfect are encountered much earlier. Indeed, the earliest source 11
It is perhaps worth clarifying here that Quackenbos does not consider the passive a ‘tense’, but his ‘passive tenses’ seems to be a shorthand for ‘the passive of the respective tenses’, as his parsing of is come makes clear.
34 Anderwald
(just outside the time span covered by the CNG) is Priestley’s grammar from 1768 (the first edition from 1761 does not have the passage yet): It seems not to have been determined by the English grammarians, whether the passive participles of verbs neuter require the auxiliary am or have before them. The French, in this case, confine themselves strictly to the former … Yet I think we have an advantage in the choice of these two forms of expression, as it appears to me, that we use them to express different modifications of the sense. When I say, I am fallen, I mean at this present instant; whereas, if I say, I have fallen, my meaning comprehends, indeed, the foregoing; but has, likewise, a secret reference to some period of time past, as some time in this day, or in this hour, I have fallen; implying some continuance of time, which the other form of expression does not. (Priestley 1768: 127f.)
Although Priestley does not use the term ‘perfect’ (or one of its eighteenthcentury equivalents), he does mention ‘time past’ and ‘the continuance of time’. In addition, he describes the alternation clearly as an alternation, and tries to interpret it in terms of a subtle meaning difference between the hav e -perfect (as a more clearly temporal form) and the b e -perfect as a stative construction. Note also that he evaluates this choice positively, and claims this gives English an advantage in particular over French. Very much in contrast to his contemporary Lowth, and later Murray, it is striking how rarely Priestley was plagiarised. In the CNG, there are no formulations in passages dealing with the b e -perfect that would indicate any direct influence of Priestley on nineteenth-century grammar writers, whereas Lowth/ Murray’s influence was pervasive, as Figure 2.11 has demonstrated. (See also Straaijer 2011, who in his careful study finds no cases of direct influence of Priestley on later grammarians.) Despite the lack of a direct influence of Priestley on British grammar writing, occasional early correct descriptions of the b e -perfect are encountered, such as Allen (1824 [1813]) or Arnold (1838). Allen writes: ‘Many verbs neuter that denote motion or a change of condition, have not only the regular form of conjugation, but another, made from their past principle [sc. participle] and the verb to be, – expressing a past sense. The following, among many others, are of this kind; to go, to come, to fall, to vanish, to arrive’ (Allen 1824 [1813]: 26). In the 1830s, Arnold observes correctly: The past participle with the verb I am, is sometimes the present of the passive voice, sometimes the perfect of the passive voice, and sometimes the perfect of the active voice … ‘Henry is come’ (perfect act. venit) … When the participle, though passive in form, is the past participle of a neuter verb, this participle with the present tense of the verb to be forms a perfect of the active voice. My servant is come (venit). / [in smaller type] (a) The neuter verbs that form their perfects in this way, are generally verbs that imply motion. (Arnold 1838: 48f.)
The BE -perfect and 19c grammar writing 35
These two examples indicate that the influence of terminology was not all-pervasive. The majority of grammar writers, however, seemed simply unable to wriggle out of the confines of their terminological straitjackets, as Figures 2.11 and 2.12 have illustrated. 2.5 Summary I have shown that the b e -perfect with mutative verbs was in terminal decline in the nineteenth century and became a truly marginal construction. American English seems to have been in the vanguard of change here, since we could observe a relegation of forms with b e to marginal status (under 10 per cent) in the first third of the century. Perhaps it is hardly surprising that grammar writers found it difficult to account for this oddity in the verbal paradigm, and we have seen that the b e -perfect violates in both directions the principle of o ne m e a ning – o ne f o rm that underlay much grammar writing (see section 2.4.4). Especially in Britain, grammar writers’ difficulties in describing the b e -perfect must have been compounded by the lack of adequate terminology. As I have argued, the nineteenth century saw a change in grammar writing from dividing verbs into act i v e – pas s i v e – n e u t er to distinguishing them in terms of t ra ns it i v e – i n t r a n s i t i v e , and more generally from a semantic to a syntactic definition of grammatical terms. My new approach of quantitative grammaticography allowed me to date the switch-over point for American and British grammars separately. Before this switch-over, the b e -perfect presented a real difficulty in descriptive terms, and the mistaken ‘solution’ of describing the b e -perfect as a ‘passive of neuter verbs’ was dominant for much of the century. While in British grammar writing we saw the quick success of an adequate description in terms of a perfect once the requisite terminology was in place, American grammar writing developed the ‘improper passive’ as a topos that became quite divorced from actual language use. I have argued that the persistence of this incorrect description can actually be linked to the fact that in American English, the b e -perfect had become extremely marginal by that time, so that grammar writers lacked the necessary intuitions about this increasingly rare form. In this rather complex way, the evaluation of at least this one construction seems bound up with the terminology of the time, and depended to a large degree on established traditions of grammar writing. It remains to be investigated whether a similar dependency on terminology can be observed for other phenomena, and which other factors play a role in negative evaluations by normative grammarians. At any rate, looking at prescriptive comments in terms of quantity as well as quality promises to result in a more nuanced view of grammar writing in the nineteenth century, and is clearly an important prerequisite for determining the potential success (or lack thereof) of prescriptive comments more generally.
36 Anderwald
Appendix to Chapter 2 Table 2.1 The decline of the be-perfect in COHA (relating to Figure 2.2) be gone
have gone
be come
have come
be fled
have fled
be arrived
have arrived
1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
127 465 863 1123 1318 1372 1490 1441 1561 1809 1803 1943 2051 1927 1894 1667 1727 2160 2704 2652
48 643 980 1265 1403 1820 2302 2652 2787 3223 3362 3867 3541 3669 3212 2819 2706 2501 2755 2853
35 102 176 207 211 144 111 141 172 129 69 76 67 44 25 30 27 15 9 9
43 536 1080 1593 1902 2353 3258 3683 4081 4666 4934 5451 4759 4639 4297 3894 3633 3948 3789 4092
8 8 19 26 15 13 9 17 18 4 9 6 5 5 2 0 3 4 2 1
14 81 159 186 169 133 154 172 151 157 156 115 97 128 96 104 93 167 114 118
6 9 15 10 11 8 2 5 4 11 4 1 3 1 1 2 0 0 0 2
17 169 288 336 329 300 321 308 318 344 324 427 339 343 324 330 364 372 357 537
1810 1820 1830 1840 1850 1860 1870 1880 1890 1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000
be become 14 38 36 47 62 27 24 31 33 44 27 14 13 17 6 7 11 9 8 6
have become 59 517 916 1232 1398 1398 1642 1813 1928 1945 1954 2099 1949 2098 2277 2457 2426 2930 2818 2989
be risen 1 10 13 8 14 7 4 3 19 13 5 3 4 10 6 8 5 6 3 2
have risen 10 96 186 232 287 272 301 321 320 392 324 383 350 398 365 362 309 353 252 282
have fallen 33 454 671 814 851 887 937 946 934 961 858 941 800 731 652 648 605 765 816 909
be departed 0 5 10 8 5 3 6 7 5 2 1 4 3 3 3 1 2 1 0 0
have departed 4 59 101 138 132 89 108 74 88 106 118 113 76 46 59 46 47 46 30 45
be fallen 3 9 19 21 23 13 4 13 22 21 9 9 6 5 11 3 5 5 7 1
The BE -perfect and 19c grammar writing 37 Table 2.2 Defining verb types (British grammars) (relating to Figure 2.6) no 1: definition active – passive – neuter 1800 1810 1820 1830 1840 1850 1860 1870 1880 1890
1 2 1
1 3 1 7.5* 5 6
2 2
2: active3: active=transitive, transitive neuter=intransitive vs. activeintransitive 1
1 1 3 2.5 4.5 4 1
1 4.5 6.5 5.5 3 1.5 1
4: 5: transitive – transitive – intransitive intransitive + neuter
2 4 1
2.5 6 8.5 13 16.5 3 4
* Note: if grammars gave two definitions, each was counted 0.5 to preserve the absolute number of grammars.
Table 2.3 Defining verb types (American grammars) (relating to Figure 2.7)
1800 1810 1820 1830 1840 1850 1860 1870 1880 1890
no 1: definition active – passive – neuter
2: active3: transitive active=transitive, vs. active- neuter=intransitive intransitive
1 1 2.5* 1 2
1 1 6 3 3.5 2.5 1
2 3 2.5 3.5 1.5 1
4: 5: transitive – transitive – intransitive intransitive + neuter 1
1 1.5
1
1
3 1
1
* Note: if grammars gave two definitions, each was counted 0.5
1
2 4.5 11 7 7 5 13 25
3
Let’s not, let’s don’t and don’t let’s in British and American English Anita Aue r
3.1 Introduction An uncommon linguistic construction that I first came across in the Kay Scarpetta detective series by the American novelist Patricia Cornwell is the let’s don’t construction, as illustrated in (1) below: (1)
a. “Let’s don’t fight about this anymore.” (Lucy, Trace, p. 38) b. “Let’s don’t go into that,” she says. (Scarpetta, Blow Fly, p. 62) c. “Let’s don’t call him that.” (Benton, The Scarpetta Factor, p.280)
The construction is frequently used in the book series, which started in the 1990s. It is found in the speech of the protagonist Dr Kay Scarpetta, who is the chief medical examiner in Richmond, Virginia, as well as in the speech of other characters, both women and men. The immediate questions raised by historical linguists are the following: (a) When was the construction first used? and (b) how and in what circumstances did the construction develop further? A Google search to investigate the productivity of the construction revealed that let’s don’t was not all that common and that it had struck a nerve with some Cornwell readers as the comments in (2) below illustrate. (2)
a. Since when does [the highly educated and once classy] Kay
Scarpetta say idiotic things like “let’s don’t”? (www.amazon.com/Book-Dead-Kay-Scarpetta-No/product-reviews/ B002PJ4HNU?pageNumber=12, accessed in July 2012) b. […] I am shocked to find some grammatical errors that should have been picked up by the editor because they are repeated so many times. For example, Scarpetta, a forensic doctor says many times “Let’s don’t do that!” For crying out loud. English is my
I would like to thank Moragh Gordon, who extracted the examples from the Corpus of Historical American English (COHA) and did the initial categorisation. Many thanks to Marianne Hundt, the workshop participants, and the anonymous reviewer for their useful comments on an earlier version of this paper.
38
Let’s not, let’s don’t and don’t let’s 39
second language and I know you should say “Let’s not do that.” Finding such errors slows my reading incredibly because I keep wanting to correct it, and it is impossible. (www.goodreads.com/ review/show/10678534, accessed in July 2012) c. I have a hard time with some bad syntax used though, like “let’s don’t fight”. Sounds weird to me. (www.goodreads.com/user_status/show/7111978, accessed in July 2012) These selected comments indicate that let’s don’t is considered to be bad syntax and a grammatical error that readers do not expect to be used by highly educated characters. The reviewer in 2b points out that one ‘should say “Let’s not do that”’ rather than ‘Let’s don’t do that’. In fact, it appears that in Present-Day English (PDE) let’s don’t is not the only alternative to let’s not, as we also increasingly come across constructions with don’t let’s. All three forms can be found both in British (BrE) and American English (AmE), albeit at different frequency rates. For PDE, Algeo (2006: 30) observes that ‘don’t let’s is 7 times more frequent in British than in American, and let’s don’t is 4.5 times more frequent than don’t let’s in American’. This raises a number of questions, namely (a) how have these three constructions developed diachronically in BrE and AmE? (b) can let’s not V, don’t let’s V, and let’s don’t V be seen as competing forms, or (c) do they fulfil different functions in both varieties of English? (d) can gender differences be observed in the use of these constructions? This chapter takes up these questions. The time span 1700–1900, i.e. the Late Modern English (LModE) period, will be the main focus of attention as it allows for a comparative diachronic study of the negation constructions in BrE and AmE. As this period encompasses the American Declaration of Independence in 1776 and thereby the notional birth of American English, we can observe whether the linguistic constructions under investigation behave the same way or differ in the two varieties of English. Considering that the LModE period is also strongly associated with the codification and prescription of the English language, both British and American, metalinguistic comments about the constructions may shed light on their development and usage. In what follows, I will first look at previous research on the construction let us and let’s. The focus will also be on the historical development of let us and let’s as well as the negation patterns. Second, I will contribute to existing discussions on the topic by providing new empirical results that are retrieved from a range of corpora; this will allow us to shed some light on the questions raised above.
40 Auer
3.2 Let us and let’s – grammatical properties, gradience, and grammaticalisation The first-person plural imperative let us and the hortative construction let’s have already drawn a great amount of scholarly attention.1 The debates have focused primarily on the grammatical properties of the constructions, i.e. on their word-class status (main verb, modal auxiliary or particle), as well as their interconnectedness with semantics and pragmatics. As pointed out by, for instance, Huddleston and Pullum et al. (2002: 924), De Clerck (2004: 213), Collins (2006: 190–91), and Krug (2008: 315–16), there is a difference between 3a and b. (3)
a. Let us move on! b. Let’s move on!
Example 3a, which I will refer to as first-person plural imperative, may be described as ambiguous in that it can be interpreted as, on the one hand, a non-inclusive request for permission (i.e. the hearer does not belong to the group referred to by us) – Allow us to move on, or on the other hand, ‘as a hearer-inclusive proposal or suggestion for joint action’, which involves both the speaker and the hearer (De Clerck 2004: 213). As regards example 3b, it is not just an informal variant of (a) that contains the abbreviated objective pronoun us; it has also lost its non-inclusive interpretation and may therefore be regarded to be more restricted in semantic scope. In the case of Let’s move on, the ‘illocutionary function is restricted to a hearer-inclusive proposal for joint action’, as in Shall we move on? What can be observed here is that the development from the imperative let us to the hortative let’s involves a process of semantic bleaching (Huddleston and Pullum et al. 2002: 924) that also affected the pragmatic illocutionary functions (De Clerck 2004: 214). In her work on grammaticalisation, Traugott has used the hortative let’s construction as one of her standard examples (see Traugott 1995: 36–37; Hopper and Traugott 2003: ch. 1). In fact, it has been shown that this construction has undergone differing degrees of grammaticalisation, which again differ from dialect to dialect (Hopper and Traugott 2003: 10). Traugott (1995: 37) also points out a more recent development in the history of let’s. Example (4) is still considered a hortative in that it conveys ‘an appeal to the addressee(s) to help make the future state of affairs true’ (van der Auwera et al. 2005: 294); at the same time, it expresses what might be called the speaker’s ‘condescending support-style’ (see also Biber et al. 1999: 1117):
1
See Seppänen (1977: 524–27), Quirk et al. (1985: 148), Davies (1986: ch. 8), Stuurman (1990: 154–67), Joseph (1992: 135–36), Denison (1993: 319–22), Traugott (1995: 36–37), Huddleston and Pullum et al. (2002: 934–37), de Clerck (2004: 213–33), van der Auwera et al. (2005: 294–95), Algeo (2006: 30–31), Collins (2006: 190–91), Krug (2008: 315–45).
Let’s not, let’s don’t and don’t let’s 41
(4)
Let’s take our nap now, George.
A different construction (see example 5) barely involves the hearer as it appears to function primarily as an indication that the speaker is aware of the hearer’s presence while the speaker is addressing him/herself. (5)
Let’s see now, what was I looking for?
Quirk et al. (1985: 148) point out pleonastic variants of let’s, namely let’s us and let’s you and me, which are primarily found in AmE. These examples strongly suggest that ’s is no longer associated with us.2 Huddleston and Pullum et al. (2002: 935) state that the latter construction is used ‘widely enough to qualify as acceptable informal style in standard English’. This is thus a further indication that syntactically the specialisation of let has been taken a significant step further. As regards the historical development of adhortative let’s constructions, Krug provides a good overview of the diachronic development in his 2008 paper (see Table 3.1). Stage (i), the imperative construction, as in ‘allow us’, goes back to Old English. (6)
c.1000 æchom II, 31–2 Of eorþan we a¯rison. Læt us on eorþan gerestan oþ ðæt god us eft a¯rære. From earth we rose. Let us on/in earth rest Until that God us afterwards raise-up. (Example from Traugott 1995: 36)
In Middle English, an old subjunctive and an old periphrastic expression uton ‘let’s’ were replaced by the hortative construction (Traugott 1995: 36). (7)
c.1000 Ags. Gosp. Matt. xvii. 4 ȝyf þu wylt, uton wyrcean her þreo eardung-stowa. If you want to, let us make here three dwelling-places (Example from the OED online; see also Krug 2008: 321)
In the sixteenth century, the modal adhortative construction (stage v) gained ground and increased rapidly. Krug’s historical trajectory fits the cline of grammaticality, as put forward by Hopper and Traugott (2003: 7): content item > grammatical word > clitic > inflectional affix. As regards the negation patterns of let hortatives, the following can be said. Imperatives such as Open the door are negated by adding an initial Don’t or Do not, thus Don’t open the door. First-person imperatives, on the other 2
This observation can be supported by the occurrence of spoken forms such as s’get as variants of let’s get, as well as s’go and s’do (p.c. with Marianne Hundt).
42 Auer Table 3.1 Historical trajectory of adhortative let constructions (adopted from Krug 2008: 337) Stage (i) (ii)
(iii)
(iv)
(v) (vi)
Syntactically transparent transitive main (or pre-modal) verb construction, 2nd ps. imperative 2nd ps. imperative (permissive/causative) let us constructions inviting inferences related to adhortative, esp. in religious contexts Construction with independent (pre-) modal verb let plus free pro-form us; ambiguous between (i) and adhortative; typical early complements: verbs of speaking Unambiguous modal adhortative let construction, as co-ordinated with subjunctive adhortative Rapid growth of modal adhortative construction: encliticisation of us with reference to first-person plural inclusive Modal particle: extension of reference to non-first-person plural (generalised imperative for all persons, still rather rare)
Example (Modern English equivalents)
Attested since
Let us (‘allow us to’) go.a
Old English
Let us (‘make us’) hate the evil and love the right
13th c.
Let us speak of lusty life in Troy. (‘Allow us to speak’ or ‘Let’s speak’)
14th c.
Let us pause of Constance for a moment and speak we (subjunct adhort) of the Roman Emperor. Let’s go.
14th c.
Let’s wash your hands. Let’s everybody stay where they are. Let’s you and I do it.
20th c.
16th c.
a Less common than permissive ‘allow us’ is the causative ‘make us’ reading. This does not have pre-modal semantics but nevertheless serves an auxiliary function.
hand, are generally negated with the insertion of not after the pronoun following let (Huddleston and Pullum et al. 2002: 831), e.g. (8)
a. Let’s not talk about this. b. Let us not talk about this.
It should be pointed out that the negation with not has been taken as evidence that let resembles marginal modals (see Quirk et al. 1985: 148, footnote). As Huddleston and Pullum et al. (2002: 831) note, informally, the negation with don’t is frequently heard. Algeo 2006 and others remark that the variant where don’t precedes let’s is associated with BrE and don’t following let’s with AmE: (9) Don’t let’s talk about this.
(10) Let’s don’t talk about this.
From a semantic/pragmatic point of view, hortative let us, don’t let’s and let’s don’t can thus be interpreted as variants. As regards the historical development of the negation patterns, the OED online dates the first unambiguous affirmative let hortative as coming from Chaucer’s Canterbury Tales, thus late fourteenth century (see Traugott 1995:
Let’s not, let’s don’t and don’t let’s 43
37). Note that Krug (2008: 322) found examples from the early thirteenth century which may be interpreted as precursors that motivate the emergence of let adhortatives. Example (11) illustrates the first unambiguous affirmative let example. (11) c.1387 Chaucer Ship’s Tale (OED online)
Com doun to-day, and lat (imp) youre bagges stonde … What, Lat (hort) us here a messe, and go (subj) we dyne.
The first negated let adhortative appears to have emerged at almost the same time, as example (12) indicates. This example was also found in Chaucer’s Canterbury Tales (see Krug 2008: 338–39). (12) c.1390 Chaucer Man of Law’s Tale, Prol. 32 (OED online)
Lat vs nat mowlen thus in ydelnesse.
As regards the variant forms, the earliest example of do negation with hortative let can, according to the OED online, be found in 1738 (see example 13). The first example in the OED with a clitic ’s is recorded in 1869 – thus about 130 years later (see example 14). (13) 1738 Swift Polite Conv. iii. (D) (OED online)
No candles yet… don’t let us burn daylight.
(14) 1869 Galaxy June, 831 (OED online)
Don’t let’s get outside of more’n a bottle apiece, and that plain whiskey.
Based on previous studies, notably Krug (2008: 339), an earlier example can, however, be found in ARCHER, which is illustrated in (15) below. (15) 1820 Serle Exchange no robbery [drama] (ARCHER, 1820serl.d5b)
Ugh no, his mother was an angel – is an angel – don’t let’s talk of that; where is he now?
Considering that the first don’t let’s example was found in 1820 in BrE, the question arises whether this can be antedated in AmE. The following section will therefore be concerned with the comparative development of let’s not and its possible variant forms, as retrieved from various historical corpora. 3.3 Corpus study 3.3.1 Corpora The corpora and databases used for this investigation cover the LModE period and contain BrE data, AmE data or both.3 Because some of the 3
The low occurrence numbers of the constructions under investigation did not allow me to carry out statistical tests of significance. Nevertheless, the results do provide an indication as to the distribution and the development of the constructions.
44 Auer
constructions under investigation are not very productive, a wide range of different corpora/databases has been scrutinised, namely A Representative Corpus of Historical English Registers 3.1, the Oxford English Dictionary Online, Literature Online, Hendrik de Smet’s Corpus of English Novels, the Corpus of Historical American English, the Old Bailey Proceedings Online, the Old Bailey Corpus, the Corpus of Late Modern English Texts, the Lampeter Corpus of Early Modern English Tracts,4 and Eighteenth Century Collections Online. As regards data retrieval, in order to ensure that all relevant constructions are retrieved from the corpora/databases, a search for let was carried out and the constructions under investigation were then manually selected. 3.3.2 Results The data have been investigated with respect to a number of external variables, notably (a) regional variety (BrE versus AmE), (b) text type, with a particular focus on more speech-based texts, and also (c) gender wherever possible. As regards internal variables, special attention was paid to the verbs occurring with the constructions and the categorisation of these verbs into semantic fields. A corpus that allows us to investigate the development of the selected constructions in both varieties is A Representative Corpus of Historical English Registers (ARCHER) as it contains texts from both BrE and AmE from the period 1650–1999. In the first instance, the focus will be on the negation variants of the hortative let’s construction.5 Table 3.2 summarises the chronological distribution. Considering the extremely low numbers, only the raw frequencies will be presented. Table 3.2 suggests that let’s not is more productive than the other variants during the period under investigation, notably more so in AmE than in BrE; however, there are only very few occurrences. A search for let us not yielded five BrE and three AmE occurrences in the entire data, of which only one example can be unambiguously interpreted as hortative let’s. (16) 1943 Hamilton The Duke in Darkness [drama] (ARCHER, 1943haml.d7b)
Let us drink – shall we?
As regards let’s don’t, no instances are contained in ARCHER. As pointed out earlier, the first and only BrE instance of don’t let’s can be found in 1820 (see example 15 and Krug 2008: 339). ARCHER also contains one American example, which is dated 1878 (see example 17). This appears to be the first AmE attestation. 4
The Lampeter Corpus of Early Modern English Tracts did not contain any negated let’s constructions. In this study I shall only be concerned with contracted let us, i.e. let’s constructions.
5
Let’s not, let’s don’t and don’t let’s 45 Table 3.2 Negation patterns of the let’s hortative in ARCHER (according to subperiods)
1710–1749 1750–1799 1800–1849 1850–1899 1900–1950 1950–1990
let’s not BrE
let’s not AmE
don’t let’s BrE
don’t let’s AmE
let’s don’t BrE
let’s don’t AmE
– – – – – 1
– – – 1 – 3
– – 1 – – –
– – – 1 – –
– – – – – –
– – – – – –
(17) 1878 Eggleston Roxy [fiction] (ARCHER, 1878eggl.f6a)
Come, Mark, don’t let’s quarrel any more.
Not surprisingly, all relevant examples stem from dialogues; in fact, six out of the seven instances can be found in drama and one in fiction. Moreover, both the British and the American don’t let’s examples are followed by verbs denoting speech situations, notably talk and quarrel, while the verbs following let’s not have a wider semantic range: ask, let, be, get ahead, wait. In view of the fact that the ARCHER query did not yield sufficient results, more data representing spoken language, notably trial records and literary genres such as drama and fiction, were scrutinised. The trial records investigated are those contained in the Old Bailey Proceedings Online, covering the period 1674 to 1913, as well as the Old Bailey Corpus (OBC, 1720 to 1913), based on the latter.6 Interestingly, the Old Bailey data do not contain any instances of the let’s not construction. There are, however, two instances of don’t let’s, one dated 1738 and one 1862 (see examples 18a and b); this means that the BrE ARCHER example in (15) above has been pre-dated by more than eighty years. (18) a. Well, says I, let’s go in and drink don’t let’s have a dry Bargain;
so we went into the House, and I ask’d him (Machell) if he had not another Mare, and if another Man was not concerned in the Property of the Mares? (OBP, October 1738, trial of John Machell and Richard Wilkinson (t17381011–1)) [indicted for stealing two black Mares] b. ‘[L]et me have it’ – they both positively denied that there was anything passed – I said, ‘Don’t let’s hurt the baby; I am determined to have it’ – (OBP, March 1862, trial of John Pitt, Jane Close, Harriet Young, Margaret Young (t18620303–297)) [Feloniously having in their possession a mould on which was impressed the obverse side of a shilling]
6
It is noteworthy that while the corpus is based on the proceedings online, it does not contain the same/full number of texts.
46 Auer
Apart from the 1738 don’t let’s example, the OBC data do not shed any more light on the possible variant constructions. It is also noteworthy that there are no do let’s examples in the database. Similarly to the OBC findings, the don’t let’s construction is the only negated let’s pattern to be found in the Corpus of Late Modern English Texts (CLMET; BrE only; 1710–1920). The latter corpus contains twelve don’t let’s instances, of which the first examples can be found in Maria Edgeworth’s collection of children’s stories The parent’s assistent (1796). A couple of do let’s constructions can also be found in CLMET; interestingly enough, they postdate the don’t let’s examples. Just as in ARCHER, there are thus no instances of let’s don’t in the OBC and CLMET, which supports previous observations by other linguists (see e.g. Visser 1963–73; Denison 1998: 253; Algeo 2006: 30) that the latter construction is strongly associated with AmE. This is indeed confirmed in a search of the database Literature Online (LION). LION, which distinguishes between poetry, drama and prose, contains three let’s don’t examples, all of which are found in American drama from the late nineteenth century (see examples 19a–c). (19) a. EVELYN: Well, now let’s don’t lose any time! (LION, William
Dean Howells (1837–1920), Parting Friends)
b. LIZ: Let’s don’t have a wrangle. (LION, Charles Hale Hoyt (1860–
1900), A Midnight Bell)
c. BONITA: Let’s don’t talk about it, Mother. (LION, Augustus
Thomas (1857–1934), Arizona)
What these examples have in common is that the playwrights are Americanborn and male and that the speakers using the let’s don’t construction are female. I will return to this observation later. As far as don’t let’s in LION is concerned, the construction occurs most frequently in drama (87 hits), followed by prose (61 hits) and then poetry (16 hits). The earliest BrE example from 1690 antedates the OBC example by forty-eight years. It stems from John Crown’s play (1640?–1712) The English frier (1690) (see example 20a). Other early English playwrights who used the don’t let’s construction were Sir John Vanbrugh in his comedy The provok’d wife (1697) and William Congreve (1670–1729) in The way of the world (1700).7 (20) a. Pray don’t let’s lose your company. (LION, John Crown (1640?–
1712), The English frier)
b. Don’t let’s mind him. (LION, Sir John Vanbrugh (1664–1726),
The provok’d wife)
c. Good, good, hang him, don’t let’s talk of him. (LION, William
Congreve (1670–1729), The way of the world)
7
Visser (1963–73: §1448) states that Congreve’s example is the earliest don’t let’s example (see also Rissanen 1999: 247–48). Visser dates the example 1696, but as the play was only premiered in 1700, Visser’s date is most likely incorrect.
Let’s not, let’s don’t and don’t let’s 47
Interestingly enough, the next English examples of don’t let’s can be found in translations of Molière’s work by Henry Baker (1698–1774) and James Miller (1706–1744), notably in The hypochondriac, The princess of Elis, and Tartuffe (1739) (see examples 21a–c). (21) a. Please don’t let’s talk about that, dear heart. (LION, Baker and
Miller (transl.), Molière’s The hypochondriac, 1739)
b. Don’t let’s interrupt him now, but turn this way to meet them again
by and by. (LION, Baker and Miller (transl.), Molière’s The princess of Elis, 1739) c. Give over; don’t let’s make things worse. (LION, Baker and Miller (transl.), Molière’s Tartuffe, 1739) Based on LION, the first female author to have used the don’t let’s construction is Fanny Burney (1752–1840) in her novels Cecilia (1782) and Camilla (1796); this is close in terms of date to Maria Edgeworth’s use of the construction. The earliest AmE example can be found in the play The village wedding (1790) by an anonymous author (see example 22). (22) Old Peasant. Come cheerly, neighbour, don’t let’s disturb ourselves
with sad thoughts, we met to be merry. (LION, anonymous, The village wedding, 1790)
After 1800, a gradual increase can be observed in the use of the construction in both BrE and AmE. Well-known British writers that used the construction are, for instance, Charles Dickens (1812–1870) and William Makepeace Thackeray (1811–1863). In contrast to the ARCHER and the OBC data, Literature Online clearly shows that let us not (hearer-inclusive) and let’s not are and remain the prevailing constructions in literary texts. While LION allows us to trace early occurrences of selected constructions and provides an impression of their developments, it is a database rather than a representative corpus and therefore less suitable for systematic searches. Thus, in order to shed more light on the differences, particularly as to the infinitive choice in the constructions, the 25-million-word Corpus of English Novels (CEN), which covers the period 1884–1992 and contains novels by both British and North American novelists, was used for a systematic search. The findings in terms of distribution are provided in Table 3.3 below. As Table 3.3 shows, there are no BrE let’s not examples in the corpus and only a few AmE examples of the construction, while don’t let’s is the dominant variant in both varieties. It is noteworthy that CEN does contain a far higher number of let us not instances in either BrE and AmE, but let us not will not be discussed here in detail. The findings in the table give the impression that the prevailing variant is don’t let’s, i.e. ninety-two instances in total. It is clearly more productive in BrE than in AmE. Even though the numbers are very low, an increase in usage of the construction can be found in both varieties, notably up to the 1900s for BrE and the 1910s for AmE,
48 Auer Table 3.3 Negation patterns of the let’s hortative in CEN (according to subperiod; raw frequencies are given in brackets; the incidence figures are per 1 million words)
1880s 1890s 1900s 1910s 1920s
let’s not BrE
let’s not AmE
don’t let’s BrE
don’t let’s AmE
let’s don’t BrE
let’s don’t AmE
– – – – –
– – 0.8 (3) 2.1 (4) 2.8 (2)
1.2 (5) 2.8 (16) 7.7 (37) 6.9 (13) –
0.9 (1) 0.87 (2) 2.5 (9) 3.7 (7) 2.8 (2)
– – – – –
– – – 0.5 (1) –
which is followed by a decline. As for the constructions in AmE, don’t let’s is more frequently used than let’s not and let’s don’t. In fact, there is only one example of let’s don’t (see 23 below), which is contained in Layman Frank Baum’s famous novel The Emerald City of Oz. (23) “Let’s don’t wait,” returned Dorothy. “I’ve heard of the Rigmaroles,
and wondered what they were like; but now I know, and I’m ready to move on.” (1910, The Emerald City of Oz (53,615), Lyman Frank Baum; CEN)
As regards the verbs used in combination with the different variants, it may be observed that verbs denoting speech situations prevail with don’t let’s in many subperiods, both in British and American novels (see Table 3.4 below). This very much resembles the situation found for the early attestations of affirmative adhortative let us constructions outlined in Section 3.1 above. Overall, in the BrE examples 39 per cent of the complements denote speech situations, and 33 per cent in the AmE examples. In both varieties, talk is the verb most frequently found with don’t let’s. Similarly, the few AmE let’s not examples contain talk and quarrel. Let’s don’t is the only construction that combines with a different verb, namely wait. As for a comparison of the constructions in BrE and AmE, the scant evidence suggests that let’s not and don’t let’s are used with similar verbs and may therefore also be seen as variants of one another, notably more so in BrE than in AmE. In order to shed more light on the let’s don’t construction, particularly its development in AmE and the verbs used, data from COHA (1810–2009) was investigated. Figure 3.1 and Table 3.5 below present the historical development of negated let’s constructions. Figure 3.1 also includes the development of the let us not construction in order to show how different its frequencies are in comparison to the three main constructions under investigation. Figure 3.1 and Table 3.5 reveal that since the 1890s let’s not is the prevailing variant in comparison to don’t lets and let’s don’t. Before then, the numbers of occurrences are so low that it is difficult to interpret the data. It may be observed, however, that don’t let’s, the form that has been more strongly
Let’s not, let’s don’t and don’t let’s 49 Table 3.4 Verbs occurring with let’s not, don’t let’s, and let’s don’t in CEN (according to subperiod) let’s not BrE
let’s not AmE
don’t let’s BrE
don’t let’s AmE
let’s don’t BrE
let’s don’t AmE
1880s
–
–
talk
–
–
1890s
–
–
go, magnify
–
–
1900s
–
make, take, worry
beat (around), beat (about) (2×), talk, say, think, forget parley, bother
–
–
1910s
–
talk (2×) quarrel make
talk, think, hang, go, pay, begin, return
–
wait
1920s
–
talk commit
say (2×), talk, have, go talk (5×), speak, quarrel, – (2×), think (2×), have, wish, make, waste, go, talk (6×), be (5×), tell, speak, quarrel (2×), remember, forget, squabble, hear, spoil, – (2×) have (2×), get, think (2×), make, go (2×), work, neglect, stand, waste (2×), burgle, play, let talk (5×), quarrel, argue, discuss, make, (h)ave, believe, make, forget –
talk mention
–
–
associated with BrE, was used more frequently in the first half of the period under investigation. One could therefore conjecture that this ‘British’ form gradually gave way to other forms in the AmE variety, notably let’s don’t, which starts being productive in the 1880s, based on COHA data. A comparison of incidence figures between don’t let’s and let’s don’t reveals that the former is used more frequently, i.e. except for the 1950s, the 1980s and the 1990s. Interestingly, the 1820s don’t let’s instance is not pre-dated by any do let’s examples. The latter occur from the 1830s onwards and are a lot less frequent than don’t let’s. In order to explain the reason for the don’t let’s order, I agree with Rissanen (1999: 322) who claims that it is most likely ‘the wish to emphasise the prohibition by an early placement of the negative particle’. The earliest let’s don’t example found in COHA, as well as in LION, is Charles Hale Hoyt’s Midnight Bell example from the 1880s (see example 19b). Interestingly, let’s do examples are present in COHA from the 1830s onwards and are a lot more frequent than the let’s don’t examples, as illustrated in Table 3.6 on page 52.
50 Auer 14 12 10 8 6 4 2
2000s
1990s
1980s
1970s
1960s
1950s
1940s
1930s
1920s
1910s
1900s
1890s
1880s
1870s
1860s
1850s
1840s
1830s
1820s
1810s
0
Period let us not
let’s don’t
let’s not
don’t let’s
Figure 3.1 Distribution and historical development of negated let and let’s constructions in COHA
Based on these findings, it may thus be assumed that let’s don’t is a result of let’s do where do was negated instead of let’s. The conclusion may be drawn that we are dealing with a conflict of different analogies here, i.e. on the one hand, that all imperatives are negated with don’t, and on the other hand, that all first-person plural imperatives start with let’s (see Denison 1998: 254). The language user is thus unsure whether to negate do, as in let’s don’t, or whether to add not after let’s, as in let’s not do. A comparison with let’s not do shows that the latter construction only became productive (in COHA data) in the 1930s. The CEN corpus results revealed that the verbs occurring with let’s not and don’t let’s were similar. This is confirmed by the COHA data, where let’s don’t follows a similar pattern. Table 3.7 provides the complements of don’t let’s and let’s don’t. Due to the great number of let’s not occurrences in COHA, namely 1,277, the examples will not be presented in a table. Until the 1870s, let’s not, as opposed to don’t let’s, does not combine with verbs denoting speech situations but verbs such as allow, eat, butchered, back out or spend. Both constructions continue to be used with many different verbs that cover a wide semantic range, but these are mostly single occurrences, while verbs such as talk and quarrel are used repeatedly. If all
Let’s not, let’s don’t and don’t let’s 51 Table 3.5 Distribution of negated let’s constructions in COHA (raw frequencies are given in brackets; the incidence figures are per 1 million words)
1810s 1,181,022 w 1820s 6,927,005 w 1830s 13,773,987 w 1840s 16,046,854 w 1850s 16,493,826 w 1860s 17,125,102 w 1870s 18,610,160 w 1880s 20,872,855 w 1890s 21,183,383 w 1900s 22,541,232 w 1910s 22,655,252 w 1920s 25,632,411 w 1930s 24,413,247 w 1940s 24,144,478 w 1950s 24,544,831 w 1960s 23,977,232 w 1970s 23,815,191 w 1980s 25,315,978 w 1990s 27,941,535 w 2000 29,567,390 w
let’s not
don’t let’s
let’s don’t
0
0
0
0.14 (1)
0.14 (1)
0
0.07 (1)
0.07 (1)
0
0.06 (1)
0.12 (2)
0
0
0.12 (2)
0
0.12 (2)
0.18 (3)
0
0.32 (6)
0.54 (10)
0
0.29 (6)
0.48 (10)
0.05 (1)
0.38 (8)
0.28 (6)
0.05 (1)
0.79 (18)
0.67 (15)
0.09 (2)
2.16 (49)
1.5 (34)
0.13 (3)
1.95 (50)
1.13 (29)
0.31 (8)
3.15 (77)
1.19 (29)
0.41 (10)
5.14 (124)
0.75 (18)
0.08 (2)
7.54 (185)
0.53 (13)
0.81 (20)
6.67 (160)
0.54 (13)
0.29 (7)
6.38 (152)
0.34 (8)
0.17 (4)
4.42 (112)
0.2 (5)
0.28 (7)
5.37 (150)
0.0 (2)
0.11 (3)
5.92 (175)
0.1 (3)
0.07 (2)
subperiods are taken together, 21.5 per cent of all let’s not examples contain verbs that denote speech situations.8 This may be compared to don’t let’s, of 8
This percentage figure is based on unambiguous examples such as ask, say, tell, talk, speak. Ambiguous verbs like fight have thus not been included here. If we break down the percentages of verbs denoting speech situations into subperiods, the following picture emerges: 1870s: 83%, 1880s: 33%, 1890s: 38%, 1900s: 38%, 1910s: 31%, 1920s: 36%, 1930s: 30%, 1940s: 20%, 1950s: 23%, 1960s: 16%, 1970s: 25%, 1980s: 23%, 1990s: 13%, 2000s: 14%.
52 Auer Table 3.6 Comparison of let’s do and let’s don’t occurrences in COHA Period
let’s do
let’s don’t
let’s not do
1830s 1840s 1850s 1860s 1870s 1880s 1890s 1900s 1910s 1920s 1930s 1940s 1950s 1960s 1970s 1980s 1990s 2000s
2 1 0 5 8 4 15 14 29 21 21 29 33 41 47 60 100 85
0 0 0 0 0 1 1 2 3 8 10 2 20 7 4 7 3 2
0 0 0 0 0 0 0 0 0 0 1 0 2 5 4 0 6 7
which 38.7 per cent of the examples contain the latter types of verbs, and to let’s don’t, where we find 31.4 per cent of these complements. The COHA data thus suggest that the constructions don’t let’s and let’s don’t are more strongly associated with speech situations than let’s not. Even though COHA makes a distinction between different genres, notably fiction, popular magazines, newspapers and non-fiction books, a subdivision into genres does not allow us to shed more light on the constructions. It has been observed, though, that the great majority of don’t let’s and let’s don’t instances are found in fiction, notably in fictionalised dialogue. At the beginning of this chapter I pointed out that the let’s don’t construction had struck a nerve with readers of Cornwell’s book series. They labelled the construction as bad syntax and a grammatical error that educated people are not expected to use. Considering that the LModE period is strongly associated with the codification and prescription (see, for instance, Beal 2004; Tieken-Boon van Ostade 2008; Auer 2012; Anderwald, this volume), one would assume that the alternative constructions to let’s not had been picked up on by grammarians and language guardians. In fact, as regards metalinguistic comments on the let’s don’t and don’t let’s negation variants, I have not been able to find any comments in LModE grammars and usage guides – neither British nor American. It ought to be noted, though, that I did not carry out systematic searches. While the database Eighteenth-Century Collections Online (ECCO) contains several don’t let us examples, I was not able to find examples with contractions nor comments
Let’s not, let’s don’t and don’t let’s 53 Table 3.7 Verbs occurring with don’t let’s and let’s don’t in COHA (according to subperiod)
1810s 1820s 1830s 1840s 1850s 1860s 1870s 1880s 1890s 1900s 1910s
1920s 1930s
1940s 1950s
don’t let’s
let’s don’t
– sboil [spoil] quarrel go be, break, do, stand, talk (2), worry get, talk, hide be, do, fight, go, hear, leave, mince (the matter), quarrel, say, talk (2), think, wander off, worry go, quarrel (2), talk (5), waken bandy about, go, have, overlook, stand, talk (8) get on, get upon, give, go, leave, say, scare, take (2) talk (5), think, work up be crazy, be in a hurry, call, fool, get into, git on, go into, go on, have, look, miss (2), play, pretend, say, take, talk (10), tell (2), think (3), try, wait, waste, worry (2) attempt, be ridiculous, bother, break, camp, do, fool, go (2), quarrel (5), settle down, spoil, talk (2), think, waste, worry be, break up, confuse, dance, dwell on, forget, get frank, gi[e]t discouraged, go into (2), have, leap off, look, run, scrap, speak, squabble, take (2), talk (10), think, try act, be (2), cross, do, drip, forget, get serious, go over, let, quarrel (3), start, talk (7), waste be beastly, begin, forget, go, have (3), leave, let, play, start, talk (2)
– – – – – – –
1970s 1980s
be, bother, fight (2), get started, go (2), quarrel, shove, take, talk (2), wake get (2), have, quarrel, talk, think (2), waste speak, do, quarrel, sit, get
1990s 2000s
get personal, waste keep, bolt, be
1960s
have lose be spoiling let, go, mix
be sure, fall out, go, talk (2), tell cross, do, fall out (2), go, say (3), start, talk do, go (2), talk (2) ask, be sorry, do, fight, get (2), get off, go back, have, keep, let (2), live, make, part, sandbag, start (2), talk (3) be (2), hassle, let, make, rush, talk go off, take, talk (3) do, fall out, fuss, get sidetracked, indict, tip, use cry, fight, go talk, put
on the constructions. Similarly, the investigation of random samples of nineteenth-century grammars published in Britain and North America did not yield any results.9 This strongly suggests that evaluative comments on the constructions are a more recent, i.e. a twentieth-century, development. Notably, in the Oxford Companion to the English Language (1992) it is noted 9
I based my choice of grammars on Lieselotte Anderwald’s Collection of Nineteenth-Century Grammars (CNG) as the basis for this investigation and consulted grammars that were freely accessible on google books or archive.org.
54 Auer
that ‘the chiefly BrE variant don’t let’s (argue) and the AmE variant let’s don’t (argue) [are] often reproved as non-standard’ (1992: 45). As part of Walt Wolfram’s American Tongues project,10 a Mississippi woman distinguished between three different ways of speaking, notably ‘cultured’, ‘white trash’ (associated with uneducatedness) and ‘Black’. She illustrates the third with the following example: (24) Let’s don’t let no stump knock no hole in the bottom of this here boat.
(Walt Wolfram – American Tongues project: Mississippi woman 2)
While one can observe that the let’s don’t construction occurs in novels set in southern parts of the United States and/or texts written by southern writers, e.g. Patricia Cornwell was born and raised in the south, COHA does not allow us to draw conclusions as to geographical and social variation at this stage. I observed earlier in this chapter, i.e. when illustrating LION data, that women first used the let’s don’t construction. The COHA data corroborate this finding and reveals that women (i.e. female characters, speakers) continued to use the construction more frequently than men. Mark Liberman, in a Language Log entry dated 22 June 2006 in which he compared Google news, yahoo news, and msn news with respect to the three negation variants, also noticed that let’s don’t is found more frequently in the American south. Notably, three out of ten instances from Google news were used by the former US president George W. Bush. Liberman makes the following observation: My impression is that ‘let’s don’t’ is not only somewhat more informal that [sic] ‘let’s not’ – which will already be tagged as informal by some because of the contraction – but also is in commoner use in the south. (There are four examples of ‘let’s don’t’ in a corpus of telephone conversations that I searched, three used by southerners and one by someone from a ‘midland’ state, a region that includes places like Oklahoma and Missouri. The 67 instances of ‘let’s not’ are from all over.) (http://itre.cis.upenn.edu/~myl/ languagelog/archives/003277.html, accessed in October 2012)
Based on Liberman’s observations, we can conclude that we are dealing with an interesting change in progress that ought to be followed in the years to come, particularly with respect to geographical, social and gender variation. 3.4 Concluding remarks In this chapter, I aimed at shedding some light on the diachronic development of the let’s not, don’t let’s, and let’s don’t patterns in BrE and AmE and 10
This project surveys ‘regional variety, standards, the influence of foreign languages, ethnic and gender differences, and presents attitudes towards and stereotypes of U.S. regional ideolects and sociolects’ (see www.uta.fi/FAST/US1/REF/tongues.html, accessed in February 2014; www.cnam.com/downloads/amt_ts.html, accessed in February 2014).
Let’s not, let’s don’t and don’t let’s 55
raised the question whether they can be seen as variant forms of the same construction. As we were dealing with low-frequency items, many different corpora and databases had to be scrutinised. Concerning the first occurrences of the don’t let’s and let’s don’t constructions in the two varieties, it was possible to show that the first don’t let’s examples are found in English seventeenth-century drama, as has also been previously observed. The first AmE example is from 1790. Even though don’t let’s is more frequently found in BrE, it is also productive in AmE throughout the period under investigation. The pattern was first used by men around 1700, followed by women approximately eighty years later. Considering that do-periphrasis with affirmative imperatives can already be found in Old English, one would expect do let’s to pre-date the don’t let’s construction, but this was not verified by the data contained in the investigated corpora. The placement of don’t before let’s may thus be interpreted as an attempt to emphasise the prohibition (see Rissanen 1999: 322). As regards let’s don’t, it was possible to observe that let’s do examples pre-date let’s don’t examples by around fifty years, which suggests that language users were not sure whether to negate do or whether to add not after let’s. The investigation of the corpora/databases also revealed that let’s don’t is only found in AmE and appears to be predominantly found with speakers from the southern states. The LION and COHA data reveal that women, i.e. female characters and female speakers, used the construction first and also more frequently than men. As regards the functions of the three constructions, a comparison of verb types in the examples has revealed that don’t let’s and let’s don’t are even more strongly associated with verbs rendering a speech situation than let’s not; nevertheless, the data suggest that we are dealing with forms that may be considered variants of one another. It remains to be seen how the different constructions will develop further in years to come and whether don’t let’s and let’s don’t will gain more ground or gradually disappear.
4
Do we got a difference? Divergent developments of semi-auxiliary (have) got (to) in British and American English Christian Mair
4.1 Grammaticalisation: layering, long-term drifts and short-term ripples Unlike grammatical subsystems such as nominal inflection, which – after undergoing dramatic changes in late Old English and early Middle English – settled into a new stable state which has lasted for around 800 years, modal verbs and related expressions have been ‘on the move’ (Leech 2003: 223) throughout the recorded history of the English language. Within the modal system, change has been particularly dynamic in the expression of obligation and necessity. From the Early Modern English period, the anomalous finite modal verb must, itself developed from the past tense of the Old English preterite-present verb motan, was complemented by have to, and subsequently (have) got to. As the modal uses of these latter two forms derived from preceding possessive ones, the interaction of subsequent grammaticalisation-related developments (such as phonetic reduction and eventual disappearance of have in have got) with possessive and modal use brings in an additional complication. In Present-Day English (PDE), these diachronic developments are reflected in a complex layered system which allows semantic, stylistic and, of course, dialectal variation, as is shown in the following quotation from a mid-twentieth-century British novel: (1) Sebastian’s life was governed by a code of such imperatives. ‘I must
have pillar-box red pyjamas,’ ‘I have to stay in bed until the sun works round the windows,’ ‘I’ve absolutely got to drink champagne tonight.’ (Evelyn Waugh, Brideshead Revisited [1945], 1960: 48)
From the vast corpus-based literature on modals and related expressions in varieties of English (e.g. Krug 2000; Leech 2003; Mair 2006: 103–08; Leech and Mair 2006; Leech et al. 2009: 71–117; Collins 2009; Seggewiß 2012, and further references therein) it becomes apparent that must has been on the decline in practically all varieties of English during the past two centuries, and that there is a near universal trend in World English for the spread of have to as the default expression of obligation and necessity 56
Semi-auxiliary (have) got (to) in BrE and AmE 57
(see Collins and Yao 2012: 47–49). This is not really surprising, as have to suffers from none of the paradigmatic deficiencies displayed by must. Unlike must, have to can be used in all tenses and has an infinitive. Over the past two centuries, do-support in questions and negation has become more common. According to Trudgill et al. (2002), this development took place first with dynamic uses of have + NP (e.g. Do they have parties every week? / They don’t have parties every week) and subsequently spread to the stative-possessive and modal uses, with North American English leading the way (Do we have any money? / We don’t have any money; Do they have to pay now? / They don’t have to pay now). More recently, the progressive has become common in the dynamic transitive and modal uses, as well: We are having several problems; We are having to revamp the website. Both developments have moved semi-auxiliary have (to) further towards the regular verbal paradigm of PDE. In view of this development, it is not surprising that the emergence and spread of (have) got (to), rapid at first in the wake of have (to) in the late eighteenth and nineteenth centuries, has slowed down considerably from the mid twentieth century; after all, the paradigm of (have) got (to) is defective in similar ways to that of must. Today, have got (to) is frequently listed as a statistical Briticism in comparisons with American English (AmE). Trudgill and Hannah, for example, state that ‘in all forms of English, stative have can be replaced by have got, although this is more common in British than in American English’ (2008: 66; see also Biber et al. 1999: 162, 216). Detailed corpus studies (e.g. Mair 2009, based on an analysis of the Santa Barbara Corpus of Spoken American English and five components of ICE, the International Corpus of English) support this claim to the extent that have got is indeed a significant presence in informal and spoken British English (BrE). In formal and written registers the form is rare also in BrE, and it never seems to have made significant headway even in British-derived New Englishes. The present chapter is concerned with these long-term drifts only to the extent that they are the background to a short-term diachronic ripple which has unfolded in North AmE during the past 150 years and which gives us instances of a different kind of layering from the one illustrated in (1) above. Examples (2) to (5) are from COCA, the Corpus of Contemporary American English and, alongside more familiar and common expressions of possession, obligation and necessity, feature the phenomenon at the centre of attention here, namely do-support with got: (2) And when asked for an explanation, Wallace only grew surly. ‘I don’t (3)
care what people say. I don’t got to explain myself to nobody, man,’ he snapped. ‘That’s the last thing I have to do.’ (COCA, Press 2007) and what do we have to do to – do I got to hug you like – what – what do we need to do – what should we do? You know? (COCA, NBC_Today 2002)
58 Mair
(4) It’s what happens. You meet. You go through the, [‘]Hey, you’re cute,’
(5)
whatever, you know what I mean? Then you’ve got to work, you’ve got to pay bills, you’ve got kids, you have to pick them up from the day care. That new girlfriend don’t have to do that. She ain’t got to do nothing but when you come over, she got on her lingerie, she got on whatever you want her to have. She’s cooking your favorite meal, she’s playing your sports. Whatever she say – whatever you say you like, believe me, when you get there, she’s going to be ready for you. (COCA, Ind_Oprah 2005) [Mr KAYSEN] So you have to be careful. The Mediterranean salad that we’re going to put with the lamb, you know, it’s something that I think, for springtime, is nice and fresh. [WRAGGE] Yeah. [Mr KAYSEN] You want to use things that are very, very fresh at this time of the year. [WRAGGE] What do we got? We’ve got a little a… [Mr KAYSEN] We have tomatoes. We have olives. We have feta cheese. [WRAGGE] These olives look fantastic. [Mr KAYSEN] Oh yeah. These are the – these are the kalamata olives. We’re just going to put all of that in. If you want to toss this, here. [WRAGGE] You got a… [Mr KAYSEN] Sure. I’m just going to add a little bit of red wine vinegar and olive oil. OK. And then as you’re going, you know, you just got to OK. And then as you’re going, you know, you just got to turn the lamb every once in a while like that. [WRAGGE] Yeah. [Mr KAYSEN] This one. OK. And you’re just going to move it around. [WRAGGE] Got it. [Mr KAYSEN] Now, the thing is, this is going to be – this going to be the final process. [WRAGGE] Got it. (COCA, CBS_SatEarly 2008)
It should be pointed out that the first three speakers represented here happen to be African Americans, i.e. members of a community in which dosupport with got has been established for some time (see e.g. the relevant entry in DARE, the Dictionary of American Regional English). Example (2) was produced by basketball professional Ben Wallace, whose don’t got to is clearly marked as non-standard through the co-presence of negative concord (don’t got to explain myself to nobody). The speaker in (3) is a member of the all-female African American rhythm-and-blues trio Destiny’s Child and uses do I got to hug you without any other conspicuous signs of nonstandard English, so that the register is best classified as informal-colloquial
Semi-auxiliary (have) got (to) in BrE and AmE 59
here. The speaker in example (4) is one of a multi-ethnic lot of ‘seven cheating husbands’ invited on to Oprah Winfrey’s famous talk show. While the passage does not feature any instance of do-support with got in questions and negations, it contains all the forms needed to understand how this phenomenon could arise. Presence of have in have got is highly variable for this speaker, and the standard negation hasn’t / haven’t / hadn’t got does not seem to be part of the repertoire, non-standard ain’t got being used instead. This interesting constellation – a standard negation which is hardly used any more in contemporary North AmE1 and two competing non-standard variants, a traditional one with ain’t and a more recent one with don’t – will be returned to below. Example (5) makes clear, however, that do-support with negated (have) got is not restricted to this demographic group any more. Both Chris Wragge, the media professional hosting the show, and Gavin Kaysen, the chef, are white. The example is quoted at length because of the particularly rich variation it displays in the use of the grammatical constructions under investigation here, and also because it is possible to check the transcription against the video recording of the broadcast, which is still available on the Web.2 The communicative context of this cooking show is relaxed and informal but clearly within the range of standard usage. To express obligation, Mr Kaysen uses have to, got to, want to and even going to, for possession he has plain have. Wragge only uses the possession senses: have is optional for him in have got, but interestingly his use of do we got does not trigger a suppression of the auxiliary have in the following we’ve got. Languagehistorically, the frequent suppression of have must have been a precondition for do-support to apply with got. Synchronically, the historical chain of developments does not seem to constrain usage for this speaker. 4.2 Do-support with got (to): diachronic origin and synchronic spread Mair (2012) explores the diachronic origins of do-support with got. The oldest mention of the construction in the reference literature was from DARE, which listed a first attestation from 1938, characterising it as used by ‘younger people, colloquial (s.v. get)’. As this very informal and, until recently, distinctly non-standard usage is very unlikely to show up in our written records, a large number of corpora, digital text archives and the 1
Note that this is the only standard negation available for (have) got. As an anonymous reviewer points out, the default standard negation in North AmE in such cases is negated have, without got, which of course is vastly more frequent in COCA than any of the forms dealt with here. 2 See www.cbsnews.com/video/watch/?id=4104646n, accessed on 25 July 2013. The portion transcribed as example (5) is found at 4 minutes 40 seconds and lasts until 5 minutes 24 seconds.
60 Mair
World Wide Web were consulted in an attempt to antedate the phenomenon. The search was successful and uncovered examples from the second half of the nineteenth century. A significant portion of the nineteenth-century attestations were from fictional representations of immigrant talk and Caribbean Creole speakers, which means that we are probably dealing with a contact-induced instance of analogical simplification. As Mair (2012) was being written, the Corpus of Historical American English (COHA) was not available, which is why I will provide and briefly discuss the relevant data from this source. A search for [do] [p*] got (i.e. any instance of do / does / did + pronoun + got) yielded seventy returns. The six produced before 1910 were all either due to typographical errors (did he got ← did he get, etc.) or otherwise irrelevant (as are many of the more recent ones). A first peak occurs in the year 1912, with eight genuine instances of the construction from Montague Glass’s (1877–1934) novel Elkan Lubliner, American (New York: Doubleday, Page & Co, 1912). A typical one is the following: (6)
‘Another thing,’ Louis Stout broke in: ‘Out in Johnsonhurst what kind of society do you got? Moe Rabiner lives there, and Marks Pasinsky lives there – and Gott weiss wer noch.’ (COHA, Fiction 19123)
The foreigner-element is still prominent in a rare example from the 1930s: (7)
‘Maybe it all be over in two, three month,’ Salvatore said. ‘Maybe if a revolution in France, England, she maka peace right away, hey?’ ‘Maybe,’ said Mr. Henkel. Salvatore’s face clouded. ‘If it last a long time,’ he said, ‘I don’ know what to do. What you think, Mr. Henkel? [If] I gotta my first citizen papers do I gotta go back home to fight?’ (COHA, Press 1930)
Examples such as these support the claim put forward in Mair (2012) that language contact played a role in the origin of this usage. Plausible examples from native-speaker usage do not occur in COHA until the 1960s (although the novel the quotation is taken from is still set in a north-eastern ethnic urban under-class milieu): (8)
3
‘Billingsgate said. ‘Hell now, why does everybody gotta go and put all the blame on … Billingsgate? I didn’t do a thing. I just slapped him silly, that’s all.’ (COHA, Fiction 1965)
That main-verb got serves as an index flagging the English of Yiddish-speaking immigrants becomes apparent in the very first sentence of this novel: ‘[“]Nu, Philip,” cried Marcus Polatkin to his partner Philip Scheikowitz, as they sat in the showroom of their place of business one June morning, “even if the letter does got bad news, you shouldn’t take on so hard. When a feller is making good over here and when the Leute im [sic] Russland hears about it” […]’ (1912: 3)
Semi-auxiliary (have) got (to) in BrE and AmE 61
The usage seems to have increased in frequency since, and also to have spread beyond its original narrow socio-demographic base. In this regard, as will be shown, COHA is quite different from its contemporary counterpart COCA, which contains substantial amounts of spoken and speech-like material. To document the increase in frequency of do-support with got and its spread among a widening community of speakers, Section 4.3 below will provide an analysis of the ‘spoken’ portions of COCA, with a view to exploring the extent to which the construction has moved from the margins of foreigner talk and ethnic vernaculars towards the centre of informal mainstream usage. Note, however, that so far any such developments seem to be confined to North American Englishes. In corpora documenting other varieties it is difficult to find examples which are not quotations from American sources, performed imitations of American speechways or foreigner talk of some kind.4 The present situation in North AmE is succinctly summarised by Pullum (1997: 89): ‘in standard English, necessity GOT only occurs as a complement to perfect have. In colloquial American English, it occurs without have as in I gotta go, with negative I don’t gotta go, polar interrogative Do you gotta go?, etc. Some speakers have zero inflection in the 3rd singular present (She gotta find another place).’ 4.3 Main verb got to / gotta: a sociolinguistic profile of its use in contemporary North American English The comparison of a large historical corpus of North AmE (COHA) and a comparable corpus of present-day North AmE (COCA) shows that do-support with got, though still a rather rare feature overall, is considerably more frequent in the latter.5 This raises the question of whether this increase in frequency is due to language change in the community, for example a spread 4
In the British National Corpus (BNC) searches for don’t / doesn’t / didn’t got to yield a total of 15 returns, most of which are spurious as got to appears to be misspelled for get to or go to. There is a single plausible example:
(SP:PS0KP) Well you’re not gonna get, be able to get […] marks if it comes to that son! (SP:PS0KS) What? Ho –, you gotta negotiate about cars and prices? (SP:PS0KP) No, you don’t got to negotiate, you’ve got to [pause] be able to speak to people on a [pause] regular basis. (BNC, spoken-demographic sub-corpus) On British examples in the Corpus of Global Web-Based English (GloWbE) see note 10 below. 5 For example, structures retrieved in searches for any form of do followed by a pronoun plus got (i.e. ‘[do] [p*] got’ in the COHA and COCA query syntax) do not exceed a normalised frequency of 0.4 in any portion of COHA. Moreover, and interestingly enough, this peak is reached in the 1910s – a sampling bias introduced by the many attestations from Montague Glass’s novel referred to above. In COCA, on the other hand, the corresponding figure for the whole corpus is 0.51 and reaches peaks of 1.5 and 0.81 in the ‘spoken’ and ‘fiction’ portions of the corpus, respectively. In absolute terms, COCA has more than three times more examples than COHA (235 vs. 70, uncorrected raw output).
62 Mair
of the form from various ethnic non-standard varieties into informal mainstream usage, or whether it is an artefact of corpus composition, for example the fact that more non-standard writing has been sampled for the contemporary corpus. Sampling bias of this type is certainly relevant. In COHA, almost the only points of entry for vernacular writing into the corpus are representations of non-standard usage in works of fiction or direct quotations of vernacular speakers in press reportage. In COCA, the copious transcripts of media broadcasts, which often feature distinctly non-standard speakers, provide a rich additional source. Nevertheless, there are three kinds of evidence to show that, above and beyond that, there has been genuine language change, as well: (a) In contemporary North AmE, do-support with (have) got is absolutely inconspicuous in mainstream usage in several syntactic environments, for example in question tags. (b) COCA (and similar corpora) yield examples of do-support with got which are produced by mainstream speakers in contexts which rule out accommodation to non-standard interlocutors or performed imitation of stigmatised non-standard usage. (c) Results from an ongoing elicitation experiment (Mair and Garley, Forthcoming) suggest that young North American mainstream speakers evaluate do-support with got less negatively than one would expect if the form was still strongly associated with stigmatised ethnic vernaculars. 4.3.1 (a) do-support with (have) got in question tags: This perfectly standard use of do with (have) got was already recognised by Trudgill (2002: 169), who argues that ‘American Standard English currently admits a new verb to got in You haven’t got any money, do you? but not (or not yet) in You don’t got any money, do you?’ Allerton (2009: 318) speculates on possible points of entry for such an innovation into BrE, mentioning I got a great chance, don’t I? as the ‘obvious choice’ when have is dropped in BrE. Searching COCA for examples of a similar construction, anaphoric tags with so, confirms this assessment. The seven genuine attestations all have do-support, even with the auxiliary have present. Consider these two typical cases: (9)
But when I look around the house, I’ve got everything I want. I’ve got three cars, and most families around here have two or three as well. I’ve got a couple of televisions, and so do most people. I’ve got a washing machine, a refrigerator, and so do all my neighbors. (COCA, NYTimes 1998). (10) We’re live at the site of the 67th Annual Oscars, the Shrine in downtown Los Angeles. You know, John Travolta isn’t the only one with
Semi-auxiliary (have) got (to) in BrE and AmE 63
Academy Award fever. His fellow nominees have it, the presenters have it, heck, we’ve got it and so do millions of people around the world. (COCA, CNN_Showbiz 1995) Have is used in the tag in two cases. As (11) is from a work of fantasy fiction with a British setting and (12) was produced by South African writer Lewis Nkosi, they are both questionable, however. (11) Besides, I can’t go – I’ve got mark to do, and so have you! (COCA,
Fiction 2006)
(12) ‘Listen, boy,’ J.B. suddenly snapped. ‘We’ve been narking too
much, you and I. I’ve got work to do and so have you.’ (COCA, Fiction 1996)
The spread of do-support with got into new dialects, registers and styles is likely to proceed from less salient environments such as the tags discussed to more salient ones. Compare the following three test frames, for example: • You got the money, ____ you? • You’ve got the money, ___ you? • Do you really need to make that trip? And ___ you got the money and the time? The first has no overt have and is thus likely to prompt the use of do even in varieties such as BrE, where do-support with got is otherwise absent. In BrE, do is certainly not expected in the second frame, where the full clause preceding the tag contains have, nor in the third one, where the only potential trigger is context-priming by a preceding do with a different function.6 4.3.2 (b) do-support with got in the ‘spoken’ portions of COCA – survey: One source which supplies a sufficient amount of relevant data for a systematic quantitative analysis are the c.90 million words of transcribed spoken English included in COCA (see Appendix for a detailed survey of the composition of the sub-corpus). Being orthographic transcriptions of radio and television broadcasts produced by non-linguists, the data are one step removed from the stylistic baseline of spontaneous face-to-face interaction. Owing to the large number of talk show and discussion programmes, however, much of the material is dialogic and even fairly spontaneous, and a fair number of non-standard speakers are represented through on-site reportage in disadvantaged communities. As has been explained, for the purposes of the present analysis the variable (have) got (to) is assumed to have the following three variants in presentday North AmE, based on the type of negation used. 6
Similar frames are being tested with American informants in a web-based elicitation experiment. See Mair and Garley (Forthcoming).
64 Mair
(i) the standard variant, in which the auxiliary have is present in questions and negations and in which its occasional absence in the affirmative is best analysed as the result of phonetic reduction in rapid speech. For illustration, consider the following example: (13) That’s the problem for the West, that we have not got the cast-
iron proof that Iran does have a peaceful program (COCA, NPR Daybreak 2006)
(ii) the traditional vernacular variant, characterised by the regular absence of have in the declarative and the use of ain’t in negations (and negated questions): (14) She can’t dance. She ain’t got no rhythm. (COCA, ABC
20/20 2012) (15) They ain’t got to go to school. They ain’t got homework. All they
got to do is box. (COCA, NPR ATC 2012)
(iii) the modern-vernacular variant, characterised by the frequent absence of have in the declarative and the presence of do-support in questions and negations: (16) That’s poppycock. It’s BS. It’s Barbra Streisand. Don’t listen to
the – do not listen to the doom and gloomers. I’m telling you. You keep watching this show. Listen to somebody positive like me. Listen to people who’ve succeeded and can tell you that it can be done and you’re doing the right thing, just don’t be impatient. Don’t think because you don’t got a $30,000 or $100,000 a year job right out of college that your life is over.7 (COCA, Ind_Limbaugh 1995)
Table 4.1 reports on the distribution of negated possessive and modal (have) got (to) in the spoken portions of the COCA corpus. The figures for donegated plain have8 are given for comparison (see page 65). The first thing Table 4.1 makes clear is how rare any form of (have) got (to) – standard or non-standard – has become in contemporary North AmE. The total of 622 for all variants of (have) got (to) is as nothing when compared to the 34,573 instances of plain have. What Table 4.1 shows for got specifically is the dominance of the possessive uses of got over the modal ones in North AmE, in all variants. Also note the expected high incidence of negative concord in the two non-standard variants. Given that it cannot 7
Note that this example contains three instances of don’t and one of do not preceding the don’t got, which raises the question of whether the use of do-support with got could have been primed by the massive presence of the form don’t in the immediate context. 8 The results were retrieved through searches for [do] + n’t + have, [do] + not + have, [do] + n’t + have no and [do] + not + have no. The 34,573 instances were not hand-classified into possessive and modal uses.
Semi-auxiliary (have) got (to) in BrE and AmE 65 Table 4.1 Standard, traditional vernacular and modern-vernacular negations of (have) got (to) in the COCA corpus Variant
Standard (have + not got) Traditional vernacular (ain’t got) Modern vernacular (do+n’t got) TOTAL do
+ not have
Proportion of possessive to modal uses
Negative concord
359
355:4
3
214
192:22
113
49
43:6
22
622
590:32
138
34,573
n.a.
142
occur in many examples (e.g. intransitive and unexpanded you don’t got to go as opposed to transitive you don’t got no money or expanded intransitive you don’t gotta go to no loan shark), the frequency of this feature is impressive. The distinction between the standard and the traditional and modern non-standard variants is not as straightforward in questions as it is in negated clauses. In negative questions, we have the same tripartite division that we have in the negative declaratives: Hasn’t she got rhythm?, Ain’t she got rhythm?, Don’t [or doesn’t] she got rhythm? – to adapt one of the examples from COCA given above.9 The situation is more complex in affirmative questions, where the traditional vernacular variant lacks a distinct form and thus merges either with the standard (Has she got rhythm?) or the modern vernacular (Do [or does] she got rhythm?). In very rapid speech, even this distinction is occasionally neutralised, as may be the case in the following example, where the directly following modern-vernacular form don’t got suggests the possibility of analysing what’s as a contracted form of what does instead of the more common derivation from what has. (17) But what’s Planck got that these guys don’t got? (COCA, NPR
Science 2009)
In questions in which an object noun phrase is fronted it is often difficult to decide whether a sequence of (have) got to / gotta instantiates modal usage or not. Example (18) presents itself as a clear case of modal usage (obligation) in context. Examples (19) and (20) are equally clear instances of the existential-presentational subtype of the possessive use in which the object 9
However, negative questions of this type hardly occur even in the very extensive spoken COCA material, which has a mere six instances each of ain’t … got and haven’t … got, several of which are irrelevant, and only two of don’t … got (searches for ain’t [p*] got, [have] n’t [p*] got and [do] n’t [p*] got).
66 Mair
noun phrase is postmodified by an infinitive. In example (21), on the other hand, even the context does not help to disambiguate between these possible readings. (18) [(CROSSTALK) BOTEACH] When people behave badly, just hold
them individually accountable. Now, let me just say one thing. Pat, the reason why many Jews – I[’]m not among them – are fearful of Christianity is, they[’]re tired of Christians saying that we[’]re a bunch of Christ killers. They[’]re tired of the lie that we killed Jesus. [DONAHUE] How far back do you got to go back in history to get this? [(CROSSTALK) BOTEACH] Pontius Pilate – I[’]m almost done. I[’]m almost done. (COCA, MSNBC_Scarborough 2004) (19) [ZEINS] What do you got to do?! [JORDAN] I[’]ve got something to do. Just call me in 15 minutes. (COCA, CBS_48Hours 2003) (20) Let’s see, he shouldn’t survive Bush winning because What has he got to talk about (COCA, Ind_Limbaugh 1995) (21) [Mr-SULLIVAN] I just want there to be a debt solution in this country. […] And every economist will tell you that. The Republicans insist… [MATTHEWS] That’s the math. [Mr-SULLIVAN] No! [MATTHEWS] That’s the math. [Mr-SULLIVAN] So what has he got to do? [MATTHEWS] OK. We asked The Matthews Meter, 12 of our regulars, including Gloria, Helene and Andrew, will President Obama’s tax the rich stand actually net him votes in the 2012 elections? (COCA, NBC_Matthews 2001) Both the existential-presentational reading – ‘what does he have that he can do?’ – and the modal one – ‘what is it that he must do?’ – make sense in (21). Apart from these problems, which derive from ambiguity in the structures we are looking for, the search for [have] [p*] got to will turn up the usual small number of entirely spurious hits, such as the following one, in which get has the meaning of ‘proceed’ or ‘move’ and to is a preposition introducing an adverbial of place: (22) Had you got to the stage where you considered him to be your friend?
(COCA, NBC_Dateline 2006)
Finally, as in the case of the negatives discussed above, we must bear in mind that, in comparing [have] [p*] got and [do] [p*] got, we are disregarding the default form, plain have (to): a search for [do] [p*] have yields 13,248 hits, one for [do] [p*] have to 2,312. Even the negative questions (search for [do] n’t [p*] have) occur 612 times.
Semi-auxiliary (have) got (to) in BrE and AmE 67
That having been said, we note that (have) got (to) occurs in affirmative questions with standard have 578 times, and with modern-vernacular do 143 times. This is raw output which, at least for do, reflects a very high degree of search precision, as only one example had to be eliminated as spurious on manual inspection. As has been pointed out above, ambiguity between possessive and modal uses cannot always be resolved in questions. A close examination of the context of the 142 examples, however, suggested a distribution of 124 possessive against 18 modal uses, which leaves no doubt as to the general predominance of the possessive use, which we were also able to confirm in the negated examples. Precision was reasonably high in the search for [have] [p*] got, as well, though the use of get in the sense of ‘move’ / ‘proceed’, and the consequent status of to as a preposition rather than an infinitival marker, created some additional problems. But let us return from the classification of concrete examples to the central issue, the current sociolinguistic status of the modern-vernacular variant. In the ‘spoken’ examples from COCA, the socio-demographic range of users contains an extremely wide spectrum, from disadvantaged minority groups at one end to high-profile media professionals, who are clearly prestigious speakers and linguistic norm-givers to the community, at the other. Starting at the bottom of the sociolinguistic scale of prestige, we still find the modern-vernacular variant associated with immigrant groups with possibly limited proficiency in English (i.e. members of communities that represent one of the putative historical models for the construction in North America). The following example was produced by a member of a Latino drug trafficking gang, with the speaker’s nickname Listo (= ‘the smart one’) suggesting competence in Spanish): (23) [LISTO] You know – they – they told me, you know, don’t be telling
him, you know, about, you know, our neighborhood’s business and, you know, what we do and our plans. Of course not. I mean – and, you know, he – he – he knows not to ask and, you know, I do the same. [RIVERA] So you don’t have pillow talk about, you know… [LISTO] But it ain’t for me to ask, so you don’t got to worry about it. [PE-WEE] Well, see, I trust Listo, see? [C-ROCK] How can you be trusted when you go into another (censored) hood? [PE-WEE] Well, they trust me, so… [C-ROCK] Oh, man. You ain’t got no real home boys [homeboys?], then. (COCA, Ind_Geraldo 1995)
African American gang members interviewed in another programme anchored by the same host even produce the form gots, in which the inflectional ending does not signify further integration of the verb into the standard paradigm but reflects traditional African American variable use of –s
68 Mair
throughout the present-tense paradigm, as similar examples feature you gots and we’s gots, as well. (24) [RIVERA] OK. Let me ask: You need a place to live?
[ASSASIN] No, I – I got a good home, huh? [SNYPER] She – she gots a – a lady that takes care of her. [RIVERA] What? A foster mother? [ASSASIN] Yeah. [SNYPER] Yeah, she’s like that – something like that. [ASSASIN] It’s cool. I got a – I got a cool home. [SNYPER] She gots a place to live. [ASSASIN] I’m – I’m – look, I’m starting to get out of the gangs right now. I’m starting to – I’m starting to go back to school. Starting to get good grades. [SNYPER] Me and her… [ASSASIN] But I just got to finish taking care of my business. I’m starting to slow down, OK? This is my homey, she’s always there for me (COCA, Ind_Geraldo 1995) (25) [DELIA] He’s the man. Of course. Everybody admires him in the block. He’s like the God from the block. Put it like that. He got mad juice everywhere. (Censored) You name it, he got it. [RIVERA] What’s mad juice? [DELIA] Mad juice mean that he got mad props. He got respect. He got money. He got cars. He got the works. He got a bar. He got a… […] He gots everything. (COCA, Ind_Geraldo 1995) (26) [5th TEEN CRIMINAL] Do I rob? [POTTER] Yeah. [5th TEEN CRIMINAL] They don’t care about you, so why should you care about them, you know what I’m saying? Yeah. [6th TEEN CRIMINAL] I don’t like tourists getting robbed and this and that, but if they dumb enough to come down here and with money, you gots to have it. [2nd TEEN CRIMINAL] When you take they money they don’t be broke, man. They still got more money than you got. (COCA, ABC_DayOne 1993) There are several attestations of do-support with got in material produced by white prison inmates interviewed for documentaries, among them the notorious killers Charles Manson (27) and Kristofer Hans. Hans’ example, quoted as (28) below, raises the question of whether you don’t got to do this is intended as obligation not to commit the murder (= ‘you must not do this’) or in the weaker sense of negating the self-perceived necessity (= ‘it is not necessary that you do this / you don’t have to do this’). (27) The women I got, I don’t got to tell them what to do. If I got to tell
them what to do, I’ll send them up on the highway and get them away
Semi-auxiliary (have) got (to) in BrE and AmE 69
from me. I don’t deal with women I got to tell them what to do. They know what to do. If they don’t know what to do, they’d better just stay away from me. (COCA, Ind_Geraldo 1996) (28) [SAWYER] You understood that you were going to shoot someone. [Mr. HANS] Yeah, I understood that. And to me it was just like the thing to do. It was like it was nothing. I went into the school and I had the gun in my backpack. I walked past the room once, and I remember just the slightest inkling saying, ‘You know, you don’t got to do this.’ I pushed that aside and went ahead and clicked the gun, and I went back and I knocked on the door. And then the teacher came up, and she just opened the door, and I pulled out the gun and shot her. […] (COCA, ABC 1992) But, as has been pointed out, the very opposite end of the social scale is also represented. Among the examples given above, (5), (16), (17), (18) and (19) were produced by mainstream speakers in informal contexts. One caveat remains, however. As many examples were produced by media professionals, we cannot rule out that their use of stigmatised non-standard forms was part of their audience design (Bell 1984) or a mediated vernacular performance (Coupland 2009). This is likely to be the case in the following three examples produced by Tavis Smiley, a well-known African American talk show host and liberal political commentator: (29) Before we advance to that, though, the other news of the day, President
Bush of course has a press conference and now, I’m not the only one who picked up on this. You don’t got to be a rocket scientist to see this dichotomy, this distinction. (COCA, PBS_Tavis 2005) (30) Well, I ain’t got a problem with it. It’s fasc – It’s a very good question. I’ll answer it if you want me to answer it seriously. It’s fascinating to me, because whenever I hear Erykah Badu is about to make an appearance, I immediately know that I want to tune in, to see what she’s gonna look like. So it works. (COCA, PBS_Tavis 2005) (31) Do people ever get in your ear – managers, agents, publicists – I don’t know if they’re on the stage or not, but anyway, people ever get in your ear and say, ‘You ain’t gotta say that. You’re Halle Berry. You’re universal. You’re crossover. Why you gotta keep bringing that to folks’ attention? Just don’t raise that.’ (COCA, PBS_Tavis 2005) Examples (29) and (30) could be regarded as an African American media professional consciously flagging his street credibility by using the nonstandard linguistic resources of his community in contexts in which they are normally excluded. Example (31) is a vernacular performance. Ironically, it is the very people who are presented as trying to talk the African American star out of emphasising her roots who are using nonstandard language.
70 Mair
No such metapragmatic justifications for the use of do-support with got are available for examples such as the following, however. Two pillars of the establishment, CNN host Larry King and US Attorney General John Ashcroft, use the modern-vernacular variant talking to each other: (32) [KING] Are you going to keep Louis Freeh?
[JOHN ASHCROFT] Well, he’s appointed by the president. He’s an individual that I’ve called to meet with. He is to stay. I mean, he’s got a term of years. [KING:] … does he got a chance to stay. (COCA, CNN_King 2001)
The communicative situation is similar in this exchange between King and fellow media professional Don Imus: (33) [IMUS] Why, Chance? That’s a great name.
[KING] We met by chance. The only Chance I know is Sweet Bertie Youth [= Sweet Bird of Youth], Tennessee Williams named the lead character Chance. That’s the only Chance I know. We met by chance. Why Wyatt? What are you, Earp? [IMUS] Well, I don’t know why Wyatt. I mean… [KING] I like the name. It’s nice. Does he got a middle name? [IMUS] His first name is Frederick, after my brother. (COCA, CNN_King 1999)
Such examples show that the modern-vernacular variant is starting to lose some of its former stigma. Some of the loss of stigma may not be directly due to language change in the community but simply reflect the fact that today’s media are more hospitable to non-standard language than in the past. Technological advances, sharper economic competition in the media sector and a prevalent climate of anti-elitism in late modern democratic societies have conspired to create a more participatory media system in which previously silenced members of the public are now given the opportunity to articulate their views in their own voices. More frequent exposure habituates mainstream speakers to non-standard usages previously excluded from prestigious domains, which is wearing down the social stigma historically attaching to them. However, there are enough examples in the corpus to show that in addition to that, there is spontaneous active use of do-support with got among mainstream speakers in informal situations. The beginnings of such active use may well be in situations of ‘crossing’ (Rampton 1995), in which standard speakers ‘perform’ or otherwise strategically use linguistic resources to which they are not ‘entitled’ by tradition and heritage. The COCA material also supplies ample evidence of this type of language use, which may well lead to spontaneous use in the wider community eventually (of which COCA also has some instances). Such a spread of do-support with got into the wider community in North America is not entirely unexpected. The
Semi-auxiliary (have) got (to) in BrE and AmE 71 Table 4.2 What-questions in the standard and in the modern vernacular in the spoken COCA material Question addressed to …
have … got
do … got
do … have
Self (= 1st ps. sg.) What have I got? / What do I got? / What do I have? 1st ps. pl. What have we got? / What do we got? / What do we have? 2nd ps. (sg. or pl.) What have you got? / What do you got? / What do you have? all affirmative questions
22
8
64
48
23
236
145
74
461
578
142
n.a.
Chi-square test have got vs. do got: p < 0.0001)
most important structural precondition for its emergence, regular loss of have in have got, has existed in mainstream usage for some time. In addition, we note that the advance of do-support with got seems to be tied to certain lexical chunks, much like the way Torres Cacoullos and Walker (2009: 322–23) have recently described the advance of will and going to in grammaticalisation: Collocations constitute an important locus of grammatical development, since they may lead in changes and constitute subclasses that contour the grammaticalisation of more general constructions … The patterns of future forms revealed in this article also point to an interaction of conventionalized expressions with general, productive constructions. Thus, rather than dismissing frequent collocations as simply lexicalized and therefore marginal to grammar, we view them as particular instances of constructions which, while formulaic, interact with their associated general construction … We conclude that collocations contribute to the shape of grammatical variation.
One such pattern which clearly emerges in the COCA data is due to the broadcasting situation. Again and again, we find media hosts asking themselves what they have (got) to offer their listeners or viewers, or asking their guests what they have (got) to offer. This may be a trivial observation, but what is not trivial is the very high proportion of the modern-vernacular variant which occurs in precisely this context, as is illustrated in Table 4.2. As can be seen in the table, 105 out of a total of 142 relevant instances of the modern-vernacular variant occur in the collocation in question. This is 73.2 per cent, as against 37.2 per cent for standard have … got. Note further that plain do … have still dominates over all forms of (have) got, but that this dominance is far less pronounced with this set of collocations than it is generally. Relatively speaking, they thus provide an environment in which
72 Mair
have got is over-represented, and within the uses of have got, it is the modern vernacular which is over-represented again. Unlike negative questions, in which a host might choose between a traditional folksy vernacular ain’t we got and the more stigmatised modern don’t we got, there is no such choice in affirmative questions. The modern vernacular becomes the only informal option, providing a platform for the more general spread of do-support with got into mainstream informal usage. 4.3.3 (c) mainstream speakers’ attitudes to do-support with got: Corpus data do not provide a direct window on speakers’ attitudes towards a possibly spreading linguistic innovation. This is why at the time of writing, Mair and Garley (Forthcoming) are carrying out a series of Web-based elicitation experiments aimed at young educated US-born whites, the target group whose views will be crucial for the future status of the modernvernacular form. The following are some statements obtained by Garley in a trial test in response to the question whether informants would use does he got a chance or do I gotta go to the store right now themselves and, if not, what type of speaker they associated with these constructions: • ‘I would totally say them. They probably signal someone younger or rural and definitely informal.’ (female, aged 34, Illinois) • ‘Yeah, I say these as well.’ (male, aged 28, Illinois) Another informant (male, aged 32, Texas) claims not to use the first form, but only the second, adding that it could be made to sound even more acceptable by inserting a degree adverb: Do I really gotta go to the store right now. As expected in the early stages of a development, the majority of informants state that they do not use the form themselves. All are aware of the usage, however, and variably regard it as typical of young, uneducated, east-coast urban or ethnic speakers. 4.4 Main verb gotta in North American English: the emergence of a regional feature from a cross-variety drift From assessing the sociolinguistic status of do-support with got in contemporary AmE I will now turn to the question why the modern-vernacular variant has so far not gained a solid foothold outside North America. Note how the form is corrected even in the very informal communicative atmosphere of a British website serving computer gamers: (34) ‘Hey, did you lot hear about what happened to the head girl today?’
‘What happened?’ ‘She’s been deprefected!’
Semi-auxiliary (have) got (to) in BrE and AmE 73
‘How…!’ ‘Yeah, tell us!’ ‘Well, it went like this you see …’ ‘Hurry it up man, we don’t got all day!’ ‘We ain’t got all day, you mean.’ ‘WhatEVER.’ (www.srcf.ucam.org/assassins/archive/2005-mich/news02.html)10 Very likely the issue here is not lack of familiarity with the American modern-vernacular form. It has been spreading globally for at least two decades in the media and in popular culture (through music lyrics, for example) and has thus become part of the store of deterritorialised non-standard and vernacular English resources which are accessible to the initiated denizens of the global ‘mediascape’ (Appadurai 1996). The gamer who utters it probably uses it as an ‘off-the-shelf ’ form in that context. The correction, from the modern to the traditional vernacular, on the other hand, is motivated by the fact that do-support for got does not integrate into present-day BrE grammar as easily as it does into AmE. This also explains the apparent discrepancy between the near absence of do-support with got in the spontaneous spoken data of the BNC (see note 4 above) and its relative frequency in the British portions of GloWbE.11 In BrE, have got (to) is firmly established and common, both in nonstandard and informal/spoken standard usage. Got to / gotta, without have, occur, but speakers are not as ready to disconnect the reduced forms from the full ones in an environment in which both remain available. This is very different from the situation in AmE, where the full form have got (to) never established itself as solidly as in BrE and has in fact been on the decline for several decades, as is shown in studies by Jankowski (2004), Tagliamonte and D’Arcy (2007), Tagliamonte, D’Arcy and Jankowski (2010) and Lorenz (2013). With the older full form on the wane, reduced got (to) / gotta becomes available for reanalysis as a monomorphemic verbal root which is grammatically treated like a lexical verb.12 10
This example was identified through GloWbE, http://corpus2.byu.edu/glowbe/). At roughly 387 million words, the British component of this corpus is about four times the size of the BNC. Searches for don’t / doesn’t / didn’t got to yielded a total of 35 returns, with typical examples in references to music lyrics, in direct quotations of American vernacular speakers in British newspapers, or in transatlantic digital conversations conducted in blogs. Note that examples of do-support with got are attested in all of the 20 national sub-corpora of GloWbE in similar contexts. This supports the argument put forward here that while active use of the construction is largely restricted to North America, some familiarity with it can be assumed for speakers of English all over the world. 12 We must not overstate this point, of course, as do-support is merely one feature of mainverb syntax. As yet, there is no trace of an infinitive to got, just as there is no newly created past tense / past participle gotted. 11
74 Mair
4.5 Conclusion The present investigation of the divergent development of (have) got (to) in BrE and AmE has shown that ‘grammaticalisation processes reside in particular varieties (of a given “language”) and not in a “language” tout court’ (Miller 2004: 229). As Miller has demonstrated for perfect and resultative constructions in spoken and non-standard English, our data also show that ‘a particular grammaticalisation may take place in one variety but not in another – that is, different groups of speakers have different linguistic behaviours; alternatively, the results of a particular grammaticalisation may become firmly fixed in one variety but in another variety be affected by later changes; and the original construction that was grammaticalised may merely survive in one variety and survive and flourish in another’ (Miller 2004: 229). Assessing modal (have) got (to) at the beginning of the twenty-first century, we cannot speak of a flourishing process of grammaticalisation, of course. The form which has been spreading at the expense of traditional must in practically all varieties of English is the structurally more versatile have to. (Have) got (to) remains established at modest levels in spoken Standard BrE, but has been recessive in mainstream North AmE for some time. What the present study has shown is that disappearance has not been along the expected pathway, through fossilisation of a residue of idiomatic chunks. Instead, got (to) / gotta has continued to undergo rapid and drastic change in its period of decline, turning a semi-modal into a main verb with a defective tense paradigm. The corpus data analysed here suggest that, sociolinguistically, do-support with got is on the verge of becoming mainstream in informal North American usage. Its spread is facilitated by the fact that there is a place for this construction in the grammatical system of PDE. As the full form have got (to) declines in frequency, reduced got to / gotta becomes available for reanalysis as a monomorphemic verb, which leads to some structural impasses. Gotta resembles the central modals in that unmarked 3rd-person singular is possible: she gotta go. The remaining defining properties of the core modals, however, are not likely to be added. There are no inverted questions (*gotta you go?); there is no enclitic negation, and no use in tags (*you gottan’t go; *you gotta go, gottan’t you). The traditional alternative, ain’t (you ain’t gotta go; you gotta go, ain’t you?), is stigmatised and, lacking a suitable question form in the affirmative, also paradigmatically defective. At least for the present tense, treating got to / gotta as a lexical verb solves all the problems, and a complete paradigm is created: you gotta go; do you gotta go?; don’t you gotta go?; you gotta go, don’t you? Limited expansion into the future seems a possibility: you gonna gotta go. Gaps in the paradigm remain in the past, but like the gaps in the future paradigm these are not serious because they can be easily filled by the most successful product of successive cycles
Semi-auxiliary (have) got (to) in BrE and AmE 75
of grammaticalisation in the possessive and modal system of English: the semi-modal have to. On the basis of the evidence analysed here, I would predict a modest expansion of do-support with got in informal varieties of North AmE. Rapid spread into other varieties of English through ordinary diffusion in dialect contact is unlikely. The construction might receive a boost in varieties of English outside of North America if, like quotative like, it became part of a dynamic of vernacular globalisation. Two conditions for vernacular globalisation of this sort to take place are met even now. Do-support with got is associated with young people’s usage and is massively present in the youthcultural media, as any Web search of song lyrics will show.13 Whether this is enough to set such a dynamic in motion is something we can only speculate on for the time being. What has been established beyond speculation, however, are the following two facts. With regard to the construction under study here, contemporary sociolinguistic variation is deeply rooted in the history of LModE grammaticalisation, and the interaction between unfolding grammaticalisation and the changing sociolinguistic ecologies provided by different communities of speakers is fascinating and complex. As has been shown here and also in Mair (2012), do-support with got may be a very recent addition to mainstream usage, but it is not a historically young construction in itself. It has roots in African American English, creoles and nineteenth-century language shift among North America’s immigrant populations. Arguably, its development was largely hidden from linguists’ view because of its origins in strongly stigmatised ethnic and immigrant vernaculars – until the advent of linguistic mega-corpora and vast digital textual resources made possible a more systematic sifting of the nineteenth- and early twentiethcentury evidence. An issue which should be addressed in a future study is whether the emergence of do-support with got in late nineteenth-century North America is a stand-alone development or part of a wider diachronic dynamic in which modal and semi-modal forms (e.g. need, dare, used to or
13
For an illustration, consider the sole ‘Nigerian’ example from GloWbE, lyrics of a rap song mixing Standard English, Nigerian Pidgin and African-American-inspired hip-hop nation language: Somebody lied They tell me say you no go blow if you no be Yoruba mehn somebody lied They tell me say you come from Jos and you no go go far mehn somebody lied They tell me say my guy you no fit compete mehn somebody lied Somebody lied oh Oh you from J-town, omo forget rap You don’t got a chance out there with your dead crap (source: www.freenaijalyrics.com/category/iceprince/feed) Informally, the pervasive web presence of don’t got can easily be established through a search for ‘you don’t got to be no billionaire,’ with its almost 10,000 returns.
76 Mair
ought to) show increasing tendencies to combine with auxiliary do (on which see Denison 1998: 206–12). Appendix COCA – sources and amounts of spoken material SECTION NAME SPOK:ABC SPOK:NBC SPOK:CBS SPOK:CNN SPOK:FOX SPOK:MSNBC SPOK:PBS SPOK:NPR SPOK:Indep
# WORDS 15,514,463 6,348,632 12,887,290 20,656,861 6,302,918 814,156 6,612,202 17,399,724 4,343,343
5
From contraction to construction? The recent life of ’ll Nadja Ne sse l hauf
5.1 Introduction In a recent analysis of the development of future time expressions in Late Modern English (LModE) (Nesselhauf 2012), one of the findings was that the form ’ll has replaced shall to a much greater degree than will in the past 250 years. This raises the question of whether the form ’ll in Present-Day English (PDE) is best regarded as a contraction of shall rather than or in addition to being considered a contraction of will or whether it might even have developed as an independent construction in PDE. Building on an analysis of the emergence and early development of ’ll (Nesselhauf 2013) this chapter investigates both the more recent development of the form ’ll, which has been almost completely neglected to date, as well as its presentday status, which is highly disputed (see Section 4.2). ‘Construction’ will be defined here as a pairing of form and meaning and is assumed to be not only a theoretical but also a cognitive entity (see also Croft and Cruse 2004: 257). Polysemous expressions are considered as single constructions (as e.g. in Croft 2001: 19 or Goldberg 1995: 38) if the meanings are closely related. Closely related forms (such as a contracted and its corresponding non-contracted form) will be considered as part of the same construction if the meaning structure is the same. Meaning patterns will be considered to be the same if, first, the same sub-senses may be distinguished, second, the same sub-senses can be considered central and, third, the same sub-senses can be considered peripheral to a form’s (or form group’s) meaning. If related but different forms (such as a full form and its cliticised form) have different meaning structures, these will be assumed to constitute two different constructions. With respect to the term ‘contraction’, it should be noted that it will be used both in its wider and in its narrower sense in this chapter (see also Westergren Axelsson 1998: 19, 24). In its narrower sense, the term only refers to the cliticised form which is attached to a host (e.g. ’ll), as, for example, in the definition given by Quirk et al. (1985: 123): ‘Contractions are phonologically reduced or simplified forms which are institutionalised in both speech and writing’ and ‘[c]ontracted forms [are] enclitic to a preceding word’. In the 77
78 Nesselhauf
wider sense of the word, ‘contraction’ refers to the combination of host plus enclitic (e.g. you’ll), as, for example, in the definition by Peitsara (2004: 28): ‘Contractions are combinations of two forms in such a way that one of the elements is reduced and the combination functions like a single word.’ When there is the danger of ambiguity, the terms ‘enclitic’ and ‘host’ will be used. 5.2 Previous research As studies on the recent development of future time expressions including the form ’ll have in most cases subsumed the form under will, our knowledge of the recent usage of this form is severely limited.1 In contrast to the lack of empirical research, statements on the present-day status of the form abound in the literature.2 Most researchers dealing with the issue state or at least imply that ’ll has to be considered a contraction of will (e.g. Biber et al. 1999: 469; Coates 1983: 204; Curme 1913: 530; Fries 1925: 989; Jespersen 1909–49: 237; Joos 1968: 156, 162; Palmer 1988: 136ff.; Quirk et al. 1985: 28; Wekker 1976: 143; Westergren Axelsson 1998: 123). In dictionaries, ’ll is also commonly listed under will (see, e.g. OED 1989, Cambridge International Dictionary of English 1995). A number of researchers, however, consider ’ll as a contraction of both will and shall (e.g. Huddleston and Pullum et al. 2002: 429; Krapp 1909: 295; Visser 1963–73: 1606) or at least allow for this possibility (Gotti 2003: 270; Mindt 1992: 27; Taglicht 1970: 204). The researchers who consider ’ll to be a contraction of will exclusively give various reasons, with the most comprehensive argumentation being provided by Coates: ’ll will be counted as a form of will rather than shall on the following grounds: (i) phonetically it is the weak form of will just as [ʃəl] or [ʃɬ] are the weak forms of shall; (ii) it occurs freely with the first, second and third person subjects (like will but unlike shall); (iii) it expresses meanings (‘Willingness’, ‘Predictability’) which are not expressed by shall (1983: 204).
The first (phonological) reason she gives is also adduced by several other writers on the topic, who sometimes seem to argue from a historical point of view, however (e.g. Fries 1925: 989, Jespersen 1909–49: 237; Wekker 1976: 143).3 Quirk et al. support the view of ’ll as contracted will with the statement that it ‘can always be expanded to will’ (1985: 136), which might 1
The only study besides Nesselhauf (2012) that considers ’ll independently of will or will and shall is Berglund 2005, whose analyses are, however, limited to the time span of 1961–91. 2 There are, however, a few studies on the present-day usage of ’ll (e.g. Westergren Axelsson 1998). 3 Fries argues for ’ll being exclusively a contraction of will by pointing to ‘[t]he very common loss of (w) in English unstressed syllables. Middle English examples are abundant of the loss of (w)’ (1925: 989).
From contraction to construction: the recent life of ’ll 79
be understood as a short version of Coates’ reasons (ii) and (iii). Curme also resorts to the semantic argument and points out that both ’ll and will can express pure prediction but shall cannot (1913: 529f.); however, this is clearly inaccurate, not only now but also in the last few centuries (see Nesselhauf 2012: 97f.). Those researchers who consider the possibility that ’ll is also a contraction of shall do not usually give reasons for this view (beyond illustrating it with potential utterances where ’ll can stand for both will and shall, e.g. Huddleston and Pullum et al. 2002: 429). To my knowledge, the only attempt at providing a reason is made in Visser, who states that ‘the extremely subtle difference in meaning between I will and promissory I shall caused I’ll to be regarded by the average speaker as also standing for promissory I shall and consequently to be used indiscriminately’ (1963–73: 1606). That there might be a third viewpoint on the present-day status of ’ll has been hinted at in three publications to date. First, Berglund, in her studies of future time expressions in the twentieth century, partly investigates ’ll separately and comes to the conclusion that ‘the use of ’ll is different from that of both will and shall, which further warrants treating the expression individually’ (2005: 22f.). Secondly, Lewis, after several sections in which he regards ’ll as a contracted form of will (2002: 115ff.), points out the following: ‘It is by no means obvious that ’ll is short for either shall or will. It is possible that ’ll is an independent form and is beginning to assume the role of ‘the pure future’ in English’ (p. 121). This is not pursued further, however, except in a different section more than twenty pages later, where he speculates that ’ll might be ‘evolving in the direction of a neutral future – an English “future tense”’, because, according to him, native speakers tend to provide I’ll hit him as the future for I hit him and because ’ll expresses ‘a concept of weak inevitability and psychological immediacy’ (p. 145). The most comprehensive discussion of the possibility that ’ll might be a construction in its own right can be found in Huddleston and Pullum et al. (2002). After considering ’ll a contraction of will and shall in declarative clauses (2002: 429), they later point out the following (2002: 632): There are contexts involving decisions and arrangements where ’ll is normally the only choice. In such cases, ’ll is best not seen as a contraction of either will or shall; it may be seen as an independent form. As an independent form, ’ll is often used to indicate an instantaneous personal decision: There’s a garage. We’ll just stop and get some petrol [...] [speaker A has just poured herself some water] A: Anyone else want some? B: I’ll have a drop, yes. Thanks.
80 Nesselhauf Table 5.1 Number of words in the relevant subperiods of ARCHER-3.1 Period (BrE)
d
h
j
m
n
s
x
f
1750–1799 1950–1999
23,962 11,068 21,843 21,003 23,087 20,565 12,091 21,329 23,727 178,675 24,450 10,190 22,225 20,794 22,920 21,308 11,259 17,577 27,518 178,241
con
Total pro
The abbreviation d stands for drama, h for sermons and homilies, j for diaries and journals, m for medical writing, n for newspaper reportage, s for science writing, x for private letters, f con for fiction conversation and f pro for fiction prose.
Independent ’ll is also used for indicating informal decisions or arrangements where will or shall would sound too direct or imposing: Okay. That’s it then. We’ll meet next week, I suppose. […] Then we’ll get a bus to Bridgford, the boys’ll get off near Asda and we’ll all walk to school.
5.3 An analysis of the usage of ’ll in Late Modern English For the examination of the more recent usage of ’ll, which is also intended to contribute towards clarifying its status, two periods from the British part of ARCHER-3.1 (A Representative Corpus of Historical English Registers) were used in the present study. ARCHER records texts from the Late Modern period and is subdivided into fifty-year periods, of which the periods 1750– 99 and 1950–99 were chosen for the present analysis. Both subperiods contain directly comparable proportions of nine different registers each (some of them written, some speech-based), with about 180,000 words per subperiod (see Table 5.1).4 With respect to the overall development of ’ll, it comes as no surprise that the form has increased during the time span investigated, from 129 occurrences in the period 1750–99 to 223 occurrences in the period 1950–99. To what degree this is due to an increase in its occurrence with particular hosts and to what degree it is due to an increase in its occurrence in particular registers is displayed in Figures 5.1 and 5.2. Figure 5.1 shows that I is the pronoun which occurs most frequently by far with the modern /l/-contraction in the Late Modern period and that it 4
For a detailed description of the version of ARCHER used (i.e. ARCHER-3.1), see Nesselhauf (2012: 88ff.). The only subcategories that do not contain the same number of words across subperiods are fiction prose and fiction conversation, as these registers in reality do not occur separately (much rather, it is the category of ‘fiction’ which is represented with equal numbers of words across subperiods, with different proportions of prose and conversation in the different subperiods).
From contraction to construction: the recent life of ’ll 81 120 I
Instances
100
you he
80
she
60
it we
40
they
20
other hosts
0 1750–1799
1950–1999 Period
Figure 5.1 Pronouns with ’ll in the eighteenth and twentieth centuries (ARCHER-3.1; absolute numbers) 140 120
drama
Instances
100
fiction conversation
80
fiction prose
60
diaries & journals newspaper reportage
40
private letters
20 0
1750–1799
1950–1999 Period
Figure 5.2 Use of ’ll in different registers in the eighteenth and twentieth centuries (ARCHER-3.1)
and hosts other than personal pronouns are very rare. This marked preference for the enclitic to co-occur with the first-person singular pronoun is a continuation of its earlier development (see Nesselhauf 2013). As the results displayed in this figure also make clear, the increase in the use of ’ll cannot be put down to its use with one pronoun or a particular group of pronouns; much rather, all pronoun plus ’ll combinations display an increase, with I’ll increasing proportionally less and the other pronouns catching up to some degree (in the period 1750–99, I occurs with ’ll in 64% of all cases; in the period 1950–99, in 44% of all cases). Figure 5.2 shows that in the period 1750–99, the contraction ’ll only occurs in significant numbers in the speech-based registers of drama and
Instances
82 Nesselhauf 140 120 100 80 60 40 20 0
semantics of future'll prediction intention intention arrangement
1750–1799
1950–99 Period
Figure 5.3 Semantic development of ’ll in ARCHER-3.1 (based on Nesselhauf 2012: 96)
fiction conversation; there is also a single instance in the category of fiction prose. In the period 1950–99, the number of occurrences in these two registers has increased (note that the increase in the category of fiction conversation is not as steep as the figure suggests, as at the same time the proportion of conversation has increased in fiction, see footnote 4). Moreover, the form has spread to other registers in the twentieth century, namely to the registers of diaries and journals, private letters and newspaper reportage, but still (since the time of its emergence; see Nesselhauf 2013) predominantly occurs in speech-based registers. In the other, more formal, registers represented in ARCHER, such as medical and science writing, and sermons and homilies, ’ll does not appear at all (see also e.g. Berglund 2005: 164). In order to determine whether present-day ’ll is still (like when it first emerged; see Nesselhauf 2013) a contraction of will or whether it might be better considered a construction of its own, its meaning development needs to be investigated. In a thorough investigation of the semantic development of all future time expressions in LModE (Nesselhauf 2012), I found that while in the period 1750–99, ’ll was predominantly used to express a prediction based on the intention of the subject referent, in the period 1950– 99, this use was superseded by its use for pure prediction (i.e. a prediction based on neither the subject referent’s nor the speaker’s intention nor on an arrangement of any kind; see Figure 5.3). Will, on the other hand, was predominantly (and with similar frequencies) used for pure prediction in both periods, whereas its use for intention decreased considerably. In the case of shall, both the (predominantly used) pure prediction sense and the intention sense decreased, with pure prediction decreasing proportionally more but still being slightly dominant in the second period. Shall was also used with the sense of speaker’s intention (as in He shall be hanged spoken by someone with the authority to decide this) in a high percentage of instances in the first period, whereas ’ll was hardly ever thus used. These differences in sense development therefore do not support
From contraction to construction: the recent life of ’ll 83 100
will /l/
Proportion of instances
90 80 70 60 50 40 30 20 10 0
end of 16th c.
2nd half of 18th c. Period
2nd half of 20th c.
Figure 5.4 Proportion of pure future uses (in relation to intention uses) in the OED
the view that ’ll in PDE should merely be considered a contraction of will (and shall). To capture the whole of its semantic development, the early semantics of both ’ll (or more precisely, of all /l/-contractions, including, for instance, ’l, ’le, and –le; see Nesselhauf 2013) and will in the same time span were also examined. This investigation is based on 100 (arbitrarily chosen) occurrences each in the OED online quotations database, from 1580–1610 (for the precise extraction procedure see Nesselhauf 2013). In Figure 5.4, the proportion of uses of the two expressions for pure prediction (as opposed to its only other major use of intention) is depicted. (Non-future uses such as predictability, as in That’ll be Dad, are disregarded, as they were in Figure 5.3.) Figure 5.4 shows that the use of /l/ for intention predominated in the earlier period and has decreased over time (first slowly, then much more quickly) in favour of its use for pure prediction, up to roughly equal use in both senses in the second half of the twentieth century. Will was also still predominantly used in its original intention sense (see also Bybee et al. 1994) at the end of the sixteenth century, though with an already higher proportion of pure future uses. In the second half of the eighteenth century, both types of use occurred with about equal relative frequency, and in the final period, will was used for pure prediction in almost 70 per cent of cases. The hypothesis put forward by Lewis that ’ll may be an independent form because it is beginning to assume the role of pure future (see section 4.3.1) is therefore flawed, because, while it is true that pure future uses of ’ll have increased and become frequent, this is also the case for will, which is used in the pure prediction sense even more frequently.
84 Nesselhauf
The results from this analysis thus indicate that both expressions follow what I have termed the Bybee-path for future time expressions (see Nesselhauf 2012: 87), that is, the path from predominantly expressing future only in connection with intention to increasingly expressing pure prediction. However, the uses of ’ll and will are dissimilar in all periods, with will being used for pure prediction proportionally more in both periods. Furthermore, the developments are not strictly parallel. In the first 150–200 years, the intention use remains by far the major use of ’ll, with a considerable increase of the pure prediction sense in the following 200 years. In particular, in the period 1750–99, therefore, the usage patterns for ’ll and will are quite different (with only about 15% pure prediction uses of ’ll and more than 50% of will). All of this additionally might be considered as evidence against regarding ’ll merely as a contraction of will, that is, merely as a variant form of the will-construction. The data presented here and in Nesselhauf (2013) suggests a different interpretation. In Nesselhauf (2013) I show that, after a time when various reduced forms of will occurred occasionally in Middle English and the early stages of Early Modern English, one of the contractions, namely / aɪl/ (mostly in the spelling Ile), became heavily entrenched or, in construction grammar terms, constructionalised. (Other existing contracted forms such as /wɪltʊ/ either did not become constructions or were only shortlived.) Evidence for considering /aɪl/ a construction around the sixteenth century are its (absolute and relative) frequency of occurrence,5 its frequent spelling as one word and the persistence of this spelling. From the numerical evidence of the written sources, it may also be assumed that it was the first /l/-pronoun combination used. Shortly after this construction was formed, it seems to have been reanalysed as a form of I plus /l/ with the meaning of ‘prediction based on an intention’, which thus constituted a separate construction (while the construction /aɪl/ at the same time most certainly persisted). This meaning assignment was possible because /aɪl/ was initially exclusively used with this meaning. Reanalysis then also allowed the extension of the use of /l/ with other pronouns (by analogy) for the expression of this meaning (when /l/-contractions started to appear toward the end of the sixteenth century, all instances of /l/ − several Ile, some weele and two theyle – appear to be instances of this meaning). Besides Ile being frequent because it was stored as a whole, /l/ was stored as a meaning element for ‘prediction based on an intention’. This element could most easily be attached to we, it could also be attached to other personal pronouns, and could not be attached to it. Such a process, of a more general construction being derived from a more specific
5
Psycholinguistic evidence strongly suggests that frequently co-occurring items are stored and processed as chunks (see e.g. Bybee 2004: 617).
From contraction to construction: the recent life of ’ll 85
high-frequency construction, has also been observed for other expressions (see e.g. Torres Cacoullos and Walker 2009: 325). After around 1600, then, the /l/-construction slowly started to appear in the meaning of pure prediction as well, with probably two mechanisms at work simultaneously. First, the new /l/-construction can be expected to have moved independently down the Bybee-path (through the known mechanisms of pragmatic inference, analogy etc.). Evidence in support of an independent development is the difference in the development of ’ll from will (see Figure 5.4). Moreover, its development is typical of a grammaticalising expression, as its increase of secondarily grammaticalised uses (i.e. pure prediction uses) coincides with a marked increase in frequency (Bybee 2004: 602): in the OED database, the frequency of ’ll roughly doubled from the period 1600–49 to the period 1750–99, and then more than doubled in the (shorter) time span from the period 1750–99 to 1850–99. As a second mechanism, the use for pure predictions will also have been supported by the fact that will already allowed that use, as /l/ at that time can also still be assumed either to have been a form variant of the will-construction or at least to have maintained a strong link to it. Evidence for the continuation of some sort of link of ’ll to will is the fact that the step from merely expressing intention to also expressing pure prediction is much faster than in an independent secondary grammaticalisation process (see Nesselhauf 2012: 101f.). To find out more about the present-day status of ’ll, the present-day semantics of ’ll was, in a final step, examined more closely (assuming that the ARCHER period 1950–99 is roughly representative of PDE). In this investigation, one particular use of the form for a prediction based on an intention was noticeable, namely its use for an intention that has only just been formed or for what could be called ‘spontaneous decision’ use (see also Nesselhauf 2007). Some examples of this use are given in (1) to (4). (1) (2) (3) (4)
‘One moment, Roper, I’ll do this!’ (1960bolt.d8b) ‘Very well. I’ll call a meeting this very evening.’ (1962mann.con8b) ‘Sherry.’ ‘Yes today. We’ll have a wee bit. Doctor Romney says I’m to leave off but I think today.’ (1968donl.con8b) ‘Help! – look at my head! that’s blood!’ ‘I’ll stop it.’ (1969bond.d8b)
This is precisely the usage that was also observed by Huddleston and Pullum et al. (an expression of an ‘instantaneous personal decision’ in their terms; 2002: 429, see section 4.3.1), where they consider ’ll an ‘independent form’. The second type of usage they mention, but which is rare in my data (‘for indicating informal decisions or arrangements where will or shall would sound too direct or imposing’ as in ‘Okay. That’s it then. We’ll meet next week, I suppose’) can easily be subsumed under the ‘spontaneous decision’ usage, as it is just an extension of this usage – something is presented as a spontaneous decision so as to make it appear less imposing. Such a usage is an aspect of what I have referred to as pseudo-democratisation (Nesselhauf
86 Nesselhauf 100% 90% 80% 70% 60%
spontaneous decision other intention uses pure prediction
50% 40% 30% 20% 10% 0%
will
‘ll
Figure 5.5 Use of ’ll and will to express a spontaneous decision (ARCHER-3.1 1950–99)
2012: 116), in other words, a linguistic attempt to pretend that social structures are less hierarchical than they actually are or, formulated more positively, to create an unauthoritarian atmosphere. Some examples from the BNC (British National Corpus) are given in (5) to (8): (5)
which is actually poorly understood, we’ll talk a bit more about that in the next lecture. (F8E 63; uttered by a lecturer) (6) They’re probably even clearer if you take this copy of my er [unclear] which I’ll leave by the door and er we’ll carry on next week. (HUM 768; also uttered by a lecturer) (7) Erm, well we’ve got, I’ll take a few notes erm during this, during the course of our conversation, I’ll take them away, study them in complete confidence and come back to you next week with my recommendations. (J3V 191; uttered by a trainer) (8) We’ll leave it at that and we’ll pick it up again next time because we’ve done enough on that. (JK5 635; uttered by a school teacher in the classroom)6 In the twentieth-century data from ARCHER-3.1, the spontaneous decision use occurs in about half of all the instances where a prediction is based on an intention (see Figure 5.5). In comparison, it occurs only very rarely, in 3 per cent of the intention uses, with will in the same time span. With shall, it occurs similarly rarely, in about 5 per cent of the intention uses (in absolute numbers even more rarely than in will, due to its overall lower use as future 6
Further aspects of pseudo-democratisation are also present in these examples, for instance, the use of the subject we with talk spoken by a lecturer (although the audience is expected not to talk) and the use of an if-clause for what is, in effect, an order (if you take this copy).
From contraction to construction: the recent life of ’ll 87
time expression in the twentieth century). This means that the meaning pattern of present-day ’ll (in addition to its still lower proportion of pure prediction uses) is clearly different from present-day will (and from shall) and that it therefore, without a doubt, constitutes a separate construction in PDE. It is important to point out, however, that this does not preclude the possibility of ’ll still being part of (or maintaining a strong link with) the willconstruction at the same time, so that in certain uses a speaker may retrieve the contracted will-form and in others, the separate construction. (Perhaps there is even a link of some sort, at least for some speakers, between shall and ’ll, due to their phonological and semantic similarity.) If this is the case, present-day ’ll will have to be considered a separate construction as well as a contraction. Moreover, I’ll, and probably a number of other highly frequent pronoun+’ll combinations, might also be stored separately (see e.g. Bybee 2004: 617) and thus constitute additional constructions; the same might apply to combinations of a pronoun + ’ll plus some other element (such as I’ll be -ed, I’ll tell, I’ll never). That I’ll, in particular, might still have a special status is supported by a comparison of all future marker uses in period 1950–99 in ARCHER-3.1, where the combination I’ll is clearly dominant, occurring more than twice as often as the next most frequent future time expression co-occurring with I, namely shall (see Nesselhauf 2010: 176f.). These findings also provide a number of insights for grammaticalisation theory. First, from a form which has already undergone secondary grammaticalisation (i.e. will),7 another form (i.e. initially a contraction) can not only split off but also (instead of inheriting the meaning pattern of its noncontracted mother form) inherit one layer of meaning. This meaning may be, moreover, the older one of the mother form. Secondly, although the younger form clearly is a phonetically reduced form of the older form, it is not (neither initially nor at any later stage in the process) more grammaticalised than the former, so that in this particular case, the process of phonetic reduction and the process of increasingly assuming more grammatical meaning, which are both commonly considered as co-occurring characteristics of grammaticalisation (e.g. Bybee 2004: 615), do not go hand in hand. Thirdly, if the above interpretation of the results is correct, it is not only future time expressions derived from lexical material that travel down the Bybee-path (as shown by Bybee et al. 1994), but also items that are already grammaticalised. Fourthly, as the rise of the new form with the older meaning layer seems to have occurred in one particular context (/aɪl/), it is not just the case that new grammatical functions first arise in particular contexts, as has been shown (e.g. Bybee 2004: 602), but also that particular contexts give rise to (a new form with) an older meaning. Finally, the further development of ’ll shows 7
Will would presumably already have been used in the sense of pure prediction for some time at the end of the sixteenth century (see also Visser 1963–73: 1692f.).
88 Nesselhauf
that a desemanticisation or ‘bleaching’ of meaning does not necessarily persist in the later stages of a grammaticalisation process (see e.g. Hopper and Traugott 2003: 94; Bybee 2004: 607); in the case of ’ll what has occurred recently is rather a further specialisation of meaning, namely the possibility to express that the prediction that is being made is based on an intention just formed. This also constitutes further support for the finding that phonetic and semantic reduction do not necessarily go hand in hand. 5.4 Conclusion This chapter has traced the development of ’ll after the form was fixed not only in spoken but also in written language. It has been shown that it is still largely used in speech-based text types and still predominantly used with I in PDE, but with the proportion of other pronouns (and other elements) on the increase. An examination of its meaning has demonstrated that ’ll initially mainly expressed intention, with the sense of pure prediction displaying a considerable increase only from the nineteenth century onwards. Its meaning development also emerged as different from will in several respects, in particular its frequent present-day use for a spontaneous intention. With respect to the status of the form, the analysis in Nesselhauf (2013) strongly suggests that during the transition from the late Middle English to the LModE period, a construction /aɪl/ emerged from the earlier, varied, contractions. This was then reanalysed into I + /l/, of which /l/ was assigned the meaning of ‘prediction on the basis of the subject’s intention’. This construction then spread to other environments (as shown in Figure 5.1) and started to develop the meaning of pure prediction. That this was at least partly an independent development for ’ll was indicated by the present analysis. Further evidence for the status of ’ll as an (at least partly) independent construction in PDE is the increasing use of ’ll in contexts in which will has not, in turn, declined, as would have to be expected if ’ll has increasingly replaced will in the course of the past few centuries. Most importantly, the form ’ll has been shown in this chapter to have developed a new use, which it does not share with will (nor with shall), namely to indicate a spontaneous decision. Thus, while ’ll might also still maintain a link to will and/or remain part of the will-construction (and, for some speakers, possibly also has a link with shall) and thus constitute a contraction in certain cases, it certainly has (also) developed into an independent construction in PDE. With respect to the newly emerging field of diachronic construction grammar (see e.g. Bergs 2012; Noël 2007), the analysis presented in this chapter has provided further evidence that such a theory must account both for the emergence of new constructions (and decrease or even loss of older ones, such as volitional will) as well as for the change in already existing constructions, on both the form and the meaning side. It has also
From contraction to construction: the recent life of ’ll 89
provided evidence that the notion of construction can be integrated well into a theory of grammaticalisation (see also Sahkai 2009). As to the question of whether the form–meaning link of a construction has to be ‘at least partially arbitrary’ (Croft 2001: 18), as, for example, suggested in Radical Construction Grammar, this seems an unnecessary prerequisite, in particular if it is assumed that constructions have cognitive reality. As more and more evidence, in particular from the area of psycholinguistics, suggests, the human mind is much better equipped for storage than for fast processing (see e.g. Partington 1996: 20). Thus, multiple representations that allow high-frequency items such as I’ll to be retrieved as a whole rather than created from two constructions each time seem highly likely (see also Bybee 2004: 617, 621, 2010). An indication for the likelihood of multiple storage is the great and prevailing preference of /aɪl/ (both preference over the use of /l/ with other pronouns and preference of ’ll as future time expression after I; see also Nesselhauf 2013). This cannot simply be explained as a consequence of meaning specialisation of /l/, so that a disappearance of this construction right after reanalysis seems improbable. Finally, the chapter has provided an insight into the complexity of grammaticalisation processes when reduced and non-reduced forms are involved simultaneously; in particular, it has shown that a grammaticalisation process may bring forth not just one or several diachronically consecutive construction(s), but two new constructions that co-exist.
6
Books that sell – mediopassives and the modification ‘constraint’ Marianne Hundt
6.1 Introduction The verb sell usually requires an agent who does the selling and a patient that undergoes the process of being sold. In the mediopassive construction This book sells for only ten pounds, the verb is used intransitively and the patient (i.e. book) has been promoted to the position of subject. Normally, the promotion of a transitive object to the function of subject requires some kind of marking, for instance with passive voice, a reflexive pronoun or a combination of a reflexive pronoun with a verb meaning ‘let’. But English mediopassives carry no marking on the VP that signals detransitivisation.1 Objects of transitive verbs also appear as subjects in ergative constructions (e.g. The door opened) and the distinction between mediopassives and ergatives is often difficult to draw.2 The most important difference between the two is that mediopassive constructions have a specific meaning: they focus on inherent properties of the patient-subject that either facilitate or hinder the process expressed by the verb (see e.g. Erades 1950 or Dixon 2005). Other properties of the construction follow from its meaning, e.g. the fact it is not eventive.3 The following examples are typical representatives of the construction:4 (1)
Morton’s newest product … turns out to sell as well as its stock did … (Brown, A26) (2) After almost four years in the Senate, Barbara Boxer doesn’t shock easily. (Newsweek, 8 September 1997, p. 38) I would like to thank Martin Hilpert and Maarten Lemmens for helpful comments on an earlier version of this paper. Lemmens’ detailed criticism, in particular, led to some rewriting. 1 Note that mediopassives can also be formed from inherently intransitive verbs (see e.g. Davidse 1992; Levin, 1993: 26; Hundt 2007: 17ff.). Næss (2007) provides a recent discussion of prototypical transitivity within a typological perspective. Lemmens (1997, 1998a) includes diachronic evidence that verbs may shift in their constructional prototype (abort was originally ergative but now belongs to the class of inherently transitive verbs). 2 See Hundt (2007: ch. 3) for a review of previous literature on this issue. 3 Mediopassives are non-eventive and have therefore been claimed not to combine with the progressive; however, progressive mediopassives are attested, but have a special semantic effect (see Hundt 2007: 117–21 and example (6)). 4 Emphasis in examples has been added throughout.
90
Mediopassives and the modification ‘constraint’ 91
(3) (4)
Scraps of fleece … may combine to make a multi-colored fleece hat … (Vegetarian Times, June 1996, p. 14) When a book will not sell, it is shrewdly guessed, that it is not worth buying. (COHA, 1827, MAG)
One of the typical features of a mediopassive construction is that it contains an element which modifies the process expressed by the verb, be it a manner adverb like well or easily (as in examples 1 and 2), modals like may and will (examples 3 and 4) or negation (examples 2 and 4). For Palmer (1988: 92), the association of the construction with adverbial modification is so central that he even calls them ‘adverbial passives’.5 Examples (5) to (7) illustrate the fact that modification can also occur outside the VP. The prepositional phrase in (5) expresses the price of the merchandise and thus an inherent property of the patient in subject position. The sentence adverbial currently limits the temporal validity of the proposition in example (6) and thus enables the mediopassive to combine with the progressive aspect, thereby violating another apparent constraint that has been postulated to restrict the mediopassive construction.6 (5) Dozens of 1960s woven purses, priced to sell at $20 or less. (COCA,
2010, MAG) Currently, black leather jeans, jerkins and men’s frilly shirts are selling. (BNC, AJJ 132) (7) It should be pretty like a picture and eat like you’ re on holiday. (Comment about a desert on Master Chef, BBC 1 Television, 2.5.2013) (6)
The aim of this chapter is to use corpus data to trace the diachronic development of the modification constraint and to discuss the role that the modification constraint plays in the modelling of the construction within a Construction Grammar (CxG) approach. As the review of previous research in Section 6.2 shows, the variable degree to which different middable7 verbs require some kind of modification has not been investigated on the basis of synchronic or diachronic corpus data. Evidence for this chapter comes from a collection of American mail-order catalogues, the Corpus of Historical American English (COHA) and the Corpus of Contemporary American English (COCA). Section 6.3 will give background information on the data and methodology employed in this study. Various case studies serve to provide quantitative and qualitative evidence on the diachronic development of the 5
The semantic effect is also evoked in Fawcett’s (1980) term ‘facility-oriented passive’, Faltz’s (1985: 16) notion of ‘facilitative’ or the more recent notion of a ‘dispositional middle’ (Alexiadou and Doron 2011). For other terms that have been used for the construction, see Hundt (2007: 22f.). 6 For a detailed discussion of apparent constraints against corpus data for Present-Day English (PDE), see Hundt (2007: ch. 5). 7 The term was introduced by Lemmens (1998a, 1998b).
92 Hundt
modification constraint in Late Modern English (LModE) (Section 6.4). This provides the basis for the discussion of the modification constraint in Section 6.5. 6.2 Previous research on the modification constraint Modification of mediopassives is the requirement that linguists have commented on most widely. Levin (1993: 26) claims that ‘[m]ore often than not, the middle construction involves an adverbial or modal element’. The argument that middles are ungrammatical without some kind of modification is mostly found in generative accounts of the construction. But even some generativists concede that this constraint is pragmatic rather than structural. Ackema and Schoorlemmer (1994: 72), for instance, maintain that modification is mostly needed in examples out of context. Similar comments can be found in Fagan (1988: 201, 1992: 189) and Rosta (1995: 132) or in the CxG-based description in Goldberg and Ackerman (2001: 805f.). Dixon (2005: 450) claims that mediopassive constructions may occur without a manner adverb ‘when the construction is in “present” tense, perhaps reinforced by an adverb such as always, e.g. I find that Easter eggs always sell (sc. well)’. Garcia de la Maza (2012) obtained acceptability judgements from thirtytwo native-speaker informants indicating that modification enhances the acceptability of mediopassives in general. However, elicitation experiments are not particularly well suited to an investigation of the importance of the modification constraint. Out of context, mediopassives with adverbial or modal modification or negation will always be more acceptable than those without. Moreover, for previous stages of the language, elicitation data are unavailable. Hundt (2006, 2007) is the first to provide corpus evidence that the modification constraint is not a grammatical constraint in the generative sense of the term but rather a pragmatic requirement. Data from the British National Corpus (BNC) for the verb sell, for instance, show that sell is frequently used with some kind of modification, namely in over 90 per cent of all mediopassive constructions. The high proportion of modified mediopassives with sell has to be attributed mostly to instances giving the price of the merchandise. A more varied collection of mediopassive constructions, on the other hand, yields a relatively high proportion of ‘bare’ mediopassives. In a corpus of 477 mediopassives, bare variants amount to 26 per cent (Hundt, 2007: 114). In late-twentieth-century advertising language, bare mediopassives even outnumber those with some kind of modification at a little over 70 per cent; instances with adverbial modification are the second most common type (17.4%), whereas modal modification and negation amount to less than one per cent (Hundt, 2006: 177–78).
Mediopassives and the modification ‘constraint’ 93
The following examples illustrate the kind of bare mediopassives that are common in advertising language: (8) […] spandex adapts to your figure […] (Sears & Roebuck, 1986: 198) (9) Leveling glides adjust. (Sears & Roebuck, 1986: 806) (10) [pump] Disassembles for cleaning. (Sears & Roebuck, 1986: 713) (11) Top bolts on. (Sears & Roebuck, 1986: 603) (12) [high chair] Converts to youth chair. (Sears & Roebuck, 1986: 266)
Verbs that are well established in the construction (like fold, mount and adjust) are much more likely to occur without some modification (in 79% of all instances). Verbs with a lower token frequency as mediopassives are less likely to be used without modification. But the degree of entrenchment in the construction is not really the decisive factor for the use of bare mediopassives. We saw earlier that a verb like sell has a lower frequency of bare mediopassives than might be expected given the entrenchment of the verb in the construction. Verb meaning and text type appear to be equally important factors. In advertising language, for example, bare mediopassives are possible and even quite common with verbs that are attested only once in a mediopassive: (13) … hamper door tips out … (Sears & Roebuck, 1986: 559) (14) Variable-speed control dials down to super quiet. (Sears & Roebuck,
1986: 730) One of the characteristics of verbs that are used to form mediopassives in advertising language is that they can be the base of adjective formation in –able (see Lemmens, 1998a: 82–84, and the corpus evidence in Hundt 2007: 121–26). The following examples of low-frequency adjectives lend additional support to previous findings. However, the verbs they are derived from are more likely to occur in a mediopassive than to form the base for –able derivation: (15) Oh, there were quite enough mentally activatable switches. (COCA,
1996, FIC)8 (16) Danny pounced on the evidence … and stuck it into his only buttonable pocket for closer scrutiny later. (COCA, 2006, FIC) (17) A foldable, zippable compact hiking shoe. (www.springwise.com/ lifestyle_leisure/foldable-zippable-compact-hiking-shoe/) Moreover, adjective formation from middable verbs does not always make use of the derivational morpheme –able but might also involve conversion, especially with verb-preposition collocations like bolt on, clamp on, lift off or cinch back: 8
Maarten Leemens (p.c.) interpreted the adjective as an example of an ergative-based derivation. However, the presence of the adverb mentally implies external, human agency, making a mediopassive-based derivation more likely, in my opinion.
94 Hundt 70% 57%
60%
51%
active passive
50% 35%
40% 30%
mediopassive passive infinitive
26%
reflexive linear (mediopassive)
20% 10% 0%
1897
1927
1957
1986
Date
Figure 6.1 Mediopassives and alternative constructions in four Sears & Roebuck catalogues (from Hundt 2007: 164)
(18) Demountable bolt-on wheels turn on high-speed roller bearings (Sears
& Roebuck, 1986: 544)
(19) For on-site work, portable clamp-on units are available … (COCA,
1997, MAG) (20) This conversation piece … features a lift-off top for storage. (COCA,
2010, MAG) (21) For the best in comfort and fit […] an elasticized cinch-back waist.
(Sears & Roebuck, 1986: 124)
Hundt (2007: 164) also shows that, in advertising copy, mediopassive constructions increase in frequency against alternative patterns from the end of the nineteenth to the end of the twentieth century (Figure 6.1). The increase is particularly marked between the late 1920s and the late 1950s. A question that has not previously been investigated is how the modification constraint develops diachronically. It is possible that the weakening of the modification constraint was lost as a result of the spread of the mediopassive construction. It might also have gone hand in hand with the spread of the construction. A third possibility is that it occurred prior to the spread of mediopassives in advertising copy. Various studies have shown that transitivity alternations are sensitive to the meaning of the verb (e.g. Levin 1993; Lemmens 1998a). In addition, the semantics of the participants involved in the construction play a role, that is, whether the non-agent subject in a mediopassive is animate or inanimate (see e.g. Lemmens 1998a). The most prominent verb types that take animate non-agent subjects in a mediopassive construction are psychological event verbs (Hundt 2007: 96).
Mediopassives and the modification ‘constraint’ 95
With respect to the modification constraint, psychological verbs like scare, frighten, shock and offend, which occur with animate non-agent subjects, may behave differently from verbs like adjust, button, or zip, which can be used to describe inherent properties of merchandise.9 In other words, it is not just the constructional meaning itself but also its interaction with both verb meaning and the semantics of the participants that plays a role in what makes an acceptable mediopassive and hence in the requirement for modification. Evidence from elicitation experiments (Legenhausen 1998) shows that mediopassives from psychological event verbs vary with respect to their acceptability: out of a set of twenty verbs, scare and frighten showed the lowest amount of inter-speaker variation and were generally found to form acceptable mediopassives; acceptability ratings for embarrass, shock and offend ranked somewhat lower. Corpus data from BNC confirm that scare and frighten make good mediopassives, but overall, the construction is still a marginal option with this kind of verb (Hundt 2007: 97f.). (22) Part of Suzy’s allure, of course, is that though she writes about a lot of
rich and famous people, she is not awed by them. ‘I don’t awe easily,’ she says. ‘I know a lot about them, and I don’t want to change places with any of them. (Woman, October 2001, p. 153)
Previous corpus-based research shows that mediopassives are a prototypically structured category. Hundt (2007: 75) models the prototype effects within a CxG approach that connects mediopassives via inheritance links to other constructions (transitives, intransitives, reflexives, etc.). The modification constraint may apply more rigidly to verbs that form marginal mediopassives (i.e. psychological event verbs with human non-agent subjects or intransitive verbs) than to those that easily enter into the mediopassive construction (i.e. transitive verbs with affected objects). 6.3 Corpus data and methodology Mediopassives occur with a relatively low overall text frequency and large amounts of data are therefore needed. Evidence for the case studies comes from three sources, a non-computerised corpus of American mail-order catalogues and two electronic corpora of American English (COHA and COCA). Due to the nature of these corpora, different approaches to data retrieval were used. The American mail-order catalogue corpus consists of four Sears & Roebuck catalogues (from 1897, 1927, 1957 and 1986). This corpus is a good source for a diachronic study on the development of the modification 9
Note that animate (cattle and other kinds of livestock) and even human subjects (slaves in historical texts) can also be merchandise (see Section 6.4.2), but these are less typical instances of mediopassive construals of sell.
96 Hundt
constraint, given that the ‘inherent property’ meaning of the construction fosters its regular use in advertising copy. The catalogues vary considerably with respect to the relative amount of text and illustrations. However, differences in sample size are irrelevant for the purposes of this study because the relative frequency of bare mediopassives can be measured against those with some kind of modification. The mediopassive constructions were collected manually from the mail-order catalogue corpus (see Hundt 2007: 138).10 This has the advantage that all instances of mediopassives can be retrieved from the corpus (i.e. there will be 100% recall and precision). Automatic retrieval of all mediopassives from computerised corpora is impossible (see Hundt 2007: 86f.). Additional data from COHA and COCA were therefore collected by searching for individual verbs. This verb-based approach is necessarily selective. The verbs I decided to focus on are sell, attach and clamp on/to. The first is prototypically associated with the mediopassive construction; attach combines with a wide range of non-agent subjects and is therefore likely to allow conclusions based on a broader range of text types than that used in the previous section. Finally, clamp on/to was chosen because it is a low-frequency verb that allows me to answer the question whether frequency in diachrony is a prerequisite to the emergence of bare mediopassives. Another case study looks at psychological event verbs because these are rare in the mediopassive, combine with an animate patient in subject position and are thus expected to require modification more than those frequently used in the mediopassive and more amenable to the ‘inherent property’ semantics. While the mediopassive construction itself is rare and therefore occurs rather infrequently even in very large corpora, the verbs that enter into the construction are very frequent and it is therefore impracticable to manually post-edit all instances of a verb to weed out irrelevant data. The lemma search for sell in COHA retrieves a total of 69,537 instances, i.e. an impossible amount of data to analyse manually. As examples (23), (24) and (25) show, forms like sold, sells, and selling are all attested in the mediopassive construction: (23) The modern English paintings sold about as well as those which were
called old. (COHA, 1828, MAG)
(24) Ohio sends to New York produce which sells for $2,000,000 (COHA,
1827, NF) (25) … cloths then selling at 25 cents, can now be had at six and one quarter
cents per yard; … (COHA, 1827, NF)
Nevertheless, the analysis in Section 6.4.2 was limited to instances of the base form of the verb and the first half of the nineteenth century. This 10
Special thanks go to my amanuenses Carolin Biewer, Heidrun Fortner and Dorothea Halbe.
Mediopassives and the modification ‘constraint’ 97
was mostly in order to limit the data to manageable amounts. At the same time, with tensed examples (especially the past tense), bare intransitive sell often can be ambiguous between a mediopassive and an ergative analysis: the book sold, given a generic interpretation of the NP, is mediopassive because it is compatible with the inherent property reading; if it refers to a specific book and thus an event in the past, it is an instance of the ergative construction. The house sold is most likely to be given an eventive, ergative reading because it is difficult to construct a context in which the house could be generic.11 The event may have been facilitated by inherent properties in the subject NP, but the focus is still on the completion of the event. Similarly, to obtain manageable amounts of data, retrieval for the second case study was limited to attach and attaches. For clamp on/to (a much more infrequent verb) all mediopassive constructions could be retrieved. Data retrieval for psychological event verbs, including bare mediopassives, necessitated a markedly different retrieval strategy. A lemma search for scare produces 5,630 hits in COHA. A comprehensive look at all uses of the verb in the BNC had shown that the mediopassive was extremely infrequent at only 1 per cent of all instances; the relative frequency of mediopassive uses of other psychological event verbs was even lower (Hundt 2007: 97). Likely contexts of mediopassive scare are the adverb easily, a preceding negative particle or a following punctuation mark. Even though this introduces a bias into the data set, a combination of search strings that include such likely mediopassive construals of psychological event verbs will have reasonable recall (including bare mediopassives) and relatively good precision. The data in Section 6.4.3 are based on such queries.12 Once the data have been retrieved from the corpora, the next step is to define the variable ‘modification’. Adverbial modification sometimes combines with negation or a modal verb, as in examples (26), (27) and (28): (26) ‘I don’t scare easily,’ the huge man said. (COHA, 1920, FIC) (27) [exercisers] Can easily carry in satchel. (Sears & Roebuck, 1897: 597) (28) Fine and soft fabric which will wash and do up handsomely. (Sears &
Roebuck, 1897: 304).
11
Note that a generic analysis is possible in the context of prefab houses, as in Even before it featured on Channel 4’s Grand Designs, the Huf House sold all over the UK. The search strings used for the psychological event verbs included the following (examples given on the basis of scare): (1) ‘[scare].[v*].|,|;’, (2) ‘[scare] ********** easy|easily, (3) ‘[scare] easy|easily’ and (4) ‘not|n’t [scare].[v*]’. With the more complex query syntax, a total of 58 mediopassive constructions with scare were retrieved from COHA (i.e. 3.3 per thousand words; this compares with 4 mediopassives retrieved by manually post-editing all 591 instances of the verb in the BNC, i.e. a relative frequency of 6.8 per thousand words). This can be considered reasonable recall if we take into account that mediopassives have become more frequent over time and that quite a few instances of the verb scare in COHA were actually mis-tagged instances of the noun.
12
98 Hundt
Modal verbs, in turn, are often negated, as in examples (29) and (30): (29) A Corset of Perfect Form that will not stretch, break, roll up or pucker
(Sears & Roebuck, 1897: 306)
(30) [footed hand lamp] … the collars … cannot work loose. (Sears &
Roebuck, 1897: 702)
In the analyses, combined modification was at times subsumed under the heading of one kind of modification (e.g. adverbial and modal under ‘adverbial’). Instances that were modified by elements which did not fit in one of the categories above were labelled as ‘other’. Typical examples are illustrated by the following instances from the Sears & Roebuck corpus: (31) [toy safe] Fastens with a lock and key. (Sears & Robuck, 1897: 99) (32) Tank fills from top. (Sears & Roebuck, 1927: 1075) (33) [telephone] Attaches to wall by four suction cups. (Sears & Roebuck,
1957: 1144) 6.4 Findings 6.4.1 The development of the modification constraint in advertising copy For the results obtained from the Sears & Roebuck corpus, it is important to note that mediopassives with adverbials include instances with additional negation and modal verbs. Similarly, negated modals are subsumed under the label of ‘modal modification’. The counts for mediopassives with negation thus include only those with a negative particle and do-support but no additional modal or adverb. Figure 6.2 gives the relative frequencies of mediopassives with adverbial modification, a modal, negation or other kind of modification, as well as the kind of bare mediopassive illustrated in examples (8)–(12) above. The evidence from the Sears & Roebuck catalogues indicates that bare mediopassives increase most markedly between 1927 and 1957, that is, simultaneously with the drastic increase of the construction itself. While there was only a mild increase in the relative frequency of mediopassive constructions towards the end of the twentieth century, the proportion of bare mediopassives increases substantially from just under 60 per cent in 1957 to slightly over 70 per cent in 1986. Somewhat surprisingly, the most frequent kind of modification in the 1897 catalogue is not a manner adverb but the category ‘other’. On closer inspection, this has to be attributed to the large number of mediopassives with the verbs sell and retail, which often combine with the price. Curiously enough, these verbs are no longer used as part of the advertising strategy in the 1986 catalogue. If mediopassives from these verbs are removed from the count, a much more intuitively plausible picture emerges, as can be seen in Figure 6.3.
Mediopassives and the modification ‘constraint’ 99 80%
71.4%
70% 59.2%
60%
adverbial bare
50% 37.5%
40%
modal negation other
30% 20%
18.6%
10% 0%
1897
1927
Date
1957
1986
Figure 6.2 Modification of mediopassives in four Sears & Roebuck catalogues (1897: N = 441; 1927: N = 496; 1957: N = 2,750; 1986: N = 1,090) 80%
71.4%
70%
60.2%
60%
40%
adverbial bare modal
45.9%
50% 30.7%
negation
30%
other
20% 10% 0%
1897
1927
Date
1957
1986
Figure 6.3 Modification of mediopassives in four Sears & Roebuck catalogues – skewing verbs removed (1897: N = 267; 1927: N = 405; 1957: N = 2,706; 1986: N = 1,090) (from Hundt 2007: 164)
The category ‘other’ is now a fairly stable factor in the modification of mediopassives at around 10 per cent. Adverbial modification remains fairly stable for most of the time (at around 27–28%) and only decreases at the end of the century. Once sell and retail are removed from the data, bare mediopassives start out as the most frequent type of mediopassive construction even in the 1897 catalogue, but compared with the initial results, the overall
100 Hundt
increase of bare mediopassives now proceeds at a steadier pace. Initially, modal verbs outnumber adverbials as the most common kind of modification. An important difference in the modified data set is that modal verbs now show the most drastic decrease as modifiers of mediopassives from 29 per cent in the 1897 catalogue to 0.75 per cent in the 1986 catalogue. The question is why it is this type of modification that decreases so markedly over time in advertising copy. A look at the following three examples with modal modification will help explain this result: (34) [husking pin] … can adjust to any size hand. (Sears & Roebuck,
1897: 229) (35) [Handy lamps] Will clamp to almost anything … (Sears & Roebuck,
1927: 655) (36) [Screwdriver] Can convert to manual return. (Sears &Roebuck
1957: 1403) Mediopassives in advertising copy are used to verbally display important design features of the merchandise. In Present-Day English (PDE), where the constructional meaning of the mediopassive is well established, a modal verb can be argued to provide redundant information. For the modern reader, the modal verbs do not add essentially to this meaning. If the modals are left out, the meaning remains the same: (34)′ [husking pin] … adjusts to any size hand. (35)′ [Handy lamps] Clamps to almost anything. (36)′ [Screwdriver] Converts to manual return.
The goods advertised in the modified examples (34)′-(36)′ are still adjustable, clampable and convertible. Omitting the modal does not result in semantically impoverished expressions. It just makes for snappier advertising copy that reads better and takes up less space on the page. However, at the end of the nineteenth century, with a lower overall text frequency of the construction, the modal semantics of the mediopassive may not yet have been as well established and modal modification therefore was less dispensable than it is in PDE. The diachronic evidence on the decrease of the modal modifiers ties in nicely with Davidse and Heyvaert’s (2003:71) analysis of mediopassives as inherently modal constructions: What typifies this construction is the non-agentive letting-function of its Subject in the process … We have stressed that the letting notion is modal in nature … and arises from the constructional link between a non-agentive Subject and an active VP.
As modal modification does not add substantially to the mediopassive semantics in PDE advertising copy, it is not surprising that this kind of modification decreases as bare mediopassives become more frequent (and thus more entrenched) in the course of the twentieth century.
Mediopassives and the modification ‘constraint’ 101
6.4.2 The modification constraint: three case studies 6.4.2.1 Sell For the analysis of modification found with sell in COHA, the category ‘price’ was defined in a fairly broad sense, including the following examples with prepositional phrases that semantically express a low price alongside actual prices quoted: (37) Who can bottle up the sunlight, to sell at retail? (COHA, 1834, FIC) (38) Such stuff would sell in York for candy. (COHA, 1845, FIC) (39) ... the buckskins will sell for a shilling apiece … (COHA, 1845, FIC)
Speakers of PDE would not expect mediopassive sell to be used with human non-agent subjects, but examples from COHA show that these are attested in earlier stages of the language, and with the same kind of modification that sell typically takes in this construction: (40) All domestic slaves … are always in great demand, and often sell at the
most extravagant prices. (COHA, 1835, FIC)
(41) I want you to brush your hair all back straight, tomorrow, said Susan.
‘What for, mother? I don’t look near so well, that way.’ ‘Yes, but you’ll sell better so.’ (COHA, 1842, FIC, Uncle Tom’s Cabin)
Note that the subject in example (40) could also be given an agentive interpretation, so background knowledge on the novel, the characters and the historical facts is necessary to arrive at the (intended) mediopassive reading. In COHA, the first mediopassive uses of sell are attested in the 1820s, and they are already quite frequent, at nineteen out of thirty uses of the base form. All mediopassives from this decade feature some kind of modification, though, with the price occurring in fifteen out of nineteen instances, often in combination with a modal: (42) ... a blanket or two yards of coarse cloth, which cost two dollars, would
sell for sixty (COHA, 1827, MAG)
(43) It rents at present for – dollars per annum: perhaps it might sell for –
thousand dollars. (COHA, 1827, FIC)
Bare mediopassives of sell, as in PDE, are extremely rare in the historical data. In fact, there is only one example from 1843 (out of a total of 101 mediopassives in the first half of the nineteenth century). (44) … provided his books sell, he probably cares little whether the public
adopt or reject his opinions. (COHA, 1843, FIC)
One could argue that the hypothetical context (i.e. the verb provided) in (44) licenses the bare use of the mediopassive because it implicitly establishes a contrast between books that sell and those that do not (see also Lemmens 1998a: 78).
102 Hundt
As in the data from BNC, a large number of mediopassive uses of sell are modified by the price of the merchandise (60 out of 101), often in combination with a modal (36 of 60). Modal modification is also very frequent (66 out of 101 uses), either in combination with the price, by itself or with an adverb. Adverbial modification (even in combination with another modifying element) is much less frequent, at an overall 24 out of 101 instances. The diachronic data on sell from COHA thus confirm the evidence from BNC: the verb is frequently attested in the construction but very rarely as a bare mediopassive. 6.4.2.2 Attach The data come from searches of attach/attaches sampled from COHA at thirty-year intervals. Initially, all intransitive uses of the verb were included in the analysis. The first intransitive use of attach is from 1816, even though intransitive attach in the meaning of ‘adhere’ (either with a reflexive sense or ‘as an appertaining quality or circumstance’) is attested from the end of the eighteenth century, according to OED (s.v. attach v. 10,11). All early examples of intransitive attach in OED and COHA are with abstract nouns such as blame, failings, reputation or interest:13 (45) In most cases of contention, considerable blame attaches to all the par-
ties concerned. (COHA, 1817, NF)
(46) … she undoubtedly had her share of the failings which attach to us all.
(COHA, 1849, MAG)
(47) … the sort of reputation which attaches to the Prince of Wales,
(COHA, 1870, MAG)
(48) But a peculiar interest attaches to
have got in connection with the preterite-present verbs. (COHA, 1907, MAG)
The intransitive uses of attach in (45)–(48) are incompatible with facilitative modal or adverbial modification, i.e. they are not mediopassive construals of the verb. The first instances of intransitive attach with the ‘attachable’ reading are from the second half of the twentieth century: (49) … a short bar which attaches to the wall (COHA, 1969, NF) (50) … the 35-oz. ‘Futuramic Strobonar’ attaches to any camera equipped
with speedlight synchronization. (COHA, 1958, MAG)
I would argue that, on semantic grounds, examples (45)–(48) illustrate a different construction from (49) and (50), the difference being one of implied agency, which is absent from the former but implicit in the latter. In other words, on the transitivity cline, (45)–(48) are somewhat more intransitive 13
Occasionally, concrete objects also attach (usually with a strong reflexive sense): ‘These are structurally continuous with the short vertebral muscles that attach to the spines of the vertebrae’ (COHA, 1937, NF).
Mediopassives and the modification ‘constraint’ 103
(and closer to ergative constructions) than (49) and (50) with their implied agents. However, (45)–(48) are non-eventive and therefore not prototypical ergative constructions, either. Interestingly, the first clear examples of mediopassive attach do not feature any kind of modification. As the mediopassive construal with the ‘attachable’ sense increases slightly in frequency, examples with modification start occurring but are still outnumbered at only five out of seventeen in the 1990s; modification is typically by a modal or a prepositional phrase: (51) … the drink tube can attach to either the spout of a water filter or a
backcountry shower spigot. (COHA, 1997, MAG)
(52) … it’s usually keyed so that it will attach only one way. (COHA,
1994, MAG)
(53) … it uses a small dish…, which can attach to your roof or the side of
the house … (COHA, 1994, NEWS)
(54) … the primary and secondary/eyepiece holder sections that attach
with long aluminum tubes. (COHA, 1995, MAG)
The important outcome of this case study is that intransitive uses of attach with abstract nouns and without any kind of modification are attested relatively early in the LModE period. While these are not instances of the mediopassive construction themselves, their availability might have facilitated instant use of bare mediopassives when the verb first entered the construction in the second half of the twentieth century. 6.4.2.3 Clamp on/to A search for the collocations clamp(s) on and clamp(s) to yielded seventy hits from COHA, including instances of the noun clamp followed by a prepositional phrase. Of the verbal patterns, there were a total of five mediopassives, and only one of them was modified by a negated modal (56); however, two more instances (55 and 58) feature modification from the category ‘other’. The remaining two (57 and 59) are ‘true’ bare mediopassives: (55) The heart of the instrument is the little ear-lobe oxygen meter, which
clamps on like an earring. (COHA, 1949, MAG)
(56) The hose attachment … wouldn’t clamp on. (COHA, 1957, MAG) (57) Most consist of rollers that clamp on to the blade. (COHA, 1976, NF) (58) This arm clamps to the pipe with set screws … (COHA, 1996, MAG) (59) … a digital readout that clamps to your bike’s handlebar … (COHA,
2006, MAG) Since all instances of mediopassive clamp in COHA were from the twentieth century, additional evidence was collected from COCA, which yielded a total of fifteen mediopassives of clamp(s) on/to (out of 184 noun/verb– preposition collocations). With the exception of one instance (60), all of these were examples of bare mediopassives. Many of them came from
104 Hundt
magazines in which properties and characteristic design features of new products (bicycles, computers, carpentry tools, etc.) were described, i.e. from contexts similar to those found in advertising copy (61): (60) The detector unit is compact and simply clamps on to the pipe enabling
installation in less than an hour … (COCA, 1993, ACAD)
(61) A tabletop art projector designed to clamp to a drawing table …
(COCA, 1990, MAG)
However, there were also a few instances from other text types, such as fiction (62) and even academic writing (63). In other words, the semantics of the mediopassive construction is not limited to advertising copy: (62) One of my jobs is keeping a large aluminum umbrella, which clamps to
the pipe, in place (COCA, 2000, FIC)
(63) The platform clamps on … (COCA, 2007, ACAD)
Taken together, the COHA and COCA data for this verb suggest that the development of bare mediopassive constructions does not depend on a previously frequently attested prototypical mediopassive use of the verb. In other words, there does not have to be a development from modified to bare mediopassives for an individual verb. At the same time, the majority of bare mediopassive construals of clamp are from the end of the twentieth century, and it might well be that the construction as such has continued to increase in frequency in the last decades of the twentieth century, thus making bare mediopassives more readily available even for low-frequency verbs like clamp. 6.4.3 Mediopassive constructions from psychological event verbs As pointed out in Section 6.3, mediopassives of psychological event verbs are very infrequent. A look at the diachronic data from COHA shows that only one of the psychological event verbs investigated in this chapter, namely scare, is attested in mediopassives that do not contain any modification within the verb phrase (additional context is quoted to show that these instances are not examples of the transitive use of scare with an omitted object). This variant is attested relatively late, however, namely at the beginning of the twentieth century:14
14
The only instance of mediopassive frighten in COHA that I found is from an author (D. A. Walcott) who is from the Caribbean and makes frequent use of non-standard language in his writing; I did not want to include this in a study of mediopassives in British and American English: ‘I am the hypocrite? That lightning would make devil sweat cold. Of course I frighten, man cannot control thunder’ (COHA, 1959, FIC) – the character in this extract uses two non-standard articles: the definite article instead of indefinite a with hypocrite and a zero article with devil.
Mediopassives and the modification ‘constraint’ 105
(64) And who told you I was scared? Forget that stuff. Take it from Annie
that she ain’t the kind that scares.’ (COHA, 1920, FIC)
(65) Homer was a cool customer. It wasn’t like him to scare. (COHA,
1921, FIC) (66) Robinson gibed. ‘You scare me, Pops – you scare me to death.’ North’s
hand reached down to bunch the gaudy sports jacket at the lapels. He pulled Robinson up toward him and grunted his words into the brown, pitted face. ‘You’d better scare, Robinson, because I get real mean when I’m sore and I’m sore now. (COHA, 1960, FIC)
Note, moreover, that (64) and (65) are negative contexts that establish a contrast and thus make the ‘bare’ mediopassive use of scare more acceptable; the same could be said of the modal-imperative context in (66). Interestingly, only two of the four psychological event verbs investigated here are attested in a mediopassive construction in OED, namely scare and shock: (67) The big depositors wouldn’t scare. (Academy 8 September 1900, 194/1;
OED s.v. scare v. 3.)15 (68) She liked to say things to shock Clarence, partly because he shocked so easily. (N. Clad, 1960, Love and Money 15, OED s.v. shock v.2, 4.)
According to OED, offend in the sense of ‘take offence’ is attested in intransitive constructions in the Early Modern period, but the examples provided would not qualify as mediopassives because they do not involve inherent properties of the non-agent subjects but refer to past or future events (i.e. in the sense of ‘take offence’): (69) A Noble man … answered, that it was a devout imagination, where-
with John Knox did greatly offend. (J. Spottiswood, 1677, Hist. Church Scotl., iii. 174; OED s.v. offend v. 5) (70) That his Majestie will offend with the lenitie alreadie shewed. (D. Calderwood. a1650, Hist. Kirk Scotl. VII. 617; iii. 174; OED s.v. offend v. 5)
Even though psychological event verbs are middable verbs, in principle, they do not seem to lend themselves (easily) to adjective derivation in –able: scareable, frightenable and offendable/offendible are unattested even in very large corpora of contemporary English such as COCA and the Time Magazine Corpus (TimeMag); shockable is only attested once with a human subject in COCA and twice in TimeMag, the earlier being an example from 1974. Moreover, they feature the kind of modification typical of the mediopassive
15
Note that there was an earlier intransitive pattern, namely to scare at, which is now obsolete but attested between 1575 and 1721 in OED.
106 Hundt
construction, strengthening the constructional link between the adjectives and middable verbs (see also Lemmens 1998a: 82).16 (71) That is a very important question because only asking Why? shocks
the establishment or beancounters, who are not easily shockable, owing to this being the sixties and all the romantics. (COCA, 2012, FIC) (72) Millett’s latest book, Flying, to be published in June, will tell all about her bisex life to an audience not so shockable. (TimeMag, 1974/10/13) But shockable is not an innovation from the last third of the twentieth century; late nineteenth-century examples are to be found in OED and COHA: (73) His parents were not of the shockable sort. (W. G. Collingwood, 1893,
Life and Work J. Ruskin I. 27; OED s.v. shockable adj.)
(74) He becomes the timid, the prudent, the shockable lover as contrasted
with the brave. (COHA, 1892; Henry James, A Change of Heart)
A relatively frequent –able adjective derived from a psychological event verb is irritable, but irritate, in turn, is not attested in the mediopassive construction, at least not as a psychological event verb with a human or animate subject,17 probably because get irritated is a well-established construction that corresponds to the mediopassive when it is used with a manner adverb like easily.18 Interestingly, frightenable and offendable are both attested in LModE, as the following examples from OED show: (75) You do not look so frightenable as my Aunt does. (F. Trollope, 1850,
Petticoat Govt., 32; OED s.v. frighten v. Derivatives).
(76) I am the least offendable of mortal men. (A. Helps. 1868. Realmah,
xvii. 482, OED s.v. offendable adj.)
On the whole, then, the diachronic evidence discussed here shows that mediopassives from psychological event verbs are infrequent and non-prototypical on several accounts. It is not surprising that modification facilitates the mediopassive construal of these verbs and that bare mediopassives of these verbs, in turn, are exceedingly rare. The fact, however, that even such non-prototypical mediopassives can occasionally do without modification
16
Note that while mediopassive equivalents of (70) and (71) would satisfy the modification constraint, the following is an example of an unmodified -able adjective from a psychological event verb: ‘The plummy clubmen, the young wastrels in spats and waistcoat, the shockable aunts, the frosty butler belong in a diorama at the Museum of Natural History, not onstage’ (TimeMag, 1983/10/30). 17 The following is a very recent example of the mediopassive use with an inanimate subject (i.e. not a psychological event verb): ‘Shaving can be a culprit too; razors create micro-cuts that irritate easily’ (COCA, 2012, MAG). 18 On the connection between the get-passive and the mediopassive, see Hundt (2007: 69–72).
Mediopassives and the modification ‘constraint’ 107
(given the right context) strengthens the argument that modification is not a constraint in the generative sense of the term. 6.5 Modification – a constraint on mediopassive formation? The case studies in the previous section yield quite different results with respect to the modification constraint on mediopassives that are relevant on two levels: a theoretical and a more descriptive one. On the descriptive level, the usage-based perspective provided here confirms that the modification constraint does not apply globally to all mediopassive constructs (where ‘construct’ is used in the CxG sense of ‘instantiation’/token of a construction/type). More importantly, it is highly context dependent: (a) certain text types (such as advertising copy) seem to license bare mediopassives more than others (see Section 6.4.1), (b) contextually implied contrast may enable an inherent property reading of a bare mediopassive (see Sections 6.4.2.1 and 6.4.2.2), or the verb itself might foster bare mediopassive use (this seems to be the case with denominal verbs like zip, button, bolt or clamp, see Goldberg and Ackerman 2001: 806). The semantics of the verb and the subject also seem to play a role, as the case study on psychological event verbs has shown: with animate (human) subjects it is particularly difficult to obtain a mediopassive reading without some kind of modification. With respect to text type as a relevant factor, the diachronic survey of mediopassives in advertising copy reveals that bare mediopassives are already relatively frequent at the end of the nineteenth century. They increase steadily throughout the twentieth century, while mediopassives with modal modifications decrease. But whereas there is a marked increase of mediopassives in relation to alternative constructions between the late 1920s and the late 1950s, bare mediopassives increase more steadily over time. Moreover, once mediopassives and their special semantics are entrenched – both as the prototype with modification and the variant without – new verbs can be used in a bare mediopassive without previously having to be used in a modified mediopassive. On a theoretical level, the results feed into a CxG taxonomy of mediopassives that complements the model provided in Hundt (2007: 75), which only dealt with the inheritance of properties between the macro-level (transitive vs. intransitive constructions) and the meso-level (mediopassives, ergatives, passives, etc.).19 The evidence provided in Section 6.4 suggests that on the micro-level, different verb classes may give rise to different kinds of mediopassive that require modification to different degrees: verbs of combining and attaching (Levin 1993: 159–64) appear to be able to do without a modifying element, 19
For levels in CxG, see e.g. Trousdale (2008b: 170).
108 Hundt Macro-level
Transitive construction
Intransitive construction
Meso-level
Mediopassive construction
Ergative construction
Micro-level
[verb class 1-mediopassive]
[verb class 2-mediopassive]
Construct
People who frighten easily ...
Dress buttons at the back.
Figure 6.4 Constructional taxonomy for English mediopassive constructions
especially if they are well established in another intransitive construction prior to their use in a mediopassive. Psychological event verbs and ergative verbs like break and open, on the other hand, require some kind of modification to the extent that it is extremely difficult to find contexts in which they are used as bare mediopassives. On the level of the construct, individual verbs like sell, finally, diverge from the prototype in that they more commonly take another kind of modification than an adverbial or a modal. The typology of mediopassive constructions in Figure 6.4 thus suggests that the mediopassive is not simply one abstract construction but that – at the microlevel – it is rather to be viewed as a family of constructional schemas.20 6.6 Conclusion The diachronic case studies of mediopassive constructions in LModE have shown that it is impossible to treat ‘the’ development of ‘the’ modification constraint in mediopassive constructions on a global scale. Instead, a CxG approach allows us to take into account the variability of the constraint and how it relates to the interaction between constructional meaning on the one hand and the semantics of verbs and the participants they typically occur with on the other.21 Variability in the modification requirement is thus found on the micro-level and the levels of individual constructs rather than on the
20
See Hilpert (2013: ch. 5) for a similar view of the historical development of concessive parentheticals. 21 In this respect, mediopassives are similar to the English at-construction as discussed in Perek and Lemmens (2010).
Mediopassives and the modification ‘constraint’ 109
more abstract macro- and meso-levels. With individual constructs it might ultimately be impossible to arrive at a mediopassive or ergative meaning out of context, even with the use of modal or adverbial modification: He will wake and He wakes easily remain ambiguous between the two readings, and only the context will indicate whether an external agent is implied or not. As far as frequency effects and a potential weakening of the modification requirement over time are concerned, the case studies provide evidence that bare mediopassives are typical of twentieth-century usage but rare in previous periods. This is most likely a result of the overall spread of the construction and a concomitant strengthening of its modal semantics. Case studies on attach and clamp show that the modification requirement does not have to weaken diachronically for individual verbs. With attach, previous nonmediopassive intransitives turned out to have paved the way for bare mediopassive use of the verb. With clamp on, a combination of factors is likely to have enabled instantaneous use of bare mediopassives: an overall prior increase in text frequency of the construction and a concomitant strengthening of the inherent constructional meaning seems to have worked in tandem with text-type effects to facilitate bare mediopassive use of the verb.
Part II Changes in the NP
7
Beyond mere syntactic change: a micro-analytical study of various and numerous Tine B re ban
7.1 Introduction This study is concerned with two concepts of syntactic change, actualisation and paradigmaticisation. Actualisation (among many others, see Timberlake 1977; Andersen 2001; De Smet 2012a) involves the emergence of new syntactic behaviour as a consequence of prior syntactic reanalysis. In reanalysis, a particular syntactic string is assigned a new underlying structural analysis, without affecting the surface structure. At this point, change has taken place, but it is not observable. It is during the actualisation process that the new analysis starts to become visible, when the surface structure displays new syntactic behaviour typical of the newly assigned function and incongruent with the old function. The second concept, paradigmaticisation, was introduced by Lehmann (1982, 1985) as one of the features accounting for the morphosyntactic changes in grammaticalisation. Paradigmaticisation refers to the incremental integration of the item into the new grammatical paradigm by assimilation of the morphosyntactic behaviour of existing members of that paradigm. Like actualisation, it takes place after the initial reanalysis. This study presents new data on the development of adjectives into quantifiers in Late Modern English (LModE) which show that actualisation and paradigmaticisation do not only or necessarily figure as syntactic change and expansion but also operate on other levels of linguistic behaviour, including collocational expansion and contextual embedding. These changes are less immediately obvious than syntactic change, but are brought out by microanalysis of actual language data, as presented in this chapter. The structure of this chapter is as follows: in Section 7.2, I will provide the necessary background to the development of adjectives into quantifiers. Section 7.3 discusses two new case studies of this development, that of various (7.3.1) and numerous (7.3.2). Section 7.4.1 summarises the findings of the two case studies, and Section 7.4.2 discusses their implications for the notions of actualisation and paradigmaticisation. Section 7.5 wraps up the discussion about actualisation and paradigmaticisation and identifies the implications of the case studies for certain tenets about grammaticalisation, including the primary role and form of semantic reanalysis. 113
114 Breban
7.2 The development of new quantifiers in Late Modern English Throughout the history of English the set of unspecific quantifiers has been subject to a considerable turnover. There are a small number of very frequent quantifiers such as many, much and few, which have been part of the English language since Old English. Fuelled by the general idea that word classes with (more) grammatical meanings are typically small and closed, this has led to an intuitive perception that quantifiers are a small and stable set. However, in recent years, several studies have discussed the development of new unspecific quantifiers. Brems (2003, 2011, 2012) and Traugott (2008a; Traugott and Trousdale 2013) focus on binominal quantifiers such as a lot of/lots of, a bit of, a scrap of, heaps of. Denison (2006, 2010) and Breban (2008, 2010) propose the existence of a productive path of development from adjectives to unspecific quantifiers, affecting, for example, various, several, divers(e). Other quantifiers have fallen out of use. Frequent quantifiers fele and sundry, for example, became obsolete in Early Modern English. Fele was entirely replaced by the already common quantifier many (see OED, s.v. †fele adv. (and n.) and adj.2). Sundry and the less frequent divers(e) lost out to the newer, and semantically similar several (Breban 2008, 2010). Previous diachronic studies identify LModE as the period in which prominent new members, such as a lot of, a bit of and several, became established quantifiers. This chapter examines the development of two new quantifiers in LModE, various and numerous, on the basis of data from the Penn Parsed Corpus of Modern British English (PPCMBE). The data have been analysed on a micro-analytical level for a wide range of linguistic features on the levels of semantics, syntax, and collocational/contextual embedding. The findings of this study are interpreted in the light of the hypotheses proposed by Breban (2008, 2010) for the development of another former adjective, several. Breban argues that several underwent a two-step grammaticalisation process, first from descriptive adjective with lexical, referential semantics to individualising postdeterminer, then from postdeterminer to unspecific quantifier, which is the main use for several in Present-Day English (PDE). The following examples from the OED illustrate the different uses for several. Examples (1–3) contain several as a descriptive adjective in a predicative construction and as adnominal adjective respectively. Here several means ‘separate, existing apart, different’. (1)
Many Mahumetans bee also found both several in sundry prouinces, and otherwise mingled with idolaters. (OED, s.v. several, a1613: E. Brerewood. Enq. Lang. and Relig. (1614) xii. 89) (2) Every scale of an onyon is a severall and differing scale. (OED, s.v. several, 1635: J. Swan. Speculum Mundi iv. §2. 66) (3) Christianity hath vsed many words in seuerall sence from the common phrase. (OED, s.v. several, 1599: Master Broughtons Lett. xii. 40)
Beyond syntactic change: various and numerous 115
In (4) and (5), several does not describe a property of the referent, but has a procedural function. It explicitly codes that an implicitly (as in 4) or explicitly (5) plural referent has to be perceived as individual instances rather than as a single referent or a plural mass. (4)
To euery dish a seuerall sawce appropriat to his kinde. (OED, s.v. several, 1583: P. Stubbes Anat. Abuses sig. Hviiiv) (5) I doe not like these seuerall councels. (OED, s.v. several, 1597 Shakespeare Richard III iii. ii. 73) (6) During which times he received severall sums of money to the value of 300l. (OED, s.v. several,?1661 in 12th Rep. Royal Comm. Hist. MSS (1890) App. V. 6) In (5), several can be paraphrased as ‘these different counsels’, with the intended meaning being ‘any of these counsels’, that is, foregrounding the individual instances in the plural referent set.1 Example (6), finally, illustrates the quantifier use meaning ‘more than one, but not very many’. The two-step hypothesis is supported by the changing relative frequencies of the different uses in the distinct consecutive historical data samples for several in Breban (2008, 2010), as well as by the presence of two sets of vague examples, between a descriptive adjective and individualiser reading on the one hand, and between individualiser and quantifier interpretation on the other. 7.3 Case studies The data studied for various and numerous are exhaustive extractions for both items from the PPCMBE. They are divided into three sets following the subperiodisation of the PPCMBE, i.e. 1700–69, 1770–1839, 1840–1914. All examples are independently analysed for (a) semantics (lexical, quantifier, etc.), (b) a range of formal (predicative, pronominal, adnominal, number of the head noun, etc.) and contextual features (presence of other quantifiers, distributive triggers such as each, every, etc.),2 and (c) the nouns they modify (collocational analysis). In this section, I present the findings of these different analyses and show how they interact for each of the adjectives. Despite the relatively restricted amount of data analysed, together these three types 1
Breban (2010: 181–88) uses Langacker’s (1991: 78–81) analysis of plural noun phrases to explain the process involved. According to Langacker, plural nouns by default denote a (heterogeneous) mass and through cognitive processes of reification foreground the mass aspect. Breban (2010: 183) proposed that if a language user wants to foreground the heterogeneity, i.e. the individuals within the mass, s/he has to overrule the default interpretation. One way of doing this is to add an individualiser to the determiner. 2 Where necessary, the attested formal and contextual patterns are checked against the Corpus of Late Modern English Texts, version 3.0 (CLMET3.0), which is a larger corpus of British Late Modern English (34,386,225 words) based on public-domain texts available on the web. It is divided into three 70-year periods, 1710–80, 1780–1850, and 1850–1920.
116 Breban
of analysis tell a coherent story, each adding different kinds of evidence. Even though various and numerous, as well as several, develop along the path descriptive adjective > quantifier, the processes of change are very different, with semantic, formal, contextual and collocational changes happening at different moments and playing different roles. 7.3.1 Various in PPCMBE This section presents the data analysis for various on the 161 examples of this word in the PPCMBE, covering the three subperiods. The data are analysed separately for semantics/function, formal/contextual features and collocational patterns. Table 7.1 summarises the results of the semanticofunctional analysis. In the first period, various is strongly associated with descriptive adjective usage, paraphrasable as ‘varied, displaying variation’. Eighty per cent of the data can only be analysed as having a descriptive interpretation, as in example (7), the remaining 20 per cent are vague between a descriptive reading and one of the procedural interpretations. In example (8), for instance, various can be read as describing qualitative variation between the degrees and kinds of putrefaction, i.e. as a descriptive adjective equivalent to ‘dissimilar, varied, diverse’. This reading brings out the qualitative differences between the types of putrefaction. On a second reading, various functions as a means of introducing the list of kinds of putrefaction; this is various as individualiser, paraphrasable as ‘respective, particular, diverse’. The function of various is procedural and signposts the list to follow. (7)
The Reader is here to survey a Character of such eminent and various Goodness, as might demand Veneration (PPCMBE period 1) (8) It would be foreign to my purpose, to observe what various degrees and kinds of putrefaction may be induced in the human body by other means, viz. by putrid ferments, or putrid substances of any kind, contagious poisons, and acrimony of different sorts (PPCMBE period 1) The analysis for the second period recognises the first clear individualiser and quantifier uses, e.g. (9) and (10), paraphrasable as ‘distinct, individual’ and ‘several’ respectively. (9)
[…] all educated Europeans conceive the relation of the various parts and lines on the terrestrial globe with a distinctness in which the rude savage or uncultured boor has no share. (PPCMBE period 2) (10) The walls, in various parts, are ornamented with circular representations either of the sun, or full moon, and with gilded crescents. (PPCMBE period 2) The most obvious trend across the three periods is the decrease in the proportion of examples with a descriptive meaning (from 80%, to 41% to 29%).
Beyond syntactic change: various and numerous 117 Table 7.1 Semantico-functional analysis of various
descriptive adjective individualiser quantifier descr. adj./individualiser individualiser/quantifier descr. adj./quantifier descr. adj./indiv./quant.
1700–1769
1770–1839
1840–1914
28 (80%)
26 (41%) 11 (17%) 10 (16%) 5 (8%) 2 (3%) 8 (12%) 2 (3%) 64 (100%)
18 (29%) 16 (26%) 8 (13%) 7 (11%) 6 (10%) 5 (8%) 2 (3%) 62 (100%)
5 (14%) 2 (6%) 35 (100%)
Individualiser examples continue to increase towards period 3 (from 0% to 17% to 26%), whereas quantifier examples do not increase between period 2 (16%) and period 3 (13%). Interestingly, the vague examples found in period 1 show vagueness of the descriptive sense with both the individualiser and the quantifier use in separate examples. This observation gains more significance if combined with the findings from Breban’s (2008, 2010: 328–29) study of various, which includes data from earlier periods as well.3 This study identifies LModE as the first period in which both the individualiser and the quantifier use are attested; there are no examples with these functions in the Middle and Early Modern English data. The development of the procedural uses of various thus appears to be the result of simultaneous rather than consecutive developments. Taken together with the new data presented here, the timing as well as the nature of the vague examples indicates that the quantifier use developed from the descriptive meaning, rather than from the individualiser ones, as was hypothesised to be the case for several. The second analysis focusing on form and contextual features (Table 7.2) sheds more light on the circumstances under which the individualiser and quantifier uses developed. The different features for which the data are analysed include various as complement in a copular clause (predicative), various as full noun phrase (pronominal) and various as modifier in a noun phrase (adnominal). Within the adnominal set, a distinction is made between noun phrases (NPs) with morphologically singular versus morphologically plural head nouns. The plural head noun examples are further categorised into indefinite versus definite NPs. The singular nouns were not further categorised, as all nouns are mass nouns in indefinite bare NPs. The plural 3
The data came from the Diachronic Part of the Helsinki Corpus of English Texts (HC) and CLMET3.0. They consist of random extractions for the word various, making up a total of 200 examples from HC, 100 for Late Middle English (1250–1500), the period in which various is first attested, and 100 for Early Modern English (1500–1710); and a total of 300 examples from CLMET3.0 representing Late Modern English, 100 for each 70-year subperiod in the corpus (1710–80, 1780–1850, 1850–1920).
118 Breban Table 7.2 Formal and contextual analysis of various DA predicative pronominal adnom. sg. N adnom. pl. N
I
Q
DA/I
DA/Q
1 indef. mass N indef. + mod. indef. + distrib. indef. + co-ord. indef. (other) def. + mod. def. + distrib. def. + co-ord. ‘all the’ ‘these and various’ def. (other)
I/Q
DA/I/Q
6
3
2
1
1 1
6 4
1
1
5
12
3 4 36
7
2 4
1
17
1 17
16
12
DA = descriptive adjective, I = individualiser, Q = quantifier; adnom. sg. N = adnominal use with singular head noun, adnom. pl. Ns = adnominal use with plural head noun; mod. = modified various, distrib. = distributive marker in the context, co-ord. = co-ordination with another modifier.
examples are also marked for the contextual presence of several additional features. The ones that proved relevant are listed in Table 7.2; they include whether various itself is modified (e.g. very various, more various, such various, as in (7) above), whether various is co-ordinated with another modifier (e.g. various and heavy and various and heterogeneous in period 1), and whether the contexts contain triggers for a distributive reading, as in (11). (11) I now proceeded to determine, as far as possible, the characters of the
various modes of vibration (PPCMBE period 3)
In this example, the occurrence of the NP in the postmodifying of-phrase to a plural head noun triggers the distributive reading that each individual mode of vibration is associated with a particular character (‘characteristic’). For the plural NPs, two additional patterns are distinguished, NPs with the complex determiner all the and the co-ordinated determiner phrase these and various. The definite and indefinite examples that do not contain any of these additional features are grouped together separately (e.g. definite (other)). The classification of these formal/contextual features is crossclassified with the semantic-functional classification in Table 7.2. The table generalises over all three subperiods. A first thing to be noted is that the adnominal singular use exclusively matches with the descriptive adjective semantics, as in (12).4
4
This finding was supported by a search of CLMET3.0 with the query various tagged as adjective followed by any singular noun, which yielded 330 examples, which all had a descriptive semantics.
Beyond syntactic change: various and numerous 119
(12) In the eleventh century it had two other names; Farria, and Fossetis-
land, which have been written with various orthography. (PPCMBE period 2)
Likewise co-ordination with another modifier proves to be indicative of descriptive semantics. This follows from one of the central claims of Bache (1978), postulating that co-ordination only takes place between adnominal modifiers with the same function, e.g. classifiers and classifiers, descriptive adjectives and descriptive adjectives. These patterns are exclusively associated with the descriptive meaning. Another illustration of the same phenomenon in the data set, the co-ordination these and various, in which various as a quantifier is co-ordinated with a determiner, is exclusive to the quantifier use. A further observation pertains to two other criteria often used to distinguish descriptive adjectives from other adnominal modifiers, gradability (by degrees of comparison or modification) and predicative alternation (see e.g. Bolinger 1967; Quirk et al. 1972). However, for various these criteria are not conclusive tests. A probable explanation is that these two tests, which are designed to distinguish different adjectival senses (see Breban and Davidse 2003), are not always able to capture the difference between descriptive adjectives and quantifiers, because certain quantifiers also allow predicative construal, e.g. her virtues are many, the problems are several, and adverbial modification, e.g. very many, very few. A second important finding concerns the distribution of the two procedural meanings over the different formal/contextual features. The quantifier use is mainly found in indefinite NPs; there are no definite examples in which various could be assigned a clear quantifier interpretation. The individualiser use is attested in both definite NPs (as in 9 and 11) and indefinite ones. In indefinite NPs, the presence of a distributivity-trigger often licenses the individualiser interpretation, e.g. the presence of another plural NP in (13). (13) I have given a number of results obtained at various dates (PPCMBE
period 3)
Indefinite examples which do not contain this feature (especially indef. (other)) display the greatest variation; all three interpretations and all possible combinations are attested. These examples can be characterised as ‘vague’ in the sense of Geeraerts (1997: 18). That is, various is not ambiguous between multiple readings in the sense that multiple readings are simultaneously and individually available for the reader/listener to pick one, and each one is supported by the context individually. Rather, the context is underspecified and (elements from) different interpretations for various fit it and no principled distinction between the interpretations can or has to be made by the reader/listener. Even though all meanings are attested in these underspecified examples, a closer look at
120 Breban
the figures reveals that the individualiser meaning is less often present in these examples, which might signify that it is typically more contextually embedded. This fits with its function, which is to make a non-default conceptual interpretation of the NP explicit, exactly in contexts that require this non-default interpretation. Breban (2008, 2010) identified several of these contexts: distributive contexts (11 and 13 above), list contexts (14), temporal/spatial sequence contexts (15–16), the NP contains a quantifier counting the instances. Except for this last context, these are precisely the typical contexts in which various is analysed as individualiser in the present data set. (14) staple the various branches of knowledge culture, including our own
language (PPCMBE period 3)
(15) In History we survey the various generations of mankind passing
(PPCMBE period 2)
(16) The other rheophore should be applied seriatim to the various ‘pain-
ful spots’ along the course of the nerve (PPCMBE period 3)
Interestingly, examples with distributive contexts are first attested only in the third period. The earlier vague/clear individualiser data mostly pattern as indef. (other) and def. (other). It thus seems that distributive contexts have not functioned as onset contexts (Traugott 2012c), triggering the semantic reanalysis of various. Rather, it appears that the presence of a definite determiner (typically the noun phrase resumes previously listed instances) constituted a prime locus of change. The more detailed study of the head nouns with which various is attested reveals a second locus of change in the form of nouns with an individualising value (see Table 7.3). The data for period 1 show the same range of head nouns for the descriptive adjective examples and the two sets of vague descriptive–procedural examples. It thus appears that the development of the new procedural meanings is not triggered by specific collocational contexts. This does not conform to the typical pattern of semantic change observed in the literature, which stipulates that a new meaning develops in a particular context, and is triggered by that context, and then spreads to more and more contexts different from the original context of change (e.g. Himmelmann 2004). Rather, in the case of various, the analyses for periods 2 and 3 indicate that typical collocational contexts develop later. This is particularly the case for the individualiser uses. Again, it is period 3 in which we see the emergence of typical patterns, such as nouns denoting a sequence in time/space, e.g. phases, painful spots along the course of the nerve (see 14–15), and binominal NPs with an individualising noun e.g. branches of, kinds of, parts of, modes of (see 9, 11, 13). For the latter, there is semantic accord between the individualising function of various and the individualising semantics of the noun. Both are explicit codings of individualisation.
Beyond syntactic change: various and numerous 121 Table 7.3 Head nouns in NPs with various Period 1 (1700–1769) DA
DA/I DA/Q Period 2 (1770–1839) DA
I Q DA/I DA/Q I/Q DA/I/Q Period 3 (1840–1914) DA I Q DA/I DA/Q I/Q DA/I/Q
Movements of mind; apprehensions; principles; degrees of acrimony and corruption; turns of the English language; afflictions; passions, appetites, affections (2); parts; goodness; diseases, readings; tumult; dispositions of mankind; turns which his mind took; subjects of conversation; deliverances; ways of acting; lights; symptoms of their disease; parts; shapes; blessings; excellencies; kindreds, friendships, and relations; affections; notes of mockery Tempers and dispositions; degrees and kinds of putrefaction; causes; dresses and turbans; tempers Unknown dangers and difficulties; parts Charms; exigencies; troubles and misfortunes; occupations; modes; descriptions (2); articles of provision; feats; gestures; coloured dresses; society; penalties; causes; pursuits; motives; lichens; customs; coloured silks; devices; conjectures; lengths; provisions; colours; success; orthography Parts and lines; parts (of N) (2); generations of mankind; revolutions of empire; authorities; European languages; nations of the Picts; anecdotes; combinations; rosaries Parts (of N) (3); et ceteras; respects; examples (2); inquiries; deltas; quarters; particulars Distances; tribes of animals; postscripts and notices; men-ofwar; objects; positions Costly snuff-boxes, and valuable trinkets; arts; senses; stratagems; streaks; subjects Inquiries; uses of improvement Kinds of corn, and oil; persons Shapes; structures; sorts; reasons; talk; kinds (of N) (3); styles; ways (2); glass bells; forms; crops (2); degrees; dodges; sizes and shapes Branches of N (2); forms; painful spots; places; exhibitions; kinds of N (2); phases; tones (2); parts (of N) (2); dates; bodies; modes of vibration Growers; classes of plants; purposes; legs; addresses; causes; parts (2) Conditions; matters; kinds of N; tones (2); partials; ways Purposes; hues; devices; dangers sentences; diseases; markets; horticulturists; combinations; gas companies Liniments and oils; things
The role of collocations in the development of the quantifier use is more difficult to gauge. Both semantic input (descriptive adjective) and output (quantifier) display great variety in the nouns with which they collocate. The individualising binominal phrase parts of is attested multiple times in periods 2 and 3. However, overall, the development appears to have been a change of focus in the semantics of various itself from qualitative variety to mere quantitative variety, irrespective of the semantics of the nouns modified.
122 Breban
A final observation pertains to the descriptive adjective use. The collocational analysis provides further evidence of its reduction. That is, by period 3, the range of nouns has severely shrunk and very often the noun itself means ‘varieties of ’ or denotes properties of entities such as size, shape and, in the earlier data sets, also colour. The general picture emerging from the three analyses is that the two procedural uses, individualiser and quantifier, develop simultaneously and at first independently. Both are driven by semantic ambiguity with the original descriptive sense. Ambiguity arises in distinct contexts for the two new uses: definite noun phrases and (later) noun phrases with an individualising head noun for the individualiser use and default indefinite noun phrases for the quantifier use. The fact that one and the same source meaning gives rise to different grammatical functions in different structural and collocational environments has been referred to as ‘structural persistence’ in grammaticalisation (Breban 2009). The outcome of a grammaticalisation process is determined not only by the source semantics (persistence in the sense of Hopper 1991), but also by the structural, and in this case also collocational, context which constitutes the locus of change. Only structural persistence can explain why the same source semantics develops two different new functions, as is the case for various. 7.3.2 Numerous The data for numerous in the PPCMBE amounts to a total of sixty-four examples. Again, the combination of semantic, formal/contextual and collocational micro-analyses presents an interesting picture. The semanticofunctional analysis (Table 7.4) distinguishes three sets of examples: those in which numerous functions as descriptive adjective meaning ‘vast, populous, consisting of many people, many parts’, as in (17–18), those in which numerous functions as a quantifier equivalent to many (19), and examples that can be read both as descriptive adjectives and as quantifiers (20). (17) a very numerous garrison, of their bravest Janizaries (PPCMBE
period 1)
(18) by the Trave on the south, Holsatia stretches its numerous woods to
Ditmarsia (PPCMBE period 2)
(19) Convinced that the present Treatise will fall into the hands of numer-
ous young men […] (PPCMBE period 2)
(20) A brisk breeze carried us swiftly on towards Cairo. Numerous mina-
rets, mosques, and other buildings rising into sight; […] (PPCMBE period 2)
In (20), the different meanings relate to two different conceptualisations that can best be elucidated with reference to the distinction between ‘summary’ and ‘sequential scanning’ (Langacker 1991: 21). In the descriptive reading,
Beyond syntactic change: various and numerous 123 Table 7.4 Semantico-functional analysis of numerous
descriptive adjective DA/Q quantifier
1700–1769
1770–1839
1840–1914
17 (85%)
20 (61%)
6 (55%)
1 (5%) 2 (10%) 20 (100%)
4 (12%) 9 (27%) 33 (100%)
5 (45%) 11 (100%)
the speaker/writer takes in the overwhelming urban cityscape consisting of a vast arrangement of buildings in a single view (summary scanning), the quantifier meaning, by contrast, involves a scanning motion counting individual towers and buildings in sequence and assessing their large number (sequential scanning). The conceptual shift underlies the subtle difference between description of vastness and assessing large numbers in plural contexts. Despite the small figures, we can observe an increase in the proportion of quantifier uses and a decrease in that of descriptive adjective uses over the three data sets. Even though there are very few instances for period 3, it cannot go unnoticed that there are no vague examples in the set, especially given that the second period contains a fair share (4 out of 33) of such examples. The data are analysed for two formal features: (a) predicative versus adnominal use and (b) morphological number/type of the noun modified by numerous. The latter feature is applied both to predicative uses, where the modified noun is the subject of the copular clause, and to adnominal ones, where it is the head noun. Examples (21) and (22) illustrate singular collective and plural predicative examples, (17–20) above different types of adnominal examples. (21) The congregation was numerous and respectable. (PPCMBE
period 2)5 (22) however numerous his separate Admirers may be (PPCMBE period 1) In a first step, I will focus on the contrast singular–plural noun. This parameter is cross-classified with the adnominal–predicative construal in Table 7.5, giving separate figures for each period. 5
It can be noted that the data discussed here represent British English, and hence show a preference for verbs agreeing with morphologically singular collective nouns to be singular, rather than plural. Seven of the adnominal examples for numerous in the data set contain verbs that are morphologically identifiable as singular or plural. Only one out of these seven has a plural verb. American English is known to prefer plural verb agreement in similar cases.
124 Breban Table 7.5 Formal features of numerous 1700–1769 singular plural
pred. 2 7
1770–1839 adnom. 6 5
pred. 1 5
1840–1914 adnom. 12 15
pred. 2 3
adnom. 6
The first aspect that draws attention is the absence of adnominal singular contexts (as in 13) in period 3. This context made up a considerable portion of the data in the previous periods (6 out of 11 (30%) and 12 out of 27 (36%), respectively). The figures further show a decrease of singular uses in general in period 3 (2 out of 11 examples (18%) versus 8 out of 20 (40%) and 13 out of 33 (39%) in periods 1 and 2). The CLMET3.0 data confirm the sharp decrease of the adnominal singular use,6 and provide supporting evidence for a decrease in the predicative singular use.7 The context that shows the main increase is adnominal plural, from 5 out of 20 (25%) in period 1, to 15 out of 33 (45%) in period 2, to 6 out of 11 (55%) in period 3. Again, this trend is confirmed in CLMET3.0.8 In Table 7.6, the formal features are combined with the semantic analysis, which adds further to the overall picture. As can be seen from Table 7.6, the quantifier uses are restricted to plural nouns, while the descriptive adjective use is attested for all four formal combinations.9 Four out of five vague examples are adnominal plural uses. This confirms the description in the OED, s.v. numerous, which associates the quantifier use with plural contexts. In terms of the overall diachronic development of numerous in LModE, the data suggest a shift in dominant usage. While in period 1 the descriptive adjective occurring in a range of formal contexts (singular, plural, predicative, adnominal) was the main usage pattern, in period 3 the adnominal 6
The query used to test this claim in CLMET3.0 was numerous tagged as adjective immediately followed by any singular noun; the results were manually sorted to retain only those examples in which the noun was the actual head noun of the noun phrase. This combination makes up 223 out of all 761 attestations for various in the 1710–80 period (29%), 112 out of a total of 825 attestations in the second 1780–1850 period (14%) and only 53 out of 1,056 in the final period 1850–1920 (5%), and hence shows a clear decrease in frequency. 7 The query any form of to be followed by numerous was used as a proxy for the predicative use and yielded 28 hits for period 1 with 3 singular uses (11%), 45 hits for period 2 with 7 singular ones (16%), and 59 for period 3 with a mere 2 singular ones (3%). 8 The query numerous followed by any plural noun resulted in 209 hits for period 1, 297 for period 2 and 401 for period 3. In terms of the overall number of attestations for numerous, the queried combination thus increased from 27% to 36% (period 2) and 38% (period 3). 9 None of the adnominal or predicative singular examples of numerous retrieved from CLMET3.0 with the queries numerous tagged as adjective followed by any singular noun (see footnote 7) and to be in any singular form followed by numerous (see footnote 8) had a quantifier interpretation.
Beyond syntactic change: various and numerous 125 Table 7.6 Formal and semantic analysis of numerous Period 1 (1700–1769) predicative adnominal Period 2 (1770–1839) predicative adnominal Period 3 (1840–1914) predicative adnominal
descriptive adjective singular plural 2 6 6 3
DA/Q singular
descriptive adjective singular plural 1 2 12 4
DA/Q singular
descriptive adjective singular plural 2 3 1
DA/Q singular
1
quantifier singular plural 1 1
plural 1 3
quantifier singular plural 2 8
plural
quantifier singular plural
plural
5
plural context gains in relative frequency and it is typically associated with the quantifier reading. In this development, the singular uses, especially in adnominal contexts, undergo a reduction. A final piece of the puzzle is added by a closer analysis of the modified nouns in terms of their collective versus singulative denotations. The morphologically singular nouns all have a collective denotation, and are often (10 out of 22 examples) accompanied by a postmodifying of-phrase containing a second plural NP, e.g. a numerous assemblage of sisters, a numerous fleet of merchantmen. For the category of morphologically plural nouns, a contrast between collective and singulative can be perceived as well. There are some nouns that designate collective sets, which are conceptualised as a homogeneous mass, e.g. bees, soldiers as in very numerous Forces (PPCMBE 1). These nouns are conceptually close to morphologically singular collective nouns, e.g. swarm or army. Other nouns individualise the referent set, e.g. instances, cases of (N), examples. Table 7.7 lists all the nouns found in the data; if the subject was a pronoun the referent it denotes is given in square brackets. Different fonts are used to indicate the different types of nouns: all morphologically singular (collective) nouns are in italics, collective plural nouns are underlined and plural individualising nouns are put in bold. The remaining plural nouns do not have any special marking. Table 7.8 provides a quantitative summary. The figures in Table 7.8 show host-class expansion (Himmelmann 2004). In period 1, 16 out of 20 nouns are collective (either singular or plural). In period 2, a new type of noun, the individualising noun, is attested, accounting for 6 out of 32 examples. The collective nouns at this point still make up half of the data set (16 out of 32). In period 3, the patterns crystallise: collective nouns are found in predicative contexts (23), whereas individualising and other plural nouns are found in adnominal contexts (24).
126 Breban Table 7.7 Nouns modified by numerous Period 1 (1700–1769) pred.
bees, admirers, they [type of animals], who [the people], family, commentators, who [this race of creatures], they [bees], family garrison, forces, progeny, forces, retinue, guard, troops, train (figurative), friends, enemies, train (figurative)
adnom. Period 2 (1770–1839) pred.
congregation, cases of distress, zero (the more numerous), cases, black flies, they [clouds] collection, trials, kinds of liquid rubefacients, meeting, instances, fleet of merchantmen, water serpents, families, throng of supplicants, assortment, artillery, population, idols, young men, examples, assembly, wants, rooms, assemblage of sisters, assembly, woods, diseases, points of identification; minarets, mosques, and other buildings; intervals of plains, concourse of people
adnom.
Period 3 (1840–1914) pred.
cavalry, they [people], infantry, domestic slaves, they [people] instances, compounds, floral emblems, executions, revolutions, dark spots
adnom.
Table 7.8 Types of nouns modified by numerous Period 1 (1700–1769) pred. adnom. Period 2 (1770–1839) pred. adnom. Period 3 (1840–1914) pred. adnom.
singular 2 6 singular 1 11 singular 2
plural collective 5 3
plural
plural individ. N
plural collective 2 2
plural
plural collective 3
plural
plural individ. N
4
2
2 2
1 9
9 11 plural individ. N 2 4
6 26
5 6
(23) The cavalry was very numerous (PPCMBE period 3) (24) retained in the numerous compounds (PPCMBE period 3)
Taken together, the three analyses tell a coherent story: numerous originally functioned as a descriptive adjective, used mainly in combination with nouns with a collective interpretation in a range of formal contexts, and in the course of LModE it developed a prolific quantifier use in adnominal
Beyond syntactic change: various and numerous 127
contexts with plural (non-collective) nouns as heads. Interestingly, the semantics of numerous remains very similar in this process, developing from a descriptive appreciation of the size of an entity in terms of its subparts, to assessing subparts and later entities as very many. What changes is the conceptualisation process behind the quantitative assessment, from an overall appreciation of vastness to sequential scanning and counting of individual instances. The different interpretations are very much linked to different nouns and different formal construal types. A crucial element in the shift from adjective to quantifier for numerous seems to be host-class expansion within the plural noun set from only collective to collective and individualising nouns. This expansion made numerous both distributionally and semantically very similar to existing quantifiers in English. Numerous could then have been ‘adopted’ to the rapidly expanding quantifier paradigm in LModE. 7.4 Discussion: actualisation and paradigmaticisation as evidenced by various and numerous in Late Modern English 7.4.1 Summary of the case studies Taking together the findings of the development of numerous and various from this chapter and those about several (Breban 2008, 2010), we arrive at a much more complex picture of the changes affecting descriptive adjectives developing into unspecific quantifiers. Breban (2008, 2010) proposed a single path of change evolving in two steps, from descriptive adjective to individualiser, and from individualiser to quantifier. Each step was accompanied by a reduction of usage contexts: step 1 by a reduction to adnominal plural contexts (the loss of predicative and adnominal singular contexts), step 2 by a further reduction to indefinite plural contexts (loss of definite ones). The individualiser use was particularly prominent in a restricted set of contexts: distributive contexts, NPs with a quantifier (e.g. two several Churches (HC 1500–1710 cited in Breban 2008: 263)), contexts where the individual members of the referent set are enumerated or make up a temporal/spatial sequence. The mechanism of change was argued to be the lexicalisation of invited inferences (Traugott and König 1991) via ambiguous contexts (Diewald 2002; Heine 2002) (see Breban 2010: 102–09, 156–58, 188–91). Semantic change is seen as driving the developments. The adjective numerous was shown to develop a quantifier use in LModE, without ever functioning as individualiser. From the analysis presented in this chapter, the development of numerous does not appear to be driven by semantic ambiguity, but to be set in motion by host-class expansion (Himmelmann 2004) from morphologically singular collective nouns via plural nouns with a collective interpretation to other types of plural nouns. The second expansion goes together with the reanalysis and adoption of
128 Breban
numerous into a new word class, that of unspecific quantifier. It was proposed that this kind of development was possible because the semantic shift from descriptive to quantifier semantics for numerous was minimal. In terms of the formal distribution, the data show an increased relative frequency of adnominal plural contexts. The changes affecting various are in some ways similar to those affecting several: they include the development of both individualiser and quantifier uses and the mechanism of change is semantic reanalysis. However, there are certain differences. Most importantly, the development of individualiser and quantifier uses is simultaneous rather than consecutive. The reason why this is possible lies in the descriptive adjective meaning of various. For several, this meaning was ‘separate, distinctive’, which allowed semantic reanalysis towards the individualiser meaning, but did not constitute a suitable source for reanalysis towards a quantifier meaning. The descriptive meaning of various ‘varied, displaying variation’, by contrast, does. Secondly, even though the changes are semantically driven, they occur in semantically and structurally underspecified contexts rather than in ambiguous contexts that invite inferencing of a new meaning by the presence of contextual triggers. The data evidence a numerical increase of individualiser uses and the gradual acquisition of contextual and collocational patterns associated with it as time advances. The quantifier use of various does not show a numerical increase and its development seems to stagnate. 7.4.2 Actualisation and paradigmaticisation as two sides of the same coin Actualisation is used in the literature on syntactic change as a cover term for changes to the surface structure which happen after reanalysis of the underlying structure and which bring the surface structure in line with the new underlying structure. It is the ‘gradual mapping out of the consequences of the reanalysis’ (Timberlake 1977: 141; see also, among others, Andersen 2001; De Smet 2012a). While actualisation has been reserved for syntactic change in the literature so far, the data in this chapter evidence the same gradual process with respect to different types of observable distributional behaviours of items after an initial change. The case study of various dates the first clear individualiser uses to subperiod 2, with vague examples occurring in period 1. In period 3, the individualiser uses of various are attested to co-occur with nouns typically collocating with individualisers and to occur in contexts that contain an explicit distributivity licenser, which have been found to be a typical context for individualisers. Thus in period 3, various has assumed these new distributional behaviours typical of its new function. I therefore propose that actualisation is not restricted to syntactic change but should be extended to other types of distributional behaviour, such as collocational patterning and contextual distributions, as well. A plausible reason why a similar extension has not been discussed before is that,
Beyond syntactic change: various and numerous 129
in comparison to syntactic change, which can be observed very straightforwardly, changes in collocational and contextual distributions require a more detailed analysis of contextualised examples (a micro-analytical approach). This might explain why, especially if there is evidence of syntactic change, researchers did not look further and other distributional changes stayed under the radar. Cases of change such as those of various and numerous are characterised by a minimal difference in syntactic behaviour between old and new uses. If anything, the original descriptive adjective use displays more syntactic options: both an adnominal and a predicative realisation. As adnominal modifiers in the NP, the only syntactic difference is one of (relative) positions. As argued in Breban (2008, 2010: 346–349), individualisers occur in postdeterminer position, that is, to the left of descriptive adjectives when the two types co-occur. This means that the adjective in its original descriptive use could follow another descriptive adjective in the linear left-to-right structure of the NP, but in its new use as individualiser it precedes any descriptive adjectives. Evidence for a positional move accompanying the new use is always negative, that is, the absence of examples in which, for example, several as an individualiser is preceded by descriptive adjectives. Syntactic change in the development from descriptive to individualiser is hence not only minimal but also difficult to substantiate. The quantifier use has been associated with distinctive new syntactic behaviour, partitive construal such as in many of the apples, several of the problems (Denison 2006: 292–96, 2010: 114). The PPCMBE data used for this study contain no evidence for the acquisition of this new syntactic behaviour by the quantifier uses of various or numerous.10 In the absence of clear syntactic actualisation, the less obvious processes of collocational and contextual actualisation draw our attention. The proposal that the distributional features affected by language change involve more than syntax has an important consequence for the recognition of individual case studies as instances of grammaticalisation. It is generally assumed that grammaticalisation involves a combination of semantic and observable, formal changes (e.g. Hopper and Traugott 2003). Mere semantic changes do not qualify as grammaticalisation. If one adheres to a restrictive interpretation of formal change as syntactic change, certain examples of change are excluded as cases of grammaticalisation. Applied to the case studies at hand, for example, even though the individualiser use has a procedural meaning as compared to the original lexical, descriptive meaning, it cannot straightforwardly be shown that the change from descriptive adjective to individualiser is accompanied by syntactic change, and its status 10
In the larger CLMET3.0, the syntactic pattern is attested for various. The query various of returned 6 relevant examples (out of a total of 4,730 attestations of various): 3 examples in period 2 and another 3 in period 3. For numerous, CLMET3.0 contained 5 occurrences of the pattern numerous of, but in all of them numerous was submodified by most (4 examples) or by an adverbial (1 example) and hence had a descriptive semantics.
130 Breban
as a case of grammaticalisation is questionable. If one accepts that formal change includes a wider range of distributional behaviour, as argued here, our view on which changes constitute grammaticalisation and which do not also changes. De Smet (2012a) is a data-based exploration of the characteristics of actualisation, and especially of the perceived non-random order in which actualisation proceeds from one syntactic change to a second to a third, etc. De Smet proposes that the course of (syntactic) actualisation is determined by local and global analogies to existing syntactic patterns. Innovations are licensed by already existing patterns. As a consequence, the syntactic structure of the locus of reanalysis plays a crucial role in the process. A second consequence is that the course of actualisation is item-specific depending on the item’s individual syntactic patterns. The spread to new collocational and contextual distributions observed in this chapter shares several of these characteristics. Numerous was shown to gradually spread from occurring with singular collective nouns, to plural collective nouns, to other plural nouns. Following De Smet, it can be argued that the first spread is licensed by the shared feature ‘collective’, the second one by the shared feature ‘plural’. The course of collocational actualisation is hence also analogically motivated. In other aspects, the findings of the case studies complement and add to De Smet’s proposals. De Smet is concerned with analogy from the perspective of the item undergoing change, because it is these analogies that determine the order of actualisation. That is, the item’s older or already acquired patterns determine which ones will emerge next. This emphasises language’s conservative tendency. The case studies presented here put analogical attraction by the target category in the foreground. As an individualiser, various assumes the collocational and contextual distributions typical of other individualisers such as several. This perspective emphasises the changing item’s tendency towards assimilation and conformation. Individualiser various takes up a range of distributions with which its new use is associated, becoming a more typical, more assimilated individualiser. Traugott and Trousdale (2013) argue that this kind of behaviour is a typical pattern in language change: new members develop at the margins of a category, i.e. as non-prototypical members, and a portion of them gradually become more central members. The idea of assimilation with existing members of a new category has been around in grammaticalisation studies since the 1980s. Lehmann (1982 [1995], 1985) discusses it as one of his six parameters of grammaticalisation, viz. paradigmaticisation (Lehmann 1985: 4). One aspect of this process is the fact that paradigms become more homogeneous in terms of their morphosyntactic behaviour. The case study of various shows paradigmaticisation at work as an important mechanism in the establishment of new uses, but, again, with respect to collocational and contextual patterns rather
Beyond syntactic change: various and numerous 131
than morphosyntactic ones. Similar to actualisation, I propose that the concept of paradigmaticisation in fact includes a wider range of distributional behaviours.11 Secondly, the case study for various shows that reanalysis and actualisation are not only item-specific (De Smet 2012a), but context-specific as well. The same item, various, developed different new uses characterised by different subsequent behaviours in different contexts. The individualiser use first developed in definite noun phrases and later in noun phrases with an individualising head noun, whereas the quantifier use developed simultaneously in underspecified indefinite noun phrases. Language change is subject to structural persistence (Breban 2009), the path of change is co-determined by the specific (structural) environment in which the item changes. Finally, it is interesting to look at the predicative examples of numerous. Predicative construal is, of course, one of the main syntactic patterns of descriptive adjectives. In addition, it is also one of the possible syntactic patterns for indefinite quantifiers, e.g. The benefits to nurses of a clinic system are many (BNC), The attractions to savers are several (BNC) (see also Huddleston and Pullum et al. 2002: 356). In the data for numerous presented here, periods 1 and 2 contained a few examples of predicative quantifier uses, but all five predicative examples in the period 3 data are clearly descriptive adjectives, that is, four of them are graded or modified (2 × very numerous, 1 × more numerous, 1 × numerous enough), and the fifth example has numerous coordinated with another descriptive adjective, numerous and willing. This suggests that the quantifier use did not break through in its different potential syntactic environments at the same time. The adnominal context appears to be the prime locus of its development.12 This then raises the question why the quantifier use numerous did not develop in the two environments simultaneously or why the adnominal context was chosen over the predicative one as locus of reanalysis. A possible explanation could lie in the strength of the attractor analogue. That is, predicative construal is only an infrequent pattern for the quantifier use. There are only thirty-six relevant hits for the search _NN2 (‘plural noun’) are many in the BNC and only two relevant hits for _NN2 (‘plural noun’) are several.13 If this were the reason for the development of numerous, it would show that the frequency of the attracting behaviour (possibly in comparison with the frequency of the behaviour for the older usage) is a 11
Diewald (2011) proposes that another aspect Lehmann’s concept of paradigmaticisation, the fact that items become part of small, tightly regulated paradigms, makes up the essence of grammatical meaning: an item has a grammatical function if it is defined by its relations to other members of a paradigm rather than its (deictic) relations to the outside world or to other elements in the text. 12 In the light of De Smet’s proposals, one might hypothesise that predicative construal of the new quantifier use of numerous is likely to emerge later, in the course of actualisation, because it is licensed by original behaviour. 13 In 15 out of the 36 hits, many is used in the fixed phrase are many and varied.
132 Breban
factor in reanalysis and actualisation as well. The role of (relative) frequencies of behaviours in change has been emphasised in the works of Bybee and Hopper (2001), Hoffmann (2004), Bybee (2007), Gries and Hilpert (2010) among others, but is by no means fully understood. De Smet takes target attraction to be a given, fixed factor in the analogy behind reanalysis and actualisation, i.e. an ‘abstract syntactic generalization such as a reanalysed item’s new word class’ (De Smet 2012a: 604) and plays down its distinctive importance. If attractor strength plays a role in the order of reanalysis and actualisation, this view has to be nuanced in favour of a more subtle interplay between analogies with existing behaviours and the target category’s typical behaviours. 7.5 Conclusion The case studies presented in this chapter have shown that the traditional interpretations of actualisation and paradigmaticisation in terms of syntactic or morphosyntactic change only are too narrow. Both processes can be evidenced in a wider range of observable changes in the distributional behaviour of language items undergoing change. The case studies for various and numerous illustrated actualisation and paradigmaticisation with respect to collocational and contextual distributions. The data presented confirm the crucial role of analogy in actualisation (De Smet 2012a), yet, they nuance De Smet’s exclusive focus on analogy with the item’s existing behaviours as sole determining force in the actualisation process. Analogies with the behaviours of members of the target category (i.e. paradigmaticisation) were argued to play determining roles in both the actual reanalysis (for numerous) and the actualisation (various and numerous) as well. Actualisation is a balancing act between the contrasting pulls towards conservativism (De Smet 2012a) and conformation/assimilation. The case studies also have implications regarding certain tenets of grammaticalisation studies. Firstly, the development of numerous shows that grammaticalisation is not always driven by semantic change but can be set in motion by collocation shifts. Secondly, even if semantic change is the driving force, it is not necessarily a semanticisation of invited inferences in ambiguous contexts (see Traugott and König 1991) but may be the acquisition of a new meaning (as individualiser or quantifier for various) in underspecified contexts. However, it has to be noted that in the cases of various and numerous, the old and new meanings are very close, which might explain their atypical developments.
8
Culturally conditioned language change? A multivariate analysis of genitive constructions in ARCHER Benedikt Szmrec s anyi, Anett e Rose nbach, J oan Bresnan and Christoph Wol k
8.1 Introduction This chapter is concerned with the development of two competing NP constructions in Late Modern English (LModE), the s-genitive, as in (1), and the of-genitive, as in (2). (1) (2)
before [the Seneschal]s [Brother] could arrive, he was secured by the Governor of Newport the Duke of Norfolk, having lately received another Challenge from [the Brother] of [the Seneschal], went to the place appointed
Historically speaking, the of-genitive is, of course, the incoming form, which appeared during the ninth century. According to Thomas (1931: 284), the inflected genitive vastly outnumbered the periphrasis with of up until the twelfth century. In the Middle English period, we begin to witness ‘a strong tendency to replace the inflectional genitive by periphrastic constructions, above all by periphrasis with the preposition of ’ (Mustanoja 1960, I:70). The Early Modern English period, however, sees a revival of the s-genitive, ‘against all odds’ (Rosenbach 2002: 184). While we know that the s-genitive is comparatively – and increasingly – popular in Present-Day English (PDE), especially American English (Rosenbach 2002, 2003), the literature about genitive variability in the LModE period is somewhat sketchy (but see Szmrecsanyi 2013; Wolk et al. 2013). In addressing these gaps in our knowledge about genitive variability in LModE, we follow recent probabilistic approaches to language (see, for example, the papers in Bod et al. 2003; Bresnan and Ford 2010) and assume that grammatical variation and change is sensitive to probabilistic (rather We wish to thank the Freiburg Institute for Advanced Studies (FRIAS) for funding the project ‘Predicting syntax in space and time’. This material is also based upon work supported by the National Science Foundation under Grant No. BCS-1025602. We also acknowledge helpful comments and suggestions by an anonymous referee. The second-named author would like to thank the University of Paderborn for providing her access to their online library resources.
133
134 Szmrecsanyi, Rosenbach, Bresnan & Wolk
than categorical) constraints, because conditioning factors may influence linguistic choice-making in extremely subtle, stochastic ways (Bresnan and Hay 2008: 246). In this spirit, we are interested in subtle changes in some conditioning factors. Specifically, we provide evidence for a diachronic weakening of the animacy constraint, for which we offer various explanations, paying particular attention to a rather novel source of change in the distributional pattern of variants: changes in the wider cultural context, such as industrialisation as a transformational period (we would thus like to emphasise right at the outset that the present study is not about e.g. cultural transmission and iterative learning). On the methodological plane, the study contributes to a growing body of literature on the probabilistic nature of the genitive alternation (see e.g. Gries 2002; Szmrecsanyi 2006; Hinrichs and Szmrecsanyi 2007; Hundt and Szmrecsanyi 2012; Shih et al. Forthcoming; Grafmiller Forthcoming). Specifically, we build on research reported in Wolk et al. (2013), a paper that investigates grammatical variation in A Representative Corpus of Historical English Registers (ARCHER), Version 3.1. (Yáñez-Bouza 2011). ARCHER covers the period between 1650 and 1999, spans about 1.8 million words of running text, samples eight different registers, and contains British as well as American English texts. The corpus design categorises all texts into seven subperiods of 50 years, but the precise year of composition for each text is typically also available. Coverage of American English is restricted to three of the seven periods: 1750–99, 1850–99, and 1950–99. Among other things, Wolk et al. (2013) investigate genitive variability in ARCHER’s British English news and letter section, drawing on a data set featuring a variety of explanatory variables, such as length of the possessor/ possessum (the principle of ‘end weight’), animacy of the possessor, definiteness of the possessor, and so on. Building on Wolk et al.’s work, we present here an extended and richly annotated genitive data set that covers over 5,000 genitives not only in British news and letters texts but also in American texts. On the basis of this extended data set – which, for culture-relevant dimensions, covers not only time (1650–1999) but also space (Britain versus America) – we calculate a logistic regression model with mixed effects which correctly predicts over 90 per cent of writers’ genitive choices by jointly considering a wide range of explanatory variables, as well as idiolectal and lemma-specific random effects. Our goal is to explicitly address the interplay between culture change and grammar change in the LModE period. We will see that, as a matter of fact, genitive variability is patterned in surprisingly similar ways in British and American English. Where culture comes in is the weakening of the possessor animacy constraint in the course of the LModE period. This chapter is structured as follows. In Section 8.2, we explain how we identified interchangeable genitives in the corpus, and canvas genitive frequencies. In Section 8.3, we discuss the language-internal and
Genitive choice: culturally conditioned change? 135
language-external conditioning factors that we used to annotate our data set. Section 8.4 presents the regression model. Section 8.5 discusses the model and compares various linguistic, conceptual and cultural explanations for the observed changes in the effect of the animacy constraint. Section 8.6 offers some concluding remarks. 8.2 Identifying interchangeable genitives In the present study, we explore variation between the s-genitive and the of-genitive as two roughly equivalent ways of saying the same thing (Labov 1972). Our definition of variable (i.e. interchangeable) genitives precisely follows the guidelines in the Wolk et al. (2013) study, which itself follows best practice in the literature on genitive variation. Identifying interchangeable genitives is a labour-intensive task. The procedure can be summarised as follows. We used *’s, *s’, of, and *s (the latter only in the first two periods, when spelling without an apostrophe was still common) as search strings, and screened the resulting hits manually to weed out irrelevant material such as plural nouns. To identify interchangeable genitives, attention was restricted to genitive constructions with non-pronominal possessors or possessums. We excluded (i) demonstrative possessums (e.g. her face is as ugly as that of a dog), (ii) constructions that are clearly fixed expressions (e.g. the Duke of Normandy), and (iii) partitive genitive relation contexts (e.g. 3 chests of wine). Further, we only included of-genitive constructions headed by the definite article (as in the use of the navy), and excluded of-constructions with modifying function (as in all persons of quality), and of-genitives expressing a clearly appositive relation (as in the number of 13000 men). As for the s-genitive, we focused on determiner (specifying) s-genitive constructions. Overall, the aim was to include genitive constructions that are interchangeable in principle rather than relying on a coder’s intuition. So rather than asking the question ‘does this genitive alternate?’ in every individual case, we relied on the application of clear and replicable inclusion/exclusion criteria as sketched above. Our coding guidelines yielded a data set consisting of N = 5,576 interchangeable genitives. Table 8.1 breaks down the numbers by genitive type, ARCHER period, and variety, Figure 8.1 plots s-genitive proportions in real time. It is amply clear that the genitive alternation exhibits robustly fluctuating variant proportions in British English (see Szmrecsanyi 2013; Wolk et al. 2013 for in-depth discussions). The frequency trajectory is such that the s-genitive was comparatively frequent in the early and late ARCHER periods, but relatively unpopular in the middle ARCHER periods. Overall, s-genitive frequencies and proportions in American ARCHER texts match those in British ARCHER texts rather well. With regard to s-genitive proportions, there is virtually no difference in the 1850–99 and 1950–99 period. Only in the 1750–99 period are American s-genitive proportions
136 Szmrecsanyi, Rosenbach, Bresnan & Wolk Table 8.1 Interchangeable genitive frequencies (raw hits in ARCHER, and variant frequencies) British English
1650–99 1700–49 1750–99 1800–49 1850–99 1900–49 1950–99
American English
of-genitive
s-genitive
Total
312 69% 364 71% 418 79% 558 89% 446 80% 435 76% 357 62%
139 31% 152 29% 109 21% 70 11% 109 20% 134 24% 221 38%
451 100% 516 100% 527 100% 628 100% 555 100% 569 100% 578 100%
of-genitive
s-genitive
Total
523 87%
77 13%
600 100%
483 79%
132 21%
615 100%
327 61%
210 39%
537 100%
45% 40% 35% 30% 25%
proportion s-genitive BrE
20%
proportion s-genitive AmE
15% 10% 5%
16 50 –9 9 17 00 –4 9 17 50 –9 9 18 00 –4 9 18 50 –9 9 19 00 –4 9 19 50 –9 9
0%
Period
Figure 8.1 Proportion of s-genitive (y-axis) against real time (x-axis).
significantly lower than their British counterparts (χ2 = 8.96, df = 1, p = 0.003). So British–American differences do not appear to be the locus of culturally conditioned genitive variation. As an interim summary, we conclude that genitive proportions in British and American LModE are quite similar. The task before us is to determine whether the same is true for how genitive choice is conditioned in the two varieties.
Genitive choice: culturally conditioned change? 137
8.3 Determinants of genitive choice Genitive variation is subject to a large number of language-internal constraints (e.g. possessor animacy, constituent weight), and to some languageexternal factors, such as real time or variety. To probe the probabilistic grammar that conditions genitive choice, we need a data set where each and every genitive observation is annotated for as many variables as possible. The annotation we use in the present study is exactly parallel to the genitive annotation discussed in detail in Wolk et al. (2013). In what follows, we briefly summarise the annotation guidelines. 8.3.1 Animacy Previous studies have reported reliable and strong effects of possessor animacy as a locus of diachronic variability (e.g. Rosenbach 2002; Jankowski 2009, 2013; Hundt and Szmrecsanyi 2012): more animate possessors favour the s-genitive, less animate possessors favour the of-genitive. Our operationalisation of animacy is based on a simplified version of the guidelines in Zaenen et al. (2004). Following Rosenbach (2008), five possessor animacy categories are distinguished: 1. Animate possessors comprise humans, higher animals and sentient human-like beings such as gods, e.g. king, horse, god, or John. 2. Collective possessors are organisations such as administration or church, as well as temporally stable groups of humans with potentially variable concord, such as delegation, family, or enemy. 3. Temporal nouns consist of both points in time and durations, for example February or moment. 4. Locatives are locations, including geographical states, e.g. Russia, this kingdom, the seas. 5. All other concrete (e.g. rock) or non-concrete (e.g. idea) noun phrases were classified as inanimate. The hypothesis is that animate possessors are particularly often coded with the s-genitive, while inanimate possessors tend to occur with the of-genitive. The other categories should cover the middle ground. 8.3.2 Definiteness and nominal expression Definiteness overlaps to some extent with discourse accessibility. Thus we assume that definite possessors tend to be discourse-old, which is why they should favour the s-genitive because the s-genitive places the possessor before the possessum, thus establishing old-before-new order (Quirk et al. 1985: 1282; Biber et al. 1999: 305). In this spirit, we straightforwardly distinguished between proper name possessors (e.g. Mary), definite possessors (e.g. the king), and indefinite possessors (e.g. a man). Note that we classified
138 Szmrecsanyi, Rosenbach, Bresnan & Wolk
as proper name only noun phrases that could be considered proper names in PDE and that were capitalised in the text. As definite we coded all noun phrases headed by a definite determiner, that, or an s-genitive. All other constituents were classified as indefinite. 8.3.3 Constituent length In languages like English, ‘heavier’ (i.e. longer and/or more complex) constituents tend to follow ‘lighter’ ones (see e.g. Behaghel 1909). Therefore, long possessums should favour the s-genitive (which places the possessum after the possessor), while long possessors should favour the of-genitive (which places the possessor after the possessum). To determine the constituent lengths of individual genitive occurrences, we manually identified the constituents of each genitive construction. Genitive possessums, which may only start with a determiner in the of-genitive, had their first word removed from the count if that word was a determiner. Subsequently, constituent length in orthographic characters (including blanks) was established. Wolk et al. (2013) offer a detailed discussion of the accuracy of character counts vis-à-vis other measures; the upshot is that character counts, word counts, and syllable counts are very highly correlated in the type of data sampled in ARCHER. Mean possessor length in the data set is about eighteen characters (as in the command of his royal highness); mean possessum length is about eleven characters (as in my wife’s concernment). In regression analysis, a logarithmic transformation was applied to the character counts to reduce skewness, and then the values were centred around 50-year period means to reduce multicollinearity and to account for possible changes in average lengths. 8.3.4 Semantic relation Genitives may encode a wide range of different relations. We followed the binary distinction between prototypical and non-prototypical possessive relations in Rosenbach (2002). Prototypical relations comprise legal ownership (3a), body-part relations (3b), kinship relations (3c), and part–whole relations (3d), while all remaining cases were coded as ‘non-prototypical’ (see the examples in (4)). (3)
Semantic relations considered prototypical a. ownership: RHODESIAN forces have increased security measures in and around [Mr Ian Smith’s cattle ranch and farm at Selukwe]+prototypical after a sharp upsurge of guerrilla activity in the Midlands region of the country. < ARCHER 1979stm1.n8b> b. body parts: The Irish came in to the house pul’d the man out of bed from his wife and murdered him; then tooke all the rest of the houshold, led them to the seaside, and threw them off the rocks; one of the Children hung about one of [the
Genitive choice: culturally conditioned change? 139
(4)
murderers legs]+prototypical, yet was pull’d off and thrown after the rest. c. kinship: It’s said [the Duke of Berwick’s Son]+prototypical is in one of the Ships, and Perth’s two Sons in the other. < ARCHER 1715eve1.n3b> d. part-whole: [The Hull of a Ship]+prototypical was seen floating between Blackness and Point and Calais and Ambeleteuse; < ARCHER 1735rea1.n3b> Semantic relations considered non-prototypical a. Christian sources in Egypt say that President Sadat has gone back on a pledge he gave some years ago not to allow Islamic law to become [the law of the country]-prototypical < ARCHER 1979stm1.n8b> b. THE new drama, ‘John Garth,’ produced at Wallack’s Theatre, New York, is spoken of by the press as the best work ever written for the American stage. [Mr. Wallack’s acting as the hero]-prototypical is greatly admired. < ARCHER 1872gla1n6b> c. However, this rule is sometimes dispensed with; and particularly since the signing of the Preliminaries of Peace, our Government has permitted [the granting of such passports]-prototypical, provisionally, for the space of a year, to ships built out of the Republic, provided that they entirely belong to natives of this country, and also fitted out here. < ARCHER 1802joh2.n5b> d. The Supreme Educational Council had given instructions to the school-masters which had established religious neutrality, and a request by the Council-General of the Seine that [the name of God]-prototypical should never be uttered in school had been rejected. < ARCHER 1883tim2.n6b>
According to the literature, prototypical relations should favour the s-genitive, while non-prototypical relations should favour the of-genitive. 8.3.5 Final sibilance Possessors ending in a sibilant, as in Alice’s child, discourage the s-genitive (e.g. Zwicky 1987). We used an automatic annotation process, relying on the Carnegie Mellon University Pronouncing Dictionary version 0.7a,1 to identify all possessor phrases ending in a sibilant (which includes possessors ending with a plural s). Tokens not included in the dictionary were coded manually. 8.3.6 Variety and real time ARCHER provides the year of creation2 of each corpus file. To ease the assessment of diachronic changes and make statistical analyses more reliable, 1
Available online at www.speech.cs.cmu.edu/cgi-bin/cmudict. A small number of texts in ARCHER are not dated exactly; these were placed in the middle of a time segment, i.e. in year 5 of a given decade or year 25 of a fifty-year period, as the case may be.
2
140 Szmrecsanyi, Rosenbach, Bresnan & Wolk
the individual dates were centred around 1800 and converted to centuries, so that a text from 1651 would count as (1651 – 1800) / 100 = –1.49, and a text from 1931 as (1931 – 1800) / 100 = 1.31. Additionally, of course, we consider whether a given text derives from the British or American section of ARCHER.3 8.3.7 Lexical effects We used the lemma of the possessor head noun (e.g. president in the president’s speech) to test for by-item effects (modelled as random effects in logistic regression). 8.4 Regression analysis To model the joint impact of the factors discussed in the previous section on genitive choice in ARCHER, we draw on logistic regression analysis. Logistic regression is a multivariate statistical analysis technique that quantifies the effect that individual explanatory factors have on a binary dependent variable, such as genitive outcomes. We utilise a modern refinement of logistic regression analysis known as mixed-effects logistic regression (Pinheiro and Bates 2000).4 In addition to so-called fixed effects – which are classically estimated predictors suited for assessing the reliability of the effect of repeatable characteristics, such as, e.g. possessor animacy – mixed-effects modelling addresses random effects to capture variation dependent on openended, potentially hierarchical and unbalanced groups. In this study, we model corpus file ID (a proxy for idiolectal differences) and possessor head noun lemmas (see Section 8.3.7) as random effects. We followed the customary model-fitting procedures. We first generated a model containing all predictors and all putatively relevant interactions. This model was then stepwise reduced by removing non-significant predictors and interactions (i.e. dependencies between predictors). Pruned models were assessed by means of the Akaike Information Criterion; random effects were evaluated by means of likelihood ratio tests. Finally, the model underwent bootstrap validation to assess the possibility of overfitting.5 The resulting model (the so-called ‘minimal adequate model’) has an 3
In principle, text-type differences are also important (Grafmiller Forthcoming), but we note that in the data set at hand the difference between news and letters texts does not happen to make a significant difference. 4 We utilised R version 2.12 (R Development Core Team 2011) and lme4 version 0.999375–33. 5 More precisely, the individual observations were repeatedly randomly re-sampled with replacement, and the model was fitted to this new data set. To ensure that each ARCHER fifty-year period has a sufficient number of observations in each run, the total number of observations per period was kept constant. All results reported as significant below are also stable under bootstrap validation; unstable predictors were removed from the model.
Genitive choice: culturally conditioned change? 141
excellent classification accuracy (Somers’ Dxy = 0.92), and correctly predicts 90.7% of all genitive tokens, a considerable increase over baseline (75.7%). Multicollinearity is not an issue (κ = 8.2). Table 8.2 reports regression coefficients (positive values indicate a favouring effect, negative values a disfavouring effect), along with rather more interpretable odds ratios (odds ratios > 1 indicate a favouring effect, odds ratios < 1 indicate a disfavouring effect), and significance levels. Predicted odds are for the s-genitive. We begin by discussing the main effects of the genitive predictors considered in the present study (headings ‘Animacy of possessor’ through ‘Variety and real time’ in Table 8.2). Briefly put, all of the main effects behave as they should, given the literature: • Vis-à-vis animate possessors (e.g. Tom), all non-animate possessor categories disfavour the s-genitive. For example, the odds ratio of 0.02 associated with outright inanimate possessors (e.g. snow) indicates that if the possessor is inanimate instead of animate, the odds for an s-genitive are reduced by a robust 1 – 0.02 = 98%. • In comparison to definite possessors (e.g. the man), proper name possessors (e.g. Tom) favour the s-genitive (under this condition, the odds for an s-genitive increase by a factor of 3.73). Indefinite possessors (e.g. a man) disfavour the s-genitive. • Long possessums favour the s-genitive, long possessors disfavour the s-genitive, fully in accordance with the principle of end weight. The complication is that both possessum and possessor length have a linear and a quadratic (squared) effect. This duality is discussed in some detail in Wolk et al. (2013); suffice it to say here that medium-length possessums and possessors are not as well behaved as a strictly linear operationalisation of length effects would predict. We will return to this issue below. • Against the backdrop of non-prototypical semantic relations, prototypical semantic relation (e.g. kinship relations, as in the king’s sister) increase the odds for the s-genitive by a factor of 2.14. • A final sibilant in the possessor (e.g. President Bush’s speech) reduces the odds for the s-genitive by 58%. • Finally, neither real time nor variety have significant main effects, but the two language-external variables partake in significant interaction with language-internal predictors, as we shall see shortly. We now turn to the meat of our analysis: interaction terms between language-internal and language-external variables in the model. How does the effect of language-internal conditioning factors evolve in real time? Do British English and American English differ with regard to these effects? To begin with, the model shows that length effects are clearly variable in space and time. Figure 8.2 presents a series of four univariate plots that illustrate this variability. As for real-time developments, observe that in
142 Szmrecsanyi, Rosenbach, Bresnan & Wolk Table 8.2 Fixed effects in the minimal adequate mixed-effects logistic regression model for genitive variation in ARCHER
(Intercept) Animacy of possessor (default: animate) collective inanimate locative temporal Definiteness of possessor (default: definite) proper name indefinite Constituent length possessum length possessum length, squared possessor length possessor length, squared Semantic relation (default: non-prototypical) prototypical Final sibilant in possessor (default: no final sibilant) possessor has final sibilant Variety and real time American English (default: British English) centuries since 1800 Interactions … involving constituent length possessum length (I: centuries since 1800) possessum length, squared (I: American English) possessum length, squared (I: centuries since 1800) possessor length (I: American English) … involving animacy animacy of possessor: collective (I: centuries since 1800) animacy of possessor: inanimate (I: centuries since 1800) animacy of possessor: locative (I: centuries since 1800) animacy of possessor: temporal (I: centuries since 1800) … involving variety and real time centuries since 1800 (I: American English)
Odds ratio
Coefficient
0.56
-0.58
**
0.10 0.02 0.03 0.14
-2.34 -4.09 -3.68 -2.00
*** *** *** ***
3.73 0.63
1.32 -0.47
*** **
1.15 2.24 0.09 0.43
0.14 0.81 -2.39 -0.85
*** *** ***
2.14
0.76
***
0.42
-0.88
***
0.74 1.04
-0.31 0.04
1.47
0.39
0.66
-0.41
*
0.79
-0.24
*
1.70
0.53
*
1.87
0.62
***
0.97
-0.03
2.24
0.81
***
1.67
0.51
*
2.11
0.75
***
***
‘I’ indicates interactions. Predicted odds are for the s-genitive. Significance codes: * significant at p < 0.05, ** p < 0.01, *** p < 0.001.
1650–1875 100%
1876–2000
Variety British American
Proportion s-genitives
80%
60%
40%
20%
0% –1
0
1 2 3 –1 0 1 Binned period-centred log possessum length
1650–1875 100%
Proportion s-genitives
80%
2
3
1876–2000
Variety British American
60%
40%
20%
0% –2
–1
0 1 2 –2 –1 0 1 Binned period-centred log possessor length
2
Figure 8.2 S-genitive rates (y-axis) as a function of possessum (top) and possessor (bottom) length (x-axis; binned period-centred lengths on a log scale) and variety (heavy smoother: British English; dotted smoother: American English). Left plots: 1650–1875; right plots: 1876– 2000. Note: constituent lengths may be negative because observed lengths have been centred.
144 Szmrecsanyi, Rosenbach, Bresnan & Wolk
later ARCHER periods (1876–2000), possessum length has a more linear effect than in the earlier ARCHER periods (1650–1875). This is another way of saying that in later ARCHER texts, medium-length possessums are not as unexpectedly hostile towards the s-genitive as in earlier ARCHER texts. (Figure 8.2 seems to suggest that a similar strengthening of linearity is observable with regard to possessor length, although the regression model does not identify this development as significant.) Second, the regression model suggests that possessor length has a weaker effect size in American English than in British English, and indeed the bottom plots in Figure 8.2 indicate that the American English smoother curve is more level than the British English smoother curve, especially in the earlier period. The regression model sketched in Table 8.2 also suggests a significant interaction between possessor animacy and real time. All non-animate possessor categories except outright inanimate possessors (e.g. table, chair) have become less hostile towards the s-genitive in the course of the LModE period. In other words, there has been a tendency to increasingly use collective possessors (e.g. government), locative possessors (e.g. island), and temporal possessors (e.g. year) with the s-genitive. Figure 8.3 is an univariate plot that nicely illustrates this development: all the curves, except for the inanimates, are on the rise, at least since the nineteenth century. The figure also shows that, overall, British and American English behave rather similarly with regard to the development of the possessor animacy constraint – and indeed, the regression model does not find evidence for significant differences between the two varieties.6 Finally, we note that the regression model suggests an interaction between real time and variety, such that – all other things being equal – the s-genitive is gaining popularity faster in American English than in British English. This interaction certainly dovetails with claims in the literature about the increasing popularity of the s-genitive in American English (Rosenbach 2002; Szmrecsanyi and Hinrichs 2008). In summary, the effects that individual constraints have on genitive choice in ARCHER are largely the theoretically expected ones. What is interesting is how the interactions between language-internal constraints and language-external factors are playing out in the data. For one thing, we saw that the very similar frequency trajectories in British and American English identified in Section 8.2 are mirrored in the regression analysis to some extent: there are few significant interactions involving variety, and the only language-internal constraint that has different effects in British and American English is possessum length (which, interestingly, echoes similar 6
A small exception pertains to the temporal possessors, where American English significantly differs from British English, presumably due to the lower number of s-genitives in the 1950–99 period. This interaction was removed from the final model as it decreases model quality according to Akaike’s Information Criterion.
Genitive choice: culturally conditioned change? 145 British animate collective locative temporal inanimate
80% Percentage s-genitives of all genitives
American
60%
40%
20%
0% 1700
1800
1900
1700
1800
1900
Real time
Figure 8.3 S-genitive rates (y-axis), as a function of ARCHER time slice (x-axis) and several animacy categories. Left: British English; right: American English. Only even-numbered ARCHER periods are shown to make the plots visually more comparable. See Wolk et al. (2013) for a graphic representation of all British English periods.
findings in Hinrichs and Szmrecsanyi 2007, obtained on the basis of a different corpus database). However, there is robust interaction between real time and possessor animacy. It is to the discussion of this interaction that we shall turn next. 8.5 Discussion: culturally conditioned change? To recapitulate, we have seen in the previous section that there is a diachronic change in the effect of animacy in the genitive alternation that can be interpreted as a weakening of that predictor, in the sense that non-animate possessors (specifically collective, locative and temporal nouns) are on the rise with the s-genitive in real time. In this section, we sketch three more or less subtly different accounts for this change: linguistic change, conceptual change, and
146 Szmrecsanyi, Rosenbach, Bresnan & Wolk
change in the cultural context. We will explore how these accounts can be tested, what the predictions would be, and how plausible the accounts are given the data at hand. To set the scene, let us consider the assumption that the weakening of the animacy effect is a change in the linguistic constraints of the s-genitive. This would certainly be a plausible interpretation, but it is not the only one. When using a regression model in the way we did, we essentially determine to what degree the observed variation is conditioned on various other factors; when modelling an interaction with real time, we test whether this relationship changes over time. The accuracy of such tests, however, crucially depends on the explanatory factors themselves (a) being adequate and (b) remaining constant. We have good reason to assume that our classification strategy broadly satisfies the first condition: not only is our approach based on a body of previous research into animacy classification (e.g. Zaenen et al. 2004; Rosenbach 2008), it also leads to significant effects that behave in essence as expected. This, however, does not mean that it is perfectly adequate: if there was a small systematic difference between our classification scheme and the one that speakers use to make linguistic choices, we would in general expect that the model would still lead to similar results, albeit with less pronounced distinctions. As a hypothetical example, consider the following two ways of distinguishing between animate and collective possessors: (a1) organisations as well as temporally stable groups of humans with potentially variable concord (a2) organisations as well as temporally stable groups of humans with potentially variable concord containing at least ten people (if there are fewer than ten people, the entity in question is ‘animate’). Our coding scheme uses (a1). What would happen to our analysis if speakers actually used (a2)? On a purely synchronic level, it would mean that some tokens that we classified as collective are actually animate, and thus that the collective group behaves more similarly to the animate group in our analysis than it actually does for speakers. If the number of affected tokens is not too large, the big picture should remain largely unaffected. Diachronically, assuming everything else remains stable, we would expect the same result: a systematic, but unchanging and usually unproblematic error. In that case, a significant change through real time would be a linguistic change (in a narrow sense). The stability of the external factors, however, cannot be taken for granted: we can imagine at least two ways in which non-linguistic changes could lead to the same result. The first is that the adequacy of type classification could have changed. Let’s assume that speakers used (a1) until a certain point in time, then switched over to (a2). From then on, we would expect more genitives with ‘collective’ possessors to be realised as s-genitives than before, as more of the things we consider collective are actually members of the ‘animate’ group.
Genitive choice: culturally conditioned change? 147
The linguistic constraints would remain unchanged, but the conceptualisation processes that govern them change. Yet another scenario is that, while the adequacy on the type-level stays the same, the overall adequacy on the token-level changes. Here, we assume that speakers have always used (a2), and therefore that there is a small error in our estimations of the difference between the two groups from the beginning. Now imagine that the number of collective possessors (according to (a1)) that consist of at most ten people increases. We would then expect the collective possessors (again according to (a1)) to start behaving more like the animate possessors, as in absolute numbers the misclassifications increase. This would be neither a change in the underlying grammar, nor a change in conceptualisation, as all types themselves behave diachronically in a stable way. The observed effect would be a change purely in the cultural context (or in the frequency context) in which speakers use genitive constructions. Let us now apply these ideas to derive possible explanations for the observed changes in our genitive data set. (e1)
This explanation offers that genitive grammars have come to feature fewer selection restrictions. More specifically, the s-genitive used to have a strong constraint such that only animate possessors appear in it. This restriction persists through all periods under study, but it has considerably weakened over time. Its weakening may be related to the change of the s-genitive from a purely inflectional marker in Old English to a possessive determiner and referential anchor in late Middle and Early Modern English (Rosenbach 2004). Couching this in grammaticalisation terms, we would say that the semantics of the constructions involved are subject to bleaching and that their host class expands. (e2) Conceptual change. There are two distinct, but related, explanations involving conceptual change. (e2a) Change in the conceptualisation of possession. This explanation holds that the relation of possession is hypothesised to go against the preference for animacy more easily. The intuitive interpretation for this would be to say that possession, which has previously been largely restricted to human beings, is something that can now also be associated with other entities, for example corporations. In exactly this spirit, Nobel Prize laureate and professor of Economics and International Affairs Paul Krugman recently pointed out7 how ‘transformational technologies’, particularly railroad transportation, enabled the rise of abstract forms of legal ownership at the expense of concrete 7
Linguistic change.
For the blog post, see http://krugman.blogs.nytimes.com/2011/01/30/transformationaltechnologies/.
148 Szmrecsanyi, Rosenbach, Bresnan & Wolk
ownership during the nineteenth century. Once such forms of possession become common enough, it would not be surprising to see an extension of general possession to classes of possessors that are not animate. Note that from a typological perspective, such modern forms of abstract ownership are not necessary for broadening the concept of possession (see e.g. Seiler 1982; Heine 1997) but, given that during Early Modern English possession in English was conceived of rather narrowly as animate possession in a rather strict sense (Rosenbach 2002), the changes in the nineteenth century may have helped to pave the way for perceiving possession more broadly. (e2b) Change in the conceptualisation of animacy. Alternatively, instead of the conceptualisation of the possession relation changing, the conceptualisation of animacy may change, leading to the same local result. To stay within the example, a corporation possessing an asset could be licensed by an extension of possession to entities not considered animate, or by extending animacy to corporations – a case of the adequacy of our classification scheme decreasing in relation to types. This does not necessarily mean that speakers become confused about the ‘true’ animacy of the entities they encounter, but that they may change their perception of how similar two classes of entities are. To give an example, vehicles such as ships are clearly inanimate, but can appear in several relations normally restricted to animate contexts (consider reference by the personal pronoun she, which is attested even in Early Modern English). Observe here that of the twenty s-genitive usages with pure inanimates in our data, at least five involve ships as possessors. It would make sense that speakers judge ships to be more like animates or collectives than they would other inanimates, such as abstract concepts – not only do ships share some of the properties of animate entities, such as movement, they are often in a metonymic relationship with the crew operating them. (e3) Change in the environment / cultural context. Finally, maybe the adequacy of our classification scheme regarding types remains constant but deteriorates overall thanks to changing token frequencies. Let us consider that some different (classes of) entities – such as ships – differ from other members of the same category; we can think of them as in-between categories, or as oscillating depending on an individual speaker’s conceptualisation. If the distribution of entities in the environment changes such that those that are less like the previous majority become more frequent, it will appear that any categories between which these entities oscillate become less distinct, even without any change in the linguistic system itself. As an example
Genitive choice: culturally conditioned change? 149
for how this might happen for the other categories, consider a city council. This entity would clearly be classified as a collective, yet an individual speaker may have interacted with a good number of its members, and may well construe it as more animate – and thus as a better (potential) possessor – than she would construe a large, somewhat faceless collective such as, say, an army. If discourse contained a large proportion of city council-type recipients or possessors, collectives would show a pattern closer to clear animates, and if army-type recipients or possessors were frequent in discourse, collectives would rather tend towards genuine inanimates. It should be noted that using a different, more detailed classification scheme would not eliminate the problem. More fine-grained distinctions would increase the fit as long as we chose a better scheme, but it would also increase the possibility of choosing the wrong class – we cannot directly access the representations used by speakers in the precise moment of utterance, and especially not for historical data. We would thus just move the problem. How do these explanations fare on the empirical results presented here? We first note that a pure version of (e1) does not satisfactorily explain the cross-constructional similarity between the genitive and the dative alternation (Tom gave Mary a present versus Tom gave a present to Mary) reported in Wolk et al. (2013). This is in contrast to the other explanations. Those under (e2) can account for the cross-constructional overlap by means of the shared semantics, be they based on possession or animacy, and (e3) easily applies to both alternations if the discourse frequency of entities is assumed to be similar for recipients and possessors. However, (e1) does seem most consistent with the observed increase in temporal s-genitives, a change that is difficult to reconcile simply with a broadening of the concept of possession or animacy (note also that temporal nouns lead the change toward more s-genitive realisations among non-animate possessors). Explanation (e3) – and to some extent also (e2) – predicts that clearly inanimate cases should not change, which is supported by our data set. In addition, many of the s-genitives with inanimate possessors in the data (such as ships and animacy-related notions such as life and soul) are good examples for oscillation between categories. This may be a limitation of the size and composition of the current data set, though – Rosenbach (2003) found a clear age-grading effect in an experimental study, such that younger speakers rated s-genitives with clearly inanimate possessors more acceptable than older speakers did, which would argue against explanation e3. We wish to add that, although the numbers are too small to make reliable statistical claims, according to our data, inanimate possessors in the s-genitive are increasing in American English, and, especially in the 1950–99 period, take possessors such as plan or system that are neither vehicles nor animacy-related nor personalised concepts. This is again compatible with the experimental results
150 Szmrecsanyi, Rosenbach, Bresnan & Wolk
in Rosenbach (2003), according to which American subjects lead British subjects in accepting inanimate possessors in the s-genitive. Explanation (e2), and especially (e2b), would lead one to predict similar changes happening to other constructions. Beside the dative alternation (which is covered in Wolk et al. 2013), Hundt (2004b) explores progressive constructions in ARCHER and finds evidence for a real-time increase of inanimate NPs in subject position, which had previously been limited to animate subjects: (5)
a. I was just leaving these Lodgings <1737anon.f3b> (Hundt
2004b: 51) (human subject) b. I had never given up my opinion that an abscess was gathering <1868bowd.m6a> (Hundt 2004b: 62) (non-human subject) The animacy change in progressives that Hundt (2004b) diagnoses appears to have started about a century earlier than in genitives, however. All together, there is circumstantial evidence for and against each of the three explanations we have offered; a summary is given in Table 8.3. While we have presented them as analytically distinct beasts, they are not by necessity mutually exclusive: the observed pattern could well have resulted from any combination of underlying changes. From the viewpoint of probabilistic, experience-based grammar, it is not implausible that the underlying causes even go hand in hand, with individual changes enabling and facilitating others. For example, let us assume, as in (e3), an increase in the discourse frequency of entities that do not distinctively belong to one category. As a result, the individual categories will, for the output of existing speakers, seem less distinct even without any actual linguistic change, as per explanation (e3). Now consider new speakers entering the speech community. The input to which they are exposed contains more conflicts than past speakers received, and thus their hypotheses about the association between grammar, possession and animacy are likely to be less strong. This change could now be considered a genuine grammatical or semantic change. As a result, these speakers would produce even more utterances violating the original constraint, especially again in the less clear cases, feeding back into the process for the next generation. Several of the explanations considered in the previous discussion presume an underlying change in the discourse environment. Being able to observe such changes in ARCHER would strengthen our case considerably. So, for the sake of describing the population of nominal animacy categories as a function of real time, we created two general noun samples (which are not limited to genitive noun phrases), one each for ARCHER’s British letters and news sections and each sampling approximately 5,000 random nouns spread out evenly over ARCHER’s time periods. We next coded the nouns in these
Genitive choice: culturally conditioned change? 151 Table 8.3 Comparing different explanations for the decreasing effect of animacy in genitive variation, dative variation, and in progressives Explanandum Explanation
Increase of s-genitive with collective and locative nouns
Increase of Increase of s-genitive with double-object temporal nouns datives with non-animate recipients
Increasing use of progressive with inanimate subjects
Change in selectional restrictions of s-genitive Grammaticalisation of s-genitive as a possessive determiner (referential anchor) Change of conceptualisation of possession Change of conceptualisation of animacy
✓
✓
✗
✗
✓
✓
✗
✗
✓
✗
✓
✗
✓
✗
✓
Cultural change: rise ✓ of collective entities
✗
✓ (for those noun classes which can be perceived of as animate) ✓ (for collective ✓ (for collective nouns) nouns)
samples for animacy according to the guidelines in Zaenen et al. (2004), subsequently collapsing categories as necessary to match those used to code our genitive data set. The area plots in Figure 8.4 hence depict the distribution of animacy categories in real time. ARCHER’s letter section (right diagram) is fairly stable over time, and we will thus concentrate on the news section (left diagram) in what follows. Observe first that there is no straightforward relationship between the distribution depicted and the frequency of genitive outcomes in the data: place nouns become less frequent, time nouns stay rather constant, and collective nouns become more frequent – yet all three categories have become more likely, as we have seen, to appear as possessors of s-genitives. That said, the increase in the frequency of collective nouns, which started during the 1850–99 period, is consistent with our conjecture about environmental and cultural changes due to industrialisation and the transformational technologies that have accompanied it. We finally note that in the 1800–49 period, the frequency of animate nouns in both news and letters decreases, coinciding with – and partially accounting for – the substantial drop in s-genitive frequencies at that time (see Figure 8.1). 8.6 Concluding remarks In this chapter, we investigated the probabilistic underpinnings of genitive choice in LModE, with a particular interest in the interplay between cultural
152 Szmrecsanyi, Rosenbach, Bresnan & Wolk 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%
time noun place noun inanimate noun collective noun animate noun
16 50 17 –99 00 17 –49 50 18 –99 00 18 –49 50 19 –99 00 –4 19 9 50 –
16 50 17 –99 00 17 –49 50 18 –99 00 18 –49 50 19 –99 00 –4 19 9 50 –
100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0%
ARCHER period
ARCHER period
Figure 8.4 Animacy classification for general noun phrase sample by ARCHER period. Left: British English news; right: British English letters.
change, conceptual change and linguistic change. We have seen that, overall, genitive variability is patterned in a surprisingly similar fashion in British English and American English – so, somewhat unexpectedly, regional differences are not implicated in culturally induced genitive variation. Instead, we have suggested that the diachronically variable structure and importance of the animacy constraint may very well have cultural reasons. The idea that culture change facilitates linguistic changes is, of course, difficult to test empirically, as past speakers’ internal knowledge is not available to us. There are, however, research avenues that would permit at least a partial investigation of such mechanics. The fourth-named author is currently exploring ways to utilise the predictive capability of regression models to generate responses for new data. Assuming that a formalisation of the factors underlying the three explanations (outlined in Section 8.5) for the volatility of the animacy constraint can be found, we could train a model that represents the experiences of a single speaker in a linguistic community. By slightly altering the data used to train the model as compared to the data used to generate new data – for example, by increasing the number of tokens with less clear animacy distinctions – we should be able to simulate linguistic, conceptual and/or cultural changes such as those proposed in the previous section. Several such models can then be used to create the input for a new generation of models, an iterative process that simulates real-time probabilistic language change under controlled conditions. A simulation along such lines can, of course, not provide hard evidence on the real history of grammar (nothing can!), but it just might provide supporting evidence and constraints on the workability and plausibility of particular explanations.
Part III Changes in complementation patterns
9
On the changing status of that-clauses Gün ter Rohde nburg
9.1 Introduction In the course of the Modern English period, English has undergone a massive restructuring of its system of clausal complementation (referred to as the ‘Great Complement Shift’ in e.g. Vosberg 2006 and Rohdenburg 2006a). It is some of these developments that the present chapter explores in terms of both general evolutionary tendencies and specific contextual constraints. In addition to the quotations in OED2, the database used for this study consists of several years of British and American newspapers from the 1990s and early 2000s, the British National Corpus (BNC), the Corpus of Contemporary American English (COCA), and a sizeable collection of historical data sets provided by Chadwyck-Healey and the Gutenberg project.1 The focus is on two superficially contrary trends involving the status of that-clauses in Early and Late Modern English, which were described in previous work. In an article devoted to directive verbs such as command, enjoin and entreat, it was shown that transitive uses featuring the presence of direct objects have been largely replaced by object + infinitive constructions over the last few centuries (Rohdenburg 1995). Crucially, the change has left unaffected any that-clause complements associated with existing intransitive uses of these verbs. Accordingly, this chapter extends the analysis of that-clauses to two other verb classes, referred to as expressives and representatives in Sections 9.2 and 9.3, respectively. Here, the expectation is that the that-clause may have been affected in similar ways with some or most of the relevant transitive uses. Moreover, given that the change involving directive verb + object can be interpreted as a process of detransitivisation. Sections 9.2 and 9.3 are guided by the search for other tendencies promoting the use of low transitivity features in the sense of Hopper and Thompson (1980). Sections 9.4 and 9.5 are inspired by the developments that have occurred with nouns like risk, where the that-clause was the last type of complement to be established (Rohdenburg 2013: 139–140). Thus, 1
A summary of the corpora and the abbreviations used to refer to them can be found in Table 9.26 in the Appendix to this chapter.
155
156 Rohdenburg
these sections are concerned with that-clauses that have undergone some form of consolidation or resurgence in American English (AmE), British English (BrE) or both. Section 9.4 deals with intransitive uses of directive verbs, comparing so-called subject-to-object raising structures (or AcI constructions) and that-clauses in the passive, and Section 9.5 investigates the contrast between finite clauses and their rival gerundial constructions after emergent or established phrasal expressions such as in (the) event ‘if ’. More generally, this chapter is throughout informed by a resolutely paradigmatic approach, which, wherever feasible, contrasts the evolution of the that-clause with that of its immediate non-finite rivals or, in the case of transitive verbs, with corresponding intransitive uses. 9.2 Expressive verbs Deviating somewhat from the definition of expressive verbs given by Searle (1976, 1979: 14–16), the class is here circumscribed by the items listed in (1). (1)
abuse, accuse, blame, bless, charge, chide, compliment, condemn, congratulate, criticise, curse, felicitate, flatter (o.s.), forgive, laud, pardon, pique (o.s.), plume (o.s.), praise, pride (o.s.), rebuke, reproach, revile, ridicule, scold, taunt, thank, upbraid, value (o.s.)
All of them are characterised by the following features. They are (or were) used as transitive verbs complemented by human objects (including reflexive pronouns) and prepositional expressions, which may be realised by gerunds as in (2) or (3). (2) They accused him of theft/of having stolen the money. (3) Several … have flatter’d themselves, with obtaining … a universal
Dissolvent. (OED, 1718)
In all of these cases, the implied subject of the simple gerund (lacking an explicit notional subject of its own) is co-referential with the object of the superordinate verb. Alternatively, the gerund may be replaced by a thatclause as in (4), which, of course, need not have a subject that is co-referential with the preceding object. (4)
Carson blamed himself that his mount had not fared better. (t90)
The limitation to transitive uses means that we will not be concerned with verbs like apologise, condole and deplore, which along with congratulate and thank are mentioned by Searle (1979: 15) as paradigms of the class of expressive verbs. Generally speaking, that-clauses represent an older stratum than gerunds, and with several classes of verbs they have indeed been largely replaced by infinitival constructions and by the increasingly popular gerund (see e.g.
The changing status of that-clauses 157 Table 9.1 The evolution of finite and non-finite complements after flatter + [+/reflexive] object in the quotations of the OED between 1650 and the present I that – clause
1650–99 1700–99 1800–99 1900–88
II gerund
(a) [+ reflexive]
(b) [- reflexive]
(c) % [+ reflexive]
[+ reflexive]
3 21 21 1
– 4 1 –
84% 95.5%
– 2 –
Vosberg 2006, Iyeiri 2010, and Rohdenburg 1995 as well as the extensive coverage of prepositional gerunds by e.g. Rudanko 1996 and 1999). This section attempts to investigate the question if and to what extent similar developments are true of expressive verbs. Using six more or less arbitrarily selected verbs (flatter, congratulate, pride o.s., upbraid, reproach, blame), I will compare the evolution of that-clauses over the last few centuries with that of rival gerunds. The discussion in this and the following section leans heavily on the theory of transitivity developed by Hopper and Thompson (1980). They conceive of transitivity as a global category relating to ‘the effectiveness with which an action takes place’ (p. 251). The prototypical transitive clause describes an asserted, punctual event in which a human agent wilfully and forcefully impinges on a patient other than himself. Analysing the infinitival construction involving causative bring, Mondorf (2010) is able to show that its decline in the course of Late Modern English (LModE) is accompanied by various manifestations of detransitivisation in the sense of Hopper and Thompson. Thus, the infinitival construction has been increasingly confined to reflexive and passive uses like (5a–b) as well as to examples containing inanimate (and typically non-agentive) subjects as in (5c).2 (5)
a. She could not bring herself to tell them. b. He was brought to tell them the truth. c. It was these incidents that brought him to change his mind.
With this in mind, consider in Table 9.1 the evolution of the verb flatter in the quotations of the OED between 1650 and the present. With the exception of the gerund exemplified in (3) above, which according to other sources is only attested for about a hundred years, all of the relevant cases are represented by that-clauses (mostly with that expressed). 2
Earlier treatments of some special effects triggered by inanimate (and non-agentive) subjects are provided by, inter alia, Hurst (1951) (for Spanish) and Rohdenburg (2006b: 57–59). For instance, Rohdenburg shows in the case of verbs like force that, unlike control uses, raising constructions are typically associated with inanimate subjects (or the absence of explicit subjects in -ing-clauses).
158 Rohdenburg Table 9.2 The evolution of finite and gerundial complements after congratulate + object and the corresponding passive in a series of historical databases
EEPF, EPD (17th century)a ECF, EPD (18th century) NCF1, EPD, LNC/B (19th century) ETC/B, BNC: wridom1 (early and late 20th century)
I that-clause
II prep –ing
III prep Pro/NP –ing
IV % that-clause
1
3
–
14 (50%)b E:8c 45 (86.7%) E:21
29 (34.5%)
25.9%
80 (52.5%)
11 E:10 8 E:3
18 (100%) E:8
52 (78.8%)
–
25.7%
33.8%
a
In the 17th-century data, the object is in part realised by a with-phrase: once preceding a that-clause (column I) and once preceding a (simple) gerund (column II). b The bracketed figures in columns I and II specify the percentages of all detransitivised examples comprising reflexive, reciprocal and passive constructions. c In this and later tables, the figures introduced by E (for Equi) indicate the number of explicit or implied subjects that are co-referential with the object of the superordinate verbs in columns I and III.
However, the evidence suggests that non-reflexive uses such as (6) were largely ousted by the predominant reflexive type as in (7) in the course of the LModE period. (6) (7)
My friends flatter me, that I pronounced those words with a tolerable accent. (OED, 1712) I flatter myself that I have preserved a philosophic poise. (OED, a1953)
Indeed, examples like (6) have become obsolete by now. Turning next to a large-scale analysis of congratulate in Table 9.2 based on data from the seventeenth century to the present day, we find that the proportion of that-clauses has remained relatively stable throughout. However, the raw figures conceal some intriguing manifestations of detransitivisation. While the use of reflexive objects and other low transitivity features is consistently increased with both that-clauses and immediately rivalling gerunds, the development in the case of the that-clause is much further advanced in the eighteenth and nineteenth centuries and reaches completion by the end of the nineteenth century. Moreover, as can be seen in column III, gerunds using additional possessive pronouns have increasingly become available to refer explicitly to non-coreferential subjects. This brings us to the survey of the verb pride, which has always been confined to the use of reflexive pronouns. The results in Table 9.3 reveal that
The changing status of that-clauses 159 Table 9.3 Finite and non-finite complements after pride o.s. in a series of historical databases including the fictional components of the BNC and COCA
BrE
AmE
EEPF (1518–1700) ECF1 (*1660–*1699) ECF2 (*1703–*1752) NCF1 (*1728–*1799) MNC/B (*1803–*1829) NCF2 (*1800–*1869) LNC/B (*1830–*1869) ETC/B (*1870–*1894) BNC: wridom1 (1960–1993) EAF1 (*1744–*1799) EAF2 (*1801–*1827) MNC/A (*1803–*1829) LNC/A (*1830–*1869) ETC/A (*1870–*1894) COCA: fiction (1990–2012)
I that-clause
II prep –ing
III to-infinitive
IV total
V % that-clause
–
–
10
–
0%
2
1
5
40%
3
–
3
–
14
–
14
0%
–
14
–
14
0%
–
36
–
36
0%
–
37
–
37
0%
–
15
–
15
0%
9 E:5 1 E:1 1 E:0 –
25
–
34
26.5%
28
–
29
3.4%
15
–
16
6.3%
10
–
10
0%
–
29
–
29
0%
6 E:4 3 E:3
25
–
31
19.4%
96
–
99
3.0%
2 E:1 –
with the exception of two instances of the that-clause from the early eighteenth century, none of the British LModE databases provides any further examples. Surprisingly enough, the narrative domain of the BNC exhibits a revival of the that-clause, which represents over a quarter of all cases. But then four of the nine instances involve non-coreferential subjects, an option that would have been impossible with (simple) gerundial constructions. The analysis of the nineteenth-century and early twentieth-century American databases appears to suggest that the resurgence of the that-clause has been spearheaded by AmE. There are two examples in the EAF, dating from 1827 and 1843, and the ETC/A, covering the beginning of the twentieth century, reaches a rate of nearly 20 per cent. However, comparing the situation in the
160 Rohdenburg Table 9.4 Finite and gerundial complements after upbraid + object and the corresponding passive in a series of historical databases including the BNC, COCA as well as British and American newspapers I that-clause EEPF ECF NCF1 MNC/B NCF2 LNC/B ETC/B BNC: wridom1 BNC (overall) t90–95, g90–95, d91–95, i93–94 COCA 12 L92–95, D92– 95, W90–92, N01
5 E:4 6 E:4 3 E:3 – – – – – – – – –
II prep –ing 12 35 12
III prep Pro/NP –ing 1 E:1 3 E:2 –
IV total
V % that-clause
18
28.6%
44
13.6%
15
20%
8 13 9 3 1 8 125
– – – – – – –
8 13 9 3 1 8 125
0% 0% 0%
39 57
– –
39 57
0% 0%
0% 0%
fictional components of the BNC and COCA, we find that AmE seems to have lost its earlier advantage over BrE. This takes us to a second group of items, which contains three verbs of blaming, upbraid, reproach and blame itself. The evolution of that-clauses after upbraid presented in Table 9.4 appears to be fairly straightforward. Compared with the EEPF, the eighteenth- and early nineteenth-century corpora display reduced rates of that-clause occurrence, and the later databases, including the BNC and COCA as well as British and American newspapers, do not produce any further instances. This suggests that the type is now obsolete. With reproach and blame documented in Tables 9.5 and 9.6, we also find that the that-clause has been receding more or less steadily over the last few centuries compared with its gerundial rival. Yet in both cases, the that-clause is found in all of the LModE data sets, and it can still be met with, although infrequently, under favourable circumstances, such as in reflexive uses of the verb, as in (4) above. A further parallel concerns the increasing use of such low transitivity features, as shown in Tables 9.7 and 9.8. In the case of blame, we can also observe a general increase of non-coreferential subjects in the that-clause, in particular. In the period up to the first half of the eighteenth century, virtually all that-clauses contain subjects co-referential with the object of the superordinate verb. In the later
The changing status of that-clauses 161 Table 9.5 Finite and gerundial complements after reproach + object and the corresponding passive in a series of historical databases including the BNC I II that- clause prep –ing EEPF (17th century)a ECF1b ECF2 NCF1 MNC/B LNC/B ETC/B BNC: wridom1 BNC (overall)
3 E:2 10 E:6 2 E:1 8 E:6 1 E:1 6 E:6 – – –
III total
IV % that-clause
7
10
30%
45
55
18.2%
27
29
6.9%
98
106
7.5%
28
29
3.4%
38
44
13.6%
4 8 28
4 8 28
0% 0%
a
Two other types of non-finite constructions have been disregarded here: (1) to-infinitives (1×) and (2) the type possessive pronoun + –ingform (3×) functioning as the (sole) object of the verb. b Column II includes one instance of the type with + possessive pronoun + –ing-form, where the (explicit) subject of the –ing-form is (also) coreferential with the object of the verb reproach.
Table 9.6 Finite and gerundial complements after blame + object and the corresponding passive as well as the type X is to blame in a series of historical databasesa I thatclause EEPF (16th century) EEPF (17th century) ECF1 LNC/B ETC/B BNC: wridom1 a
5 E:5 9 E:9 8 E:7 6 E:3 5 E:3 1 E:0
II gerunds (a) prep –ing 23 63 74 77 44 122
III total
IV % that-clause
(b) prep Pro/NP –ing 1 E:1 2 E:2 –
29
17.2%
74
12.2%
82
9.8%
1 E:0 –
84
7.1%
49
10.2%
3 E:0
126
0.8%
The analysis excludes a sprinkling of infinitives found as late as ECF1 and a few nonprepositional gerunds scattered across three databases: the 17th century of the EEPF (3×), LNC/B (1×) and wridom1 (2×).
162 Rohdenburg Table 9.7 Long-term detransitivisation trends with reproach measured in terms of the combined percentages of (a) reflexive objects, (b) passives and (c) inanimate subjects (including nouns such as heart or conscience denoting body parts or attributes of a person)
1 2a 2b
EEPF (17th century) + ECF + NCF1 MNC/B + LNC/B + ETC/B MNC/B + LNC/B + ETC/B + BNC: wridom1
I that-clauses
II all gerunds
43.5% N=23
38.4% N=177
100% N=7 100% N=7
30% N=70 33.3% N=78
Table 9.8 Long-term detransitivisation trends with blame measured in terms of the combined percentages of (a) reflexive objects, (b) passives and (c) the type X is to blamea
1
EEPF + ECF1
2a
LNC/B + ETC/B LNC/B + ETC/B + BNC: wridom1
2b
I that-clause
II prepositional gerunds
18.2% N=22 (21/1) 63.6% N=11 (6/5) 66.7% N=12 (6/6)
28.2% N= 163 (163/0) 35.2% N=122 (121/1) 29.1% N=247 (243/4)
a
The bracketed figures distinguish between subjects that are co-referential with the object of the superordinate verb and all others, which are non-coreferential.
databases, by contrast, the share of non-coreferential subjects reaches the halfway mark. Significantly, a similar, though much weaker, trend extends to the category of gerunds. This indicates that the pressure to select a noncoreferential subject is no longer sufficient to produce a that-clause. 9.3 Non-central representatives Unlike expressives, that-clauses after direct or directly linked objects constitute a much more prominent feature with so-called representatives, as in (8). (8) They advised/notified/reminded us that the train might be delayed.
Relevant verbs that possess or used to possess this option include those in (9).
The changing status of that-clauses 163
(9)
acquaint, advise, answer, assure, certify, confirm, inform, let X know, notify, remind, satisfy, tell, warn
As is pointed out in e.g. Searle (1976, 1979), there is some overlap between representatives (or assertives) and directives. In contradistinction to directives, including advise, remind, tell and warn, representatives cannot occur in object-infinitive constructions such as (10). (10) They warned me to get there on time.
In fact, representatives are simply concerned with the transfer of information, which – with some of the verbs listed in (9) – might conceivably be false. In addition to that-clauses, a further feature characterising almost all of the verbs in (9) is or was the use of of-phrases as in (11). (11) You need be under no Apprehension, Sir, I shall satisfy Mr. Allworthy
perfectly of the Matter. (OED, 1749)
With five of these verbs (acquaint, answer, certify, confirm, satisfy), the thatclause has now either been phased out or is severely restricted in its syntactic range. In the following, the evolutionary pathways of these verbs, which for various reasons may be regarded as less central than, say, tell, inform and let X know, will be described in turn. The analysis of acquaint summarised in Table 9.9 distinguishes between that-clauses and all other complements after personal pronouns as in (12a–b). (12) a. I acquainted her that I spoke the Mandarin language … which
they constantly use at the Court. (OED, 1697)
b. Have you acquainted her with the defeature Of the Carthaginians.
(OED, 1623)
By comparison, the occurrence of direct objects not followed by other complements used to be very rare, even in earlier stages of the language. The evidence in Table 9.9 leaves no doubt that the dramatic decrease of the that-clause during the eighteenth and nineteenth centuries mirrors a general decline of the verb. However, the retreat of the that-clause is much more extensive than that of the remaining kinds of complement, most of which survive to this day. This is true, in particular, of the reflexive use of acquaint governing with-phrases in the sense of ‘familiarise oneself with’. In addition, comparing BrE and AmE, we find that the decline of both the that-clause and other complements occurred noticeably earlier in AmE than in BrE. We turn next to the verb answer, whose evolution is portrayed in Tables 9.10 and 9.11. The analysis has here been confined to two types of context representative of active and passive uses, (a) personal pronouns
164 Rohdenburg Table 9.9 The distribution of that-clauses and other complements after acquaint + object (realised by a personal pronoun) in a series of historical databases including the BNC I that-clause
BrE
AmE
EEPF (17th cent.) ECF1 NCF1 MNC/B NCF2 LNC/B ETC/B BNC: wridom1 BNC (overall) EAF1 EAF2 MNC/A LNC/A ETC/A
II other complements
III % that-clause
IV that-clause: frequency per million words (pmw)
≈ 135
20.90
138 83 3 6 4 – –
≈ 397 22 60 17 8 5
12% 9.1% 19% 0%
–
17
0%
11 6 3 – –
101 32 22 16 17
25.8%
26.90 7.30 0.28 0.23 0.19 – – –
9.8% 15.8% 12% 0% 0%
0.69 0.32 0.14 – –
Table 9.10 The evolution of that-clauses after answer + object and the corresponding passive in the quotations of the OED between 1550 and the present
1550–1599 1600–1649 1650–1699 1700–1749 1750–1799 1800–1849 1850–1988
I active + Pro (excluding it)
II passive involving human subject
III total
3 6 2 1 1 – –
– 1 – 1 1 3 –
3 7 2 2 2 3 –
immediately following the active verb, and (b) passives immediately preceded by a form of the verb be and immediately followed by a that-clause or an agent phrase and a subsequent that-clause. Relevant examples are provided in (13a–b). (13) a. He answer’d me that he had a thousand such sort of liquors,..Old
Pharaoh, Knockdown, Hugmatee [etc.]. (OED, 1698)
The changing status of that-clauses 165 Table 9.11 The evolution of that-clauses after answer + object and the corresponding passive in a series of historical databases including the BNC I active + Pro (excluding it) EEPF (17th century) ECF1 ECF2 NCF1 MNC/B NCF2 LNC/B ETC/B BNC: wridom1 BNC (overall)
II passive involving human subject
III total
IV frequency per million words (pmw)
71
3
74
11.46
14 10 9 2 ≈8 14a 5 – –
8 7 14 5 5 1 1 – –b
22 17 23 7 ≈13 15 6 – –
4.29 3.72 2.02 0.65 0.50 0.72 1.25 – –
a b
The figure excludes two instances involving to + Pro in the object slot. The analysis excludes a passive example ascribed to the theologian Richard Hooker, who died in 1600. (BNC, file CFF)
b. She was answered that some twenty minutes ago a vehicle had
passed the barrier containing the deputy M. le Chapelier and the Paladin of the Third Estate, Mr. Moreau. (ETC/B, 1921)
A pilot study carried out on the OED quotations (see Table 9.10) suggests that the that-clause may have begun to decline in the second half of the seventeenth century and that it became obsolete in the first half of the nineteenth century, with the passive surviving longer than the active. While the first observation is confirmed by the large-scale analysis in Table 9.11 focusing on narrative corpora, the second needs to be qualified. The that-clause after direct objects is still found to occur on a regular basis as late as the early twentieth century (see the data in the ETC/B), but further evidence for the delayed survival of the passive is lacking so far. Crucially, however, the that-clause after the intransitive use of answer remains as firmly established as ever. This brings us to two other verbs, certify and confirm, after whose intransitive uses the that-clause continues unabated. In the case of certify, let us start by taking a look at the pilot study in Table 9.12 based on the quotations in the OED. The data suggest that – discounting a solitary example of the that-clause dating from 1878 – the previously well-established option was dropped abruptly in the second half of the seventeenth century. In Table 9.13 devoted to narrative corpora, a more delicate analysis is provided.
166 Rohdenburg Table 9.12 The evolution of that-clauses after certify + object and the corresponding passive in the quotations of the OED
1550–1699 1700–1899 1900–1988
I object + that-clause or corresponding passive
II objectless verb + that-clause
III % transitive uses as in I
8a 1 (1878) –
6 15
57.1% 6.7%
6
0%
a
The figure excludes one instance involving unto + NP in the object slot, dating from 1656.
Table 9.13 The evolution of that-clauses after certify + object and the corresponding passive in a series of historical databases including the BNCa
EEPF (17th century) ECF NCF1 MNC/B NCF2 LNC/B ETC/B BNC: wridom1 BNC (overall)
I object + thatclause
II III passive objectless verb corresponding to I + that-clause
IV % transitive uses as in I + II
22
5
3
90%
1 – 2 3 – – – –
3 – 2 – – – – –
6 4 7 15 6 – 5 74
40% 36.4% 16.7% 0% – 0%
a
The analysis excludes several instances involving unto/to + NP in the object slot scattered across various databases: EEPF (2×), NCF1 (1×), wridom1 (1×), BNC (4×).
While it confirms the prominent status of the that-clause after direct objects for the Early Modern English period, it supplies sufficient evidence for assuming a threefold periodical division. In the seventeenth century, that-clauses after verb + object clearly predominate over those after intransitive certify, which are comparatively rare. After a sharp drop in its frequency of occurrence in the eighteenth century, the that-clause after direct objects continued for roughly 150 years at a reduced average rate of 25 per cent of all relevant cases, and from the middle of the nineteenth century onwards it is no longer found in the databases investigated. By contrast, that-clauses after intransitive uses of certify survive to this day. Interestingly, the phasing out of that-clauses after certify + direct object is accompanied by three further kinds of detransitivisation involving the increased use of (a) passives,
The changing status of that-clauses 167 Table 9.14 The evolution of that-clauses after confirm + object in a series of historical databases a I object + that-clause EEPF (17th century) ECF1 EPD (17th century) EPD (18th century) NCF1 MNC/B NCF2 LNC/B ETC/B EPD (19th century) BNC: wridom1
II objectless verb + thatclause
III frequency of I pmw
11
1
1.70
11 10
1 2
2.14 1.64
12
1
1.16
2 – – 2 – –
1 – – 1 – –
0.18 – – 0.10 – –
–
145
–
a
The analysis excludes several instances involving to + NP in the object slot for ECF1 (5×) and wridom1 (5×).
(b) reflexive pronouns and (c) inanimate subjects. The three categories are exemplified in (14a–c). (14) a. You’re a hired smockster; here’s her letter, In which we are certi-
fied that you’re a bawd. (OED, 1607)
b. But the Wharton lawyer by use of reckless telegrams had certified
himself that the infant was a girl, and Everett was the hero of the day. (NCF, 1876) c. She paused, and saw him enter, flushed, and with a look that certified her that something had happened. (NCF, 1856) In the EEPF data, the combined total of these low transitivity features makes up 22.2 per cent of the total comprising actives and passives. By contrast, in the following period preceding the absence of that-clauses after verb + object or its passive counterpart, the percentage of low transitivity features rises to 90.9 per cent of the total. Owing to the vast number of irrelevant examples encountered, I have left the survey of that-clauses after confirm + object in a state that is, admittedly, less than ideal. Table 9.14 includes two categories of that-clause after verb + object: (a) all examples with that expressed and (b) all examples after verb + personal pronoun (excluding it) with that left unexpressed. These are compared with all examples of explicit that-clauses after intransitive uses of the verb in column III.
168 Rohdenburg Table 9.15 The evolution of the that-clause after satisfy + object in a series of historical databases I human subject (explicit or implied) + nonreflexive object EEPF (17th century) ECF1 ECF2 NCF1 NCF2 LNC/B ETC/B BNC: wridom1
II inanimate subject
III reflexive object (after human subject)
IV total (frequency pmw)
V % II + II in IV
4
2
1
7 (1.08)
42.9%
13 10 if: 1 12 if: 2 14 if: 1 unless: 1 3 if: 1 1 if: 1 1 unless: 1
5 6
10 3
28 (5.46) 19 (4.16)
53.6% 47.4%
14
41
67 (5.89)
82.1%
44
88
146 (5.61)
90.4%
9
49
95.1%
3
13
61 (2.93) 17 (3.54)
1
24
26 (1.38)
96.2%
94.1%
For what it is worth, the evidence suggests that the that-clause must have been distinctly more common – in the seventeenth to eighteenth centuries – after verb + object than after intransitive uses. In the nineteenth century, the that-clause after the transitive verb experiences an abrupt decline, and its absence in the narrative domain of the BNC allows us to assume that it has been phased out altogether. By contrast, that-clauses after intransitive uses of confirm can be observed to be still flourishing. Finally, consider the evolution of satisfy as summarised in Table 9.15. Compared with the other representative verbs, the fluctuation over time of the textual frequency of its that-clause is seen to be relatively small. After a modest start during the seventeenth century, for about 150 years the that-clause maintains a rate near or exceeding five instances per million words. The late nineteenth century then witnesses a decrease in its frequency of occurrence, which in the narrative domain of the BNC falls to as low as 1.38 examples per million words. Much more important, there are some drastic changes in the syntactic orientation of the superordinate verb, which has always been associated with a direct object. Starting in the EEPF, we can observe two wellknown manifestations of detransitivisation, the use of (a) inanimate subjects and (b) reflexive objects. These are exemplified in (15a–b). (15) a. He … cannot comprehend the arguments which satisfy men of
sober sense that his views are erroneous. (OED, 1856)
b. We have at all events satisfied ourselves that [etc.]. (OED, 1860)
The changing status of that-clauses 169
Again, as in the case of flatter and acquaint, the passive has been ignored for a very simple reason. In construction with the verb be, the perfect participle has been grammaticalising into a predicative adjective, making the identification of bona fide passives extremely difficult. As is indicated in column V, there is an abrupt change between the second part of the ECF and the first part of the NCF, with the proportion of low transitivity features as in (15a–b) increasing from 47 per cent to 82 per cent. This orientation has been further strengthened, and by now the proportion of detransitivised uses is close to 100 per cent. Furthermore, in the declining number of examples featuring human subjects along with non-reflexive objects, non-assertive uses of the verb in if- and unless-clauses appear to be assuming an increasingly prominent role. The most striking change, however, which begins to assert itself in the late nineteenth century, concerns the distribution of the two major detransitivising devices themselves. While their combined totals hover close to 100 per cent, reflexive pronouns as in example (15b) have increasingly come to exhaust the low transitivity spectrum at the expense of inanimate subjects as in (15a). 9.4 Raising structures and finite complements with directive and related verbs In an earlier study (Rohdenburg 1995), I have shown that with directive verbs the constellation object + that-clause as in (16a) has been largely replaced over the last few centuries by object + to-infinitive complements as in (16b). (16) a. *They ordered the guard that he should release the prisoner. (see
Hudson 1976: 369)
b. They ordered the guard to release the prisoner.
The purpose of this section is to draw attention to that-clauses after directives and related verbs not involving direct objects as in (17a), which continue to be well established. However, it can be demonstrated that compared with passive raising structures such as (17b), that-clauses like (17a) may have undergone drastic frequency changes over the last two centuries. (17) a. They ordered that the meal should be cooked by six o’clock. b. They ordered the meal to be cooked by six o’clock.
Let us begin by surveying in Table 9.16 the present-day situation with a representative set of six verbs in a large database comprising both British and American newspapers. The search has been restricted as follows: (a) the (passive) complements contain the item be (functioning as subjunctive or infinitive), (b) both the subject of the that-clause and the (so-called derived) object of the raising construction are represented by the-initial noun phrases, and
170 Rohdenburg Table 9.16 The distribution of that-clauses and raising structures (AcI constructions) in the passive after directive and related verbs (in the active) in British and American newspapers I II that-clause AcI (subj./should/others)
III total
IV % that-clause
BrE: t90–95, g90–95, d91–95, i93–94, m93–95 prefer 18 (12/6/0) 110 require 126 (78/48/0) 344 urge 181 (121/60/0) 11 recommend 103 (69/34/0) 4 request 124 (112/12/0) 9 direct 75 (44/26/5) 6
128 470 192 107 133 81
14.1% 26.8% 94.3% 96.3% 93.3% 92.6%
AmE: L92, D92–95, W90–92, N01 prefer (33/0/0) require 203 (197/3/0) urge 166 (164/2/0) recommend 43 (43/0/0) request 110 (110/0/0) direct 37 (37/0/0)
35 294 166 43 110 38
94.3% 69.0% 100% 100% 100% 97.4%
2 91 – – – 1
(c) with the exception of the complementiser that itself (which is occasionally omitted), there are no elements intervening between the superordinate verb and the subject or object phrases concerned. The results in Table 9.16 indicate that there are some clear-cut differences between the two national varieties. Generally speaking, AmE displays a stronger affinity for the that-clause, and with two of these verbs, notably prefer and require, BrE is seen to have very largely suppressed the finite option at the expense of the raising structure. It is only in the case of require that – at 31 per cent of the total – AmE makes liberal use of the raising structure as well. In addition, it is apparent that the two regional varieties tend to favour the subjunctive to different degrees. While even BrE displays a clear preponderance of subjunctives in this area, AmE uses them almost exclusively. Let us now take a look at the evolutionary pathways of three of these verbs in Tables 9.17–9.19, whose results are based on all relevant noun phrases including any items intervening between superordinate verb and subject or object phrase. With prefer, the rivalling constructions are not found until the end of the eighteenth century in the narrative corpora consulted, and their subsequent evolution up to the end of the twentieth century is sketched in Table 9.17. In this case it has been both possible and revealing to include other uses of the verb be alongside the passive. Such non-passive uses of be are illustrated in (18a–b) for both the that-clause and the raising construction.
The changing status of that-clauses 171 Table 9.17 The evolution of that-clauses and raising structures involving be-predicates after (active) prefer in a series of historical databases
BrE
AmE
I that-clause (passives/ other uses)
II AcI (passives/ other uses)
III total
IV % that-clause (passives/other uses)
EEPF ECF NCF (*1752–*1829) NCF (*1840–*1858) LNC/B (*1846–*1867) ETC/B (*1873–*1882) BNC: wridom1
– – 16 (8/8)
– – –
– – 16 (8/8)
– – 100%
12 (9/3)
12 (0/12)
24 (9/15)
50% (100%/20%)
2 (2/0)
18 (5/13)
20 (7/13)
t90–91
8 (8/0)
81 (44/37)
89 (52/37)
10% (28.6%/0%) 9% (15.4%/0%)
EAF (*1789–*1806) LNC/A (*1833–*1860) ETC/A (*1875–*1885) W90–92
16 (10/6)
–
16 (10/6)
100%
13 (5/8)
5 (1/4)
18 (6/12)
72.2% (83.3%/66.7%)
42 (28/14)
21 (4/17)
63 (32/31)
66.7% (87.5%/45.2%)
(18) a. You probably will prefer that the funeral solemnities should be at
Trinity-Church. (EAF, 1830)
b. It will be really a gift, but I prefer it to be as a loan for many rea-
sons. (LNC/B, 1905)
It is seen that in both BrE and AmE the verb prefer starts out by using finite clauses exclusively with both functions of be. It is only in the second half of the nineteenth century that the first examples of the raising construction begin to crop up, with BrE clearly taking the lead. Initially, though, the rise of the raising construction is virtually confined to other than passive uses of be. At present, the passive is still lagging behind other uses of be in both BrE and AmE. Yet while in present-day BrE the passive raising construction also predominates over the that-clause, it has never gained a firm foothold in AmE. Comparing the corpora of the Brown family, Leech et al. (2009:146–52, 164) find that the be-passive has generally declined in written English, with AmE being clearly further advanced than BrE. Could the scenario portrayed in Table 9.17 be interpreted as anticipating this regional contrast observed for the second half of the twentieth century? I do not think
172 Rohdenburg Table 9.18 The evolution of that-clauses and raising structures in the passive after (active uses of) require in a series of historical databases I that-clause BrE
AmE
EEPF (17th cent.) ECF NCF1 NCF2 (*1800-*1829) NCF2 (*1830-*1849) LNC/B BNC: wridom1 EAF1 EAF2 LNC/A
II AcI
III total
IV % that-clause
1
3
4
14 19 17
5 12 10
19 31 27
73.7% 61.3% 63.0%
4
11
15
23.5%
5
9
14
35.7%
28 22 30
14 12 9
42 34 39
66.7% 64.7% 76.9%
the data in Table 9.17 warrant such a conclusion. Crucially, in comparison with other uses of be, the overall ratio of the passive does not seem to have declined in this area over time. The decisive difference between the two varieties concerns the relative decline of the that-clause after prefer, with AmE being more resistant to the introduction of the raising construction than BrE. However, the contrasting behaviour of passives and other uses of be is a phenomenon shared by both varieties. With require and request, the two other verbs singled out for a detailed historical treatment, the data are not sufficient to justify the inclusion of non-passive uses of be. In the case of require, the history of the two rival constructions is presented in Table 9.18. After peaking at over 70 per cent of the total in the first half of the eighteenth century, the that-clause is found to hover round 60 per cent for over a hundred years in both BrE and AmE alike. It is not until the second half of the nineteenth century that the two regional varieties are seen to diverge dramatically. In BrE, there is an abrupt drop of the that-clause to less than a third of the two competing constructions, a rate not far off that found in the narrative domain of the BNC. By contrast, AmE displays a slightly increased rate of 77 per cent, which is close to the percentage found in American newspapers of the 1990s and early 2000s (see Table 9.16 above). As is suggested by the evidence in Table 9.19, the raising structure after request possibly represents a LModE innovation. Interestingly, the minority type is no less common in AmE than BrE. Thus, it is reasonable to assume that the contrast between the two national varieties seen in Table 9.16 must have evolved in the course of the twentieth century. Furthermore, it may be surmised that the strengthening of the that-clause at the expense of the
The changing status of that-clauses 173 Table 9.19 The evolution of that-clauses and raising structures in the passive after (active uses of) request in a series of historical databases I that-clause
II AcI
III total
IV % that-clause
BrE
ECF NCF1, MNC/B NCF2, LNC/B ETC/B BNC: wridom1
9 42 40 1 4
– 3 5 – –
9 45 45 1 4
100% 93.3% 88.9%
AmE
EAF, MNC/A, LNC/A ETC/A
55
9
64
85.9%
1
1
2
raising construction is somehow due to the dramatic revival of the mandative subjunctive in AmE (see e.g. Övergaard 1995 and Hundt 2009: 30–31). In the case of passives after demand, for instance, the beginnings of this revival can be traced back to the middle of the nineteenth century (Rohdenburg 2009: 319–22). 9.5 Finite adverbial clauses and prepositional gerunds Similar contrasts between BrE and AmE are found with a number of (partly grammaticalised) prepositional structures. These include the items in (19), which are commonly treated as (emergent or established) complex conjunctions. (19) on account, for fear (Rohdenburg 2008: 328; Rohdenburg and Schlüter
2009: 407–09), in (the) event ‘if ’ (Biber et al. 1999: 844; Rohdenburg 2013), on/upon (the) condition (Quirk et al. 1972: 746; Biber et al. 1999: 844; Schlüter 2009)
Unlike earlier treatments, which generally do not refer to any competing gerundial constructions, this chapter again focuses on the contrast between the two global paradigmatic choices available within most of such complex structures. Consider, for instance, the finite and gerundial alternatives illustrated in (20)–(22). (20) a. They rejected the plan for fear (that/lest) they/their friends
would be cheated.
b. They rejected the plan for fear of Ø/their /their friends being
cheated.
(21) a. In the event (that) the teams/we are delayed we will have to make
alternative arrangements.
b. In the event of the teams/us/our/Ø being delayed we will have to
make alternative arrangements.
174 Rohdenburg Table 9.20 Finite, gerundial and infinitival clauses after for fear in a series of historical databasesa
BrE
EEPP (1600–1649) ECF1 NCF1 (*1760–*1799) MNC/B LNC/B ETC/B
AmE
EAF1 (*1760–*1799) MNC/A LNC/A ETC/A
I finite (that/Ø/lest)
II of (Pro/ NP) –ing (Pro/NP vs Ø)
III IV to-infinitive total
V % finite
32 (6/18/8)
46 (1/45)
10
88
36.4%
172 (2/170/0) 41 (3/38/0)
127 (7/120) 44 (10/34)
– –
299 85
57.5% 48.2%
80 (3/73/4) 136 (25/85/26) 14 (0/13/1)
83 (14/69) 100 (4/96)
– –
163 236
49.1% 57.6%
14 (1/13)
–
28
50%
62 (0/61/1)
43 (2/41)
–
105
59.0%
24 (5/19/0) 148 (157/131/2) 147 (37/110/0)
13 (1/12) 65 (4/61)
– –
37 213
64.9% 69.5%
43 (3/40)
–
190
77.4%
a
The analysis disregards any examples containing elements between for fear and the preposition of with gerunds or between for fear that/Ø/lest and the subject of finite examples.
(22) a. They accepted the solution on condition (that) they/he receive an
unspecified lump sum right away.
b. They accepted the solution on condition of Ø/their/his receiving
an unspecified lump sum right away.
In what follows, the pathways traced by the three kinds of contrast will be presented in turn. Table 9.20 – devoted to for fear – shows that, starting at just over a third of the relevant total in the early seventeenth century, and despite a certain amount of fluctuation later on, the finite clause in BrE consolidates its position at 57.6 per cent in the second half of the nineteenth century. A clearer picture emerges in the corresponding American data. There is a steady increase in finite uses between the first half of the nineteenth century and the beginning of the twentieth century and, compared with BrE, AmE consistently displays higher rates of the finite clause in each of the comparable nineteenth- to early twentieth-century databases. In addition, AmE does not parallel the striking, though short-lived, British revival of lest within the category of finite choices. It does, however, exhibit a markedly higher percentage of the zero option than BrE. As is apparent from Table 9.21 dealing with British and American newspapers, both regional varieties must have experienced a clear-cut decline of the finite options.
The changing status of that-clauses 175 Table 9.21 Finite and gerundial clauses after for fear in British and American newspapers
BrE AmE
t90, m93–95 W90–92, D92–95
I that-clause (that/Ø)
II of (Pro/NP) –ing (Pro/NP vs Ø)
III total
IV % that-clause
152 (94/58) 347 (115/232)
382 (19/363) 368 (6/362)
534 715
28.5% 48.5%
Table 9.22 The use of the emergent conjunction in (the) event (that) ‘if ’ and rival gerundial options in a series of historical databases
BrE
AmE
I finite (that/Ø)
II of (Pro/NP) –ing (Pro/NP vs Ø)
III % finite
EEPF ECF NCF1 MNC/B NCF2 (*1800–*1829) LNC/B NCF2 (*1830–*1869) ETC/B BNC: wridom1
– – 1 (1/0) – –
– – 19 (16/3) 39 (39/0) 99 (96/3)
– – 5% 0% 0%
– –
28 (27/1) 14 (14/0)
0% 0%
– 8 (7/1)
10 (10/0) 12 (12/0)
0% 40%
EAF1 EAF2 MNC/A LNC/A ETC/A
4 (4/0) – 2 (1/1) 5 (4/1) 10 (5/5)
47 (42/5) 38 (37/1) 5 (4/1) 22 (22/0) 7 (7/0)
7.8% 0% 28.6% 18.5% 58.8%
Even so, the contrast between BrE and AmE has been consolidated in several respects. In addition to the striking contrast involving the categories of finite and non-finite clauses, we note two more subtle kinds of divergence. Compared with BrE, AmE uses many more zero options, but fewer (complex) gerunds containing their own notional subjects. Considering the structures governed by in (the) event (NPs, gerunds and that-clauses), it has been shown by means of OED quotations that the that-clause was the last type of dependent element to become established (Rohdenburg 2013). This finding is corroborated in Table 9.22 by the evidence drawn from the available fictional databases. Apart from one solitary example in the first part of the NCF, the thatclause is not found in any of the British databases up to the beginning of the twentieth century. By contrast, AmE supplies four examples in the EAF, and in the ETC/A, towards the beginning of the twentieth century, the type is
176 Rohdenburg Table 9.23 The use of the emergent conjunction in (the) event (that) ‘if ’ and rivalling gerundial options in British and American newspapers
BrE AmE
t91–92, g91–92, d91–92, m93–95 L92, D92, W90
I that-clause (that/Ø)
II of (Pro/NP) –ing (Pro/NP vs Ø)
III total
IV % finite
93 (90/3)
380 (355/25)
473
19.7%
209 (79/130)
3 (3/0)
212
98.6%
already overtaking its immediate gerundial rival. The divergence must have continued throughout the twentieth century. As seen in Table 9.23 focusing on British and American newspapers from the 1990s, the gap between the two varieties has widened tremendously. Compared with gerunds, the that-clause in BrE is used in about 20 per cent of the total. In AmE, by contrast, gerunds have been virtually phased out at the expense of that-clauses. The consolidation of the that-clause in AmE is also mirrored in the relationship between finite examples with that expressed and those where it is omitted. Again, it is found that present-day AmE uses a markedly higher ratio of zero options than BrE. In the case of on/upon (the) condition, the evolution of that-clauses compared with rival gerunds and earlier infinitives is far from unidirectional. Table 9.24 shows that after peaking at 100 per cent in the EEPF, the type experiences a sharp decline to reach a low of 51 per cent in the second half of the eighteenth century. From then on, we can observe a reversal of the trend, with AmE taking the lead towards the middle of the nineteenth century. The trend must have continued in the twentieth century. The data in Table 9.25 leave no doubt that the development observable towards the beginning of the twentieth century has virtually gone to completion in both varieties. Again, it is seen that AmE uses more that-clauses with that unexpressed than BrE and possibly favours the simple gerund over the complex one containing its own notional subject. 9.6 Conclusion As hypothesised in the introduction, expressive and representative verbs have shown similar evolutionary trends to those characteristic of directive verbs. With both verb classes, the that-clause after direct objects has indeed restricted its range of application to a greater or lesser extent. Thus, the analysis of three of the six expressive verbs selected for this chapter (upbraid, reproach, blame) and of several others not detailed here (e.g. thank and praise) suggests that the that-clause has typically been phased out completely or become an obsolescent feature in non-stereotyped uses (see Section 9.2). However, there are at least three verbs where the that-clause has been
The changing status of that-clauses 177 Table 9.24 Finite, gerundial and infinitival clauses after on/upon (the) condition in a series of historical databasesa
BrE
AmE
I that-clause (that/Ø)
II of (Pro/ NP) –ing (Pro/NP vs Ø)
III IV to-infinitive total
V % thatclause
EEPF (17th cent.) ECF1 ECF2 NCF1 (*1728–*1759) NCF1 (*1760–*1799) MNC/B LNC/B ETC/B BNC: wridom1
51 (39/12)
2 (0/2)
3
56
91.1%
52 (42/10) 49 (37/12) 20 (16/4)
4 (2/2) 11 (7/4) 19 (14/5)
1 – –
57 60 39
91.2% 81.7% 51.3%
43 (30/13)
20 (11/9)
–
63
68.3%
34 (25/9) 80 (63/17) 8 (7/1) 28 (21/7)
20 (9/11) 16 (6/10) 1 (1/0) 2 (2/0)
– – – –
54 96 9 30
63.0% 83.3% 88.9% 93.3%
EAF1 (*1760–*1799) MNC/A LNC/A ETC/A
59 (46/3)
30 (16/14)
–
89
66.3%
28 (17/11) 56 (37/19) 8 (7/1)
4 (2/2) 2 (0/2) –
– – –
32 58 8
87.5% 96.9% 100%
a
The analysis disregards any examples containing elements between on/upon (the) condition and the preposition of with gerunds or between on/upon (the) condition that/Ø and the subject of finite examples.
Table 9.25 Finite and gerundial clauses after on/upon (the) condition in British and American newspapers
BrE AmE
t90–95 W90–92, L92
I that-clause (that/Ø)
II III of (Pro/ total NP) –ing (Pro/NP vs Ø)
IV % thatclause
633 (536/97) 496 (347/149)
2 (2/0) 4 (0/4)
99.7% 99.2%
635 500
maintained at LModE levels of usage (flatter, congratulate) or even been revived in the nineteenth and twentieth centuries (pride o.s.). But then there is little doubt that here the that-clause owes its unusual status to the fact that the verb is either obligatorily reflexive (pride o.s.) or has increasingly adopted this option along with other manifestations of reduced transitivity, such as the use of inanimate subjects and passives. Moreover, such low transitivity features are also typical of verbs like reproach and blame, which represent the evolutionary majority pattern. In addition, with several verbs, the decline of
178 Rohdenburg
the that-clause is reflected in the fact that it is increasingly found to contain subjects not co-referential with the object of the superordinate verb, a possibility which is not available to the (simple) gerund lacking its own notional subject. In the case of the five non-central representatives analysed in Section 9.3, it is seen that, with the exception of satisfy, all of them abandon the that-clause in the course of the LModE period. Crucially, any existing thatclauses after objectless uses (answer, certify, confirm) have not been affected by this tendency. Furthermore, the phasing out of the that-clause after verb + object may be accompanied by an increase in other low transitivity features, as can be clearly observed with the verb certify. The history of satisfy is particularly illuminating. Much more important than its decline in terms of its frequency of occurrence is the fact that there is a dramatic shift away from high transitivity in the shape of human subjects associated with nonreflexive objects to two low transitivity features, the use of reflexive objects and inanimate subjects. With inanimate subjects becoming less and less common, we can predict a time when the (active) that-clause will be invariably associated with reflexive pronouns. With directive and related verbs used without object (see Section 9.4), we find that the (well-established) that-clause in the passive is replaced to some extent by the raising structure (or AcI construction). The replacement process is seen to be much more extensive in BrE than AmE. What is more, in the past century, AmE has clearly shown a more pronounced tendency than BrE to revert to the that-clause. It seems possible that this reversal has been encouraged by the increasing popularity of the mandative subjunctive, whose revival in AmE can – at least in the passive – be traced as far back as the nineteenth century. Concerning the contrast between finite and gerundial options after phrasal structures such as in (the) event ‘if ’, the three items under scrutiny display very different evolutionary profiles (see Section 9.5). Yet in all cases, AmE is found to have displayed a distinctly greater affinity for the finite clause (or the conjunctional use) as long ago as the middle of the nineteenth century. This is also true of on/upon (the) condition, where the global contrast concerning the categories of finite and non-finite options between BrE and AmE has by now been obliterated by an ongoing trend towards the use of finite clauses. An earlier study (Rohdenburg 2008) evaluates the entrenchment of emergent and declining adverbial conjunctions – apart from their textual frequencies – in terms of the ratio of the finite zero option (or the that-clause with that left unexpressed) vis-à-vis its counterpart containing that. The lesson to be drawn from this chapter is that the assessment of the conjunctional status of an emergent or established conjunction should also (and perhaps primarily) be based on the overall ratio of all relevant finite clauses vis-à-vis any non-finite options. On this basis alone, we may conclude that, of the three items analysed, the conjunctional force appears
The changing status of that-clauses 179
to have been declining since the nineteenth century in one case (for fear), while with the two others (in [the] event, on/upon [the]condition) it has been consolidated. How do the findings in Sections 9.2–4 fit in with the widely held view that in their systems of clausal complementation all varieties of English have been affected since the eighteenth century by ‘a general tendency for nonfinite subordinate clauses to spread at the expense of finite ones’ (Mair 2005: 541)? In Sections 9.2 and 9.3 dealing with expressives and representatives, the assumption is directly supported only by the contrast between thatclauses and gerunds after expressive verbs. In the case of (transitive) representatives, the general lack of rival infinitival or gerundial constructions means that the demise or contextual contraction of the that-clause with the five verbs analysed cannot be compensated by an increase in non-finite alternatives. Moreover, the recent expansion of the that-clause at the expense of non-finite options after nouns like risk (not detailed in this chapter) and directive and related verbs used without object (see Section 9.4), in particular in AmE, would appear to provide two classes of clear counterexamples to the general trend. Appendix Table 9.26 Overview of corpora Corpus
Description
Word count
BNC = British National Corpus
1995. Version 1.0. BNC Consortium/Oxford University Computing Services 1990–2012. Created by Mark Davies, Brigham Young University on CD-ROM 1991–1995. Chadwyck-Healey/ProQuest on CD-ROM 1992–1995. Knight Ridder Information Inc. 2000. Chadwyck-Healey First part of the EAF containing only those authors born in the 18th century (*1744–*1799) Second part of the EAF containing only those authors born in the 19th century (*1801–*1827) 1996 Chadwyck-Healey
100,000,000 words
COCA = The Corpus of Contemporary American English d91–95 = Daily Telegraph and Sunday Telegraph D92–95 = Detroit Free Press EAF = Early American Fiction EAF1
EAF2
ECF = Eighteenth-Century Fiction ECF1
First part of the ECF containing only those authors born in the 17th century (*1660–*1699)
450,000,000 words 127,906,347 words 102,989,512 words 34,634,660 words
9,702,696 words, omitting duplicates
180 Rohdenburg Corpus
Description
ECF2
Second part of the ECF containing only those authors born in the 18th century (*1703–*1752) 1997–2000 Chadwyck-Healey In association with the Salzburg Centre for Research on the English Novel SCREEN 1996–1997. Chadwyck-Healey A selection of British and American writings by authors born between 1870 and 1894. Source: Project Gutenberg. Compiled in the Research Project ‘Determinants of Grammatical Variation in English’, University of Paderborn American writings in the ETC British writings in the ETC (including The Observer 1994– 1995) on CD-ROM 1990–1995. Chadwyck-Healey/ProQuest on CD-ROM 1993–1994. ProQuest on CD-ROM 1992–1995. Knight Ridder Information Inc. A selection of British and American writings (complementary to the EAF and the NCF) by authors born between 1830 and 1869. Source: Project Gutenberg. Compiled in the Research Project ‘Determinants of Grammatical Variation in English’, University of Paderborn American writings in the LNC British writings in the LNC A selection of British and American writings (complementary to the EAF and the NCF) by authors born between 1803 and 1829. Source: Project Gutenberg. Compiled in the Research Project ‘Determinants of Grammatical Variation in English’, University of Paderborn American writings in the MNC British writings in the MNC on CD-ROM 1993–1995.Chadwyck-Healey on CD-ROM 2001. ProQuest 1999–2000. Chadwyck-Healey
EEPF = Early English Prose Fiction EPD = English Prose Drama ETC = Early Twentieth Century Corpus
ETC/A ETC/B g90–95 = Guardian i93–94 = The Independent and the Independent on Sunday L92–95 = Los Angeles Times LNC = Late Nineteenth Century Corpus
LNC/A LNC/B MNC = Mid NineteenthCentury Corpus
MNC/A MNC/B M93–95 = Daily Mail and Mail on Sunday N01 = The New York Times NCF = Nineteenth-Century Fiction
Word count
9,562,865 words
26,454,639 words 16,351,681 words
11,550,273 words 4,801,408 words 177,317,814 words 74,903,107 words 320,016,164 words 47,677,728 words
26,859,926 words 20,817,802 words 17,347,730 words
7,264,854 words 10,082,876 words 60,902,168 words 52,132,979 words 37,589,837 words
The changing status of that-clauses 181 Corpus
Description
NCF1
First part of the NCF containing only those authors born in the 18th century (*1728 -*1799) Second part of the NCF containing only those authors born in the 19th century (*1800 -*1869) on CD-ROM1992 (Version 1.10). Edited by John A. Simpson and Edmund S.C. Weiner. Oxford: Oxford University Press on CD-ROM 1990–1995. Chadwyck-Healey/ProQuest (including Insight on the News 1990–1992) on CD-ROM1990–1992. Wayzata Technology fictional/narrative component of the BNC
NCF2
OED on CD-ROM = The Oxford English Dictionary (Second Edition) t90–95 = The Times and The Sunday Times W90–92 = The Washington Times wridom1
Word count
232,730,527 words 93,889,488 words
18,863,529 words total
10
Variability in verb complementation in Late Modern English: finite vs. non-finite patterns Hubert Cuyc kens, F r au k e D ’ h o e dt and Be ne dikt Szmre c s a ny i
10.1 Introduction In Present-Day English (PDE), the following alternation pattern with the verb remember is commonly attested: (1)
a. He remembered to thank her for everything. b. I remember reading about it in the newspaper. (Declerck
1991: 511) As can be seen, the verb remember may take a to-infinitive clause (1a) or a gerundial –ing-clause (1b) as clausal verb complement. Language users cannot choose freely between one or the other pattern; rather, the variation can be characterised as functional differentiation, whereby the to-infinitive encodes a situation which is not yet actualised at the time of remembering, whereas the gerundial –ing-clause encodes the situation as actualising before or at that time. However, in the alternation pattern in (2) with regret (that-clause vs. gerundial –ing-clause), preference for one or the other verbal complement is far less categorical; actually, it would appear that both variants are freely interchangeable. (2)
a. I don’t regret helping her start out. b. I don’t regret that I helped her start out.
It is this type of non-categorical or probabilistic variation, which is less well understood, that is the focus of this chapter. Our aim is to offer a corpus-based analysis of complement-clause variation (in particular, The research for this chapter was carried out during the first author’s research stay the English Department of the University of Santiago de Compostela. He would like to thank the Research Foundation Flanders (FWO) and the University of Leuven for granting, and financially supporting, his sabbatical leave. He also acknowledges the financial support of the Spanish Ministry for Science and Innovation and the European Regional Development Fund (grant HUM2007–60706/FILO, co-ordinated by Teresa Fanego). All authors are grateful for useful suggestions from the audience at the workshop on Late Modern English Syntax, organised by Marianne Hundt, University of Zurich, 19–20 August 2012, and for careful feedback from the volume editor and an anonymous reviewer. The usual disclaimers apply.
182
Finite vs. non-finite verb complementation 183
finite that-clauses alternating with non-finite gerundial –ing-clauses and to-infinitive clauses) with the complement-taking predicates remember, regret and deny. Focusing on the Late Modern English (LModE) period, we detail the changing patterns of complement choice over time. We also identify the various (changing) factors determining this complementclause variation, thus probing the multivariate and probabilistic nature of complement choice. Another aim is to examine to what extent changing/varying distributions of finite vs. non-finite complement clauses can inform more general hypotheses about complement choice. The chapter is organised as follows: Section 10.2 presents a short survey of clausal verb complementation, which serves as a backdrop to our study of complement-clause variation. Section 10.3 details the goals of this chapter, while Section 10.4 presents data and methodology. Section 10.5 is devoted to the multivariate and probabilistic analysis of finite vs. non-finite complement choice. In Section 10.6, we contrast the results of our analysis with earlier claims in the literature on (changing) complement-clause preferences. Section 10.7 summarises and details the contribution our study makes to the body of previous research. 10.2 A short survey of clausal verb complementation Clausal verb complementation (i.e. structures of the type [complement-taking predicate + to-infinitive clause/gerundial –ing-clause/that-clause/for … to-infinitive clause]) has been an important research topic within generative as well as cognitive-functional linguistic frameworks. Indeed, complementation phenomena have been a concern of generative linguists (Rosenbaum 1967; Bresnan 1970) since Chomsky’s seminal Aspects of a Theory of Syntax (1965), and important work in this domain has continued ever since (Bresnan 1979; Warner 1982; Chomsky 1986; Rizzi 1990; Radford 1997; Felser 1999). From the 1980s onwards, the scope of complementation research was expanded by functional-typological linguists (e.g. Givón 1980; Noonan 1985; Dixon 1991). While the generative literature on verb complementation has largely focused on synchronic, syntactic issues, such as the constituent structure of different complement types, synchronic research within the cognitive-functional tradition to date has been concerned with the question of how complement clauses are distributed over the various complement-taking predicates (or main/matrix verbs). Much of this cognitive-functional research is semantic in orientation and has been informed by Noonan’s observation that across languages ‘complementation is basically a matter of matching a particular complement type to a particular complement-taking predicate’ (1985: 90) (see Wierzbicka 1988; Duffley 1992, 1999; Langacker 1991; Achard 1998; Smith and Escobedo 2002). However, as De Smet (2013: 20–33) points out, a satisfactory synchronic account of complementation needs to envisage additional, non-semantic determining
184 Cuyckens, D’hoedt & Szmrecsanyi
principles, such as the role of information structure (Noël 2003), the horror aequi principle (Rudanko 2000; Vosberg 2003), the cognitive complexity principle (Rohdenburg 1995), social and regional stratification (Mair 2002, 2003) and register (Mindt 2000). Significantly, though, this synchronic work has tended to neglect the fact that the synchronic matches between complement-taking predicate (CTP) and complement clause (CC) can be subject to change over time. It is only in the last fifteen or twenty years that diachronic studies have appeared which present broad accounts of change and variation in complementation patterns in different periods of the history of English (see Fischer 1995; Fanego 1996, 1998; Rudanko 1998, 2006, 2010, 2012; Miller 2001; Los 2005; Rohdenburg 2006a; De Smet 2008, 2013), attesting to a distributional reorganisation of CCs over time.1 It has thus been shown that new complement types have made their way into the English language (e.g. the gerundial –ing-CC and the for … to-infinitive construction are relatively recent), and that long-existing types have spread to new CTP-contexts (e.g. the toinfinitive) or, conversely, have become increasingly restricted (e.g. the bare infinitive). These changes have led to competition between CC-types; in particular, in some CTP-contexts, one CC-type may have been replaced by another (e.g. with verbs of volition, the that-clause has largely been supplanted by the to-infinitive; see Croft 2000; Los 2005); in other contexts, a situation of variation may have come about. Sometimes, as in (1), this CC-variation results in functional differentiation, whereby the domain of operation of one CC-type can be differentiated in a clear-cut, categorical, fashion from that of the other. Often, though, this variation is non-categorical and unstable, in that CC-types co-exist (e.g. regret patterns with a that-clause a gerundial –ing-clause, and a to-infinitive; verbs of ‘like’ and ‘love’ alternate between the to-infinitive and the gerundial –ing-form) and show varying/shifting patterns of preference across speakers and over time. While replacement of CC-types as well as CC-variation characterised by functional differentiation have been well documented, the type of competition between CC-types resulting in unstable variation (or changing patterns of preference over time) – though attested in the literature – is still largely underexplored, and it is this non-categorical variation that this chapter is concerned with. 10.3 Goals In this chapter we wish to examine non-categorical or probabilistic variation with the verbs remember, regret and deny, as exemplified in (2) above and in (3): (3) 1
a. I do not remember my saying I knew the thief. (Old Bailey Corpus
(OBC), t17640222–37)
Most linguists, including those of the generative persuasion, are now agreed that the diachronic perspective is an important facet of a satisfactory account of complementation.
Finite vs. non-finite verb complementation 185
b. I don’t remember I said your Evidence would go before Askin’s.
(OBC, t17440912–48)
What characterises this type of variation is indeterminacy of choice; this indeterminacy does not occur at the abstract level of grammar – indeed, each of the CC-types with remember in (3) can be independently motivated (see, for instance, the semantic characterisation of CCs in Langacker 2008: 429–45) – but at the usage level, that is, at the level of actual, on-line choices speakers make in discourse (see De Smet 2013a: 27–29). Our concern is with these online choices: (i) we explore the patterns of CC-preference (in terms of frequency distributions) that emerge from actual usage, as observed in corpora, and (ii) by fitting a logistic regression model, we aim to isolate the factors significantly predicting, in terms of odds or probability ratios, this CC-variation. Finally, the frequency distributions observed and, importantly, the factors that turn out to be instrumental in predicting CC-choice feed into an assessment of (some of the) earlier claims made in the literature on (changing patterns in) CC-choice (see Section 10.2; De Smet 2013a: 20–33). Thus, we are able to explore to what extent CC-choice is affected (i) by Noonan’s semantic characterisation of CCs in terms of ‘dependent time reference’ vs. ‘independent time reference’ (1985: 92), and (ii) by various types of structural complexity consonant with Rohdenburg’s (1995, 1996) notion of ‘cognitive complexity’. Further, we examine whether any changes in CC-choice observed over time tie in with Denison’s claim that ‘a long-term trend in English has been the growth of nonfinite complement clauses at the expense of finite clauses’ (1998: 256), and whether any such increase of non-finite complements can inform our views on grammaticalisation processes. 10.4 Data and methodology 10.4.1 Data selection In our selection of matrix verbs, we chose verbs that allow variation between finite and non-finite complement clauses, and these are relatively few in number. We selected the verbs remember, regret and deny from Quirk et al.’s (1985: 1182–84) list of factual verbs which allow the finite/non-finite CC alternation.2 Finite complement clauses are introduced either by the complementiser that or by ‘zero’, as in (4a) and (4b), respectively: (4)
a. I remember perfectly well that it was at the prisoner’s suggestion.
(OBC, t-18720226–267)
b. I remember Boswell and Ausser were both at my house. (OBC,
t17670715–45)
2
Quirk et al.’s (1985) list also includes suasive and emotive verbs, which will not be considered here. Even with the suasive and emotive verbs, the number of verbs allowing both finite and non-finite CCs remains limited.
186 Cuyckens, D’hoedt & Szmrecsanyi
Non-finite complement clauses include (i) the subjectless (controlled) –ingCC (as in (5a)), (ii) the –ing-CC with expressed subject (5b), (iii) the subjectless to-infinitive-CC (5c); (iv) the to-infinitive-CC with expressed subject (5d). In addition, care was taken that matrix verbs occurred sufficiently frequently with the CC-variants to allow statistically significant results. (5)
a. Do you remember at any time going to the prisoner’s house?
(OBC, t17730217–52)
b. Do you remember a green cart coming up? (OBC, t17970920–62) c. I do not remember ever to have heard a word from you before.
(1810–1813, Corpus of Late Modern English Texts (extended version) (CLMETEV), George Byron, Letters) d. I do remember this circumstance to have happened but to one man. (OBC, t17950218–46) Data were extracted from two corpora covering the entire LModE period: the Old Bailey Corpus (OBC, Huber et al. 2012), which contains court transcripts and is therefore ‘speech-based’ (see Culpeper and Kytö 2010: 16–17)3 and the Corpus of Late Modern English Texts (extended version) (CLMETEV), which mainly contains formal prose.4 We extracted all attestations of the matrix verbs (including those with NP complements). After manual pruning of all spurious hits, we were left with the following number of [CTP–CC] patterns: • remember: 4,371 observations • regret: 280 observations • deny: 1,138 observations Observations were then coded, as described below. 10.4.2 Coding of the data Each corpus attestation, consisting of [subject + remember/regret/deny + CC], was entered into an Excel database under the descriptor observation.5 Each token was then coded for a number of factors determining CC-choice and describing characteristics of the complement-taking predicate/matrix verb, 3
Culpeper and Kytö (2010: 17) classify trial proceedings as ‘speech-based’: ‘“Speech-based” genres, such as Trial proceedings, are those that are based on an actual “real-life” speech event. There is no claim here that such genres involve the accurate recording of a speech event. In the absence of audio or video recording equipment or even full systems of shorthand, most speech-based texts are reconstructions assisted by notes.’ 4 A new version of the corpus, CLMET 3.0, was not yet available when the research for this chapter was carried out (see https://perswww.kuleuven.be/~u0044428/clmet.htm). 5 Each observation received the following additional descriptors: (i) n: number of the attestation in the Excel table; (ii) concordance: the entire concordance comprising the corpus attestation (in order to provide the necessary context); (iii) lemma: remember, regret or deny; (iv) year: year of the attested example; (v) file: corpus in which the example was attested.
Finite vs. non-finite verb complementation 187
the complement clause or the combined [CTP–CC] structure. Importantly, our selection of potentially significant factors of CC-variation was informed by the relevant literature (see Section 10.2; De Smet 2013a: 20–33) and comprises semantic, structural, genre-related and periodisation-related factors; it also includes a number of factors not typically discussed in the literature.6,7 10.4.2.1 Semantic factors The following semantic factors were coded for: (i) the meaning of the matrix clause, (ii) the meaning of the complement clause and (iii) [ctp– cc] time 8 reference. (i) The meaning of the matrix verb (verb meaning) The values making up this factor are distinguished at the level of the lemma’s submeanings; not all submeanings of a verb will show the same CC-pattern.9 For remember, three meanings/values were distinguished: • ‘remember1’ = ‘recall’, ‘bring to mind’. This meaning is typically exemplified in (6a), where the subject recalls a past event/action/situation. In addition, one can also recall/bring to mind something that is a general truth or a situation still holding at the time of the matrix verb, as in (6b); or one can recall that one still has to do something, as in (6c) • ‘remember2’ = ‘not to forget to do something’, as in (7) • ‘remember3’ = ‘keep in mind’, as in (8). (6)
a. I remember a detective coming and speaking to me. (OBC,
t-18811212–125)
b. The Prisoner said in his Defence, that coming in a Chaise to the
Vine Tavern, and remembering his Pistols were Charged, he went to discharge them ... (OBC, t17211206–42)
6
Information-based factors (as discussed in Noël 2003) were not included, as the present study focuses on [CTP–CC] patterns only. Neither was horror aequi (Rudanko 2000; Vosberg 2003) or factors relating to geographical or social stratification (Mair 2002, 2003). 7 The information in small caps indicates the factor as it was labelled in the Excel database. 8 Our set of semantic factors potentially determining CC-choice does not include such dichotomies as ‘internal’ vs. ‘holistic’ (meant to capture the variation between gerundial -ing-clauses and to-infinitives; see Smith and Escobedo 2002; Langacker 2008: 439) or ‘occurrence/process’ vs. ‘proposition/grounded process’ (meant to capture the variation between non-finite and finite CCs; see Langacker 2008: 441–42). These dichotomies are meant to account for the distribution of CCs in a categorical fashion. Moreover, they are very general and therefore hard to operationalise; it would seem, then, that they only play a minor role, if any, in speakers’ on-line choice of CCs. 9 Careful inspection of the database of [CTP–CC] tokens shows that not all meanings of the matrix verbs allow finite vs. non-finite variation. As such, in our analysis, we will not consider the tokens with semantic values ‘remember2’ and ‘remember3’, which exclusively take a toinfinitive and that-clause, respectively. We will return later to these sets (Section 10.6), and try to account for the fact that remember in these meanings exclusively takes one type of CC only.
188 Cuyckens, D’hoedt & Szmrecsanyi
c. (All of a sudden), he remembered that he (still) had to do the
dishes. (non-attested example)10 (7) Remember to lay your Spitchcot Eels near the edge of the Dish. (1732, CLMETEV, Richard Bradley, The Country Housewife and Lady Director) (8) a. Remember, you are upon your oath. (OBC, t17980704–44) b. remember he is come of gentle blood, … (1824, CLMETEV, Catherine Gore, Theresa Marchont) For regret, two meanings were distinguished: • ‘regret1’ = ‘to feel sorry about something one has done and that one should have done differently or about a state of affairs one is involved in or responsible for and that one wishes was different’, as in (9) • ‘regret2’ is a more ‘polite’ use of regret where the speaker says that he/ she is sorry or sad about a situation, usually one that he/she is not directly responsible for, as in (10). (9)
a. She regretted having made the appointment. (1891, CLMETEV,
George Gissing, New Grub Street)
b. He seemed to regret that he had not finished the boy. (OBC,
t18980913–621)
(10) a. I much regret your leaving my office in such a hurry yesterday.
(OBC, t18960323–332)
b. We regret the envelopes are not the same as per specimen sent
you. (OBC, t19100426–40)
Deny has been interpreted as follows: ‘to say that you did not do something that someone has accused you of doing, or to say that something is not true’. (11) a. He denied being charged with robbery. (OBC, t-18680504–460) b. He denied that any bad money was found upon him. (OBC,
t-18730407–314)
(ii) Meaning of the complement clause (meaning CC) The complement clause may denote a state, as in (12), or an event/action, as in (13). (12) She had very much regretted being from home. (1811, CLMETEV,
Jane Austen, Sense and Sensibility)
(13) In short, to sum you up, you have all the makings in you of an ideal
wife combined with faults sufficient to make a Socrates regret he’d
10
This meaning was not attested in our corpus examples but is included here for the sake of completeness.
Finite vs. non-finite verb complementation 189
ever married you. (1909, CLMETEV, Jerome K. Jerome, They and I) (iii) [ctp–cc] time-reference This factor designates the relation between the time reference of the complement and the meaning of the CTP, and has the following values: • ITR (independent time reference) indicates that ‘the time reference of the complement … is in no way logically bound by the time reference of the CTP’ (Noonan 1985: 92); as such, deny in (14) combines with CTPs with anterior (14a) as well as simultaneous time reference (14b). (14) a. Prisoner denied that he knowingly bought stoled Goods. (OBC,
t17210301–56)
b. He absolutely denied that he knew any thing of the matter. (OBC,
t17350911–14)
• DTR (dependent time reference) indicates that the time reference of the complement ‘is a necessary consequence of the meaning of the CTP’ (Noonan 1985: 92), as in (15). (15) Remember to keep your Fire gentle and clear. (1732, CLMETEV,
Richard Bradley, The Country Housewife and Lady’s Director)
10.4.2.2 Structural factors The following set of factors pertain to structural complexity: (iv) Type subject CC This factor has the following two values: • ‘complex NP’, comprising [noun + postmodifier] and [noun + noun] • ‘other’, comprising the following types of subject: pronoun, noun, there, (pro)noun in the genitive, proper name and no subject (when no subject is expressed in control environments). (v) complexity CC The CC may show the following types of predicate structure: • verbs without argument/modifier: ‘V’ • verbs with one argument/modifier: ‘Vobjadv’ • verbs with an argument + modifier, or two arguments, or two modifiers: ‘Vobj+adv’. (vi) Intervening material in words This indicates the number of words between the CTP and the subject of the complement (when the complement is finite) or between the subject and the non-finite verb form.
190 Cuyckens, D’hoedt & Szmrecsanyi
10.4.2.3 Genre-related factor The following genre-related factor was distinguished: (vii) medium The attestation belongs to the ‘speech-based’ or the ‘written’ genre. Attestations from the CLMETEV, which contains formal prose, were classified as ‘written’, and attestations from the OBC as ‘speech-based’. 10.4.2.4 Periodisation-related factor Periodisation is captured by the following factor: (viii) period This factor locates the attestation within two time bands of the LModE period: • early (between 1710 and 1780) • late (between 1781 and 1920). We first experimented with three 70-year time bands (1710–80, 1781–1850, and 1851–1920), which are the time bands used in the CLMETEV corpus. As it turned out that the difference between the middle and the late periods was generally not significant, we collapsed the middle and the late period. 10.4.2.5 Additional factors Finally, the following additional factors were distinguished. (ix) The type of subject of the main clause (type clause) The subject of the main verb may be: • • • • •
subject main
pronoun1 = first-person pronoun pronoun2 = second-person pronoun pronoun3 = third-person pronoun noun none (when an explicit main clause subject is missing, as in Remember his giving away every thing when he set out upon his grand expedition (1794, CLMTEV, William Godwin, The Adventures of Caleb Williams)).
(x) CC subject animacy The subject may be: • animate • inanimate. (xi) voice verb CC Three values are distinguished: • active • passive • copular.
Finite vs. non-finite verb complementation 191
(xii) denotation Two values are distinguished: • ‘different’: main clause subject and CC-subject denote different entities (as in (16)) • ‘same’: main clause subject and CC-subject denote same entities; in this condition, the CC-subject may be controlled by the matrix subject (as in (17a)) or not (as in (17b)). (16) Do you remember a gang coming there in Oct. 1746? (OBC,
t17500711–31)
(17) a. Debbenham
denied ever knowing Mr. Athill. (OBC, t17970712–71) b. He denied his ever giving such an Order to the Prisoner. (OBC, t17280228–45)
(xiii) temporal relation This indicates the temporal relation between the CC and the time of the CTP: • anterior (18a) • simultaneous (18b) • posterior (18c). (18) a. I remember a detective coming and speaking to me. (1881, OBC,
trial id = t-18811212–125)
b. My lady and Miss Rachel regret that they are engaged, Colonel.
(1868, CLMTEV, William Collins, The Moonstone)
c. Remember that I shall see you in the summer. (1746–71,
CLMTEV, Philip Chesterfield, Letters to his Son)
(xiv) Complementation type Four types of CC are distinguished: • • • •
–ing-CC (includes present as well as perfect gerundials) to-infinitive-CC (includes present as well as perfect infinitives) that-clause zero-complementiser clause (a that-clause without a that-complementiser).
10.5 A probabilistic analysis of complement choice 10.5.1 Charting the distribution of CCs in LModE Table 10.1 presents the distribution of finite vs. non-finite CCs with the CTPs remember, regret and deny across our two LModE subperiods. It can be seen that the aggregate frequencies show an increase in non-finite
192 Cuyckens, D’hoedt & Szmrecsanyi Table 10.1 Distribution of finite vs. non-finite CCs with the CTPs remember, regret and deny across the two LModE subperiods
Early (1710– 1780) Late (1781– 1920)
Totals
remember
regret
deny
nonfinite
446
49.12%
4
33.33%
186
53.75%
636
50.24%
finite nonfinite
462 2,206
50.88% 76.01%
8 75
66.66% 28.1%
160 433
46.25% 54.67%
630 2,715
49.76% 68.53%
finite
696
23.99%
192
71.9%
359
45.33%
1,247
31.47%
vs. finite CCs: from a slightly higher share of non-finite CCs in the Early period (50.24% of all CCs) to a substantially higher share in the Late period (68.53%). For the individual verbs, however, distributions vary: while there is an marked increase in non-finite CCs with remember (from a 49.12% share of non-finite CCs in the Early period to 76.01% Late period), it is finite CCs which outnumber non-finite CCs with regret, while deny shows stable variation. 10.5.2 What determines the choice of non-finite CCs? It can be observed that none of the factors/values proposed uniquely conditions one outcome or the other (that-clause or non-finite clause). Consider in this respect (19) and (20). While the condition denotation = ‘same’ triggers an –ing-CC in (19), it is also a condition that allows a that-CC, as in (20). (19) I remember on one occasion asking Judd if Mr. Vyse was in. (OBC,
t-18741123–39)
(20) I remember when I was a boy myself I was once asked to take tea at
a girls’ school. (1903, CLMETEV, Samuel Butler, The Way of All Flesh)
Similarly, the choice of an –ing-CC in (21) can be hypothesised to be attributed to its simple VP structure, consisting of a verb only, without an additional argument or modifier. However, attestations such as (22) show that the condition complexity CC = ‘V’ is not sufficient to predict the –ing-CC. (21) I don’t remember any other persons calling. (OBC, t-18710227–190) (22) I do not remember he ran. (OBC, t17700221–19)
Conversely, that-clauses can be hypothesised to be associated with the condition denotation = ‘different’ (i.e. different denotation of matrix subject
Finite vs. non-finite verb complementation 193 Table 10.2 The distribution of finite vs. non-finite CCs for the factor intervening material in words Complementation type
0 words
> 1 words
finite non-finite
1,521 3,106
329 245
Table 10.3 The distribution of finite vs. non-finite CCs for the factor meaning cc Complementation type
event/action
state
finite non-finite
1,044 2,441
833 910
and CC-subject) and with a complex VP structure in the CC (condition complexity CC = ‘Vobj+adv’), as exemplified in (23): (23) a. I do not remember that he has made my payments before. (OBC,
t-18610408–350)
b. I don’t remember that I saw Miles Barne at the fire. (OBC,
t17670115–32)
However, our data show that each of these conditions can also be observed in non-finite –ing-CCs, as in (24): (24) a. I remember a detective coming and speaking to me. (OBC,
t-18811212–125)
b. Do you remember asking the prisoner after breakfast to watch while
you got to the bottom of the street ... (OBC, t-18360404–919)
In other words, the factors identified above are predictors11 of the nonfinite as well as the finite CC-types following remember/regret/deny. In particular, each condition within a particular factor/predictor favours nonfinite as well as finite CCs. Importantly, these patterns of preference vary across factors; in other words, one factor’s relative weight or impact on the choice of finite vs. non-finite CC-type may differ from another. Compare Table 10.2 representing the distribution of finite vs. non-finite CCs for the factor intervening material in words with Table 10.3 representing this distribution for the factor meaning cc. In Table 10.2, the condition intervening material = ‘0 words’ has a preference for non-finites whereas the condition intervening material = ‘>1 words’ has a preference for finites. 11
In a statistical model (such as the one described below), the factors triggering a particular outcome are said to have a predictive value (see Gries 2012).
194 Cuyckens, D’hoedt & Szmrecsanyi Table 10.4 The distribution of finite vs. non-finite CCs for the factor temporal relation Complementation type
anterior
simultaneous
finite non-finite
1,498 3,000
244 306
In Table 10.3, non-finites are preferred over finites in both conditions, but there is a greater preference for non-finites in the condition meaning cc = ‘event/action’. At first sight, then, it would appear that frequency distributions might be an indicator of differential impact of the factors, in other words, of the relative weight of the conditioning factors. However, when comparing Table 10.3 with Table 10.4, it can be seen that frequency distributions for two factors can also be fairly similar. Indeed, Table 10.3 shows that in the condition meaning CC = ‘event/ action’, there is a higher probability of the non-finite CC being chosen than when the CC denotes a state (70% vs. 52% preference for non-finite). Table 10.4 shows a similar distribution in that the condition temporal relation = ‘anterior’ favours non-finite in 66% of the cases and temporal relation = ‘simultaneous’ in 55% of the cases. As such, it is not clear to what extent the higher proportion of non-finite vs. finite CCs is to be attributed to the ‘event/action’ meaning of the CC or to the ‘anterior’ meaning of the CC; in other words, it is not clear what the relative strength of either condition is. In light of this, it would appear that frequency distributions alone cannot sufficiently inform us about the relative strength of each of the variables/ factors. In a situation such as this, where we are dealing with probabilistic complementation choice in a context where various factors are at play, it is advisable to fit a regression model. To probe the multivariate and probabilistic nature of complementation strategy, we fitted a binary logistic regression model with mixed effects (Pinheiro and Bates 2000).12 Logistic regression modelling is the closest a corpus analyst can come to conducting a controlled experiment: the technique models the combined contribution of all the conditioning factors considered in the analysis, systematically testing the probabilistic effect of each factor while holding the other factors in the model constant. Ideally, the factors should be independent of each other, but in practice a certain degree of overlap (in technical parlance, collinearity) is admissible. The random effects in the model take care of non-central and not necessarily repeatable factors such as author idiosyncrasies, register variation and differential 12
We use the lme4 package in the statistical software package R (R Development Core Team 2011).
Finite vs. non-finite verb complementation 195
propensities of different verb lemmas and meanings to co-occur with either complementation type. Our response variable is binary, distinguishing between finite complementation (that + zero) and non-finite complementation (all other observations). Predicted odds are for non-finite complementation. As was pointed out earlier, right at the outset, the ‘remember2’ and ‘remember3’ observations were excluded due to lack of variability. We observed the customary steps to obtain a minimal adequate regression model: we began by fitting the maximal model including all potentially important language-internal factors. Subsequently, the model was simplified by removing factors lacking significant explanatory power (for instance, we removed the predictor [ctp– cc] time-reference, because turned out not to have a significant effect). The resulting model is of acceptable quality.13 We begin by briefly addressing the three random effects in the model. medium (variance: 1.77, standard deviation: 1.33) accounts for most of the variance not accounted for by the fixed effects in the model. Specifically, speech-based texts favour non-finite complementation, while written texts favour finite complementation. As for verb lemmas and verb meanings (variance: 0.85, standard deviation: 0.92), we find that ‘remember1’ favours non-finite complementation, while the other verb lemmas and meanings distinguished in the present study (‘regret1’, ‘deny1’ and ‘regret2’ – in that order) disfavour it. Finally, speaker/writer idiosyncrasies – approximated here by modelling corpus file ID as a random effect – is only of minor importance (variance: 0.62, standard deviation: 0.79). Table 10.5 summarises the fixed main effects in the model. The figures reported in the table are so-called odds ratios (henceforth: ORs), which quantify the magnitude and the direction of the effect of each factor on the odds that non-finite complementation is chosen. In the case of categorical predictors (such as meaning CC), ORs measure an effect against the backdrop of an (arbitrarily chosen) default value; for continuous predictors (such as intervening material in words), ORs measure the effect of one-unit increases. Because odds ratios can take values between 0 and ∞, two cases can be distinguished: 1. OR < 1, the predictor makes the outcome ‘non-finite complementation’ less likely 2. OR > 1, the predictor makes the outcome ‘non-finite complementation’ more likely. 13
Somers Dxy, a rank correlation coefficient between predicted outcome probabilities and observed binary outcomes, is 0.81, which indicates that the model discriminates fairly between complementation types. The model correctly predicts 84% of all complementation outcomes in the data set, which is a significant increase over baseline (64.1%) predicting the overall most frequent value. The model’s condition number κ is 21.7, which is another way of saying that the model exhibits moderate though not harmful (Baayen 2008: 182) collinearity.
196 Cuyckens, D’hoedt & Szmrecsanyi Table 10.5 Fixed effects (main effects only) in the minimal adequate mixedeffects logistic regression model. Predicted odds are for non-finite complementation. Odds ratio (OR) (Intercept) meaning CC
state
(default: ‘event/action’)
not significant 0.36
***
0.73
***
3.57 4.26 11.06 5.09
*** *** *** ***
0.56
*
0.18
***
0.47
***
1.90 2.25
*** ***
0.37 not significant
***
intervening material in words
1-unit increase
type subject main clause
noun pronoun1 pronoun2 pronoun3
type subject CC
complex NP denotation
different
(default: ‘none’)
(default: ‘other’)
(default: ‘same’)
CC-subject animacy
inanimate
complexity CC
Vobj+adv Vobjadv
(default: ‘V’)
temporal relation
anterior posterior
(default: ‘animate’)
(default: ‘simultaneous‘)
(default: ‘active’) copular passive period (default: ‘late’) early voice verb cc
not significant 2.15
***
0.50
**
N = 5,228. Significance codes: * significant at p < 0.05, ** p < 0.01, *** p < 0.001.
Thus we find that meaning CC has a robustly disfavouring effect on complementation choice: if the value is ‘state’ and not ‘event/action’, the odds for non-finite complementation decrease by a factor of 0.36 (i.e. by 64%). Intervening material in words also discourages non-finite complementation (for every additional word, the odds decrease by a factor of 0.73, i.e. by 27%), and so do complex NP subjects of complement clauses (type subject CC): if the subject is complex, the odds for non-finite complementation decrease by a factor of 0.56. Also disfavouring is (i) the denotation = ‘different’ condition (OR: 0.18), (ii) an inanimate CC-subject (CC-subject animacy) (OR: 0.47), (iii) a temporal relation of anteriority (as opposed to simultaneity) between matrix and CC clause (OR: 0.37) and (iv) the early period (OR: 0.50), which spans the years between 1710 and 1780. This is another way of saying that non-finite complementation is more common after 1780 than before 1780 (and that the other main effects sketched in Table 10.5 describe the late period, thanks to the default level of period being set to ‘late’).
Finite vs. non-finite verb complementation 197 Table 10.6 Interactions between language-internal constraints and real time period in the minimal adequate mixed-effects logistic regression model (note: only significant interaction terms and predictor levels are shown). Predicted odds are for non-finite complementation. odds ratio (OR) … involving type subject CC complex NP * period=‘early’ … involving denotation different * period=‘early’ … involving temporal relation anterior * period=‘early’
4.01
**
1.47
*
0.23
***
N = 5,228. Significance codes: * significant at p < 0.05, ** p < 0.01, *** p < 0.001.
Which factors favour non-finite complementation? For one thing, with respect to type subject main clause, all categories of the factor favour non-finite complementation with respect to the default ‘none’ – most so ‘pronoun2’, which increases the odds for non-finite complementation by a factor of about 11, and least so ‘noun’, which increases the odds for nonfinite complementation by a factor of ‘only’ 3.57. Second, consider complexity CC: in comparison to CC verbs without argument/modifier, both verbs with one argument/modifier (‘Vobjadv’) and verbs with an argument + modifier, or two arguments, or two modifiers (‘Vobj+adv’) favour nonfinite complementation, with ORs of about 2. Third, vis-à-vis voice verb CC, the value ‘passive’ increases the odds for non-finite complementation by a factor of 2.15. Finally, we turn to the issue of how the probabilistic effect of various conditioning factors has changed during the LModE period. To this end, Table 10.6 reports significant interaction terms between real time period and three language-internal constraints. (Note that the figures reported in the table are factors by which we have to multiply the main effects in Table 10.5 to gauge their effect in the early period.) First, observe that according to Table 10.5, type subject CC = ‘complex NP’ disfavours non-finite complementation at 0.56. What we learn from Table 10.6 is that in the early period, type subject CC = ‘complex NP’ actually favoured non-finite complementation, at 0.56 × 4.01 = 2.25. Put differently, type subject CC has reversed its effect in the course of the LModE period. Second, denotation = ‘different’ disfavours non-finite complementation throughout, but more so in the late subperiod than in the early subperiod. Thus, the denotation constraint is subject to a diachronic strengthening in LModE. Third, temporal relation = ‘anterior’ disfavours non-finite complementation throughout, but more so in the early period
198 Cuyckens, D’hoedt & Szmrecsanyi
than in the late period. The temporal relation constraint is hence subject to a diachronic weakening in LModE. The main results from our regression analysis can be summed up as follows: • The only semantic factor significantly predicting non-finite vs. finite CCs is whether the CC expresses a state or an event. • Structural factors play an important role in determining CC-type: the factors intervening material in words and type subject CC each significantly disfavour non-finite complementation. • The speech-based genre is more favourable to non-finite comple mentation. • Over time, there appears to be a significantly higher preference for nonfinite CCs. An additional interesting result is that the condition denotation = ‘different’ disfavours non-finite complementation. This is not all that surprising because in 72.5 per cent of the cases (1,845/2,548) that meet the denotation = ‘same’ condition, the CC-subject is controlled by the matrix subject, that is, it is left unexpressed, and as such requires a non-finite complement. In line with these findings, the condition CC-subject animacy = ‘inanimate’ also disfavours non-finite complementation; indeed, inanimate CC-subjects, on the whole, differ from the matrix subject, which is a condition disfavouring non-finite complementation. In our regression model, the factor [ctp–cc] time reference turned out not to have a significant effect on CC-choice. This does not mean, however, that the frequency distributions associated with this factor are uninformative (see Section 10.6). 10.6 Interpretation and discussion In this final section, we explore how the results of this study bear on earlier hypotheses about CC-variation. Let us first consider how the factors relating to structural complexity impact Rohdenburg’s complexity principle. In various publications, Rohdenburg (1995, 1996) claims that CC-variation can be accounted for in terms of the ‘cognitive complexity’ principle. This principle states that in a cognitively more complex environment, more explicit grammatical options tend to be favoured (see Rohdenburg 1996: 151). A cognitively more complex environment, in turn, is hypothesised by Rohdenburg as requiring more processing effort. Grammatical manifestations of cognitive complexity are, for instance, constituents of sizeable length and great distance between head and dependent (notably because of insertions), in other words, structurally complex entities or environments. If we take structural complexity as a measure of cognitive complexity, it follows that these structurally complex entities/environments will tend to
Finite vs. non-finite verb complementation 199
favour the more explicit finite that-CCs (and less complex environments the less explicit non-finite CCs, namely gerundial –ing-clauses and to-infinitives) – that-CCs can be seen as more explicit grammatically because they carry tense/mood distinctions and obligatorily have an expressed subject. Our analysis has shown that the factors intervening material and type subject CC, each associated with structurally complex environments, significantly disfavour non-finite complementation. As such, our results strengthen Rohdenburg’s earlier findings: they are not only consonant with his observations (with respect to the variation to-infinitive vs. that-clause) that intervening material has a disfavouring effect on non-finite complementation, but they reveal additional complexity factors having a similar disfavouring effect. At the same time, however, it has been seen that complex CC predicates (complexity CC) in fact favour non-finite CCs (and to the extent that passive verbs can be considered to be more complex than active verbs, a similar favouring effect can be observed). In other words, Rohdenburg’s proposed disfavouring effect cannot be generalised to all structural complexity factors, and in that sense, Rohdenburg’s complexity principle does not apply as generally as commonly held. Second, Rohdenburg (1995) finds that in the course of the seventeenth and eighteenth centuries, that-clauses are increasingly replaced by to-infinitives. In a recent study on the evolutionary path of that-clauses vs. gerundial –ingclauses, however, Rohdenburg (Chapter 9, this volume) sketches a more qualified picture, in that any replacement of that-clauses by gerundial –ing-clauses varies with individual matrix verbs. Now, even though the focus of our study is on shifting/changing patterns of preference rather than on the phasing out of that-clauses (as in Rohdenburg’s studies), our results are largely in line with Rohdenburg’s observations.14 Indeed, developments for the individual verbs may vary (as can be inferred from the effects of the random factor verb meaning, i.e. differential propensities of the different verb meanings with either CC-type), but aggregate figures for the matrix verbs show an increase of nonfinite CCs vs. finite CCs (see Section 10.5). The question, then, is whether the results of our investigation can be tied up with results from grammaticalisation research. If indeed, there is an increase of non-finite complement clauses relative to that-CCs, the question is whether this path of development – from a relatively loose concatenation of [matrix verb + that-clause] to a to-infinitive or –ing-CC showing more unified/bonded clause combining – can be considered as a case of grammaticalisation? The question is justified in the light of Lehmann’s (1995) characterisation of grammaticalising structures as losing autonomy. Further, it seems 14
Denison has made a similar claim (‘A long-term trend in English has been the growth of nonfinite complement clauses at the expense of finite clauses’ (1998: 256); this claim is, however, not based on extensive corpus study).
200 Cuyckens, D’hoedt & Szmrecsanyi
to tie in with Lehmann’s (1988) and Hopper and Traugott’s (2003) suggestion that clause linkage and, in particular, the development of complementation patterns can also be seen as a grammaticalisation process, in that verb complements may be subject to ‘decategorialisation’ (in a way similar to the decategorialisation of lexical items) when being integrated with a main clause. As such, verb complements may show a path of development ‘from a loose, paratactic concatenation via syntacticization into non-finite embedding’ (Givón 1979: 214), whereby the non-finite shows a reduction in tense-aspect morphology and lack of subject agreement, may (at least for English) have become reduced to a to-infinitive or gerundive –ing-clause and may thus show more unified/bonded clause combining (see Hopper and Traugott 2003: ch. 7). While an increase, over time, in the percentage of non-finite CC tokens over that-CC tokens always represents a path of gradually stronger integration of the complement clause into the matrix, it is at first sight doubtful that this shift from a finite (looser) that-CC to a more bonded non-finite CC should be seen as a case of grammaticalisation. For one, the non-finite and finite structures have different origins; as such, one might want to talk of replacement of the finite CC by a non-finite CC rather than of grammaticalisation through decategorialisation (see Fischer 2007: 221–22). Further, a decategorialised, that is, reduced, clause is still as much a grammatical structure as a full clause; in other words, decategorialisation here does not correlate with loss of lexical characteristics. What may, to some extent, salvage the idea of grammaticalisation is that, as the finite clauses are gradually being replaced by non-finite clauses, as can be seen with ‘remember1’ (see Table 10.1), the non-finite becomes gradually more obligatory. In other words, the choice between the members of a paradigm becomes more constrained, in that once the speaker has decided on a particular matrix verb with a particular meaning, the choice of the CC becomes more obligatory (for a similar view, see Haspelmath 1989: 298; see also Lehmann 1995 on the loss of paradigmatic variability).15 Additional cases of loss of paradigmatic variability can be observed in the alternation between non-finites whose subject is controlled by the matrix subject (and hence, whose subject is the same as the matrix subject and remains
15
In fact, the use of the notion of ‘obligatoriness’ is somewhat surprising, and may even seem contradictory, in a study of non-categorical variation. However, to the extent that the preference for non-finites with a particular matrix verb becomes more pronounced over time, one can talk about ‘increasing obligatoriness’, and hence ‘increasing grammaticalisation’. So, in our study of non-categorical variation, there seems to be some categorical distribution of variants that would provide an argument in favour of grammaticalisation, but other instances that argue against it. Rather than treat this issue as verb meanings grammaticalising with a particular CC-type, one might more successfully deal with it as constructionalisation (see e.g. Trousdale 2012; Traugott and Trousdale 2013) (thanks are due to Marianne Hundt for raising this point).
Finite vs. non-finite verb complementation 201 Table 10.7 The distribution of non-finites under the condition denotation = ‘same’ Early controlled subject identical, non-controlled subject
310 51
Late 85.87% 14.13%
1,532 28
98.2% 1.8%
nexpressed) and non-finites whose subject is the same as that of the matrix u subject (but is expressed). An example is (25): (25) a. I remember my going into the House. (OBC, t17370114–1) b. Do you remember at any time going to the prisoner’s house?
(OBC, t17730217–52)
Table 10.7 shows that the relative proportion of non-finites with an expressed subject that is identical with that of the matrix becomes increasingly smaller: in other words, these types of expressed subject are gradually ousted from the paradigm of CC-subjects. In sum, the analysis of change and variation with remember, regret and deny shows that the grammaticalisation from lexical items/constructions cannot simply be transferred to processes involving sentential/clausal units (in particular to verb complementation and clause linkage). Evidently, further analysis of several sets of matrix verbs is needed to reach more definitive conclusions. As such, the results found in this study seem at least to tie in with Fischer’s uneasiness about the (alleged) grammaticalisation path from that-clauses to non-finite to-infinitives with verbs of volition (see Fischer 2007: 221). Finally, we will explore whether our observed preferences in CC-choice can be usefully linked up with Noonan’s hypothesis about reduced complement clauses. Noonan (1985: 100–01) points out that ‘reduced complements, which are likely to lack tense distinctions … are typically associated with predicates whose complements have DTR [dependent time reference]’. As such, he puts forward that ‘infinitives … are frequently restricted to DTR contexts … [and that] indicative complements are normally excluded from DTR contexts since they are typically coded for tense, and therefore the expression of tense in such cases is redundant’. A similar view is expressed by De Smet (2012b: 142). Noonan’s and De Smet’s views are supported by our study. While the factor [ctp-cc] time reference did not have a significant effect on CC-choice, the frequency distribution associated with this factor may be instructive. The matrix verb ‘remember2’ = ‘not to forget to do something’ (which was left out of the analysis because it does not show CC-variation), specifies a temporal domain in the future with respect to the time of the matrix verb, and categorically takes a to-infinitive.
202 Cuyckens, D’hoedt & Szmrecsanyi
(26) Remember to make a great difference between companions and
friends. (1746, CLMETEV, Philip Chesterfield, Letters to his Son)
Then again, the matrix verb ‘remember3’ = ‘keep in mind’ (which was left out of the analysis as well) combines exclusively with a finite that-clause and has ITR complements (i.e. complements may refer to anterior, simultaneous or posterior time domains with respect to the time of the matrix). (27) a. ‘anterior’: Remember, I’ve known and liked him steadily for nearly
three years. (1910, CLMETEV, Edward Forster, Howard’s End)
b. ‘simultaneous’: But you must remember Aldous is no boy. (1894,
CLMETEV, Humphrey Ward, Marcella (Vol. i))
c. ‘posterior’: Remember that I shall see you at Hanover next sum-
mer. (1746, CLMETEV, Philip Chesterfield, Letters to his Son)
Of course, our data do not always show a 1:1 relation between DTR and non-finite or between ITR and finite. However, when the semantics of the matrix verb specifies the temporal and modal status of the event in the complement clause, there is at least a higher proportion of non-finite CCs. This is the case with ‘remember1’ = ‘recall’, which typically specifies a past event/situation, and can therefore often do without a full tense specification in the CC; the aggregate figures for ‘remember’ (which is largely identical to ‘remember1’) attest to this. Conversely, if the matrix verb is neutral with respect to the CC’s temporal and modal status, there is a higher proportion of finite CCs (compared to the share of finite CCs with CTPs such as remember). This situation holds for deny, which does not specify any particular temporal domain, and whose share of finite vs. non-finite CCs is 45 per cent and 55 per cent, respectively. 10.7 Conclusions In this chapter, we examined the probabilistic or non-categorical variation in LModE of finite vs. non-finite complement clauses with the matrix verbs remember, regret and deny. To probe the multivariate nature of complementation strategy choice, we fitted a binary logistic regression model with mixed effects to the data, thus shedding light on the factors favouring non-finite complementation and on the issue of how the probabilistic effect of various conditioning factors changed during the LModE period. Finally, we considered the relevance of the observed preferences (i) for Rohdenburg’s (1996) complexity hypothesis, (ii) for Denison’s (1998) and Rohdenburg’s (1995, this volume) views on the replacement of finite that-clauses by non-finite clauses, and concomitantly for our position on the grammaticalisation of sentential/clausal units, and (iii) for Noonan’s (1985) views on reduced clauses. Our study shows that Rohdenburg’s complexity principle needs to be qualified, in that not all types of structural complexity have a discouraging
Finite vs. non-finite verb complementation 203
effect on non-finite complementation. While developments for individual verbs may vary, aggregate figures for the matrix verbs showed an increase of non-finite CCs vs. finite CCs, thus largely corroborating observations by Denison (1998: 256) and Rohdenburg (1995, this volume) that there has been a growth of non-finite complement clauses at the expense of finite clauses. Finally, support was also found for Noonan’s claims on reduced clauses. While our analysis is, on the whole, not at odds with earlier claims on the differential patterns of preference and differential evolution of finite and non-finite CCs, it is important that it has made use of state-of-the-art methodology – to our knowledge, this is the first study applying statistical modelling to the CC-variation – and involves a large data set, thus putting these earlier analyses on much more solid empirical ground. Further, while the focus in this chapter has been on non-categorical CC-variation, and on statistical modelling as a way to get a handle on this variation, a comprehensive account of CC-choice in LModE also needs to give due attention to categorical CC-variation (which cannot be examined in terms of probabilistic modelling) – in fact, the two kinds of study complement each other. Finally, it is also good to keep in mind frequency distributions still have a role to play in a study of CC-variation.
Part IV Category change
11
Opposite developments in composite predicate constructions: the case of take advantage of and make use of Eva Berl age
11.1 Introduction Research on language change is often driven by the assumption that syntactic structures which look alike also behave alike. This is particularly true for studies on grammaticalisation and lexicalisation, where formally similar constructions are shown to undergo the same kind of language change (for a good overview, see Brinton and Traugott 2005). For the purpose of illustration, let us look at the prepositions during, failing and pending, all of which have a verbal origin and a participial ending. Here, structurally parallel constructions undergo the same kind of language change: while during, failing and pending initially both followed and preceded their complements as in (1) and (2), they are now categorically preposed, as is illustrated in (3). Þei bene heyest in grece in August alle the moneth during.1 (1410; Visser 1966: 1219; quoted from König and Kortmann 1991: 114) (2) Durynge that persecucioun Silvester … flig out of the citee. (1387; MED, s.v. during; quoted from König and Kortmann 1991: 114) (3) During a weekend break in this beautiful Italian city I had chocolate, hazelnut, raspberry and even cream ice cream. (BNC) (1)
This chapter is concerned with the reverse observation: patterns that look alike in terms of their structure show a very different syntactic development. I will explore this relation on the basis of the composite predicates take advantage of and make use of, both of which consist of a light verb, an NP-complement and a preposition (for more information, see Section 11.2). As corpus-based investigations will reveal, the syntactic evolution of these constructions is considerably different despite their structural similarities. In accordance with Berg (2012: 2), I will interpret the degree to which ‘two given units are integrated into one larger structure or preserve their independence’ in terms of a construction’s cohesiveness. Since the term I would like to thank Thomas Berg for many inspiring discussions on previous versions of this paper. All remaining mistakes are, of course, my own. I would also like to thank my little baby daughter Freya, who slept so well that I was able to finish off the article in time. 1
Emphasis in this and all the following examples is my own.
207
208 Berlage
captures both the semantic and syntactic aspects of integration, it is preferred over such terms as ‘idiomaticity’ or ‘fixedness’, which focus either on semantic (idiomaticity) or syntactic (fixedness) aspects of a construction’s cohesiveness. The central questions guiding my research will be: how cohesive are verb and complement in the two constructions take advantage of and make use of? Does their degree of cohesiveness change in the course of time? Finally, to what extent are these changes different for the two constructions? The chapter is organised as follows. In Section 11.2, I will discuss the formal and semantic parallels between the constructions take advantage of and make use of. I will then proceed to outline in what ways the constructions can be more or less cohesive (Section 11.3). To that end, I will introduce the noun phrase status (NP status) of the complements (in the sense that they may be more or less noun-phrasy) as an indicator of the constructions’ cohesiveness. Changes in the NP status of the complements will then be explored by means of corpus-based investigations in Section 11.4. In Section 11.5, I will argue that we can interpret the changes affecting take advantage of as a process of lexicalisation, while changes involving make use of should rather be understood as a case of delexicalisation. I will also provide a tentative answer to the question why the two constructions behave so differently. The underlying motivations discussed here are speakers’ striving for flexible modification patterns and their need for economy of expression. An outlook on future research is provided in Section 11.6. 11.2 The constructions under investigation The two constructions take advantage of and make use of have been chosen for closer investigation because they can both be classified as composite predicates. In terms of their structure, they are composed of ‘verb+NP+preposition+ complement’, as in [take advantage of [Sunday opening]] and [make use of [the reservation form which hotels include in their hotel brochure]].2 With respect to their semantics, the verbs take and make can be called ‘light verbs’. While they originally have different semantics, with take meaning ‘seize, grasp, capture, catch’ and make ‘produce, construct, create’ (OED, 3rd edition, online version; for take, see also Berlage 2012: 40), they fulfil the functions of socalled ‘light verbs’ in the constructions take advantage of and make use of. 2
Nunberg et al. (1994: 520) treat the of PP in take advantage of NP as a complement of the intransitive verb take advantage for the reason that the of PP may be left out, which is illustrated in example (i) from the BNC. (i) This is the kind of inconsistency that encourages players to take advantage.(BNC) Since no comparable examples can be found for make use of, this may be interpreted as a structural difference between the two constructions. This difference will, however, be irrelevant for the investigation presented in this paper because I will be limiting the analysis to those cases where take advantage is complemented by an of PP.
Opposite developments in composite predicates 209
In the literature, ‘light verb constructions’ are discussed under various headings, such as ‘eventive object constructions’ (Quirk et al. 1985: 750–52), ‘expanded predicates’ (e.g. Algeo 1995), ‘composite predicates’ (e.g. Brinton and Traugott 2005) or ‘verbo-nominal combinations’ (e.g. Claridge 2000), each of which has slightly different connotations (for a fuller overview, see Hoffmann et al. 2011). In this chapter, I will apply the semantic definition of light verbs that is given in Huddleston and Pullum et al. (2002: 290; see also Berlage 2012): light verbs are semantically ‘light’ in the sense that their contribution to the meaning of the predication is relatively small in comparison with that of their complements … The main semantic content is located not in the light verb, but in the noun functioning as the head of the direct object.
According to this definition, the verbs take and make should give up some of their specific lexical meaning, becoming semantically bleached. This is illustrated in example (4), which gives us a rare occurrence of the construction make advantage of in Present-Day English (PDE). Without much difficulty, make can here be replaced by take. (4)
‘American Idol’ is such a huge phenomenon because it’s the first show that’s really made advantage of all the new technology, the cell phones, the Internet. (COCA, spoken domain, 2007)
Having specified the syntactic and semantic parallels between the two composite predicate constructions, critics may argue that the two constructions also involve an important structural difference in the sense that only make use of contains a deverbal noun (use). In Section 11.5 of this chapter I will discuss whether there are reasons to assume that composite predicates which have simple verb alternants (like make use of) show a syntactic development different from those without such counterparts (like take advantage of).3 11.3 Testing for the cohesiveness of the constructions In this chapter, I will be looking at the closeness of the link between take and advantage and make and use and ask whether this link has become stronger or weaker in the course of time. In order to measure the constructions’ degree of cohesiveness, I will diachronically explore the NP status of the complements. This criterion can be considered a good indicator of the degree of cohesiveness of these constructions for the following reason: if the complements have a high NP status, they will, inversely, show a low degree of integration into the verb complement structure (preserving a relatively high degree of their syntactic independence). 3
While advantage is listed as a verb in the OED, its meaning can be paraphrased as ‘to profit, benefit, to be of advantage to’. The semantics associated with the simple verb are thus clearly different from those that go with the composite predicate construction take advantage of.
210 Berlage
The NP status of the complements advantage and use will be tested on the basis of the following two criteria: the modifiability of the complements and their ability to be passivised. These criteria have been chosen for closer inspection because, in several tests, Ross (1973b, 1995) has confirmed their status as defining characteristics of prototypical (‘purebred’) NPs. In terms of passivisation, it is interesting to note that certain complements (in verb complement constructions) can occur in two different kinds of passive. This will be illustrated with the help of sentence (5), which contains three NPs, functioning as the subject (the bosses), the complement of the verb (advantage) and the complement of the preposition (the wage freeze). (5)
The bosses took advantage of the wage freeze. (Ross 1973b: 123) NP1 NP2 NP3
If we want to transform this sentence into the passive voice, two different options suggest themselves. We can either move NP2 (advantage) into the subject position of the passive sentence, resulting in sentence (6), or we can make NP2 (the wage freeze) the subject of the passive sentence, which gives us sentence (7). 4
(6) Advantage was taken of the wage freeze (by the bosses).
(7) The wage freeze was taken advantage of (by the bosses). Since, in the active construction, NP2 is closer to the verb (take) than NP3 and therefore holds a more central position inside the sentence structure than NP3, which is placed at its periphery, the passivisation strategy in (6) will be called the ‘inside’ and the one in (7) the ‘outside’ passive (for the terms ‘inside’ and ‘outside’ passive, see Ross 1973b: 123). The idea behind using the inside passive as an indicator of NP status is that NPs require a particular threshold of noun-phrasiness in order for them to occur as subjects of a passive sentence (see e.g. Ross 1973b; 1995: 408–9). With respect to the complements advantage and use, the question is thus whether they are noun-phrasy enough to occur in the subject position of a passive sentence (that is in the inside passive). Only if they are noun-phrasy enough should the sentences in (8) and (9) be acceptable. (8) Advantage was taken of the situation.
(9) Use was made of the situation.
Ross (1973b: 127) argues that if the complements advantage and use have a low NP status, the passive can ‘overlook’ them when it looks for the first NP to the right of the verb. If this is the case, we will find that the speaker turns to the outside passive instead, producing sentences like (10) and (11). 4
Since by-phrases are optional in the passive, they are enclosed in brackets.
Opposite developments in composite predicates 211
(10) The situation was taken advantage of.
(11) The situation was made use of.
In the quantitative investigation presented in Section 11.4.1, the proportion of outside passives will be taken as an index of the NP status of the complements. Since a higher NP status of advantage and use is associated with the inside passive, higher ratios of the outside passive should correspond to a lower degree of noun-phrasiness of the complements.5 In terms of the cohesiveness of the constructions take advantage of and make use of this means: the more noun-phrasy the complement is, the less cohesive the construction will be because the NP will be perceived of more strongly as a separate constituent.6 The second criterion attesting to the NP status of the complements is their modifiability in active constructions. While Ross (1995: 399) subsumes both the use of modifiers and determiners under the label ‘modifiability’, I will distinguish between the use of determiners, such as the definite or indefinite article, and the use of modifying adjectives. An example of each construction is provided in (12) and (13) below. (12) Now as this was but a terror raised by ourselves, we take the advan-
tage of it, and so buy in when others are selling out 20 per cent. (ECF, 1760) (13) Let me, then, make immediate use of my privilege. (ECF, 1778)
Since prototypical nouns take determiners and can, typically, also be modified, I will assume that the degree of modification and the use of determiners positively correlate with the NP status of the complements. For a quantitative study, this means: the higher the degree of premodifying adjectives and/or the use of determiners, the more noun-phrasy the complements. In terms of cohesiveness, the higher the proportion of determiners and/or premodifiers preceding advantage and use, the less cohesive verb and complement will be.
5
There is yet another syntactic test that can be applied with respect to inside and outside passives. In accordance with Ross (1973b: 125, 1995: 408–9), we can assume that complements with a low NP status can only occur in the inside passive if they are modified (since modification boosts their NP status). In other words, the more noun-phrasy the complement, the higher should be the share of inside passives without modification. Within the confines of the present paper, this test will not be applied. For further information, see footnote 7. 6 Bolinger (1977: 67) argues that the NP which occurs as the subject of a passive sentence must be genuinely affected by the action of the verb, i.e. conceived of as ‘a true patient’. According to Van Valin (1999: 151–53), a ‘patient’ can be understood as a generalised semantic relation which subsumes groups of more specific semantic relations, such as themes or locatives.
212 Berlage Table 11.1 Choice of databases and corpora used for the analyses presented in Section 11.4 Corpora Early English Prose Fiction Corpus (EEPF) Eighteenth-Century Fiction Corpus (ECF) Nineteenth-Century Fiction Corpus (NCF) Imaginative Prose Section of the BNC (wri1) The Corpus of Historical American English (COHA) The Corpus of Contemporary American English (COCA)
Time span depicted a
Size per million words
1518–1700
9.9
1705–1780
10.3
1782–1903
39.7
1960–1993
20
1810–2009
400
1990–2012
450
a
Only a few publications come from the early sixteenth century.
11.4 Data analysis In this section, I will test for the cohesiveness of take advantage of and make use of, focusing on changes in the NP status of the complements. While Section 11.4.1 explores the ratios of inside and outside passives associated with the complements of the constructions, Section 11.4.2 is dedicated to the changing modification patterns of advantage and use in the constructions take advantage of and make use of. The corpora that will be used for the analyses presented in this section are summarised in Table 11.1. The table is organised as follows: the first column gives the names of the corpora, the second one the time spans they cover and the third one their sizes per million words. The reader will find that different corpora are used in Sections 11.4.1– 4.2 depending on the frequencies of the collocations searched for and the research question under investigation. More specific information on which of the corpora presented in Table 11.1 are used and why will be given in the respective sections. 11.4.1 Test 1: outside passives in the eighteenth and late twentieth centuries The null hypothesis states that take advantage of and make use of will undergo the same kind of language change because they have a similar formal and semantic composition. If the constructions become more cohesive in the course of time, this will show in the fact that the NP status of advantage and use is lower today than it was in former stages of English. If the constructions become less cohesive, this will be obvious from an increase in the NP status of the complements. The hypothesis will be tested by comparing the ratios of outside passives in fictional data representing the eighteenth and late twentieth centuries (ECF and BNC wri1, see Table 11.1).
Opposite developments in composite predicates 213 Table 11.2 Inside and outside passives with take advantage of and make use of in the ECF and the BNC (wri1)
18th century take advantage of make use of 20th century take advantage of make use of
Inside passive
Outside passive
Total
% of outside passive
9 21
4 54
13 75
30.8 72
2 1
6 2
8 3
75 66.7
Table 11.3 Inside and outside passives with take advantage of and make use of in the BNC
take advantage of make use of
Inside passive
Outside passive
Total
% of outside passive
45 309
25 33
70 342
35.7 9.6
The comparison presented here will provide a first overview of how the two constructions are developing in the course of time, thereby either confirming or rejecting the null hypothesis. On the basis of the findings provided here, a second hypothesis will be set up referring to the degree of modification in active constructions. This will subsequently be tested against empirical data in Section 11.4.2. Table 11.2 provides the ratios of outside passives for the constructions take advantage of and make use of for the eighteenth and twentieth centuries. As in Table 11.3, the complements advantage and use are either preceded by a determiner and/or premodifier as in (12) and (13) or they remain unmodified as in (8) and (9).7 In our discussion, we will now first focus on the ratios of outside passives in the eighteenth century and then move on to the twentieth century. Given that there are no more than thirteen examples of take advantage of in the eighteenth century, the outcome has to be treated with caution.
7
Due to the small number of examples (in particular for the construction take advantage of), we do not distinguish between the presence and absence of determiners and/or premodifiers in inside and outside passives. A preliminary investigation of the frequencies of these modification patterns in the BNC produces the following results for take advantage of: if advantage occurs in the inside passive as in advantage be taken of x, it is modified and/or preceded by a determiner in 44.4 per cent of all cases. If it occurs in the outside passive, however, as in x be taken advantage of, modification and/or determination is reduced to a mere 8 per cent of all cases.
214 Berlage
On the basis of the data analysed, we witness a strong difference in the ratios of outside passives for the two constructions. While the outside passive is not used in more than 30.8 per cent of all cases in the case of take advantage of, it represents the majority option for the construction make use of, with a ratio of as many as 72 per cent of all cases. This results in a highly significant difference (Fisher’s Exact test, p < 0.01). To summarise, use in make use of has a much lower NP status than advantage in take advantage of, which, in the majority of all cases, occurs in the inside passive.8 In terms of cohesiveness, the results suggest that the construction take advantage of is less cohesive in the eighteenth century than the construction make use of. With regard to the twentieth century, it is obvious that the analysis yields only a very meagre overall output, leaving us with a total of eleven examples. It goes without saying that the results produced here cannot be taken at face value but, at best, represent tendencies. For advantage in take advantage of, we can observe that the 75 per cent ratio of outside passives is much higher than it was in the eighteenth century (30.8%), which suggests that its NP status has changed since the eighteenth century: the complement advantage is much less noun-phrasy than it used to be, meaning that the construction has increased in terms of its cohesiveness. For make use of, the difference is rather small and, on the basis of three examples, we should not draw any conclusions. For a general overview of the situation in the late twentieth century, we will therefore turn to the figures retrieved from a search of the entire BNC. Although these figures will not be directly comparable to those from the ECF (because they contain both spoken and written texts and, with respect to the written text type, are not restricted to fiction), they still give us a reasonable comparison of the NP status of our two complements in the late twentieth century. The results clearly show that, for both constructions, the outside passive is by far the minority option. Yet, they also reveal a statistically very highly significant contrast between the ratios of outside passives (χ² = 30.52; df=1; p<0.001): the outside passive is much more frequent for take advantage of than for make use of. Referring back to our hypothesis, this can be interpreted in the sense that the NP status is lower in the case of advantage than in the case of use. Summarising the empirical evidence presented in Tables 11.2–3, the data have suggested that take advantage of has become more cohesive in the course of time (see Table 11.1), while make use of has 8
I would like to thank David Denison for suggesting that, in a next step, I will need to take into consideration the length of NP2 and NP3 so as not to distort the results by possible weight effects. Other factors to test for would include the animacy of the prepositional complement. A multivariate analysis should include both of these factors in a model predicting the variation. In order to conduct such an analysis, however, the database would have to be extended beyond the corpora listed in Table 11.1.
Opposite developments in composite predicates 215
become less cohesive (compare the findings for make use of in Tables 11.2 and 11.3).9 Despite the fact that additional fictional data are needed to corroborate these findings, it is clear that the null hypothesis has not been borne out by the data. Section 11.4.2 will now provide further support for the claim that take advantage of and make use of have been developing in opposite directions. 11.4.2 Test 2: modification patterns of the complements advantage and use in the constructions take advantage of and make use of Ross (1973b: 124) argues for a positive correlation between the NP status of a complement and its degree of modifiability, saying: ‘It appears to be the case that the noun-phrasier the NP2 of a given idiom is – that is, the higher up NP2 appears […] [on the scale of noun-phrasiness] – the more easily it can be modified’. Let us take this correlation as the basis for the ensuing analysis, combining it with the findings from Section 11.4.1. The previous tests have shown that advantage has become less nounphrasy over time while use has become more noun-phrasy. If the second test is to confirm this, we should find a decrease in premodification for advantage but an increase for the complement use. This hypothesis will be tested against corpus data of fiction spanning the sixteenth to twentieth centuries. The extension of the database compared to the database underlying the investigations in Section 11.4.1 is motivated by the following reasons: the comparison of inside and outside passives in the context of take advantage of and make use of has suffered from data sparseness. Given that the investigation of the modification patterns associated with the two constructions is, for reasons of homogeneity, limited to active occurrences only,10 this makes a larger database even more necessary than in the case of the previous analysis. Including data from the sixteenth and seventeenth centuries will finally provide a more comprehensive picture of when major changes in the modification pattern took place. We will start with a quantitative overview of how the complement advantage in the construction take advantage of has developed in the course of time, paying attention to its degree of modification. The columns in Figure 11.1 denote the proportions of advantage preceded by a determiner, a modifier or 9
Although the data may still involve a genre effect (because we are comparing fiction-only texts with fiction and non-fiction texts), this effect should not be able to account for a contrast of more than sixty percentage points. 10 The functions of active and passive use and inside and outside passives should be kept apart. On the basis of preliminary investigations (see footnote 7), we can expect to find a much higher ratio of determiners and/or premodifiers in the inside passive than in the outside passive. Since modifiers and/or determiners boost the NP status of the complement in the inside passive, we may also expect different ratios in active and passive uses of the constructions.
% of det. and/or premod.
216 Berlage 100% 80% 60%
13/31 = 41.9%
81/185 = 43.8%
40% 20% 0%
1518–1700
69/557 = 12.4%
29/264 = 11.0%
1782–1903
1960–1993
1705–1780 Period
Figure 11.1 Historical development of determiners and/or modifiers preceding advantage in the construction take advantage of
both. What is left to complete the 100 per cent level is, accordingly, represented by unmodified uses. The columns in Figure 11.1 reveal that we have a fairly stable situation between the sixteenth and the eighteenth centuries: here, modification accounts for 41.9–43.8 per cent of all cases. Moving from the eighteenth to the nineteenth centuries, however, we see a massive change in the data: modification drops to a mere 12.4% of all cases, with the difference being very highly significant in statistical terms (χ²= 84.87; df = 1; p<0.001). With a slight decrease of modification to 11.0% of all cases, there is, again, not much change between the nineteenth and twentieth centuries. In other words, corpus data have confirmed the hypothesis that modification of advantage decreases over time. Advantage in take advantage of has lost part of its NP status in the course of time, with the strongest decline happening in the LModE period, that is, between the eighteenth and nineteenth centuries. This confirms the findings presented in Section 11.4.1. Thus, with two tests showing that the NP status of the complement is decreasing, the construction has become considerably more cohesive. In a next step, I will investigate in which contexts we get the strongest decline in terms of premodification.11 Is it (a) in the case of determiners,12 (b) in the case of premodifying adjectives or (c) in those cases where we have a combination of determiners and premodifying adjectives? In order to answer this question, I have subcategorised all instances of premodifiers into one of the three sets (a)–(c). The result of this investigation is presented in Figure 11.2. 11
The analysis does not include any postmodification of advantage or use. In all the examples analysed, the preposition immediately follows the complement. 12 Determiners may be definite as in (12) or indefinite as in taking an Advantage of these Calamities (ECF, 1710).
Opposite developments in composite predicates 217 100% % of det.+advantage 80%
% of premod.+advantage % of det.+premod.+advantage
0%
1518–1700
1782–1903
1705–1780
7/ 26 4 21 /2 6 1/ 4 26 4
27 /5 57 15 /5 5 27 7 /5 57
5/ 18 5 10 /1 85
0/ 31 1/ 31
20%
66 /1 85
40%
12 /3 1
60%
1960–1993
Period
Figure 11.2 Historical development of determiners, modifiers and determiner+modifier preceding advantage in the construction take advantage of
Figure 11.2 reveals that it is clearly the determiner which has decreased most markedly. While it still accounts for 36–39 per cent of all cases between the sixteenth and eighteenth centuries, it is reduced to no more than 3–5 per cent in the nineteenth and twentieth centuries. The question we should address, therefore, is what the functions of determiners and of premodifying adjectives are and to what extent they are different. Research on the NP has shown that determiners like the or this typically add reference to an NP (see e.g. Lyons 1977: 177–97, 208; Adamson 2000). By saying the child is in the garden we select one particular child out of the general class of children and refer to it in the real world. Yet, reference is context dependent. This means: while an NP like Mary is typically referential, it will not be so in contexts like There are two Marys in my class (Huddleston and Pullum et al. 2002: 400). While determiners can add reference to an NP, the general function of premodifying adjectives like immediate in (13) is to characterise the noun further by pointing to specific attributes of the referent that seem important in a particular context (see e.g. Adamson 2000: 56). If we now witness a loss of determiners, as in the case of take advantage of, this means that the possibility of adding reference to a noun is decreased. If, by contrast, we remove or add premodifying adjectives, the reference of the NP is left untouched. It follows that the presence of premodifiers may be interpreted as a syntactic indicator of the NP status of the complement, while the presence of a determiner is a semantic property indicating that advantage should be interpreted
218 Berlage
% of det and/or premod.
100% 80% 60% 219/591 = 37.1%
40% 20% 0%
49/111 = 44.1%
190/692 = 27.5% 65/481 = 13.5%
1518–1700
1705–1780
1782–1903
1960–1993
Period
Figure 11.3 Historical development of determiners and/or modifiers preceding use in the construction make use of
as a noun. Figure 11.2 has now revealed that it is the semantic feature which is decreasing most strongly. Let us now turn to the evolution of use in the construction make use of. Here, our hypothesis is that its NP status has clearly increased in the course of time, which should be evidenced in an increasing degree of premodification. The results of the corpus-based analysis are presented in Figure 11.3. We see that the findings clearly confirm our hypothesis, showing a continuous rise in the use of determiners and/or premodifiers. Both the increase between the first and second time spans and that between the second and third time spans are statistically very highly significant according to the chisquare test (first/second time spans: χ² = 32.43; df = 1; p<0.001; second/ third time spans: χ² = 13.53; df = 1; p<0.001). In contrast to the case of advantage in take advantage of, the changes are more gradual, resulting in a steady decrease of the construction’s cohesiveness. In parallel to the case of take advantage of, we will ask whether the change in NP status mainly correlates with the increase in one of the functions summarised under the label of premodification. In Figure 11.4, the category of premodifiers is therefore split up into determiners, premodifying adjectives and a combination of both. The situation depicted in Figure 11.4 is very different from what we saw happening in the case of take advantage of: here, the development of the determiner is not standing out from the evolution of the other two modification patterns. Rather, we see a peak for the pattern determiner+premodifier+use between 1705 and 1780 and a rapid increase for the construction premodifier+use for the time span 1960–1993, while, at the same time, the use of the determiner is slightly decreasing (compared to
Opposite developments in composite predicates 219 100% % of det.+use % of premod.+use
80%
% of det.+premod.+use
0%
1518–1700
1705–1780
1782–1903
4/ 11 1
10 /1 11
10 4/ 59 1
50 /5 9 65 1 /5 91
32 /6 27 92 /6 92
20 /4 8 21 1 /4 8 24 1 /4 81
20%
13 1/ 69 2
40%
35 /1 11
60%
1960–1993
Period
Figure 11.4 Historical development of determiners, modifiers and determiner+modifier preceding use in the construction make use of
the previous time span). Comparing the sixteenth to the twentieth centuries, there is an overall increase in the use of the determiner and of premodifying adjectives. Summarising, we can conclude: if the loss of reference is typical of constructions which gain in cohesiveness, an increase in terms of reference and premodification should be associated with constructions that lose part of their cohesiveness. 11.5 Theoretical discussion In this section, I will show how we can assign the syntactic changes affecting take advantage of and make use of to the processes of lexicalisation and delexicalisation (Section 11.5.1). In Section 11.5.2, we will then look at a possible explanation behind the divergent developments. 11.5.1 Lexicalisation and delexicalisation As constructions increase in terms of cohesiveness, the link between their syntagmatic units grows stronger. For the case of take advantage, this means: the construction occurs with fewer and fewer intervening elements between verb and complement. The absence of intervening material between verb and complement can be interpreted as an indicator of the increasing decategorialisation of the nominal, which Trousdale (2008a: 57) defines as a typical indicator of lexicalisation. A further indicator of the lexicalisation process is
220 Berlage
the growing fossilisation of the construction (Trousdale 2008a: 57). For the case of take advantage of this is obvious from the fact that the complement decreasingly occurs as the subject NP of a passive sentence (see the decreasing ratios of inside passives in Tables 11.2 and 11.3, Section 11.4.1). These two indicators pair with the semantic status of the collocation as an idiomatically combining expression in the sense of Nunberg et al. (1994): the meaning of the construction – although distributed among the parts of the composite predicate (here, among take and advantage) – is conventional. On the basis of the semantic and syntactic reasons provided, I will argue that the construction take advantage of is on its way to becoming lexicalised.13 The situation is different in the case of make use of. Here, the empirical findings presented in Sections 11.4.1–4.2 attest to an increasing loss of cohesiveness between verb and complement. Decategorialisation and fossilisation, the two indicators of lexicalisation appropriate to describe the development of take advantage of, do not apply. By contrast, a construction which is initially relatively fixed becomes syntactically much looser in the sense that its individual parts (verb and complement) gain in terms of their syntactic independence. While make use of, like take advantage of, can be interpreted as an idiomatically combining expression, I here want to claim that the progressive syntactic freedom of the complement, which shows grammatical properties typical of NPs, is indicative of a delexicalisation process affecting the construction.14 On Brinton and Traugott’s cline of lexicality (2005: 94), the changes take place on the L1 level, that is, with the beginning and the end construction being partially fixed phrases. What changes, however, is the construction’s degree of cohesiveness: it moves from a more to a less lexical phrase. 11.5.2 A possible explanation Since I have not investigated more than two single examples of composite predicate constructions as manifestations of lexicalisation and 13
We should note here that the construction take advantage of also involves a productive element, which is the verb take. Take can combine with a series of different complements, as in take a walk, take care of, take into consideration, constituting three different subtypes of verbo-nominal combinations (see Claridge 2000: 70). Different suggestions have been made as to how to interpret constructions which involve a ‘mixture of productivity and fixedness’ (Trousdale 2008a: 58). While Brinton (2008b) claims that composite predicates which involve a light verb like take follow the path of grammaticalisation, Berlage (2012) argues that only the verb take in such constructions has become (partially) grammaticalised. An alternative interpretation is provided by construction grammar (see Trousdale 2008a). Interpreting a construction in terms of a hierarchy of constructional schemas, we may argue that grammaticalisation takes place at the macro-level of the construction (the superordinate level, where we have the pattern ‘composite predicate construction involving the verb take’) but not at the micro-level where we find our particular lexical entry take advantage of. 14 As in the case of take advantage of, we may argue that the productive element make suggests a grammaticalisation process affecting the construction or part of the construction (see the discussion provided in footnote 13).
Opposite developments in composite predicates 221
delexicalisation processes, any explanation for why constructions that are formally so similar undergo different types of language change has to be tentative. Without any doubt, future research will have to take into consideration a series of other composite predicates to see whether the interpretation suggested here extends beyond the two constructions investigated. In order to understand the contrast between take advantage of and make use of, let us first look at the semantics of the light verb construction make use of and its simple verb alternant use. Different authors (e.g. Prince 1972; Brinton 1996b; Vogt 2002; Huddleston and Pullum et al. 2002; Trousdale 2008a) argue that the main semantic difference between simple verbs like look and their corresponding light verb construction (e.g. take a look) is an aspectual one: an activity (e.g. of looking) is turned into a (quasi-)accomplishment, receiving a temporal interpretation (see Brinton 1996b: 200). If this contrast can be applied to the case of make use of/use, make use of should denote a temporal event, while use should be non-temporal. Given that such subtle semantic differences exist between the light verb construction and the simple verb, my claim is that they may be overridden by discourse-functional constraints, which we will turn to for the explanation suggested here. Various authors (e.g. Poutsma 1926: 395–96; Jespersen 1942: 117; Nickel 1968: 15–16; Live 1973: 34; Brinton 1996b: 194; Kytö 1999: 179; Huddleston and Pullum et al. 2002: 291) suggest that one of the advantages of light verb constructions over simple verbs is that they can make fuller use of modification patterns than are possible with their simple verbs. Thus, Nickel (1968: 15) states that noun modification is easier to manipulate and more versatile than verb modification. This is illustrated in examples (14)–(17). While (15) cannot be considered the exact correspondence of (14), example (17) is longer than the adnominal variant in (16). (14) ‘Have a good laugh at them, Jones. They won’t take it amiss.’ (Greene,
quoted from Stein 1991: 20) (15) *Laugh well at them, Jones. They won’t take it amiss.’ (Stein 1991: 20) (16) The aim is to encourage people to make better use of their qualifications […] (BNC, 1985–94) (17) The aim is to encourage people to use their qualifications in a better way […] On this basis, we may expect a functional division of labour to take place between the periphrastic construction make use of and the simple verb use in terms of modification such that the light verb construction is preferred where the speaker wants to add modification. Such a functional split, by contrast, is not available for take advantage of, which does not have a simple verb counterpart with the same meaning (see footnote 3). While this argumentation accounts for differences between the constructions take advantage of and make use of, it still leaves us with the question of why we find
222 Berlage
an increase in the use of premodification patterns for make use of from the sixteenth to the twentieth centuries. For an answer, we will turn to a more general syntactic change that affects pre- and postmodification patterns in the English NP. In a series of empirical studies, Biber (2003b), Biber et al. (2009), Leech et al. (2009), Biber and Gray (2011) and Westin (2002) have revealed that newspaper writing shows an increasing tendency to compress information by means of premodification. Speakers thus tend to prefer premodifiers such as continuous-time feedback systems (Biber et al. 2009: 183) over postmodifiers such as systems which provide feedback continuously.15 Since the amount of information that newspapers convey is steadily growing but the space that is available to convey the information (e.g. number of lines in a newspaper) is not increasing to the same extent, this mismatch results in a massive increase of pre- as opposed to postmodification patterns in the NP. Although previous research has confirmed such developments primarily for newspaper language (see e.g. Leech et al. 2009: 210–11, 234),16 the present study seems to suggest that the tendency to replace post- with premodification also pervades fictional writing.17 11.6 Conclusion and outlook In this chapter, I have compared two composite predicate constructions which are syntactically and semantically parallel: take advantage of and make use of. Both are introduced by a so-called light verb; both are followed by an NP and a preposition. It is all the more surprising, then, that these two constructions represent such diverse syntactic developments that they are best understood as two different types of language change: while take advantage of has become associated with a process of lexicalisation, make use of has been shown to go down the path of delexicalisation.18 I will conclude this chapter by making some predictions based on the functional explanation advanced in Section 11.5.2: constructions such as take note of and make mention of, which have simple verb counterparts (note, mention), should undergo a process of delexicalisation, while such constructions as give birth to and make fun of should become lexicalised. If the correlation between the presence of a simple verb alternant and the process of 15
We should note, however, Hundt et al. (2012: 230–32), who show that non-finite postmodification involving participle clauses also increases in frequency between 1700 and 1999. Expecting that the quest for economy does not govern fictional writing, future research needs to find out why a tendency characteristic of the newspaper genre spreads over to fiction. 17 In a next step, we will have to contrast the modification patterns for make use of and use empirically in order to test the hypothesis that adnominal modification has been growing at the expense of adverbial modification. 18 For the discussion of whether the processes of lexicalisation and delexicalisation apply to the entire construction or just to the complement, see footnotes 13 and 14. 16
Opposite developments in composite predicates 223
delexicalisation can be confirmed on a larger basis, we may conclude that discourse-functional motivations like speakers’ need for flexible modification patterns and their striving for economy win the day in language history. Abstracting from composite predicate constructions, it will be intriguing to see which other structures can be interpreted in the spirit of this chapter: constructions which share the same structure attest to a very different syntactic evolution. A possible candidate is the preposition notwithstanding, which has a formal make-up similar to during, failing and pending. In contrast to the other three prepositions, however, it not only allows for both pre- and postposed placement relative to its complement in PDE (e.g. notwithstanding the bad weather/the bad weather notwithstanding), Berlage (2009) shows that its proportion of postposing has even been increasing in the course of time. As in the case of composite predicate constructions, we may suspect that such opposite developments can be accounted for in discoursefunctional terms.
12
Constrained confusion: the gerund/ participle distinction in Late Modern English He ndrik De Sme t
12.1 Introduction English –ing-clauses, as in (1), are non-finite clauses built around a verbal head in –ing. Although they all look suspiciously alike, it can be argued that –ing-clauses divide into gerunds and present participles (e.g. Jespersen 1940; Declerck 1991). (1)
a. He was ruthless, and as to being a villain – he probably was. (1993,
BNC HGD)
b. Until he told her what it was she couldn’t even deny having it.
(1992, BNC JY2)
c. Ruth bit her lip and stared at him, not knowing what to say or do.
(1993, BNC JY4)
d. And even if she had cried out, who takes any notice of anyone
shouting, even screaming? (1988, BNC GV2)
The crucial difference is that gerunds are nominalisations, while present participles are not. This means that gerunds, unlike present participles, occur in the positions typically reserved for noun phrases. Being a villain in (1a) occupies the slot following a preposition; having it in (1b) functions as direct object to a verb. Therefore, both are gerunds. In contrast, present participles occupy adjectival or adverbial positions. Thus, not knowing what to say or do in (1c) functions as an adverbial clause, and shouting in (1d) functions as a restrictive relative clause to anyone. Still, the distinction between gerunds and present participles is not generally accepted, at least not for Present-Day English (PDE). In the reference grammars by Quirk et al. (1985) and Biber et al. (1999) the distinction is simply not applied. In Huddleston and Pullum et al. (2002) it is explicitly taken issue with, gerunds and present participles being replaced by a single The research reported in this paper has been made possible by the generous financial support of the Research Foundation – Flanders. Further, I would like to express my gratitude to Marianne Hundt and an anonymous referee for their detailed comments on an earlier manuscript, as well as to the participants of the Late Modern English Syntax workshop held in Zurich in August 2012 for their valuable suggestions.
224
Constrained confusion: gerund vs. participle 225
category of ‘gerund participials’. Regardless of whether the distinction makes synchronic sense, however, it has an uncontested basis in historical fact, as gerunds and present participles derive from historically distinct sources – Old English deverbal nouns in –ing/–ung on the one hand, and present participles in –nd- on the other. Indeed, the divided opinions between reference grammars can be thought of as reflecting a situation of syntactic ambivalence that is simply the most recent stage in a gradual but incomplete syntactic merger (De Smet 2010). In its most straightforward form, this ambivalence subsumes vagueness and ambiguity at the level of specific ‘constructs’ (Traugott and Trousdale 2013) – that is, the status of an –ing-clause in a specific utterance as either gerund or present participle may be uncertain. However, since vagueness is pervasive in some grammatical contexts, while other contexts continue to provide consistent evidence of the gerund/participle distinction, ambivalence extends to the more abstract levels of the constructional hierarchy. It is as if PDE supports two competing generalisations, one with the gerund/participle distinction intact, and one without that distinction. The goal of this chapter is to give an overview of categorial collapse in the English system of gerunds and present participles and then to link it to the situation found in LModE. As is to be shown, morphosyntactic changes that involve confusion between gerunds and present participles were few and isolated during the Late Modern period. Their very occurrence naturally testifies to a blurring of the lines between the two clause types – and of course the changes further contributed to the blurring during the Late Modern period. But the evident constraints on their occurrence show that the gerund/participle distinction was certainly not completely defunct in Late Modern English (LModE). In what follows, Section 12.2 discusses the overall history of gerunds and present participles in more detail. Against that background, Section 12.3 presents a number of changes in LModE. The focus is on complement constructions involving gerunds and present participles, as this domain seems particularly to lend itself to confusion between the two clause types. Even though changes in this domain reflect a certain leakage between the categories of gerunds and present participles, closer analysis shows that they typically involved some additional trigger that was independent of the gerund/ participle merger. Section 12.4 summarises the argument and draws from it a more precise picture of the status of gerunds and participles in the Late Modern period, and of the English gerund/participle merger in general. As a terminological note, the following discussion uses the terms ‘gerund’ and ‘present participle’ where the data support (or at least do not contradict) the distinction. But the terms are only meant as descriptive shorthand for clauses in nominal and adjectival/adverbial slots respectively, without further analytical claims attached. Where the clauses in question fill a syntactic slot that fits neither label, or where there is ambiguity, the more neutral term ‘–ing-clause’ is used.
226 De Smet
12.2 A crumbling distinction Both gerunds and present participles developed from phrasal sources – a noun phrase in the case of gerunds, an adjective phrase in the case of present participles.1 The source structure for gerunds is illustrated in (2). Đæra sacerda blawunge is a noun phrase, complementing a preposition ðurh. Its head blawunge is a deverbal nominalisation that behaves fully as a noun, taking the nominal dative singular ending in –e (ðurh governs dative case). The agent of blawunge is expressed as a nominal modifier and marked with genitive case, rather than as a nominative case clause subject as it would if blawen had been the verbal head of a finite clause. (2)
ðurh ðæra sacerda blawunge toburston ða weallas through the.gen priests.gen blowing.dat burst the walls (c.1000, quoted from Visser 1963–73: 1165) ‘through the priests’ blowing the walls burst’
The source structure for present participles is illustrated in (3). Byrnendum is a deverbal form but functions as an adjective premodifying the noun ofne. In agreement with its head, it takes the adjectival dative singular ending in –um. (3) Ðas þri cnihtas het se cyning the three knights commanded the king awurpan into byrnendum ofne throw into burning.dat oven.dat (c.1000, OED s.v. burning) ‘The king ordered those three knights to be thrown into a burning oven’ The noun phrase and adjective phrase structures in (2) and (3) above developed into full-blown non-finite clauses – though the original nominal and adjectival patterns largely remained in existence alongside the clausal ones. The syntactic reconfiguration from adjective phrase to participle clause was already well underway in Old English (Swan 2003); the emergence of gerund clauses from noun phrases took place later, with truly clausal gerunds first appearing in the Middle English period (Jack 1988; Fanego 2004). But by the beginning of the Late Modern period, around 1700, both gerunds and present participles had developed into full-fledged non-finite clauses, comparable to the English infinitive. This means they could potentially select subjects, objects, adverbial modification and the auxiliaries be and have (as shown in (4a) for gerunds and (4b) for present participles). 1
The ultimate source for present participles might be an agentive nominalisation, but few traces of this (if any) are still to be found by the beginning of the Old English record (see Nickel 1966).
Constrained confusion: gerund vs. participle 227
(4)
a. so that though firm and strong in the greatest part of it [i.e.
a bridge], yet by its being ruined in the most needful, it is become impassable. (1672–81, PPCEME) b. The next day they receiv’d order to ascend the Hill into the Castle, the Rajah having appointed an House for them (1672–81, PPCEME)
These clausalisation processes rendered gerunds and participles syntactically more alike, while at the same time dissociating them from their phrasal sources. The similarity between gerunds and present participles was additionally reinforced by the fact that, for reasons not well understood, present participles in the course of the Middle English period switched from –ndto the –ing-ending and became inflectionally identical to gerunds (compare (3) and (4a) above) (Mustanoja 1960: 547–48). Not surprisingly, the subsequent history of gerunds and present participles shows evidence of mutual historical influence between the two clause types and of growing lenience in their general distributional constraints (De Smet 2010). For example, in Early Modern English, patients marked by of optionally appear in present participles, as in (5a), in all likelihood because at the time patients are variably marked with or without of in the gerund, as in (5b–c) (De Smet 2010: 1173). (5)
a. And whyle the kynge there was shyppynge of his people, dyuers of
his lordes […] were there arrestyd for treason (1516, PPCEME) ‘And while the king was there shipping his people, several of his lords were arrested for treason’ b. By eating of Garlike, a man may the safelier goe into a suspected aire, and by stinking places. (1563, OED s.v. suspected) c. I haue knowen three within the space of one yere in high Germany that toke the falling sicknesse by drinking much newe Rhenishe wine (1568, PPCEME)
Similarly, passival participles as in (6a), with active form but passive meaning, were probably modelled on the gerund, which used to be voice-neutral, as in (6b) (De Smet 2010: 1174). (6)
a. hym thought som treasone was ymagenyng (a.1500 (a.1450),
quoted from Denison 1993: 390–91) ‘it seemed to him that some treason was being dreamt up’ b. the whilke, ilke man and woman þat of elde es, awe for to rescheyue anes in þe ȝere, […], O payne of doynge owte of Haly Kyrke (c.1440, HC) ‘which each man and woman that is of age must receive once a year, upon pain of being put out of the Holy Church’
Influence also ran in the opposite direction. For instance, LModE extraposed gerunds, as in (7a), underwent the influence of participial adjuncts,
228 De Smet
as in (7b), as both construction types developed a very similar discourse function, with the matrix clause evaluating the experience of the main participant of the –ing-clause (De Smet 2010: 1177). (7)
a. And it was funny walking up there and nobody being there
(s.d., CB)
b. You’ll have great fun choosing a name for your duck (s.d., CB)
In addition, throughout the modern period, –ing-clauses emerge that are neither clearly gerunds nor clearly present participles (De Smet 2013a: ch. 6). For instance, the –ing-clauses in (8) are not nominalisations, since they do not alternate with other functionally equivalent noun phrases, but they also lack the alternations with adjectives, prepositional phrases or past participles that are characteristic of present participles. (8)
a. ’twill be this hour ere I haue done weeping. (1591, Visser
1963–73: 2209) b. As you walk with a high German officer, the salutes of passing
soldiers come so aimlessly that you wonder why he bothers taking them. (1942, TIME)
At the same time, some differences between gerunds and present participles have never been subject to confusion. Most notably, possessively marked subjects, as in (9a), remain a constructional property exclusive to gerunds, just as subject-case subjects, as in (9b), combine exclusively with participles. As a rule, no gerunds are attested with a subject-case subject,2 and no participles are attested with a possessively marked subject. (9)
a. in the event of my having misjudged the situation (1989, BNC AR3) b. They met on the Ponte Trinita, she seeing and calling out to him
first. (1990, BNC ADS)
To be sure, as Huddleston and Pullum et al. (2002) argue, subject marking is not necessarily a good differentiator between gerunds and present participles, because the majority of –ing-clauses simply have no subject, and because both gerunds and present participles invariably allow the alternative of oblique-case subjects, as in (10a-b). (10) a. What do you mean by us being set in our ways? (1991, BNC A0F) b. He knows how I feel, me being an only child … (1991, BNC AC7)
Nevertheless, the resilience of subject marking to confusion indicates that language users continue to be able to reliably distinguish gerunds from present participles at least in some cases some of the time. The same ability to differentiate between gerunds and present participles seems to manifest itself – under the guise of probabilistic constraints – in 2
But see De Smet (2010: 1174) for some possible counterexamples.
Constrained confusion: gerund vs. participle 229
patterns of formal variation in English dialects. For example, Labov (1989) (based on Houston 1985) states that where there is variation between an /ɪn/realisation of the –ing-form and the standard /ɪŋ/-realisation, the former tends to be ‘favored most in progressives and participles, less in adjectives, even less in gerunds and least of all in nouns like ceiling or morning’ (1989: 87). Similarly, studying dialectal data from Suffolk, Braeken (2011) shows that the optional a-prefix with –ing-forms, as in (11a), is almost exclusively attested with present participles – (11b) offers a rare counterexample of a gerund with a-prefix. (11) a. And we anchored off here and of course I could hear some of the
lads a-playing about on the beach. (FRED, quoted from Braeken 2011: 24) b. and you used to go to and fro, to and fro with that rope all the time like that till you got up to the top. And by so a-doing that you could get forty pockets on the wagon (FRED, quoted from Braeken 2011: 27)
Together, all these observations can only lead to mixed conclusions. Even though the data reveal constant and growing leakage between gerunds and present participles, they do not support the view that the two clause types have completely merged. 12.3 Leakage and resilience in LModE complement constructions The Late Modern period sees a number of changes that clearly reflect the ambivalent status of the gerund/participle distinction in English outlined in the previous section. I will focus here on one area of potential confusion between gerunds and present participles and survey the evidence of actual confusion occurring. Where there is recurrent confusion at construct level, this is indicative of a loss of distinction at the schematic level. As such, the evidence confirms the picture of a partial merger between gerunds and present participles. The area specifically at issue confronts us with two historically distinct complementation patterns that are often formally indistinguishable. One is the participial object complement construction, as in (12a). The other is the gerundial complement construction, as in (12b). (12) a. As they advanced they heard the sound of music, and saw youths
and virgins dancing in the grove (1759, CLMET3.0)
b. Even now […] people remember Lady O’Dowd performing a jig at
Government House, where she danced down two Aides-de-Camp, a Major of Madras cavalry, and two gentlemen of the Civil Service (1847–8, CLMET3.0)
230 De Smet
As the examples in (12) show, both constructions most typically appear as a sequence of a verb, noun phrase and –ing-clause. For ease of reference, the noun phrase between the matrix verb (saw, remember) and –ing-form will here be referred to as NP2 because it is the second noun phrase in the syntagm – these are the underlined noun phrases in (12).3 Semantically, NP2 provides the main participant of the verbal process denoted by the –ingform, and the situation jointly construed by NP2 and the –ing-clause provides the secondary participant to the matrix verb (i.e. that which is seen in (12a), or that which is remembered in (12b)). Next to these similarities, there are also differences between the two constructions that betray their different origins. For a start, the –ing-clause in the participial object complement construction alternates with the usual participial alternates, such as past participles or adjective phrases, as in (13a), whereas the –ing-clause in the gerundial complement construction alternates with noun phrases, as in (13b). (13) a. They saw the youths and virgins excited by the music / insane
with joy.
b. Even now people remember Lady O’Dowd’s jig at Government
House.
Further, the participial object complement construction can be passivised, promoting NP2 to the subject position of the matrix clause, as in (14). (14) Youths and virgins were seen dancing in the grove.
In contrast, the gerundial complement construction predictably allows possessive marking on NP2 (which functions as the gerund’s own subject), as in (15a). It also allows omission of NP2 (triggering control by the matrix subject over the gerund), as in (15b), where the gerund is headed by the form watching, whose (implied) subject now corresponds to the subject of the matrix verb (people). (15) a. Even now people remember Lady O’Dowd’s performing a jig at
Government House.
b. Even now people remember watching Lady O’Dowd’s jig at
Government House.
Finally, De Smet and Heyvaert (2011) show that present participles tend to retain time-stable semantics – a feature imposed by the adjectival slot they 3
Depending on the specific construction, NP2 may also occur in other positions when the matrix clause is passivised or when it is extracted. For clarity of exposition, in such cases I will still refer to it as NP2. Further note that the label NP2 generalises over underlying syntactic differences. As argued below, sometimes this terminological generalisation appears to be justified by a genuine loss of syntactic distinctions, but often it is not, in which case NP2 is simply a convenient descriptive label. The following discussion specifies where underlying syntactic differences are lost and where they are not.
Constrained confusion: gerund vs. participle 231
occupy. In the participial object complement construction this typically results in a progressive reading for the event denoted by the complement (see Declerck 1981; Dixon 2005: 271; Gisborne 2010: 195–97). Gerunds are not subject to this constraint, allowing both perfective and imperfective interpretations. Compare the corpus examples in (16), where the participle in (16a) conveys progressive aspect while the gerund in (16b) is aspectually neutral. (16) a. a friend, calling on him one day, found him reading a little slip cut
from a newspaper (1895, COHA MAG)
b. had they contemplated my reading them, they would doubtless
never have written them (1876, COHA MAG)
Returning to the two examples in (12) above, then, the syntactic evidence suggests different underlying structures, which can be roughly represented as in (17). (17) a. They [saw [youths and virgins]DO [dancing in the grove]oc]vp. b. Even now people [remember [Lady O’Dowd performing a jig at
Government House]DO]vp.
Note, however, that most of the evidence of this syntactic distinction resides in alternation patterns, which means that language users must draw on second-order generalisations, which are a step removed from immediately observable distributional evidence (Croft 2004). Note further that some of the crucial alternates have been growing less frequent – most notably possessively marked subjects in the gerund, but also adjectival object complements. As a result, the evidence available to language users from which to infer the distinction between the two construction types has been getting less salient. The semantic evidence is more direct, but because gerunds are not strictly incompatible with progressive interpretations, it provides no truly distinctive criterion (see (16) above). As might be expected, then, the syntactic distinction between participial object complement constructions and the gerundial complement construction has given way in some specific contexts, just like the overall gerund/participle distinction. This is reflected in a series of innovations, which, however, have so far failed to completely upset the old system. Specifically, there are a handful of innovations in LModE that clearly violate original gerundial or participial syntax, but which occur in very specific lexico-grammatical environments and which are mostly triggered by independent developments. In what follows, four such developments are considered. Each change occurred in the context of a specific matrix verb or set of matrix verbs – want and need (Section 12.3.1), prevent and stop (Section 12.3.2), fancy (Section 12.3.3) and see, hear and feel (Section 12.3.4).
232 De Smet
12.3.1 Want and need Want has long been used with passival gerunds, as in (18). In this pattern, the matrix clause subject provides the patient of the gerund, which can therefore be paraphrased with a passive infinitive (e.g. ‘to be polished’ in (18a), ‘to be shaken up’ in (18b)). (18) a. It was but rough hewen by one of the prentises, and wanted sum
polishing by the forman (1574, quoted from De Smet 2013a: 181)
b. I have written, I should think, about thirty chapters of the
South Sea book; they will all want rehandling, I dare say. (1890, CLMET3.0)
In the eighteenth century, want also entered the object complement construction, with past participles as object complement, as in (19). (19) [he] called for a pint of beer, and wanted a steak broiled, which was
done. (1744, OBC)
Soon after, want appeared in a new pattern illustrated in (20a). Here, the –ing-form is passival, but NP2 behaves as if it is the object of want in a participial object complement construction. The new pattern allowed extraction under relativisation, as in (20b). (20) a. If … you Messrs. Apothecary and Taylor want your bills paying
(1759–67, quoted from Visser 1963–73: 2362)
b. he was to make some memorandum with respect to some Dutch
names that he wanted putting on the watches (1788, OBC)
The timing of the change indicates that the new pattern in (20) arose as a blend between the passival gerund and past participle (De Smet 2013b).4 The new construction survives into PDE (21a) and has extended to need, as shown in (21b). (21) a. Look, no Richard! I don’t want things stuffing down that hole. Put
it in the bin. (s.d., BNC KB8)
b. Perhaps the gas cooker’s broken and they need it mending (s.d.,
BNC KB8)
The result is a pattern that violates both gerundial and participial syntax. Its passive meaning is (particularly from a PDE point of view) at odds with the default active semantics of the present participle.5 Note, moreover, that the –ing-form lacks the time-stable (progressive, stative, habitual …) semantics usually associated with participial constructions. Semantically, then, the –ing-clauses at issue are closer to gerunds. However, their syntactic behaviour is that of a participle, and decidedly not that of a nominalised clause. 4
A construction that looks like the opposite blend also occurs in some dialects, as in the car needs washed (see e.g. Murray and Simon 1999). Hundt (2004a), however, shows that passival progressives still marginally occur in PDE.
5
Constrained confusion: gerund vs. participle 233
Therefore, examples as in (21) would be hard to justify from the point of view of a grammar that maintains a strict distinction between gerunds and present participles. 12.3.2 Prevent and stop The verb prevent appeared with gerund complements in the seventeenth century. Predictably, NP2 could appear as an oblique form, typically when it is a full noun phrase, as in (22a), or with possessive marking when it is pronominal, as in (22b). (22) a. the dagger did not absolutely miss him, but wounded him in
the left arm; but Octavio’s youth, too nimble for Clarinau’s age, snatching at the dagger as it wounded him, at once prevented the hurt being much, and returned a home blow at Clarinau (1684, CEMET) b. these laths are to be tied round about the Pike’s body, from his head to his tail […] to prevent his breaking or falling off from the spit. (1656, CEMET)
In the eighteenth century, however, prevent also made its appearance in what looks like a participial object complement construction, as shown in (23). Examples (23a–b) demonstrate promotion of NP2 to the subject position of the matrix clause under passivisation, while (23c) shows that NP2 could be a reflexive form, indicating that it was not (necessarily) interpreted as the subject of the –ing-clause. These operations leave the remaining –ing-clauses (doing any more damage in (23a), telling me her story in (23b) and falling in (23c)) in slots that no longer qualify as ‘nominal’ since they would no longer alternate with an ordinary noun phrase – compare *they were prevented any more damage, *she had been prevented the completion of her story or *to prevent myself a bad fall. (23) a. but [they] were prevented doing any more damage by the Watch
(1718, OBC)
b. she was sorry […] that she had been prevented telling me her story
(1768, CLMET3.0)
c. I was obliged to jump over the box to prevent myself falling.
(1779, OBC)
The pattern again survives in PDE, as shown in (24), though it is more common now with the verb stop, as in (24b). (24) a. Further, Spenser helps us to discern cultural issues which
Elizabethan institutions prevented being openly represented, notably the uneasiness generated by a woman ruling within a patriarchal culture. (1992, BNC HH4) b. Wherever you are grabbed, you cannot be stopped using at least some of these weapons, and every time one hits home, the attacker’s
234 De Smet
pain will increase until he or she is forced to loosen the grip. (1989, BNC CDT) What caused the change was probably not just language users’ inability to distinguish between gerundial and participial syntax. There was a third construction that is likely to have acted as a catalyst. This was the pattern with an object-controlled gerund introduced by from illustrated in (25). When NP2 is in its default position following the matrix verb, as in (25a), the construction differs superficially from a gerundial complement with oblique subject only in the respective presence or absence of from, even though it allows extraction of NP2, as in (25b). (25) a. and then they mutually labour to cover it with the same sand, to
prevent it from being devoured by other fish. (1656, CEMET)
b. He renews his vows to her of eternal love, and that he would per-
form what they were so unluckily prevented from doing this morning (1684, CEMET)
Given this situation, from could be reinterpreted as an optional element in the gerundial complement construction and this optionality could then extend to contexts with extracted NP2, resulting in the innovative pattern in (23)–(24) above.6 Regardless of its immediate cause, however, the change in the complement constructions with prevent left another dint in the gerund/participle distinction. Since NP2 in (23)–(24) above behaves as a separate constituent, the remaining –ing-clause cannot be a gerund, because it is not in a nominal slot. But a participial analysis is not supported either, because the usual alternation patterns (with past participles, adjectives and prepositional phrases) are lacking, and because the –ing-clause does not have time-stable semantics. 12.3.3 Fancy Unlike want and prevent, fancy started out in the participial object complement construction. Its NP2 was at first consistently an oblique form, as in (26a), and it sanctioned adjectival object complements, as in (26b). (26) a. You can fancy me ascending Monte Cavallo, leaning against the
pedestal which supports Bucephalus (1783, CLMET3.0)
6
Such a reinterpretation may have been facilitated by the fact that by the beginning of the eighteenth century, the status of from as a preposition had become somewhat obscured, at least in the construction with prevent, as from rarely combined with other noun phrases than gerunds. Moreover, NP2 often lent itself to being reinterpreted as the subject of the following -ing-form, e.g. in In the evening Toobouratomida and his Wife, and a Man belonging to Tootaha, would needs lay all Night by the Casks to prevent any more from being taken away (1768–71, CLMET3.0).
Constrained confusion: gerund vs. participle 235
b. One, who is inflamed with lust, feels at least a momentary kind-
ness towards the object of it, and at the same time fancies her more beautiful than ordinary (1739–1740, CLMET3.0)
Later, fancy also began to combine with gerundial complements, as is apparent from possessive marking on NP2. The first examples are found with the imperative form fancy used as a marker of surprise, as in (27a–b). Indeed, in American English the gerundial construction remains restricted to this context. In British English, it spread to non-imperative uses of fancy as well, as in (27c). (27) a. No right to interfere! Venetia, my little fellow-labourer, no right to
interfere! Why all is yours! Fancy your having no right to interfere at Cadurcis! (1837, CLMET3.0) b. Fancy his being taken for an Englishman! Still more, fancy his wishing to be so mistaken! (1868, COHA MAG) c. She fancied their falling into the hands of some speculator, who, if he did not break the mother’s heart by putting up a gasometer, would certainly wring it by building hideous cottages, or desirable marine residences. (1865, CLMET3.0)
Why it was precisely fancy that made the crossover from the participial object complement construction to the gerundial complement construction is hard to tell, though again, there may have been special circumstances to facilitate the change. Specifically, the use of imperative fancy as a marker of surprise had just arisen (De Smet 2013a: 232) and was particularly compatible with the gerund. Because the gerund invites a factive reading (Heyvaert 2008), it reinforces the mirative function of imperative fancy – an unusual situation is all the more surprising if it is a fact. The effect is particularly clear in (28) (note, though, that the same effect is certainly missing in (27c) above, again suggesting a subsequent generalisation of the use in British English). (28) a. [in response to the addressee unwittingly giving the speaker an
important piece of information] ‘H’m. Fancy your being able to teach me what you don’t know yourself!’ (1891, CLMET3.0) b. My, what a pretty necklace! It’s new, isn’t it? Pearls, too! […] Real pearls! Angela, fancy your having a string of real pearls! isn’t that wonderful! (1920, COHA FIC)
Of course, if these functional-semantic considerations account for language users’ choice of gerundial complements with fancy, it testifies to their awareness of the difference between these and participial object complement constructions rather than a neglect of it, particularly considering that factivity is a direct reflex of the gerund’s status as a nominalisation (Heyvaert 2008). Even so, the change seen in fancy sets the stage for the later development in see, hear and feel.
236 De Smet
12.3.4 See, hear and feel See, hear and feel are some of the most typical verbs used in the participial object complement construction. Recently, however, a subtle change has set in that obscured the syntactic status of NP2 in the object complement construction, precisely with these verbs. This is evident from the construction’s behaviour in PDE. For a start, it is sometimes doubtful on semantic grounds whether NP2 in a participial object complement construction with see, hear or feel is really the object of the matrix verb, as in (29) (where what is heard is not the crowd but the crowd being addressed). (29) Fortunately, because I wanted no part of any event featuring gospel
singers, I was not present to hear an incredulous crowd being exhorted to ‘stamp on Satan’. (s.d., CB)
Further, as (30) shows, the construction is now attested with existential there, which is usually considered a subject-slot filler (see Breivik and MartínezInsua 2008). (30) a. We couldn’t see there being any problem. (s.d., CB) b. Can’t see there being much work here (2010, COCA FIC)
Finally, the whole sequence of NP2 and –ing-clause can be the focus of a pseudo-cleft, as in (31). (31) a. The last thing he felt was it plunging into his chest. (1991,
COCA FIC)
b. if you listen back to that tape now all you’ll hear is them twittering
away (s.d., BNC KBX)
All this indicates that NP2 in these examples might actually be the subject of the –ing-clause rather than the object of the matrix verb (see Declerck 1982 for a similar argument). If so, the sequence of NP2 and –ing-clause must be interpreted as a single constituent, akin to a gerund. If gerunds really provide the model for change, the obvious next step would be for NP2 to appear with possessive marking. This has in fact happened, as in (32), but just as when gerunds with possessively marked subjects first appeared with fancy (see above), the use comes with a factive reading. In terms of frequency, the occurrence of possessively marked subjects with perception verbs is very rare. Functionally, there is no evidence yet of a more generalised and semantically neutral use of possessive marking on NP2 with see, hear and feel. (32) a. Ile looked at her. The impact of being away ten months struck
softly home. She saw his staring and said […]. (1963, COHA FIC)
b. Baryshnikov is one of the world’s outstanding dancers and per-
formers, and watching him in anything is a privilege. It is no use bemoaning the sad fact that his days as a brilliantly athletic
Constrained confusion: gerund vs. participle 237
virtuoso in classical ballet seem to be over. Time inevitably moves on. What is fascinating is to see his mastering a new style and technique, and still managing to rivet our attention as of old. (1992, BNC AK4) This casts some doubt on whether the change in the complements of see, hear and feel is simply due to confusion with gerund complement constructions. There is a general tendency for NP2s to develop from matrix verb objects to complement-clause subjects. This is, for instance, what gave rise to Exceptional Case Marking constructions (Los 2005) – compare (33a–b) (see also footnote 3 above). (33) a. He dede comand..Douke..and kniȝt To diȝt her hors. (c.1330
(?a.1300), MED s.v. commaunden) ‘He ordered Duke and knight to prepare their horses.’ b. He commaunded false Edrik forto be slayne (a.1387, MED s.v. benefe¯t) ‘He ordered that false Edrik be slain.’
In addition, for the perception verbs there is another possible model, namely the AcI construction, illustrated in (34), whose NP2 likewise is closer to being the subject of the subordinate verb than the object of the matrix verb. (34) Olive felt her temper rise. (1991, BNC A0U)
Even so, the changes in the participial object complement construction with see, hear and feel violate original participial syntax and as such contribute to the general decline of the gerund/participle distinction. 12.4 Conclusions The evidence discussed in the previous two sections makes for a complex picture. It is clear that innovations occurred that caused departures from the historically motivated behaviour of gerunds and participles. The fact that such departures are possible in the first place indicates that a degree of fuzziness has crept into participial and gerundial syntax. Against the more general background of a disintegrating gerund/participle distinction, this is probably as expected or at least answers to a more general diachronic trend. But it is also clear that innovations are not only the result of a decline in language users’ awareness of the gerund/participle contrast. On the contrary, some of the changes even suggest continued awareness of the difference. More generally, the fact that no widespread confusion arises indicates that, despite superficial formal similarity, gerunds and participles most of the time continue to be recognised as such, at least within the lexico-grammatical environments delineated by the gerundial complement construction and the participial object complement construction.
238 De Smet
This ambivalent situation is instructive of what ultimately drives the decline of the gerund/participle distinction in English. Here, too, the picture is complex. Sometimes, changes seem to be triggered simply by confusion and competition resulting from functional similarity. The emergence of a blended complement construction with want is probably a case in point of confusion between competing constructions (Section 12.3.1). At the same time, there is a functional side to change. At its most straightforward, change exploits the functional potential of established constructions, as when gerunds appear with fancy presumably to reinforce its mirative meaning (Section 12.3.3). More generally, when originally phrasal structures develop into clauses they end up being subject to conflicting functional pressures. On the one hand, they still occur in phrasal environments, but on the other, they are exposed to the general functional demands on clauses. For example, one core function of a clause is to predicate a state of affairs of a subject. Tellingly, some of the changes observed above precisely result from the recruitment of an external noun phrase as subject of a non-finite clause, as in the development of the participial object complement construction with see, hear and feel (Section 12.3.4). Similarly, it is a core function of clauses to provide the template to profile central discourse participants in ways that accord with the information flow in a text. The increasing mobility of clausal participants observed in the complement structures of prevent answers precisely to that functional need (Section 12.3.2). We can therefore see the partial collapse of the English gerund/participle distinction as resulting from a complex interplay between different forces. There is simple confusion, working at both the construct level and the more abstract levels of the constructional hierarchy. With Traugott and Trousdale (2013) and Trousdale (2013), it can be assumed that the syntactic consequences of confusion gradually work their way up the constructional hierarchy. In addition, there are functional pressures that alienate non-finite structures from their historical phrasal sources and thereby in turn feed a blurring of the lines between historically distinct clause types. This process primarily happens locally, in specific lexico-grammatical environments with specific functional demands to satisfy, as well as specific structural opportunities to exploit.
13
‘You are a bit of a sneak’: exploring a degree modifier in the Old Bailey Corpus Claudia Cl aridge and M e rja Ky tö
13.1 Introduction In Present-Day English (PDE) a bit (of (a)) is a well-established and grammaticalised downtoner, functioning as diminisher and minimiser, e.g. a bit (= ‘somewhat’) tired (see Bolinger 1972: 17, 22, 124; Quirk et al. 1985: 598; Brems 2007: 294; Paradis 2008: 321). Additionally, depending on the context, the form may also indicate great(er) quantities and degrees (Channel 1994: 105), e.g. a bit of a trek from Soho, but it’s worth it, as well as emphasis, especially in negative environments (not a bit of it!, Brems 2007: 305). Furthermore, it has hedge uses (e.g. isn’t there a bit of a double standard there?, Brems 2007: 306). It is rather versatile in its syntactic behaviour, accompanying predicative adjectives, adverbs, nouns and verbs. It is overwhelmingly found in spoken data (Paradis 2008: 321) and it has a colloquial touch. The origins of the a bit-construction apparently lie in the Early Modern English (EModE) period, but no data-intensive study on its development has been carried out so far. Traugott’s discussions (2008a, 2008b, 2010) are data-based, making use of the OED, MED and the Chadwyck-Healey databases as well as various corpora. Apart from illustrative examples for each developmental stage, she does not provide any frequencies or document the linguistic situation in more detail, however. Brems (2007, 2011) provides some figures, but deals with the synchronic state only. It is therefore our main aim to supply a corpus-linguistic foundation for the suggested stages of development (see Section 13.2), either corroborating or modifying them. While a corpus-linguistic approach to language change is promising in many respects, it can also be problematic for features with oral affinities like bit. In corpora commonly used for EModE, uses of bit are fairly rare: two in the Helsinki Corpus, three in the Lampeter Corpus, eight in the Corpus of Early English Correspondence, and only twenty-four even in the Corpus of English Dialogues 1560–1760. By contrast, the large Old Bailey Corpus (OBC), with its great wealth of transcribed speech from courtroom interactions, presents sufficient data spread over almost 200 years to enable informed statements on the development of syntactic and semantic uses. In 239
240 Claridge & Kytö
a previous study (Claridge and Kytö 2014), we investigated the uses of a bit (and pretty) in the period 1732–1834 in the OBC (the version of the corpus available then did not yet include earlier or later material). All constructional variants known from PDE were present in the data, with a bit of dominating. Both ‘literal’ and degree uses, as well as intermediate types, were found, but the literal ones were still in the clear majority. Thus, while a bit was developing grammaticalised uses, this process was not very advanced. With regard to overall frequencies, we found that a bit-constructions were also not clearly increasing, but dipping especially in the last decades. Thus, our research left open questions, which we intend to pursue here with the help of the extended version of the OBC, starting from the 1720s and ending in 1913. The questions we intend to address here are: • Will a bit-constructions (all of them or individual ones) decline, increase or stabilise after the 1830s? • Syntactic change: how will the proportion of constructional variants develop? Will a bit of remain the most common variant or will it be overtaken, perhaps by a bit? Which contexts play a role in the establishment of variants? • Semantic change: how will the distribution of different readings (quantifying, degree, hedging, others) develop? In particular, to what extent will the concrete uses decrease? While the main aim of this paper is not to make new theoretical contributions, we will align our findings with claims made by grammaticalisation theory. 13.2 From a bite to a bit (of) The main component of the construction1 under consideration, the noun bit, is a long-established item in the English language. From ‘a bite’, something prototypically small, it developed the meaning ‘small piece/portion/ quantity’, which flows into the uses a bit of ‘small amount / specimen’ and a bit ‘a very small measure or degree’ (OED s.v. bit n.2). The first clear degree use quotation for a bit in the OED is dated 1675 (‘I had not wrong’d the Gods a bit.’ (Cotton, Poet. Wks., OED)). As the meanings ‘bite’ and ‘piece’ presuppose the question ‘bite/piece of what?’, a fairly natural occurrence of bit is within an NP of NP construction frame, or, to be more precise, within a construction (D) (M) N1 of (D) (M) N2 (Brems 2011: 2), where N1 is a form of bit, and D and M stand for 1
We are not working within a construction grammar framework and thus use the term ‘construction’ in a general, theory-neutral sense. The more specific sense only applies when we quote authors working in this framework (e.g. Brems).
A bit of (a): exploring a degree modifier 241
optional determiners and modifiers. Brems (2011: 73) characterises uses of bit relevant here as size-noun (SN) constructions, which for her are a subtype of the partially filled constructions of Construction Grammar (CxG), in her terms ‘collocationally constrained constructions’, with colligation and semantic prosody playing a role as well. In one perspective, the size noun itself is a collocate of a discontinuous collocational framework (a concept Brems takes over from Renouf and Sinclair 1991), here an +? + of, and it will often be more frequent within the construction than outside of it (Brems 2011: 74–75); this is not always true in cases of grammaticalisation (e.g. Hundt 2001: 62). In another perspective, the size noun functions as a collocational node, co-selecting its collocates and colligates such as the second noun, determiners or modifiers. The collocational pattern strongly interacts with the syntactic status of the constructional parts, that is, with what is to be seen as head and modifier (Hundt 2001: 73). Two possible alternatives are head + postmodifier or quantifier + head, i.e. bit can either be head or quantifier depending on which interpretation the collocational pattern favours (see also, e.g. Traugott 2007, 2008a, 2008b). For Brems (2011: 37) ‘the synchronically coexisting uses of SNs … constitute semi-stable layered (Hopper 1991) systems of head, modifier and ambivalent uses’. Traugott (2010: 46–48, see also 2007: 533 ff., 2008a: 29) charts the historical development as follows: Stage
Type
Example
Date
0
Pre-partitive
I
Partitive
(OE/) ME ME
II
Extended partitive
III
Quantifier
IV
Adverb degree modifier
V
adjunct
In to the pyne of helle .. for the bytt of an Appel (c. 1400 Ancr. Recl. 22/25 [MED bite 3.b.]) He badd tatt gho shollde himm ec / An bite braedess brinngenn (‘a bite/bit of bread’) (c. 1200 Orm 8640 [MED bite 3.a.]) The fragments, scraps, the bits, and greazie reliques, of her ore-eaten faith (1606 Shakespeare, Troilus and Cressida V.ii.159 [OED bit 2, 3.a.]) If so be as ow you ʼont blab, I’ll tell you a bit of a secret (1833 Clifford, The Highwayman of 1770, I.ii. [LION]) Your beauty is a little bit of a jilt (1771 S. Foote, Maid of Bath [OED bit 2, 4.b.]) If you be a lord, it must be a bit of bastard business (1810 W. Hickey, Mem [OED bit 2, 4.h.]) I would not be a bit wiser, a bit richer, a bit taller, a bit shorter, than I am at this Instant (1723 Steele, The Conscious Lovers III.i [LION]) I tell ye, I zee’d un gi’ Susan a letter, an’ I dan’t like it a bit (1800 Morton, Speed the Plough II.iii. [LION]) A. Hear me. B. Not a bit (1739 Baker, The Cit Turn’d Gentleman [LION])
17th cent.
18th cent.
17th cent. 18th cent.
242 Claridge & Kytö
By Stage II, host-class extension (or ‘collocational range expansion’ in Brems’ terms), to non-food and abstract nouns is already found. Stage III also sees syntactic and semantic expansion, making two syntactic analyses possible (head vs. quantifier) and leading to semantic bleaching of bit. Subjectification is also found as early as Stage III (all quantifying uses), with intersubjectivity in evidence in Stage IV (hedge uses, e.g. I’ve been a bit of a fool again) (Traugott 2010: 49). As Stages I to III clearly illustrate, it is the binominal construction from which the development takes its origin and in which guise the first non-literal or extended uses arise. In this respect, the construction is not alone, but forms part of ‘a whole class of binominal Quantifiers’ that is said to come into being around the eighteenth century, such as a bit of, lot of, (a) sort/kind of (Traugott 2010: 53).2 Not all binominal constructions move out of the nominal sphere, but bit does, losing its preposition but retaining its determiner (adverbial a bit). This further syntactic expansion takes places in Stages IV and V (Traugott 2010). According to Brems (2011: 248 ff.), the fully grammaticalised uses of Stages III to V are present in PDE, as well as evaluative and subjectified uses (in particular with a bit of a). The development just sketched is based on data reported in previous studies, but not on a comprehensive diachronic corpus study. The OBC used for the present study includes the period when uses situated on Stages III to V emerge and begin to spread through the language. It will remain to be seen which of these uses appear in which quantities and in which sequence in our data. 13.3 Data The material used for the present study is drawn from the OBC, a corpus based on the Proceedings of the Old Bailey (1674–1913) (Huber 2007). The Proceedings total over 200,000 trials, or c.134 million words and are accompanied by a search engine that is not particularly suited for the purposes of linguistic study. The OBC, on the other hand, though less extensive in size than the Proceedings, has been annotated especially for the purposes of linguistic investigation. The OBC is also of such substantial size that it makes the material particularly useful for the study of low-frequency items, as we will see below. The corpus is also well suited for studies within historical 2
Note, as Traugott (2008b: 226) points out, that the NP1 of NP2 string subsumes ‘a wide array of functions, among them locative the back of the house, partitive a piece of the plate, approximative (a) sort of a frog, emotionally charged epithet an idiot of a teacher, subjective genitive the singing of the diva, and objective genitive a portrait of a hunter’. Regarding semantic functions, ‘sort of (= ‘subtype’) in a (special) sort of rose is in a taxonomic (‘kindof ’) rather than partitive (‘part-of ’) relationship to NP2’; on this point, Traugott refers to Croft and Cruse (2004: ch. 6).
A bit of (a): exploring a degree modifier 243 Table 13.1 OBC – periods, word counts, and data
words all bita normalisedb
1720–1759
1760–1799
1800–1839
1840–1879
1880–1913
3,581,463 152 4.2
3,554,756 315 8.9
3,720,812 267 7.2
3,408,928 183 5.4
2,973,892 197 6.6
a
Only singular occurrences. There are 139 instances of the plural bits spread thinly across the 200 years, of which 59 are lexical head uses and 80 occur within the bits of N frame. Of these, 78 co-occur with a concrete N2, and only two with potentially indeterminate nouns, namely conversation and poetry. The plural thus does not play a significant role in the development of the item towards grammaticalised uses. b All data is normalised to occurrence per 100,000 words; the same applies to subsequent graphs with normalised figures.
pragmatics approaches as it mainly comprises speech-related data, being based on everyday language taken down ‘verbatim’ in courtroom situations (however, for caveats, see e.g. Kytö and Walker 2003). Degree modifiers have been shown to be more numerous and more varied in spoken (especially conversational) language (Biber et al. 1999; Paradis 2008: 321), which also makes it likely that changes regarding them arise and are earlier observable in spoken or speech-related language. As degree items have been identified as a field of substantial change and renewal in the Early and Late Modern periods (Peters 1993; Méndez-Naya 2008), the material seems highly suitable because of both its type and its time frame. The version we use here is the corpus release of July 2012, comprising the period 1720 to 1913 and a total of 17,239,851 words. We have subdivided the corpus into five periods spanning four decades each (see Table 13.1). The word counts are based on the calculations by WordSmith Tools (version 5.0), which was also used to retrieve the data. Needless to say, possible spelling variants were taken into consideration in all the searches, but in fact only the modern standard form bit occurred. Only data occurring within transcribed speech was analysed, which led to the exclusion of one instance used in a scribal comment. Furthermore, one fragmentary item had to be left out of consideration. The remaining bit instances were manually post-edited with regard to irrelevant hits, such as forms of the verb bite, specialised lexical (‘mouthpiece of a horse’s bridle’) or technical meanings (‘the cutting or biting parts of a tool’, see OED bit n.1) as well as idioms like bit by bit. The first two of those listed occurred relatively frequently. Otherwise all uses of bit, indicating piece, size, quantity, degree and the like, were included in order to gain as comprehensive a picture as possible of the linguistic context of bit. After all the above exclusions, a total of 1,114 instances (of the originally retrieved 1,460 items) remained for analysis. These are distributed across the five periods as shown in Table 13.1.
244 Claridge & Kytö
13.4 Findings Table 13.1 already partly answers one of the questions posited in Section 13.1: constructions involving bit indeed show a further decline after the 1830s, thus confirming the trend indicated in Claridge and Kytö (2014) (the results given in Table 13.1 are statistically significant: chi-square = 68,297; p < 0.001; df = 4). However, the decrease is apparently halted in the last period, when the numbers pick up again, although not reaching previous heights. Brems (2011: 239) provides quantitative evidence from PDE for comparison: bit in the NP of NP construction alone occurs in the Collins Wordbanks Corpus with a frequency of 86.68 per million words. This is closest to our highest period (1760–99), which includes all relevant uses of bit, however. Thus, bit is generally less frequent in the OBC than in Brems’ present-day data and the data indicates a somewhat uneven frequency development towards the modern period. But looking at all uses of bit together is ultimately uninformative; we need to investigate the possible variants more closely.3 13.4.1 The types of bit-construction We start with a classification into three different uses or constructions, the lexical head use, the NP of NP construction, and the adverbial modifier construction (with the latter subsuming ‘adverb degree modifier’, ‘adverb’ and ‘adjunct’ uses in Traugott 2010: 48). In the lexical head use, bit occurs as an independent noun (i.e. is the head in its NP) and is not followed by an ofphrase, as illustrated in (1). (1)
a. My Sister asked me for a Bit to make her a Petticoat, I bid her take
some, but not too much. (173610134) b. the Prisoner, toasted a Red Herring, whereof I did eat a Bit with him: (17460409)
Bit here is a direct (b) or oblique complement (a). Its meaning, ‘(small) piece or quantity’, is identical to that feeding into the NP of NP construction, which is one reason for including this use in the study. What also shows the closeness of the two uses is the fact that cases like (1) can also be interpreted as contextual ellipsis, i.e. a ‘bit of fabric’ (1a), fabric having been mentioned much earlier in the text, and ‘a bit of herring’ (1b). A further reason for inclusion is that this allows us to verify whether the independent
3
We are aware that the comparison involves very different kinds of data, but as there is no PDE investigation that is closer to the OBC, we think it can nevertheless serve as a useful, albeit rough, indication of the different state of affairs. 4 The numbers referring to OBC files indicate year, month and day, thus this example is dated 13 October 1736; we follow Huber (2007) in this practice.
A bit of (a): exploring a degree modifier 245
use declines as the more grammaticalised forms increase. The NP of NP construction occurs in the examples in (2): (2)
a. we found this ticket and a bit of cloth; (18000115) b. I awoke about twelve o’clock, and saw a bit of a light in the room
where my money was, (18390513)
c. ‘Give me a bit of that sheep’s head.’ (18780311)
The meaning is as described for the lexical use, i.e. ‘a piece/small amount of ’ cloth (2a). While (2a) shows the standard modern quantifier form, (b) and (c) are followed by a determiner, here a and that, but other forms occur as well. Thus, it will be necessary to further subdivide this category in a second step (see below). The adverbial modifier a bit, finally, lacks the preposition and modifies verbs (3b), adjectives (3a), adverbs and prepositional phrases. (3)
a. He said he was not a bit afraid of Death; (17471014) b. the policeman’s ear was bleeding a bit. (18871212)
As Figure 13.1 (and Table 13.2 in Appendix 1) show, the distribution and development of the three variants differ considerably (we give information on the statistical significance of the results for this and the ensuing tables and graphs in Appendix 1).5 The lexical head use is a minor one throughout the two centuries, and also declines towards the end. This might mean that it is pushed to the fringe by the other constructions becoming more established and grammaticalised.6 It also illustrates the claim quoted in Section 13.2 that lexical items taking part in certain constructions are more frequent within than outside of them. The NP of NP construction, in contrast, is the numerically dominant use, but one which is declining sharply in the nineteenth century, being in the end almost equalled by the adverbial modifier construction. This construction, in turn, exhibits a slow and constant rise until 1879, but then almost trebles in frequency in the last period. Frequency, more specifically increasing frequency, has been called a factor in grammaticalisation (as cause or effect), and as both of the latter constructions are grammaticalised in Modern English one might expect a parallel, not a contrary development as present in our data (see e.g. Hundt 2001: 56–57 and references therein). Also, it seems as if the a bit-construction was on the verge of overtaking a bit of; however, to confirm this, one would need data going beyond 1913.
5
We are indebted to Christer Geisler for running the statistical tests on our data with R (www.r-project.org). The mosaic plots given in Appendix 1 were produced using R’s vcd package (Visualizing Categorical Data). 6 Indeed, it is possible that the development of bit- constructions was influenced by the development of other NP1 of NP2 constructions; however, further research is required before any definitive conclusions on this point can be reached.
246 Claridge & Kytö
Frequency per 100,000 words
8 7 6 5 4 3 2 1 0
1720–1759
1760–1799
lexical head
1800–1839 Period NP of NP
1840–1879
1880–1913
modifier a bit
Figure 13.1 The development of three uses of bit (normalised)
13.4.2 Bit in NP of NP constructions As indicated above, NP of NP constructions appear in three different forms, which are best kept apart. The form without a determiner in the second NP (2a above) is the typical form of the modern grammaticalised quantifier, while a bit of a (with institutionalised indefinite article preceding the second noun) is the established form for modern intersubjective uses of various shades (hedging, evaluative). For those to become routinised, one would expect them to be the more frequent of the three uses, and this is indeed the case, as Figure 13.2 (and Table 13.3 in Appendix 1) show. A bit of N is indeed clearly the most frequent one, but it also shows the same declining pattern as seen in Figure 13.1 overall; it is clearly the form responsible for the overall decrease. A bit of a N does not exhibit a very clear development, but the last two periods might indicate that it had reached a stable level. What is also of note with regard to the last period is that the variant is by then half as frequent as a bit of N, which makes it comparatively much more prominent than before. This also means that subjectified uses might then be more noticeable. A consistently very small role is played by a bit of det N constructions, where ‘det’ stands for any kind of determiner apart from the indefinite one (e.g. the, this, my). Apart from (2c) above, other relevant examples are given in (4). These uses often go together with a partitive meaning (very clear in 2c and 4b), and only take on partitive-quantitive meanings in circumstances like (4a). (4)
a. I never took a bit of his bacon. (17770115) b. Gadsdell had but a very little bit of the dumpling, not bigger than
a nut; (18150405)
A bit of (a): exploring a degree modifier 247 6 5 4 3 2 1 0
1720–1759
1760–1799 a bit of N
1800–1839 Period a bit of a N
1840–1879
1880–1913
a bit of det N
Figure 13.2 Distribution of the three NP of NP constructions (normalised)
A bit of a cases merit somewhat closer inspection as they are not a unified construction. Some instances, like (5a), might be partitive or indeterminate in nature. A bit of a room could be understood as a separable ‘part’ of a room or it could be indicating a non-prototypical room, i.e. working like a hedge similar to kind of; it could even be a small room (quantifier use). Others are clear hedge uses to be found in the data, but whether they increase or decrease the applicability of the following noun is not always clear. In (5b) a decrease may be intended, as one may not want to admit to a large exaggeration in court. See below (Section 13.4.4) for further discussion and for potentially upscaling examples. (5)
a. she keeps a bit of a room ready furnished, which she hires.
(17621020)
b. I do not think I said that I saw six travellers with luggage turned
away in one afternoon. I do not remember whether I have seen it; it might be a bit of an exaggeration. (19070910)
The second noun in all NP of NP constructions is very relevant to their overall development. What we see quite clearly in Figure 13.3 (and in Table 13.4 in Appendix 1) is that nouns denoting concrete entities, such as candle, cloth, glass, iron, paper, are in the majority across the whole period, albeit with a pronounced decline after 1840. With regard to this group, it is also noteworthy that the proportion of food nouns (e.g. bread, cheese, meat, supper) wavers between almost one-third and one-fifth, thus making up a remarkably
248 Claridge & Kytö 100% 3 2 95% 1 90%
3 2 1
11 8 2
21
85%
0 6
80% 75% 117 70%
36
195
235
65%
105
3
60%
6
55% 50% 1720–1759
53 1760–1799 concrete
1800–1839 Period abstract
1840–1879
indeterminate
1880–1913
action
Figure 13.3 Semantic categories of nouns occurring in all three NP of NP constructions (tokens, proportions; raw frequencies in the chart)
substantial portion. Given the ultimate origin of bit in bite, the long-standing frequency of this particular group may be of special interest; it is in line with the persistence effect posited in grammaticalisation theory (Hopper 1991). Action nouns (e.g. laugh, push, sleep, singing, wrestle) increase noticeably in this period (this category contributes the most to the statistical significance of the results, see Table 13.4 and Mosaic plot 13. 3 in Appendix 1), while abstract nouns like passion, difference, doubt, (ill) will, clearly increase, but nevertheless make only a modest contribution overall. The small group of indeterminate nouns includes cases like confusion, fun, life, which refer to abstract concepts, mental states and the like but which have or can have visible/audible effects, thus making them similar to concrete/action nouns. Complete lists of the N2s can be found in Appendix 2, split up according to which constructional variant they accompany. A bit of is found with 178 nominal types, many occurring only once, but some as frequently as 102 (paper), 41 (bread) or 18 (candle) times. In all, 112 types complement a bit of a and 34 a bit of det. Various nominal types occur with all three constructional variants. There is overlap between a bit of and a bit of a (11 types), a bit of and a bit of det (8 types), and between a bit of a and a bit of det (3 types). While the noun often occurs only once in either group, there are also cases where it is frequent in one group and only an isolated occurrence in the other, thus indicating that nouns have a preference for certain constructions. One example will illustrate the kind of difference that can be produced by the different choices. In (6a) trouble indicates a state, a mental
A bit of (a): exploring a degree modifier 249
characteristic whose duration is not known or important (paraphrase: ‘troubled’ adj.). (6b) with the indefinite article, by contrast, refers rather to an individuated event/action. This seems to be a general difference: a bit of a can force an ‘action’ reading for the noun (paraphrase: ‘engaged in trouble’). While both speakers signal a certain amount of uncertainty in their statements, a bit in (6a) is nevertheless rather downgrading, while a bit in (6b) has hedging nuances: the speaker is not sure whether trouble is the correct description. Again, this reflects a tendency common to the two constructional variants. (6)
a. She came up. She was in a bit of trouble, I think, by the look of
her. She said, ‘Will you not open the carriage door to me?’ He said, ‘Do you not see we are not alone?’ He told me before this that she took fancies into her head. I was under the impression at that time that she was not quite right in her mind. (19130304) b. I came down because I heard that James Groom was in a bit of a trouble – I do not know whether he was fighting – I was going to bed, and my sister-in-law knocked at my door, and I went down to get James Groom away, because I was living in the same house, and I thought he was in a row – (18920725)
A considerable proportion of nouns following simple bit of are non-count or mass nouns (around 71–75 per cent across all periods), whereas a bit of a fairly unsurprisingly selects predominantly count nouns or, more rarely, nouns that have either mass or count uses. The latter form also shows a greater amenity to nouns denoting actions and abstract things; in fact, action nouns almost only occur in this form. There are almost no instances that refer to persons, or more generally to agents (metaphorical personification included), attributing some quality to them with the help of this construction, in the sense of Traugott’s (2007: 543) a bit of a hypocrite. OBC yields the following three occurrences spread widely across the entire sample investigated here: (7)
a. I live in a little concern of my own of land; a bit of a farmer.
(17870523)
b. I turned round and tapped him on the shoulder and said, ‘Hallo!
You are a bit of a sneak to take my book after giving you a night’s lodging’ – (19020909) c. She had a coach-horse, an iron-grey gelding, it was a bit of roarer – [italics original] (18360404)
The head nouns in both (7a) and (b) denote people, with a bit working in a hedging and evaluative manner: ‘not a prototypical farmer’ probably, and emphatic ‘what/quite a sneak!’. Both include the indefinite article, which is what would also be expected in PDE. Roarer in (7c), interestingly without indefinite article, indicates a horse typically making a certain unhealthy
250 Claridge & Kytö
breathing sound (see OED s.v. roarer n.1, 2), so that an evaluative characterisation is implied. 13.4.3 Adverbial modifier a bit Figure 13.1 above has shown that this construction is very rare to begin with, rises very slowly and only takes off in our last period. Figure 13.4 (and Table 13.5 in Appendix 1) shows the modification patterns and the development over time in the OBC. Assuming the nominal use to be the original one, then it seems to be the modification of verb phrases that is crucial in leading the extension to the adverbial function. Interestingly, verbs show an unclear and uneven development: after a clear rise in the first three periods, there is a drastic decrease in the fourth followed by another increase, but not quite to the former level, in the last period. Adjectives as modified targets come in fairly slowly and not consistently, but multiply almost six-fold in the last period. The rising modification of adverbs is perhaps stimulated by the adjectival increase. The modification of prepositional phrases remains a minor pattern throughout the period (e.g. I know his hand-writing – it is not a bit like it [18200112]). The great majority of verbs used are like the ones found in (8), i.e. they are (change of) motion/position verbs, which allow a temporal or spatial reading of a bit (see OED sense 6a, s.v. bit n.2). Stop is a fairly common verb (8a), others include stay, go, stand, ride, sit, walk, lie, move, come, step back, run. Other action verbs (e.g. physical or verbal action as in (9)) allow a temporal reading as well, but they also permit a quantity or degree reading: the extent/intensity of whipping or swearing. Thus, they could be regarded as bridging contexts. (8)
a. and Joseph Bell, I, and Littlepage went round, he told Littlepage
to stop a bit, (17830430)
b. Bowden hesitated a bit, and then said I might do my worst, for he
would not go; (18000115)
c. I felt the tail of my coat move – I raised up a bit, and saw the pris-
oner sitting next to me – (18341205) a. I beg’d the Justice to have her whipt a bit and let her go. (17350226) b. they had a dispute – the prisoner swore a bit, and told Holford that ho [facsimile: he]7 was always going to do this and that, (18811121) (10) a. she said he frighted me a little bit, but he has not done me any Harm (17470715) b. I said ‘I am not sure whether he said ‘I am sorry for it,’ or ‘I am sorry to hear it’’ – the solicitor seemed to prefer the work (9)
7
This refers to the original printed texts of the proceedings, which are accessible on the Old Bailey website and help to clear up the occasional transcription error.
A bit of (a): exploring a degree modifier 251
Frequency per 100,000 words
1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0
1720–1759
1760–1799 VP
1800–1839 Period AdjP
1840–1879
1880–1913
AdvP + PP
Figure 13.4 Items modified by adverbial a bit (normalised)
[facsimile: words], ‘to hear it,’ and I thought it did not signify a bit. (18811121) c. She seemed to realise what she had done and she pulled herself together a bit. (19120109) The instances in (10) with verbs denoting internal sensations/emotional or mental states (a, c) and abstract (b) aspects represent clear degree uses, meaning ‘frighten/pull together to some extent/somewhat’ and, in combination with negation, ‘signify nothing at all’. Verbs like those in (10) are mostly found from the late 1880s onwards ((10a) is exceptional), while before that, motion/position verbs clearly dominate. Thus, clear degree uses seem to be established slowly in this area. Adjectives modified by a bit are insignificant until the middle of the nineteenth century, with only six instances (sober, afraid, more,8 amazed, excited).9 It is actually only from the 1870s onwards that adjectives become more common. As (11a–d) illustrate, adjectives denoting a negative quality as such or in the particular context are in the majority (other examples: stunned, drunk, frightened, bloodshot, paralyzed,10 hysterical) (see Bolinger’s 8
More as in a bit more cake was counted as an adjective here. Twenty-one adjectival instances up to the 1830s were found in Claridge and Kytö (2014) with the help of a larger corpus, presenting still only a tiny fragment of the data but proving their earlier presence. 10 As these examples indicate, deverbal adjectives are not uncommon; they make up 37% of all adjectives. However, they do not necessarily cluster in the earlier periods, so they cannot be said to lead the expansion into the adjective phrase. 9
252 Claridge & Kytö
1972: 50 ‘unfavorable’ contexts). Neutral or positive adjectives (e.g. warm, smart, soft in (11c)) usually turn out to be intended negatively. (11) a. the baby was a little bit poorly; it had a little cough, but nothing of
any consequence – (18721216)
b. we called for a drink, and the prisoner’s friend got a bit quarrel-
some with us – (18920208)
c. ‘Sometimes we used to think he was a wee bit off’ – I thought him
a wee bit soft, non-compos mentis – (18980913)
d. With regard to Robertson’s evidence, the horse was a bit lame
when I took it out on the Saturday (19070910)
e. this piece is a bit thinner than the piece found round the child’s
body (19000212)
Claridge and Kytö (2014) had indicated a clear propensity for modifying comparative adjectives as in (11e), but this is found relatively rarely in the present data (six instances). Perhaps such uses paved the way, and the fact that most adjectives in our data are late shows the expansion beyond the early context. 13.4.4 Semantics and functions So far, we have mentioned readings of individual examples here and there. This is the place for a somewhat more systematic discussion of the semantic aspect, as it is relevant for the steps involved in the grammaticalisation of this construction. The development of a bit of sketched in Section 13.2 already introduced most of the functions of the construction, i.e. (i) partitive, (ii) quantifying, and (iii) degree modifying (adverb and adjunct). In addition, there are more subjective uses which convey an attitude of the speaker, whether it be with regard to (proto-/stereo–)typicality (hedge) or other characteristics (evaluative). The latter two are taken together in what follows (iv: evaluation). function (i) only applies to binominal uses (a bit of (a)), while (ii)–(iv) are possible with all constructional variants. The analysis has been carried out taking into consideration as much context information as possible, both the specific lexico-grammatical choices within the immediate (sentential/utterance) context as well as the overall content of the trial. Classifications have been based on the most likely semantic readings and interpretations of speakers’ intentions. Where no single most likely reading emerged, the category ‘indeterminate’ was used. We have already presented some of these indeterminate cases above. A bit of a room in (5a) is potentially three-ways indeterminate (partitive, quantifying or hedging), while the examples in (9) are indeterminate between quantifying and degree marking. A few more examples are presented in (12):
A bit of (a): exploring a degree modifier 253
(12) a. When they were gone there being nothing in the House but a bit
of Cheese, I went out to get something for Supper. (17350226)
b. there in the street, under a bit of a shutter, were three more loaves,
which he delivered to me. (17650227)
c. I laid hold of Armfield, and in bringing him along he made a bit of
a rustling, with his hands behind him; (18000528)
(12a) is either a partitive (‘one piece of cheese’) or it denotes an unknown, but small quantity (‘a little amount of cheese’). The shutter in (12b) is either a small one (quantifying reading) or it is an atypical exemplar (hedge interpretation). The rustling in (12c) could either be of short duration (quantity) or it could be more or less intense (degree). While there are criteria for distinguishing partitive from degree constructions (see Traugott 2007: 531), these proved not very helpful for our historical data;11 also, they do not necessarily distinguish between quantifier and partitive, or quantifier and degree function. Some examples will illustrate the functions more clearly. The instances in (13) were classified as partitives as the meaning is ‘a (small) piece/part’ of an object or material. The partitive meaning was seen as the most probable one if the reference was to some kind of identifiable unit. Candle (13a) is fairly consistently construed as a mass noun in the OBC, as something from which smaller segments can be taken away. Paper, another mass noun, exists in variously sized fragments of which the smaller in (13b) is modified by bit. While partitive meaning is most commonly found with mass nouns, it is not impossible, though rarer, with count nouns accompanied by a determiner, as in (13c). (13) a. and we had a candle, and a neighbour called me in, and asked
for a bit of candle; I took out my knife, and cut her off a bit, (17920113) b. the tea was in a bit of paper by itself, and it was all done up in paper – (18451027) c. Mr. Holton picked up a parcel, in a rag, tied round with a bit of a ribbon; (17720909)
11
The most objective criteria are the syntactic ones (agreement patterns within the NP [these bits of glass, these bit of rags] as well as between NP and VP [a bit of cheese and bacon are for sale], and preposing of the second NP [of an apple a bit]). While relevant instances were found with bits (those bits of cheese are laid for me), there was no formal number variance between both nouns with singular bit in our data. Preposing was also not found. Another criterion is replaceability of the modifier in degree constructions by a single word (a bit of a hypocrite > rather/quite a hypocrite), but not in partitive constructions. This criterion we have in fact made use of. Another criterion is that only degree constructions may have adverbial properties (a bit green) – according to this, degree uses are present from the beginning of our time frame (see Figure 13.1), but this does not necessarily help in the classification of individual examples.
254 Claridge & Kytö
In (14) the most likely paraphrase is ‘some’, indicating a certain, presumably not very large, amount. Thus, they were classified as quantifying instances. The small amount can consist of diverse individual items, e.g. three products bought. Adverbial uses as in (8) above, where bit refers to temporal or spatial extent, were also categorised as quantifying, as well as negated instances referring effectively to zero quantity (14c). (14) a. I gave him a bit of victuals for his supper: (17650417) b. they went up to the stores and did a bit of shopping
there – (18811121)
c. when we helped him over the bank, we could not see a bit of him
for blood, his face and his head were so much bruised; (17770514)
The degree meaning (‘to some extent, somewhat’) is illustrated by the examples in (15), and is found almost exclusively with the adverbial form with verbs and adjectives. (15) a. He seemed a bit stupefied, and sorry for what had taken place.
(18910209)
b. children do fret a bit when they are away from their
mothers – (19000212)
c. The prisoner, in his defence, stated, on oath, that he was a little bit in
drink, [italics in the original] (19010225)
Of great interest are the more subjectified instances illustrated in (16), which typically occur in the a bit of a form; see also the examples in (7) above. As Brems (2011b: 250) remarks for modern examples, various functions are often subtly intertwined in these cases, including emphatic evaluation, understatement, irony, hedging and face work. In each case a speaker attitude is clearly conveyed. (16) a. there was a bit of a boggy place, a ditch, or what you may please to
call it, (18000528, our italics)
b. I see these articles all lay one by one, the first thing was the ham-
mock, then a bit of an old towel, then a bit of a shift, and a bit of an old waistcoat, with the flannel side outward; I picked them up, and was going to sell them for old rags, and when this man took hold of me, (17940219, our italics) c. I know a person named Kinghorne by seeing him, and I was in his company about two years ago, when I used to conduct a little bit of a concert room of an evening (18570615) d. I live in the lower Room under the Prisoner: There was a Sort of a Bit of a Supper that Night, and the Prisoner went up to Bed between Twelve and One, (17420115, our italics)
A bit of (a): exploring a degree modifier 255
e. I was in a bit of a passion myself – after I took way [facsimile:
waay, C and K: > ‘away’] the knife she stack [facsimile: struck] me again – (18410920)
(16a) shows uncertainty about the correct naming of a place, also signalled by the following reformulation and the or-clause. Bit in (16b) is also intended to call into question the description of the items as shift, waistcoat etc., as the defendant is eager to point out their rag-like nature, which would justify his mistaking them for rags and taking them. (16c) is spoken by a brush maker and shop-keeper, not by a musician or somebody engaging in entertainments: bit of a may here mark the activity referred to as both insignificant (compared to what the person normally does) and as amateurish in nature. Both (16d) and (16e) could be upscaling in nature. (16d) is spoken by a witness, who seems to remember the night in question perhaps because something unusual was going on (i.e. the people upstairs may not have had a normal supper, but a longer, noisier affair). Both the late time of retiring to bed and the double marking (a sort of) may be taken to indicate that bit here means more than simply quantifying or hedging. However, the context for (16d) is not sufficient to be absolutely certain about bit’s function. By contrast, the context of (16e) speaks more clearly for an upscaling and/ or emphatic meaning. The witness-cum-victim tells how the accused first attacked her children and then herself; she mentioned that she was ‘aggravated’ and later that she actually ‘pushed her [the accused] down upon the ground’ and ‘called her a good-for-nothing old b–h’. We present all constructions together, as there is a functional continuum stretching across them; see Figure 13.5 (and Table 13.6 in Appendix 1). The indeterminate category is consistently between 10 and 16 per cent of the data, pointing to problems inherent in any semantic analysis, but more significantly to the importance of genuinely indeterminate uses serving as bridges between the various functions. The partitive clearly loses ground in the last two periods and becomes a minority use. By contrast, the degree use expands considerably, and with 35.5 per cent becomes almost as frequent as the quantifier use (38.5%) in the last period (for the salient contributions of the category of degree uses to the statistical significance over the two final periods, see Mosaic plot 13.5 in Appendix 1). The quantifier use fluctuates somewhat but is of a substantial presence throughout the period. Interestingly, the evaluative uses are very rare and declining in the last period. Looking at the different bit-constructions separately, the following picture emerges. Within the binominal construction, partitive meanings remain high until the fourth period (53, 49, 65, 46 per cent) and only drop to 19 per cent in the last. Quantifier meanings are correspondingly slow to rise, only reaching 49 per cent in the last period and staying (well) below 30 per cent in the
256 Claridge & Kytö 100% 80% 70% 60%
35
22
90%
4
5
3
9
108
39
25 6 3 76
28 17
31 3
5 72
64
50% 40% 30% 20%
67
126
78
131 61
10% 0%
19 1720–1759 partitive
1760–1799 quantifier
1800–1839 Period degree
1840–1879 evaluation
1880–1913 indeterminate
Figure 13.5 Functions of bit-constructions (all variants, raw frequencies within the columns). The results are statistically significant: chisquare = 20,101; p < 0.001; df = 4; however, as some expected values are less than 5, the chi-square result may not be valid
earlier periods. The degree meaning remains unimportant in this construction, with 1, 0.4, 0, 1, and 0 per cent across the periods. The subjective evaluative meanings fare somewhat better with the following development: 3, 3.5, 3, 4 and 2 per cent. The indeterminate category here remains high throughout (reaching as much as 30% in the last period), as indeterminate cases are almost only found among binominals. For verbal and adjectival modification, only quantifier and degree meanings are relevant. Quantification dominates with verbs (between c.70% and 95%), with almost equal distribution quantifier/degree (10 vs. 9 instances) only reached in the last period. There are a total of four indeterminate verbal cases. Adjectives are slow to rise as we have seen (Figure 13.4), making the last two periods statistically interesting. The quantifier/degree distributions found there are 2 to 7 and 1 to 51.12 The great expansion of the degree category in the last period shown in Figure 13.5 is due to adverbial uses with verbs, adjectives and adverbs alone. 13.4.5 Modifying bit About 20 per cent of all occurrences do not simply contain a bit, but some other form of determination or some additional premodification. 12
For this particular two-by-two table, the result is statistically significant; chi-square = 6,761; p-value = 0,010; df = 1; with Fisher’s Exact Test, p-value = 0,054; however, as some expected values are less than 5, chi-square may not be valid.
A bit of (a): exploring a degree modifier 257 Table 13.7 Items modifying bit, raw frequencies and (percentages)
a
determiners premodifiersb
1720–1759 1760–1799 1800–1839 1840–1879 1880–1913
Total
3 (2.4) 18 (13.5)
53 (6.5) 198 (19.4)
19 (7.4) 41 (14.6)
21 (10.4) 44 (18.2)
9 (6.8) 54 (31.0)
1 (1.0) 41 (21.7)
a b
Binominal patterns only. All variants except lexical heads (per cent calculations based on relevant groups only).
Occasionally used determiners are the, this, that, personal pronouns, every, many, one, and another. These only occur in the binominal frame, as in (17). The deictic determiner clearly individualises bit and makes only a partitive reading possible. (17) I don’t know no more of the thing than this bit of green leaf that lies
here. (17700425)
Adjectives premodifying bit in our data are little, good, goodish, short, small, fine, dirty, strong, considerable, old, nice and wee – most of them expressing an inherent notion of size. They can occur in all patterns, but are more frequent with NP of NP constructions. The distribution across the periods is shown in Table 13.7. While determiner use is relatively negligible and decreases after a peak, premodification is somewhat more prominent and is actually more frequent at the end than at the beginning of the period under investigation. In the last two periods, more than a third and a fifth, respectively, are accompanied by determiners and/or modifiers. On the one hand, this may detract from the cohesion of the construction as a whole as it highlights the independence of bit. On the other hand, some type of modification may actually strengthen the quantifier/degree meanings. In this respect, it is remarkable that 140 premodification instances are realised by little alone, as illustrated in (18) (see e.g. Ghesquière 2010 on two or more semantically similar items combining to trigger new, emphatic meanings). (18) a. and at the end of one case which was being heard, there was a little
bit of a bustle [italics original] in the Court, and he made a rush towards the door – (18530103) b. I have a little bit of garden before the door, (17931204) c. he then got pretty fast hold of me by my shoulders; I contested a little bit, and did not chuse to be robbed by him, (17970712)
Another twenty-nine examples contain good or goodish. These, together with wee, considerable, small, nice, actually reinforce quantity meanings, either by further downtoning or by positive appraisal of the quantity. Adjectives like fine, dirty, strong, in contrast, represent isolated occurrences.
258 Claridge & Kytö
13.5 Discussion and conclusions Bit has been shown to be clearly much more frequent within its constructional frames than in its independent use (lexical head) in the OBC data. The latter use is definitely negligible at the end of the period investigated here. The NP of NP use increases dramatically in the eighteenth century, which may be an accompanying effect of an important step in the grammaticalisation of this item taking place at this time. From a functional perspective, the quantifying and evaluative uses are also expanding and prominent in the second half of the eighteenth century. The last few decades of the eighteenth century also exhibit an interesting expansion of noun types (action and abstract nouns, the first agent noun) used in particular, but not exclusively, with the a bit of a construction. The rise may show users experimenting with the ‘new’ form, becoming familiar with it – and afterwards using it more moderately again, thus entering a consolidation phase for the newcomer. About half the readings of the NP of NP construction in the second half of the eighteenth century are other than partitive (quantifying, evaluative, degree or indeterminate). These uses, which make a bit of bread similar to some/a little bread, make reanalysis from [NP1 [of NP2]] to [[NP1 of] NP2] very likely. Our data thus confirm earlier research (see Traugott 2010: 53): the eighteenth century, especially its later decades, is of crucial importance for the grammaticalisation of this particular binominal quantifier. The new analysis consolidates as more and more NP2 nouns disallow a partitive reading (i.e. are non-concrete). For NP of NP as a whole, this process of host-class expansion (Himmelmann 2004) is a fairly slow one, but in our last period, non-concrete nouns approach 50 per cent. In the constructional variant with the indefinite article, the expansion takes place faster, action nouns being already very prominent at the end of the eighteenth century, and with its clear preference for count nouns this variant has a very clearcut profile. As to the adverbial a bit-construction, Traugott (2010: 48) thinks it likely that it came about through analogy with items such as very or quite. This works well with the modification of adjectives, but not with uses accompanying verbs. However, as we have seen, the initial expansion outside the nominal sphere takes place with verbs in the eighteenth century (Figure 13.4). Adjectival modification only becomes relevant from the latter half of the nineteenth century onwards. We thus hypothesise that the expansion to verbs was responsible for the establishment of the adverbial construction a bit. Either bit itself is taken to be simply polysemous, meaning ‘(short) while’ (see OED s.v. bit n.2, sense 6a., there attested from 1653) with motion/position verbs, or the verbal use is treated as a form of contextual ellipsis (see walk a bit ‘of the way’, stay a bit ‘of time’). Nonelliptical forms are apparently not attested, so that the polysemy assumption is perhaps to be preferred. By analogy, this use was then transferred
A bit of (a): exploring a degree modifier 259
to other types of verb, which in turn made a degree rather than a quantity reading more likely. As we have seen, there are types of verbs in the data which enable this (and a few more were found in Claridge and Kytö (2014), such as recover a bit, alter sth a bit in the 1780s and 90s). A second analogical transfer might then have taken place between verbal and adjectival modification, perhaps supported by the analogy to very and other intensifiers. Parallels or similarities between verbs and adjectives (see fright v. – afraid, both attested in the 1740s in our data) may have encouraged the transfer. The expansion to the modifying adverbs and verb phrases seems to be late, according to our data, which might mean that degree-indicating a bit has developed somewhat differently than even and barely (Traugott 2006), in whose cases examples involving PPs and NPs appear earlier in the development (for the left-peripheral expansion of the English NP from EModE onward, see Van de Velde 2011). Bit-constructions overall thus show a considerable expansion of their syntactic range. The original lexical meaning of bit ‘morsel, small piece’, however, has been bleached (desemanticised); as part of the grammaticalisation process, the noun has gained a more abstract, schematic meaning instead (quantification, degree marking) by way of invited inferencing (see Traugott 2007: 534–36). In some uses, it has also acquired evaluative meaning (subjectified, hedging uses), thereby producing a state of polysemy. What Himmelmann (2004: 33) called semantic-pragmatic extension in grammaticalisation has thus clearly taken place here. The original meaning lingers on for a considerable time (persistence), as one can see with regard to the partitive, and even with the prominence of food nouns in the data (‘bite’ meaning). The individual items in the construction become both more fixed (fusion) and increasingly decategorialised. Bit loses a considerable amount of its noun characteristics: it cannot be pluralised apart from in lexical head uses or in partitive uses, it increasingly disallows anything other than the indefinite article preceding it in quantifying and other grammaticalised uses, and it mostly restricts modification to quantity-reinforcing adjectives. Neither the indefinite article nor the preposition in the fully grammaticalised construction fulfil their usual grammatical functions any more. Finally, turning to the implications of our study and the prospects for further work, our results show that even a corpus of over 17 million words of speech-related language was not big enough to yield data on bit-constructions in quantities that would have enabled full-scale statistical significance analyses. There is clearly a need for even larger-scale corpora of material that would allow for the study of the development of low-frequency items such as certain degree modifiers in the history of English. Our study has also highlighted the need for testing, in empirical terms, recent advances in language theory, notably grammaticalisation theory. Our study lent support to many of the assumptions presented in the literature but also
260 Claridge & Kytö
yielded contrary evidence. Regarding future research on bit-constructions, we would benefit from having access to material representative of a more many-faceted collection of speech-related and speech-like genres to see how the different categories of construction would, for instance, align on the authentic vs. constructed language use axis (e.g. trial proceedings and witness depositions vs. drama and fictional speech). Bringing in various written genres representative of informational and fiction writing would help gain a further nuanced picture of the variation and change in the life cycle of these constructions. Appendix 1 In this appendix, we present further statistical information on the results obtained. Following each table that has served as the basis for a figure (Figures 13.1–13.6) in the main text, we present the corresponding mosaic plot depicting the contributions of the categories distinguished to the statistical significance of the results. In the mosaic plots, shaded areas represent standardised residuals of chi-square tests. The fields outside ±2 are of interest (for the scale, see the Pearson residuals columns in the plots). The white areas in the plots represent statistically less interesting parts of the data; the shaded areas represent statistically salient categories that contribute to the statistical significance of the table in question.
Table 13.2 The development of three uses of bit, raw (and normalised) frequencies, per 100,000 words
Lexical head NP of NP Adv. modifier
1720–1759
1760–1799
1800–1839
1840–1879
1880–1913
18 (0.5) 123 (3.4) 10 (0.3)
33 (0.9) 256 (7.2) 24 (0.7)
25 (0.7) 201 (5.4) 41 (1.1)
9 (0.3) 132 (3.9) 42 (1.2)
8 (0.3) 98 (3.3) 91 (3.1)
The results are statistically significant: chi-square = 145,955; p < 0.001; df = 8.
A bit of (a): exploring a degree modifier 261 Adv 1720
LexH
NP of NP
10
1760
24
1800
41
Pearson residuals: 8.91
123
18
256 33
1840
4.00
42 2.00 201 0.00 25 –2.00
1880
91
132 9
–4.00 –4.52 98
8
Mosaic plot 13.1 The development of three uses of bit
Table 13.3 Distribution of the three NP of NP constructions, raw (and normalised) frequencies
a bit of N a bit of a N a bit of det N
1720–1759
1760–1799
1800–1839
1840–1879
1880–1913
104 (2.9) 13 (0.4) 6 (0.2)
201 (5.7) 47 (1.3) 8 (0.2)
165 (4.4) 22 (0.6) 14 (0.4)
93 (2.7) 33 (1.0) 6 (0.2)
60 (2.0) 32 (1.1) 6 (0.2)
The results are statistically significant: chi-square = 33,690; p < 0.001; df = 8; however, as there are expected values of less than 5 in the table, the chi-square result may not be valid.
47
bi
to
fd bi
a
1760
a
13
to
fa to bi a 1720
fN
et
N
N
262 Claridge & Kytö
6
104
Pearson residuals: 3.37
8 2.00
201
1800
22 14 0.00 165
1840
33 6 93
1880
32
–2.00 –2.40
6
60
Mosaic plot 13.2 Distribution of the three NP of NP constructions
Table 13.4 Semantic categories of nouns occurring in all three NP of NP constructions (raw token frequencies) The fifteen instances of pronouns (it, him, something, anything) attested up to the 1850s were classified according to the noun they stood in for.
Concrete % food/concrete Abstract Indeterminate Action
1720–1759
1760–1799
1800–1839
1840–1879
1880–1913
117 31.6 1 2 3
235 21.3 2 8 11
195 24.1 1 2 3
105 27.6 6 0 21
53 28.3 6 3 36
The results (for the categories of concrete, abstract, indeterminate and action uses) are statistically significant: chi-square = 154,906; p < 0.001; df = 12; however, as some expected values are less than 5, the chi-square result may not be valid.
1720 1
3
1760 2
11
1800 1
3
Indeterminate
Concrete
Action
Abstract
A bit of (a): exploring a degree modifier 263
2
117
Pearson residuals: 9.04
235 21
8
4.00
1840 6
2.00 195 0.00 2 36
–2.00
1880 6
105 –3.59
3 53
Mosaic plot 13.3 Semantic categories of nouns occurring in all three NP of NP constructions
Table 13.5 Items modified by adverbial a bit, raw (and normalised) frequencies
VP AdjP AdvP PP AdvP + PP
1720–1759
1760–1799
1800–1839
1840–1879
1880–1913
7 (0.2) 3 (0.08) 0 (0) 0 (0) 0 (0)
20 (0.6) 0 (0) 3 (0.08) 1 (0.03) 4 (0.1)
32 (0.9) 3 (0.08) 3 (0.08) 3 (0.08) 6 (0.2)
15 (0.4) 9 (0.3) 12 (0.4) 6 (0.2) 18 (0.5)
21 (0.7) 52 (1.7) 14 (0.5) 4 (0.1) 18 (0.6)
The results (for the categories of VP, AdjP, and AdvP + PP uses) are statistically significant: chi-square = 75,919; p < 0.001; df = 8; however, as some expected values are less than 5, the chi-square result may not be valid.
264 Claridge & Kytö AdjP 1720 1760 1800
3
1840
9
AdvP+PP
VP
3
Pearson residuals: 4.19 4.00
7
4 6
20
2.00 18
32 0.00
1880
52 15 –2.00 18 21
–3.19
Mosaic plot 13.4 Items modified by adverbial a bit
Table 13.6 Functions of bit-constructions (raw frequencies)
partitive quantifier degree evaluation indeterminate
1720–1759
1760–1799
1800–1839
1840–1879
1880–1913
67 39 5 4 22
126 108 3 9 35
131 76 3 6 25
61 64 17 5 28
19 78 72 3 31
The results are statistically significant: chi-square = 261,181; p < 0.001; df = 16; however, as some expected values are less than 5, the chi-square result may not be valid.
3 3
1840
17
4
22
67
9
Quantifier
Partitive
5
1760 1800
Indeterminate
1720
Evaluation
Degree
A bit of (a): exploring a degree modifier 265
39
Pearson residuals: 11.80
108
35 126
4.00 25
76
6 1880
0.00
131
72
28
64
–2.00 –4.00
5 31 3
2.00
61
78
19
Mosaic plot 13.5 Functions of bit-constructions
–6.76
266 Claridge & Kytö
For Table 13.7 see section 13.4.5. The results are statistically significant: chi-square = 20,101; p < 0.001; p = 4; however, as some expected values are less than 5, the chi-square result may not be valid.
Determiners 1720
3
1760
19
Premodifiers 18
Pearson residuals: 1.96
41
44
1800
0.00
21 54
–2.00 1840
9
1880
1
41
Mosaic plot 13.6 Items modifying bit
–2.64
A bit of (a): exploring a degree modifier 267
Appendix 2 Types of nouns occurring after a bit of (numbers in brackets refer to number of tokens). a bit of N (178 types) acrobating and singing beef steak (3) blue (2) bread and butter (6)
anything (2) bees-wax blue pigeon bread and cheese (9)
broomstick cambrick (3) chalk cheese and bacon colour cotton rag diaculum plaister dish edging fish and taters food (2) glass (7) hair-shag hay-band holly India rubber it (10) lamb leaf life lint mode muslin (3) paper parcel peace poker rag (8) roast pork satin shirt front silk (3) soap sponge strapping stuff (8) tape tinder twine (2) wainscot(ting) (2) white (=silk) worsted
butter and cheese candle (18) charcoal chicken copper (8) covering difference dispute excrement fishing fun gold (9) hatband herring horse-racing indigo (2) ivory lark leather (5) lin(n)en (4) liver and bacon money (2) mutton (2) parting pitch (2) pomatum rent rope scantling shoe silver (11) something steak (4) straw (3) suet tea and sugar toast veal and pork washing (3) wire (2) yellow (=fabric)
a bit of a (112 types) affidavit blemish box
argument (2) boggy place brick place
bacon (10) binding brass (4) bread and tea and sugar cake canvas check (fabric) clay cord (3) damask (2) dinner (9) doubt fire (3) flannel (2) furniture good hat-binding (2) him (2) ill will ink jam lawn (4) leg of pork lining meat (11) mortar (2) newspaper partition plaster (2) pork (6) ribbon (9) salmon (2) scarlet shoe or stocking skin (2) sparing steel (2) string (5) sugar (4) thing tobacco (5) velveteen wax (4) wood (14)
beef (11) blood (2) bread (41) breakfast
bag bother (2) bundle
blanket (2) bowl of a spoon bustle
callico (3) carpet (2) cheese (5) cloth (7) cotton (3) deal dirt drink fish flesh gauze ground (3) hay (5) holiday impression iron (10) lace (9) lead (7) letter link metal (3) moulding paper (102) paste plaster of Paris print (3) roarer (=horse) sand sealing wax (2) shopping sleep spite (2) stick (6) string or a lace (2) supper (8) thread trouble (2) victuals (11) whalebone work (7)
268 Claridge & Kytö a bit of N (178 types) cape cord direction exaggeration fuss halt hedge job laugh (2) match noise partition petticoat quarrel (4) room scratch shift (=dress) shutter spree stone stumble thong trouble waistcoat (3) word
chip counterpane discourse faint game (4) ham hit lamb light mattrass note (2) passage pocket rag row (6) scuffle (6) shirt skirmish (2) spurt stoop supper tobacco-pipe trunk waistcoat-piece wrestle
concert room court disturbance farmer garden handle hole lane link memorandum notice passion poker (2) ribbon rustling scull shoe sneak (=person) squable stop table toss tug walk (2) yard (2)
confusion darn estate frolic grievance hawser huggle lanthorn lottery bill newspaper pair of stays path push rod sack (3) shed (2) shop spoon stab struggle (4) tea-leaf towel tussle (3) wall yarn
a bit of (det) (34 types) apron (3) charge counterpane flesh money pipe reward shawls things
back of the neck cloth crotchet brace gown note pork sack sheep’s head (2) way
bacon colouring dumpling hair (2) paper proceeds same (=sex) skirting
bar counter face (3) handkerchief pattern property satin stairs
Part V The syntax–pragmatics interface
14
If you choose/like/prefer/want/wish: the origin of metalinguistic and politeness functions Laure l J. Brinton
14.1 Introduction In Present-Day English (PDE), a variety of if-clauses containing secondperson subjects and verbs of ‘choice’, including choose, like, prefer, want, wish, and will, may function parenthetically, often expressing metalinguistic meaning, as in the following examples from the Corpus of Contemporary American English (COCA): (1)
a. Call this trouble, if you choose. But when it comes to exercise,
I’ll take ponies over goldfish, cats, and dogs any old day. (2004 COCA: MAG) b. As an exercise, try taking your series’ main characters and standing them at the edge of the abyss, face to face with the ultimate power of evil, the devil, if you like (2011 COCA: ACAD) c. ‘We’re more alike than you think, him and me. A matching pair. Fat boy and specs.’ He crossed his eyes behind his round glasses. ‘Spotty and swatty, if you prefer.’ (2004 COCA: FIC) d. they think it’s a unique approach. It’s not totally, but you could call it a huge subsidy if you want. You could call it a clever way of getting through the process. (2006 COCA: NEWS) e. So, this dirty play, if you wish, is something the ARENA party has done in these elections, (1994 COCA: SPOK) f. And it’s at that moment when HIV becomes a human epidemic, starts moving down the rivers and into the birthplace of the epidemic, if you will, in Central Africa. (2012 COCA: SPOK)
For two of these verbs, the metalinguistic uses are recognised in the Oxford English Dictionary (OED): the OED notes (s.v. will v.1, def. 17a) that if you will ‘is sometimes used parenthetically to qualify a word or phrase’, equivalent to ‘if you wish it to be so called’, ‘if you choose or prefer to call it so’; similarly, it observes (s.v. like v.1, def. 6b) that if you like can mean ‘if you wish to phrase or consider something in a particular manner; often used as a vaguely intensive expression’, equivalent to ‘indeed’ or ‘perhaps’. Brinton (2008a: 163–66) discusses if you will in more detail, noting that it is used metalinguistically 271
272 Brinton
with metaphors or figures of speech, with foreign words, and with words which are inappropriate stylistically (and thus enclosed in quotation marks); it may occur when the speaker is searching for the correct formulation, is uneasy with the characterisation (but may think the hearer prefers it), or is unsure whether the hearer will accept the characterisation; and finally it may serve as a more general hedge (‘if you wish to call it that’ > ‘if I may be allowed to call it that’). Pullum (2003) argues that if you will is a formal register, older (‘old fogey’) variant of like, namely ‘a way to signal hedging about vocabulary choice – a momentary uncertainty about whether the adjacent expression is exactly the right form of words or not’.1 In the only detailed study to date, Claridge (2013: 167, 181)2 observes that if you like in the British National Corpus (BNC) is more common in speech than in writing, is strongly favoured by male speakers, and is most common in the educational/informative and business domains. In this chapter I discuss the related clauses – if you choose, if you like, if you prefer, if you want, and if you wish. I begin by categorising the types of second-person if-clauses with these verbs (Section 14.2). I then focus on indirect conditional, elliptical forms of these clauses in PDE (Section 14.3), with attention to their characteristics (Section 14.3.1), their metalinguistic and politeness functions (Section 14.3.2), and their frequency (Section 14.3.3). Section 14.4 focuses on the syntactic and semantic development of these forms, which appear to arise in the Late Modern English (LModE) period. For example, the earliest entry for metalinguistic if you like in the OED (s.v. like v.1, def. 6b) is 1875: (2) They were placed in churches by simple faith, or credulity if you
like, but not by willful fraud (1875 T. E. Bridgett, Our Lady’s Dowry II.xi.336; OED)
14.2 Types of second-person if-clauses with choose/like/prefer/ want/wish If-clauses with a second-person you subject and a verb of ‘choice’ in the simple present tense may be complete (non-elliptical) clauses in which the obligatory infinitive or noun phrase/clause complement of the verb is present: (3)
1
a. If you prefer to exercise solo, bring company for your mind:
music, downloaded podcasts, foreign language tapes, or audiobooks. (2008 COCA: MAG)
Liberman (2003), in a response to Pullum’s posting, points to differences between the two expressions, not only in register and age of user, but also in frequency and meaning. Like, he argues, has additional meanings, such as the weakening of strong determiners, not found with if you will. As I have treated if you will elsewhere (Brinton 2008a), it will be omitted from the current discussion. 2 Claridge (2013) is a diachronic comparison of if you like, so to speak/say, and as it were.
If-ECs: the origin of metalinguistic functions 273
b. If you prefer fresh garlic, carry along a clove and saute it, (1998
COCA: MAG)
c. But if you prefer that it glisten, polishing it will not diminish its
value. (2007 COCA: MAG)
However, the clauses that constitute the subject of study here – what I term if-elliptical clauses (if-ECs) – have ellipsis of the nominal, infinitival, or clausal complement of the verb. If-ECs fall into two types. In the first type, the if-clause serves as the protasis of a conditional, occurring with an explicit apodosis. It expresses ‘literal’ conditional meaning; that is, it refers to an action that the hearer may choose to perform (‘if you prefer/want to do X’), as in: (4)
a. Fire me if you want. But I’m not giving you a free ticket. (2009
COCA: NEWS)
b. If you choose, you can drive small screw hooks into the rail for
hanging keys. (2010 COCA: MAG)
The apodosis is generally a direct imperative (4a ‘fire me’) or a secondperson declarative clause recommending a course of action (4b ‘you can drive small screw hooks …’). These are cases of ‘normal contextual ellipsis’ (Claridge 2013: 162). In the second type, the if-clause exists on its own; an apodosis is not explicitly expressed but must be inferred from context. These forms occasionally have literal meaning, as in the following examples from the Strathy Corpus of Canadian English (Strathy):3 (5)
a. I’ve got the Latin copy of his statement if you wish, but he
makes – he makes the point – (2001 Strathy: SPOK)
b. your real ‘employers’ can be deceived, if you choose, by the
adoption of an assumed (male) name (1982 Strathy: ACAD)
(5a) must be interpreted as ‘if you wish, you may see the Latin copy of his statement’ and (5b) as ‘if you choose, you can take this course of action (i.e. deceive your employers)’. But, overwhelmingly, the conditional meaning of the elliptical clause is attenuated, as in the examples in (1). Quirk et al. (1985: 1095) term these types of if-clauses ‘indirect conditions’: that is, they are ‘dependent on an implicit speech act of the utterance … and are therefore style disjuncts’; Quirk et al. mention if you will and if you like in this context. In the weakening of conditional meaning, this type of clause also resembles what Dancygier and Sweetser (2005: 217–19, 229) call ‘monoclausal’ if-clauses, structures involving the independent use of as if and if only clauses (As if it made any difference! If only he would stop drinking!). In the if only case, for 3
The Strathy contains approximately 50 million words of written and spoken Canadian English, see www.queensu.ca/strathy/projects.html.
274 Brinton non-elliptical (ex. 3)
elliptical (if-EC)
(nominal or clausal complement)
(missing obligatory complement)
explicit apodosis (ex. 4)
no explicit apodosis
direct condition (ex. 5)
indirect condition (ex. 1)
Figure 14.1 Typology of clauses: if you choose, like, prefer, want, and wish
example, the non-conditional meaning (e.g. the ‘wish’ aspect) has become the ‘sole conventional’ meaning, and the clause serves as a performative (= ‘I wish he would stop drinking’) (218–19). There is a similar resemblance to ‘insubordinated clauses’4 such as Wenn Sie sich vielleicht die Hände waschen möchten? ‘If you would maybe like to wash your hands’ (Evans 2007: 373) and If you could give me a couple of 39c stamps please (Evans 2007: 380; Heine 2012).5 If-ECs expressing indirect conditional meaning belong to a larger set of forms called ‘disjunct adverbial clauses’ by Espinal (1991: 726–27), ‘supplements’ by Huddleston and Pullum et al. (2002: 1350ff.), or ‘adverbial clauses/clausal adjuncts’ by Kaltenböck (2007: 30). Quirk et al. (1985: 910n) observe that subordinate clauses such as if you like with ellipted infinitives are ‘formulaic, and resemble comment clauses’, though they do not list these forms among their type (ii) comment clauses (i.e. ‘like an adverbial finite clause’) (1985: 1112ff.). In my view (Brinton 2008a), they function as ‘pragmatic markers’ (see also Claridge 2013). The types of if-clauses can be summarised as in Figure 14.1. Typically non-elliptical clauses occur with explicit apodoses, and both non-elliptical clauses and elliptical clauses with explicit apodoses are direct conditions.6 4
An ‘insubordinated clause’ is ‘the conventionalized main clause use of what, on prima facie grounds, appears to be formally subordinate clauses’ (Evans 2007: 367; italics original). 5 In the cases under discussion here, of course, we are concerned not with syntactically independent clauses, but with syntactically unbounded parentheticals. 6 Non-elliptical forms, when they contain an explicitly metalinguistic word, may lack an explicit apodosis and have indirect meaning, as in the following example: (i) how can we accept Thesis B or follow Leibniz’s assertion that ‘the last thing in the analysis of substances is that there are simple substances, namely souls or, if you prefer a more general word, Monads which lack parts?’ (1998 COCA: ACAD)
If-ECs: the origin of metalinguistic functions 275
14.3 Indirect conditional if-ECs in PDE 14.3.1 Characteristics The if-ECs exemplified in (1) – i.e. those containing no explicit apodosis and expressing indirect conditional meaning – share the following characteristics. First, they are syntactically independent from the ‘host’ clause (or ‘anchor’; see Huddleston and Pullum et al. 2002: 1351), usually set off by commas and presumably prosodically independent. Occasionally they constitute a separate utterance. Thus, by most definitions, they are ‘parenthetical’ (see Dehé and Kavalova 2007: 1–22 and Brinton 2008a: 7–14 for summary discussions of parentheticals). Second, the host clause is complete (self-sufficient) without the if-EC. Third, the internal structure of the if-EC is elliptical; that is, the verb is missing an argument required by its valency: choose, like, prefer, want, and wish typically function as transitive verbs. Thus, they resemble ‘reduced parenthetical clauses’ in the sense of Schneider (2007: 77–78).7 Note that elliptical parenthetical clauses with to also exist, in the case of if you want to and infrequently if you choose to, but because they always express literal conditional meaning, they are excluded from this study. Fourth, if-ECs may have literal or non-literal (metalinguistic) meaning. Fifth, you does not function in a truly referential manner (Claridge 2013: 162). Finally, they have (limited) mobility: they generally follow, but may sometimes precede the word(s) they refer to: (6)
a. And James is playing between the, ug, if you like, the lights and
darks of moments of realization (2002 Strathy: SPOK)
b. a spa–very spa–very special, or if you like, special or specific
sequence as a template (2001 Strathy: SPOK)
Indirect conditional if-ECs have the characteristics of what Kaltenböck et al. (2011) call ‘conceptual theticals’: that is, they are syntactically independent, prosodically set off from the rest of the utterance, positionally mobile, and internally built upon the principles of sentence grammar, but elliptical (2011: 853). Borrowing a concept from Huddleston and Pullum et al. (2002: 1350–62), Kaltenböck et al. describe the semantics of theticals as ‘nonrestrictive’, by which they mean that they are not semantically part of the host clause but concern the ‘situation of discourse’ (2011: 856).
See also the examples in (23) below. But as non-elliptical structures are otherwise (i.e. without an explicitly metalinguistic expression) always direct, I do not feel it necessary to subdivide this category. 7 Because of the presence of the complementiser if, however, Schneider’s criterion (2007) that there be ‘no overt syntactic link’ between the host clause and the parenthetical does not hold in this case.
276 Brinton
14.3.2 Metalinguistic and politeness functions Quirk et al. (1985: 1095–96) point out that ‘indirect conditions’ may serve a number of functions. First, in a politeness function, the indirect condition ‘makes the speaker’s utterance seemingly dependent on the permission of the hearer’ (1985: 1095) (e.g. if I may say so). Second, in a metalinguistic function, the indirect condition suggests either ‘that the wording is not quite precise or that it should not be misunderstood in some sense not intended by the speaker. It explicitly or implicitly calls for the hearer’s agreement’ (1985: 1095) (e.g. if one may put it so). Third, it may point to the speaker’s uncertainty about the correctness of the wording or his/her uncertainty about the hearer’s interpretation of the wording (1985: 1096). Furthermore, such forms, like metalinguistic comments generally, can be co-opted as hedges (1985: 618) or used as emphasisers, drawing attention to the hyperbolic nature of the language used (1985: 619). The if-ECs considered in this study serve a variety of metalinguistic and politeness functions consistent with Quirk et al.’s general description: (a) In rare instances, the expression is explicitly marked as metalinguistic, as with the phrase ‘for want of a better term’ or ‘which sounds less threatening’ in (7): (7)
a. Then you’ve got, if you like, for want of a better term, new wave
cheesemakers (1998 COCA: SPOK)
b. Known to every householder today – rot; or, if you prefer,
decay, which sounds less threatening (1992 BYU-BNC: W [instructional])
(b) An explicitly metalinguistic ‘call it X’ construction, as in (8), is fairly common: (8)
a. Call me an enlightened defeatist if you wish (1993 BYU-BNC: W
[fic])
b. If the intent is to somehow regulate (call it canonize, if you pre-
fer) what gets read as ‘American Indian literature,’ … previously unquestioned notions of group membership must be teased to the surface and examined. (1997 COCA: ACAD) c. call it ‘consciousness-raising’ if you like (1982 Strathy: ACAD) (c) At times, the speaker provides an alternative way of naming an item. The alternative formulation may differ in register or even language, but does not seem strongly expressive. Rather, it is used for variation or sometimes humour: (9)
a. biological clock, or self-cycle clock if you like (2001 Strathy:
SPOK)
b. Fresh sea urchin wrapped in seaweed. Or ‘nori’ if you prefer,
(1999 COCA: FIC [movie])
If-ECs: the origin of metalinguistic functions 277
c. ask your butcher for the tail vertebrae of cattle – ox tail, if you
prefer (1986 Strathy: MAG)
d. The Kingston area’s fourth largest and fastest growing industry is
unemployment insurance – pogey or, if you wish, the dole. (1976 Whig-Standard [Kingston, ON] 6 Jan. 1/6; OED)
Thus, the two terms ‘biological clock’ and ‘self-cycle clock’ in (9a) do not express clear connotational differences. The use of ‘tail vertebrae of cattle’ for ‘ox tail’ in (9c) may be for purposes of humour or mild shock. In (9d) ‘pogey’ is a recognised informal term for ‘unemployment insurance’ (see Canadian Oxford Dictionary, s.v. pogey).8 (d) The speaker may use an if-EC if the concept is metaphorical or is expressed in a flamboyant or exaggerated manner and if he or she is perhaps unsure of the reception of the language used: (10) a. the honour of the assassins as a group – as a guild, if you like –
would be tarnished (2000 Strathy: FIC)
b. What is cancer but ‘new growth’ – ‘creation’, if you like (1990
BYU-BNC: W [non-fic])
c. very much the same as trade effluent but the sewers were open, if
you like (1993 BYU-BNC: W [acad])
Discussing the use of literally, Quirk et al. (1985: 619) point out that metalinguistic meaning is ‘inextricably mixed up with expression of degree’: ‘It would seem that the speaker wishes to emphasise the extraordinary nature of what he is describing – ‘Believe it or not!’ –, as well as to draw attention to the hyperbolic language used to describe it.’ A similar sense of presumed disbelief may accompany the examples in (10), though not the more valueneutral examples in (9). (e) If the speaker is struggling with how to name something or expresses some qualification on the way it is named, an if-EC may be inserted: (11) a. Michael Freeman once wrote that a style becomes a mark of
quality or territorial signpost – a trademark, if you wish (2006 COCA: ACAD) b. I don’t know, it’s a fascination. It’s my hobby, if you like. (1975 BYU-BNC: W [fic])
(f) If-ECs are frequently used when the speaker’s designation for something differs from that of the hearer (or others). In some cases, it would seem that both the speaker and the hearer would reject the alternative description; the speaker’s mentioning of this other (but unacceptable) term is a means for finding common ground (i.e. 8
A reviewer suggests that (9d) exemplifies a common strategy for referring to recent loans.
278 Brinton
‘Others might describe it this way, but both you and I know that this is wrong’) as in (12). Or it may be that the speaker and hearer actually conceptualise and label the entity differently. Here, there are two possible effects of using the if-EC. On the one hand, the speaker, by mentioning the alternative expression, acknowledges the differing viewpoints of speaker and hearer and expresses deference to the hearer (negative politeness) (‘I acknowledge that you and I see this matter differently and your point of view is/may be legitimate’) as in (13). As Claridge notes, if you like leaves it ‘partly up to the addressee to decide’ upon the applicability of a particular phrasing and is hence strongly interpersonal (2013: 162, 176). On the other hand, the speaker may be emphasising the differing viewpoints and expressing defiance towards the hearer (flouting negative politeness) (‘You and I see this matter differently, and I don’t accept your point of view’) as in (14). (12) a. the Memphis mansion where Elvis lived from 1959 until
his death – or, if you wish, disappearance – in 1977 (2001 COCA: NEWS) b. He specifically criticized Bush’s cautious, or, if you prefer, lame response to the upheaval in Eastern Europe (1990 COCA: ACAD) (13) a. a species designed by the Almighty, or by evolution if you prefer (1996 Strathy: ACAD) b. He, or She, if you prefer, could do better than that (2001 COCA: FIC) c. Education needs to be and now has the capacity to be catholic – ‘multicultural’ if you wish. (1995 COCA: ACAD) (14) a. we, the dominant culture – ‘the boys,’ if you like – were indeed guilty (2001 Strathy: SPOK) b. the public has decreed that capital punishment is a just punishment (call it revenge, if you like) for crimes committed where an innocent’s life is taken (1999 COCA: NEWS) Whether the speaker is abiding by the rules of negative politeness (13) or flouting them (14) is difficult to distinguish without a larger context. The use of scare quotes in (14a) suggests, however, that the writer does not accept the representation of the dominant culture as ‘the boys’, and the use of the emotionally charged ‘revenge’ in (14b) also suggests the writer’s discomfort with this characterisation of capital punishment. (g) Finally, the if-EC may serve as a general hedge, equivalent to ‘indeed’ (see the OED definition of if you like cited above; Quirk et al. 1985: 618; Claridge 2013: 162, who notes the ‘basic hedging function’ of if you like):
If-ECs: the origin of metalinguistic functions 279 90.0 % if -ECs
80.0
% if -ECs that are indirect
70.0 60.0
%
50.0 40.0 30.0 20.0
if you choose
if you wish
if you like
COCA
BYU–BNC
COCA
Strathy
BYU–BNC
Strathy
COCA
Strathy
if you want
BYU–BNC
COCA
BYU–BNC
Strathy
COCA
BYU–BNC
0.0
Strathy
10.0
if you prefer
Figure 14.2 Percentage of second-person if-ECs with choose, want, wish, like, and prefer (black columns), and percentage of these that are indirect or metalinguistic (grey columns) in three corpora of PDE9
(15) a. Such if you wish are the ringings of this great painter’s brush
(1988 Strathy: MAG)
b. Well, now, there’s a dispensation of Providence if you like. (1990
BYU-BNC: W [fic prose])
c. Said to remember nothing about it. That’s interesting, if you like.
(1990 BYU-BNC: W [fic prose])
d. Does it really mean anything to say that a flock of thousands of
crows or, if you prefer, hundreds of thousands of crows noisily fluttered around in the skies of Samarkand (2003 COCA: FIC)
9
COCA and the BYU-BNC were searched for the relevant strings followed by a period, comma, and close parenthesis. Random samples of 100 were investigated for if you want, if you wish, and if you like. The COCA search was undertaken on 13 March 2012.
280 Brinton
14.3.3 Frequency The frequency with which the individual verbs choose, want, wish, like, prefer occur in if-ECs, and with indirect (metalinguistic) meaning, varies widely. A rough comparison of these verbs in three dialects, British English (the British National Corpus), Canadian English (the Strathy Corpus of Canadian English), and American English (the Corpus of Contemporary American English), shows that the verbs fall into two classes, with like, prefer, and wish occurring frequently as parentheticals with metalinguistic meaning and want and choose being rarely so used (see Figure 14.2): • if you like is elliptical 50–81% of the time, with 27–46% of these being metalinguistic; if you prefer is elliptical 30–41% of the time, with 25–57% of these being metalinguistic; if you wish is elliptical 32–46% of the time, with 2–16% of these being metalinguistic. • if you choose is elliptical 6–24% of the time, with 0–2% of these being metalinguistic, and if you want is elliptical 5–8% of the time, with 0–4% of these being metalinguistic. Although the figures are too tentative to draw any firm conclusions, it would seem that Canadian English shows the highest frequency of usage, followed by American English, and then British English. 14.4 The development of indirect conditional if-ECs Studying the historical development of if-ECs poses a number of problems, primarily because of their low overall frequency and late appearance. It was necessary to make use of a wide variety of historical corpora and text collections, which are not directly comparable. With these limitations in mind, however, I searched for the earliest elliptical uses of these clauses in the Corpus of Historical American English 1810–2009 (COHA), the Corpus of Late Modern English Texts, Extended Version (1710–1920) (CLMETEV), the ChadwyckHealey English Drama (ED) and Eighteenth-Century Fiction (ECF) text collections, The Proceedings of the Old Bailey – London’s Central Criminal Court, 1674 to 1913 (OBP), the OED quotation database, and The Modern English Collection, University of Virginia Electronic Text Center (UofV).10 14.4.1 Earliest elliptical forms with if you choose/like/wish/prefer/want Corpus research suggests that elliptical structures with if you choose/like/ prefer/want/wish are a LModE phenomenon, beginning to appear in the 10
For full details on these corpora and text collections, see the List of Data Sources (Corpora). The COHA searches were undertaken using the same search strings as in COCA (see footnote 9). For the other corpora, text collections, and OED quotation databases, all examples of the relevant clauses were collected and then manually processed using punctuation as a guide but keeping in mind that punctuation in early texts may not be identical to modern
If-ECs: the origin of metalinguistic functions 281
second half of the eighteenth century. These have literal (conditional) meaning and date from 1763 to 1848, as shown in (16): (16) a. I said no, gentlemen, they are copper, but if you choose, you
shall have them (1763 John Routh, Violent Theft; OBP)
b. Assuredly the Thing is to be sold, Go with me, if you like, (1723
Johnson, Love in a Forest; ED)
c. to deliver to you the whole set, that is the four large Corinthian
and ten Ionic ones, in the month of October 1820, together with their bases if you wish, (1819 Jefferson, Letters to and from Jefferson; UofV) d. Go, – join St. Leger to devastate your native land, or join, if you prefer, the savage brigands and their more savage allies (1848 COHA: FIC) e. You, I know, have a thousand ways to get off still; or if you want, that Gentlemen can supply you (1677 D’Urfey, A Fond Husband; ED)11 Note that these clauses are elliptical but not parenthetical: they are still syntactically linked to their host clause. They serve as the protasis linked to an explicit apodosis (e.g. ‘if you choose, [then] you shall have’ [16a]). The earliest examples of elliptical forms with clear metalinguistic meaning are given in (17)–(21). They appear in the hundred years extending from the first quarter of the nineteenth century to the first quarter of the twentieth century. Note that the most speech-based corpora used, The Proceedings of the Old Bailey (extending back to 1674) and the Chadwyck-Healey English Drama collection (extending back to the late thirteenth century), which would be expected to provide the earliest instances, yield no early metalinguistic forms.12 (17) I ought to occupy the foreground of the picture; that being the hero of
the piece, or (if you choose) the criminal at the bar, my body should be had into court. (1822 De Quincy, Confessions of an English Opium Eater; CLMETEV) (18) a. The humbleness, meanness if you like, of the subject, together with the homely mode of treating it, brought upon me a world of ridicule by the small critics (1823 COHA: FIC)13 punctuation practices. CLMETEV was searched using AntConc 3.3.5. Since the research for this paper was undertaken, CLMETEV has been superseded by CLEMET 3.0. 11 The next example of if you want that I have found dates from over 150 years later: (ii) So if you want, you can draw more upon the cider, and may be my good woman will give you a bit of bread and cheese (1824 COHA: FIC) I do not know what to make of this very large temporal gap. No examples of ellipted if you like, if you want, or if you prefer, either literal or metalinguistic, occur in the earlier Corpus of English Dialogues (1560–1760), nor are there any examples at all of if you choose or if you wish in the corpus. 13 Although she uses a somewhat different set of corpora, Claridge’s dating of the first instance of indirect conditional if you like is identical to mine (2013: 174). She sees it as ‘clearly a 12
282 Brinton Table 14.1 Dating of if-ECs
if you choose if you like if you wish if you prefer if you want a
Earliest elliptical form
Earliest metalinguistic parenthetical
1763 1723 1819 1848 1677/1824a
1822 1823 1902/PDEb 1900 1934
See footnote 11. b See footnote 14.
b. who are hospital nurses without wages – sisters of Charity, if you
like, without the romance and the sentiment of sacrifice (1847–48 Thackeray, Vanity Fair; CLMETEV) (19) ‘Very well, Madam!’ the manager replied quickly. ‘You shall be “peerless” if you wish. Every fence shall proclaim it; every post become loquacious with it.’ (1902? COHA: FIC)14 (20) ‘In England? Why especially in England?’ ‘In English-speaking – in Anglo-Saxon lands, if you prefer. The Anglo-Saxon public is beautyblind. (1900 COHA: FIC) (21) a. Well, you’ll probably be dropping into the shop some day. I can not know you from Adam – and Eve, if you want,’ he said, (1934 COHA: FIC) b. Their roots were in something simple, and vital, Primitive, if you want, (1954 COHA: FIC [play]) In their metalinguistic meaning, the clauses serve as indirect conditionals, as the protasis of an implicit speech act (Quirk et al. 1985: 1095). For example, in (17) an apodosis such as ‘you may call it so’ must be supplied: i.e. ‘the hero of the piece, or (if you choose [you may call it so]) the criminal at the bar …’. These forms are now syntactically detached from their host clause and thus function parenthetically. Table 14.1 summarises the appearance of the first elliptical uses (Quirk’s so-called ‘indirect conditional’) with literal meaning as well as those with metalinguistic meaning. 14.4.2 Semantic development As Table 14.1 shows, metalinguistic meanings consistently arise (60– 100 years) later than literal content meanings.15 Semantically, then, if-ECs nineteenth century form’. In the American data, there is a slight rise in the use of if you like over the century (178). This is a somewhat uncertain example. Instances of metalinguistic if you wish do not occur prior to PDE. 15 Whether these meanings arise gradually via a process of grammaticalisation or instantly via a process of ‘co-optation’ (see Kaltenböck et al. 2011: 874–75; Heine 2013) is not addressed here. 14
If-ECs: the origin of metalinguistic functions 283
follow the expected development from referential/propositional > nonreferential (pragmatic/procedural/metalinguistic) meaning, following the pathway formulated by Traugott and Dasher (2002: 40, 225): content > content/procedural > procedural meaning Loss of ‘contentual’ conditional meaning is accompanied by the acquisition of metalinguistic meaning; the latter is procedural because it acts as ‘instructions or linguistic “road signs”’ (Hansen 1998: 199) about ‘how to “take” … mental representations, what context to process them in, what conclusions to draw from them’ (Rouchota 1998: 98). As Traugott and Dasher (2002: 10) observe, procedural meanings ‘are primarily indexical of [the speaker/writer]’s attitudes to the discourse and the participants in it; they index metatextual relations between propositions or between propositions and the non-linguistic context’.16 If-ECs also follow a second pathway of semantic development set out in Traugott and Dasher (2002: 40, 225): non-subjective > subjective > intersubjective Subjective meanings ‘are recruited by the speaker to encode and regulate attitudes and beliefs’ (Traugott 2010: 35). Thus, when speakers use an if-EC to express doubt about the validity of a linguistic formulation, to mark the formulation as metaphorical or exaggerated, or to allow for the possibility of alternative formulations, they are imbuing the form with subjective meaning. Intersubjective meaning encodes ‘meanings centred on the addressee’ (Traugott 2010: 35), so the use of if-ECs to seek common ground, to express deference to a hearer who might formulate matters differently, or even to contradict a hearer’s point of view, all constitute clearly intersubjective uses.17 (See Section 14.3.2 above on the metalinguistic and politeness functions of if-ECs.) 14.4.3 Syntactic development The change from conceptual to metalinguistic meaning seems to coincide with a shift from more tightly integrated adjunct adverbial (conditional) clause to more loosely integrated parenthetical disjunct. For example, Brinton (2008a: 168–71) finds that if you will derives from an adjoined adverbial clause with referential meaning (‘if you are disposed to [do X]’]. Thus we would look for the historical source of if-ECs in complete clauses with literal meaning functioning as the protasis in a conditional structure. Such a derivation from complete structures implies that if-ECs are elliptical in 16
See also Kaltenböck et al.’s (2011: 856) notion of ‘non-restrictive’ meaning (given in Section 14.3.1 above). An excellent overview of (inter)subjectification is López-Couso (2010).
17
284 Brinton
two senses: first, they are missing the complement structure required by the verb’s valency, and second, they are missing their main clause (the apodosis). The following sections will explore whether there is diachronic evidence for underlying non-elliptical structures. 14.4.3.1 Existence of obligatory complement structures? Synchronically, it has been suggested that elliptical clauses without obligatory verbal complements come from underlying fully formed clauses. For example, Quirk et al. (1985: 909–10) include if you want/wish among verbs of saying or thinking that show ellipsis of their to-infinitive (possibly with stranding of to), and Huddleston and Pullum et al. (2002: 1529) describe if you want/prefer/like/choose as ‘special cases’ of ellipsis following verbs that normally do not permit it. In a diachronic context, however, derivation of clauses without obligatory complements from structures in which the complements are present would necessitate that (a) the specific syntactic form of the complement be easily reconstructable, and (b) the full complement structures be of high frequency (see e.g. Bybee and Hopper 2001: 14–16 on the importance of frequency on the formation of constructions). However, this set of verbs is historically followed by a wide variety of complement structures, thus making syntactic reconstruction of the ellipted complement difficult: (22) to V: if you wish to preserve me … (1678: ED)
NP: if you prefer a suit … (1594: ED) PP: and if you like of it … (1664: ED) S: if you like what here at first she shows … (1696: ED) NP to V: if you want witnesse to confirme the deed … (1602: ED) it: if you wish it … (1658: ED) V–ing: If you like better turning it to money … (1640: ED)
Moreover, forms with complete complements – and metalinguistic meaning – occur only very rarely: (23) a. no otherwise assume the Apostolate of England (if you choose to
use the phrase) than I assume the Apostolate of all Europe (1748 J. Wesley, Wks. (1872) XII. 102; OED) b. so edging herself a little more towards him, and raising up her eyes, sub-blushing, as she did it – she took up the gauntlet – or the discourse (if you like it better)18 and communed with my uncle Toby, thus. (1769 Sterne, Tristram Shandy; ECF)
18
Anaphoric it occurs with noticeable (but not overwhelming) frequency in these constructions. Although parentheticals are generally thought to be syntactically independent of their hosts, it has been shown that they may be bound by antecedents (see Dehé and Kavalova 2007b: 4).
If-ECs: the origin of metalinguistic functions 285
c. the prosecutor and Spooner were drunk, and if you wish to
have it so, very drunk (1840 Sarah Maher, Theft; OBP)
d. Those fertile plains … were my forefathers’ heritage. – A southern
robber – a conqueror, if you prefer the title, came with mercenary bands, and reft them of the land. (1811 Morton, The Knight of Snowdoun; ED) e. You might term them ‘volatile’ and ‘sessile’; or, after Homer, ‘winged’ and ‘unwinged’; or, if you want Greek, ‘planetic’ and ‘aplanetic’. (1894 C. S. Peirce, Let. 1 Jan. in R. B. Perry Thoughts and Char. W. James II. 414; OED) Thus, derivation from full clauses is questionable, whether one assumes the rise of metalinguistic meaning before the ellipsis of the complement structure or the reverse.
14.4.3.2 Existence of apodosis (then-clause)? As noted above (Section 14.2), metalinguistic if-ECs bear a resemblance to ‘monoclausal if-clauses’ or ‘insubordinated clauses’. These have been postulated to derive historically from bipartite clausal structures. For example, Evans (2007: 371–75, 430–31) argues that insubordinated clauses begin in a subordinate construction with an overt main clause. Ellipsis of the main clause ensues. In the early stages, it may still be possible to restore a grammatically compatible main clause (Evans 2007: 372, 430). For example, an insubordinated clause such as German ob diese Wortstellung zulässig ist ‘whether this word order is permissible’ might suggest a main clause such as ‘seems doubtful to me’ or ‘This article makes clear, as to’. But Evans admits that it is difficult to say whether there are any limits on which ellipted clauses can acceptably be reconstituted. The clause then acquires greater semantic specificity and only a subset of ellipted main clauses can be restored. The ellipted main clauses can be very general (It would be nice …) or, on account of the restriction of the interpretation of ellipted material, very specific (2007: 372–73). Finally, the original subordinate clause is fully nativised as a main clause and assumes pragmatic meaning (374–75). Heine (2012) likewise argues that the insubordinated clause is historically derived from a full conditional construction, in the course of which change the apodosis main clause is ellipted via ‘co-optation’ (2012a: §§3, 5). In the case of if you choose/prefer/like/want/wish can we find historical evidence for the existence of a then-clause, an apodosis? Both Evans and Heine admit to certain difficulties in ‘reconstruction’ of the ellipted apodoses of insubordinated clauses. According to Evans (2007), ‘exactly which main clause is restored is determined by a process of conversational implicature’19 (2007: 374) and once constructionalisation has occurred it may not 19
Dancygier and Sweetser (2005: 218) note that with monoclausal if-clauses, the apodosis is left ‘to the hearer’s contextually prompted construction’.
286 Brinton
be possible to restore any ellipted material (2007: 374).20 Heine (2012a) too admits that reconstruction is ‘hypothetical’ (2012a: §3), since in most cases it is not possible to reconstruct the exact form of the ellipted matrix clause (2012a: §5). Importantly, Heine (2012a: §5) admits that there is no diachronic evidence for the reconstruction: there is only synchronic evidence. We find explicit apodoses when if you choose/prefer/like/want/wish has literal meaning, as in the example in (24a) (also see (4) above), but even here the apodosis may be omitted (24b): (24) a. However if you choose, you could testify (1994 COCA: SPOK) b. Maybe we can leave my door open tonight, Lynn? ‘Of course. If
you like.’ (1990 COCA: FIC)
When the construction has metalinguistic meaning, we find examples such as (7) above, where the if-EC is accompanied by an expression such as ‘for want of a better word’, and very occasionally, there is something approximating an explicit apodosis referring to the use of a particular expression, as in ‘is a word you could use’ (25a) or ‘or shall we say it’ (25b). (25) a. Anonymous. Free is a word you could use if you like (2008
COCA: FIC)
b. Cla. (humming) ‘Groves of Blarney’)
A cottage-orny! with a little lawn-y! Bab. Is that rhyme? Cla. It is words and music mine own. Thus ever and anon, the poetsoul welleth up into spontaneous song. Or shall we say, if you prefer it, a semi-detached vill? (1879 Melville, The lady of Lyons; ED)
But it is fair to say that explicit apodoses are not normally possible when the meaning is metalinguistic: (26) I am tenebrous, or tenebrious if you prefer [*then you can call it so],
it’s all the same (1997 COCA: FIC)
Thus, unlike mono- or insubordinated clauses, if-ECs are not easily traced back to bipartite structures. 14.4.3.3 Alternative analysis Historically, then, it proves difficult either to reconstruct the complement of the if you choose/prefer/like/want/wish or, in its metalinguistic meaning, to postulate an adjoined apodosis. However, a possible alternative diachronic source for if-ECs presents itself in the ‘call it X’ sentence exemplified (in (1a 20
Note that for Evans, ‘constructionalisation’ is understood as ‘conventionalization of the whole construction’, in which the ‘newly independent clause acquires a more specific constructional meaning’ (2007: 375). This involves the closing down of meaning, or ‘depragmaticisation’. For another sense of ‘constructionalisation’, see below, Section 14.4.3.4.
If-ECs: the origin of metalinguistic functions 287
and d, 8)) above. Here the apodosis (call it X), referring to the speech act of naming, is explicit and the meaning is overtly metalinguistic. Originating in a separate construction such as this would be consistent with the independent or ‘parenthetical’ (Schelfout et al. 2004) analysis of parentheticals (as opposed to the ‘extraction’ analysis) in which the parenthetical begins as a syntactically independent utterance and is then inserted into the host clause with which it bears no syntactic relationship (see further Brinton 2008a: 11–12). Examples of the ‘call it X’ sentence with if you will occur quite early (27), while the earliest examples with prefer/like/want/wish/choose date from the mid to late nineteenth century and coincide with the other metalinguistic uses (see Table 14.1). The verbal complement may be present (28a) but is usually absent (28b–d). (27) That slyme was a fatnesse that issued out of the earth, like vnto tarre;
and thou mayst call it cement, if thou wilte. (1530 Tyndale Prol. 5 Bks. Moses in Wks. 6/2; OED) (28) a. ‘Call it Holland, then, if you prefer that country most.’ (1831 Cooper, The Water-Witch; UofV) b. Call it cruelty if you like, not mercy (1860 COHA: FIC) c. The Red Queen shook her head, ‘You may call it ‘nonsense’ if you like,’ she said, (1865 Carroll, Alice in Wonderland; CLMETEV) d. this intelligence is not blind, but ‘absolutely clairvoyant and infallible’. We may call it, if we choose, a ‘supra-consciousness’. (1879 Unitarian Rev. Oct. 387; OED)
The ‘call it X’ structure presents a more plausible source for the metalinguistic if-EC than the bipartite structure hypothesised in (26) or the structures with complete complements shown in (23).21 Such an origin would suggest that metalinguistic meaning does not derive from (non-linguistic) actional meaning but is present in the original form, which refers explicitly to the speech act of ‘calling’ or naming something. That is, we must consider the function of the if-clause as a metalinguistic marker to be already present before ellipsis of the main clause occurs.22 The development of more (inter)subjective politeness functions remains consistent with the pathway discussed in Section 14.4.2. 21
Claridge (2013: 173) points to both the ‘call it X’ construction and forms with explicitly metalinguistic complements such as ‘if you like the term’ or ‘if you like to say so’ as possible sources for if you like. She notes that the metalinguistic item is ‘easily omissible, as the concept is inherent in any speech act’ (173). However, she seems to propose that if you like originates in an elliptical structure: call it truck, if you like Δ … (177). 22 The original presence of metalinguistic meaning provides at least partial support for Heine’s (2012) view concerning the rise of pragmatic functions in insubordinated clauses such as If you could give me a couple of 39c stamps please. Heine (2012) questions at what point ellipsis of the main clause occurs – before, during, or after its ‘co-optation’ as a polite request – concluding that co-optation must coincide with ellipsis. Evans (2007: 374–75), on the other hand, seems to see ellipsis as preceding the rise of the pragmatic meaning; he
288 Brinton
14.4.3.4 Construction Grammar analysis The historical survey in this chapter has suggested that there is no nonelliptical direct conditional source for the metalinguistic if-EC, i.e. no source structure which then develops by a simple set of syntactic changes (ellipsis of the complement structure, ellipsis of the apodosis clause). Moreover, there would appear to be multiple and diverse sources. These facts make a straightforward grammaticalisation account difficult to postulate. A Construction Grammar approach (see Traugott and Trousdale 2013) might thus prove to be a more fruitful means to understand the development of metalinguistic if-ECs. In this view, there would exist a number of specific constructs with verbs of ‘choice’ in second-person if-clauses which lead to a more abstract micro-construction via the process of ‘constructionalisation’.23 These constructs include: • elliptical constructs with explicit apodoses and literal meaning – see examples in (4) and (16) • elliptical (and non-elliptical) constructs with explicit apodoses and metalinguistic meaning of the ‘call it X, if you “choose”’ type (encompassing constructs such as call it X if you prefer, say X if you choose, use word X if you like) – see examples in 28; these also encompass a variety of complement structures as shown in (22) • non-elliptical constructs with implied apodoses and explicit metalinguistic complements of the ‘if you “choose” X’ type (encompassing constructs such if you choose to use the phrase, if you like it better, if you prefer the title) – see examples in (23). What are likely not important sources (which would be proposed in a grammaticalisation framework) are non-elliptical if-clauses with explicit apodoses and literal meaning such as those in (3) (If you prefer fresh garlic, carry along a clove and saute it) and elliptical if-clauses with explicit apodoses and metalinguistic meaning, such as the form hypothesised in (26), but not found to occur (I am tenebrous, or tenebrious if you prefer [*then you can call it so], it’s all the same). ‘Constructionalisation’ is understood here as ‘the ways in which ambient linguistic patterns may give rise to the emergence of new constructions of different degrees of schematicity and substantivity’ (Hoffmann and Trousdale 2011: 12). Specifically, we have here a case of ‘grammatical constructionalisation’, involving movement towards the grammatical pole, notes that once a subordinate clause becomes independent, it becomes available for pragmatic interpretation. On the relation of grammaticalisation to constructionalisation, see e.g. Noël (2007) or Gisborne and Patten (2011). Noël believes it is necessary to distinguish the initial formation of constructions (the association of meaning with a particular (morpho)syntactic form), which he calls ‘schematisation’ or ‘constructionalisation’, from ‘grammaticalisation’, the possible later changes of a construction into a ‘more grammatical’ one (2007: 195).
23
If-ECs: the origin of metalinguistic functions 289
with the concomitant development of procedural meaning.24 ‘Procedural’ encompasses, among other meanings, speaker stance and perspective, such as modality or metatextual markers (Traugott 2012a, 2012b). The ‘necessary and sufficient’ conditions for grammatical constructionalisation include: • an increase in generality or abstractness • an increase in productivity • a decrease in compositionality (see Trousdale 2008a: 55–56, 2012: 175; Traugott 2012b; Bergs 2012: 1637– 42). Discussing the development of the modals as auxiliaries, Trousdale (2012: 176; see also Trousdale 2008a: 52–56) describes the process of change as beginning with a series of constructs (attested tokens), some with divergent formal behaviour but with relative semantic coherence. As speakers identify commonalities among constructs, a series of micro-constructions (individual construction types) begin to crystallise. Over time, the microconstructions entrench (become more frequent) and converge on a mesoconstruction (a set of similarly behaving micro-constructions, i.e. modals), out of which a macro-construction (high-level schema, i.e. Aux) develops.25 The meso- and macro-constructions may also license a series of extensions (new specific constructs), which inherit properties from these more abstract and schematic constructions (see Hoffmann and Trousdale 2011: 4; Trousdale 2010: 55; Bergs 2012: 1635). In the present case, we can propose the following course of development (although historical evidence for some of the steps is incomplete). We begin with a series of (primarily) elliptical second-person if-clause constructs with verbs of ‘choice’, conveying meanings ranging from literal to metalinguistic; initially, metalinguistic meanings always occur in the context of explicit markers (call it X; this title, to use this term, etc.). The meaning of the if you V becomes more abstract and metalinguistic (allowing for the ellipsis of explicit metalinguistic markers); it also becomes more inferential as the main clause apodosis (call it X) may also be ellipted. Overall, there is increased schematicity. Micro-constructions crystallise around metalinguistic if you choose and if you like (the earliest forms to appear). These constructions license extensions, with resulting increase in productivity, as other verbs (wish, prefer, and want) enter the construction – what Himmelmann (2004) calls host-class expansion. The micro-constructions converge into a mesoconstruction (metalinguistic if-EC). Over time, there is an increase in both type and token frequency, though it would appear that the verb choose falls out of favour as wish and prefer take over, along with like. Want, a very late entry in the construction, is still of low frequency. Note that decreasing 24
On the distinction between grammatical constructionalisation and lexical constructionalisation, see Trousdale (2008a, 2012) and Traugott (2012a, 2012b). The division into levels is based on the work of Traugott, see Traugott (2007: 525).
25
290 Brinton
compositionality also results from the loss of the explicitly metalinguistic elements and the increasingly routinised and fixed nature of the meaning. A macro-construction or high-level schema (i.e. indirect conditionals) ultimately develops (see Quirk et al. 1985: 1095–97). 14.5 Conclusion In conclusion, elliptical if-ECs – if you choose/like/prefer/want/wish – all serve metalinguistic and politeness functions in PDE; like and prefer are particularly common in these functions. The clauses make their appearance in the LModE period. Semantically the metalinguistic meanings of if-ECs can be seen as developing from the concrete propositional meaning ‘if you choose to do X’. The syntactic derivation of these clauses is problematic, however, because it is impossible to trace their origin back to bipartite (protasis–apodosis) structures where the valency of the verbs is complete. Reconstruction of the complement of the verb in the if-clause proves difficult because of the varied complement structures found historically and the rarity of such structures with metalinguistic meaning. While if-ECs in their propositional meaning typically occur with an explicit apodosis (if you want, open the window), apodoses never occur when the if-ECs function metalinguistically (it’s large, even gargantuan, if you wish [*you may say so]). Despite its rather late attestation, a more plausible source for the if-EC is the ‘call it X’ pattern, which is inherently metalinguistic. Metalinguistic if-ECs may result from constructionalisation of the ‘call it X’ pattern in conjunction with explicitly metalinguistic if you V-clauses.
15
Epistemic parentheticals with seem: Late Modern English in focus María José López-Couso a n d Belén Mé nde z- Naya
15.1 Introduction In the Cambridge Grammar of the English Language, Huddleston and Pullum et al. (2002: 895) define parenthetical clauses such as I think or you know as ‘expressions which can be appended parenthetically to an anchor clause but which also have a non-parenthetical use in which they take a declarative content clause as complement.’ Clausal parentheticals are syntactically and semantically independent and are typically detached from their host or anchor clause1 by comma intonation. In recent years, a great deal of scholarly attention has been devoted to parentheticals in general (see, for example, the studies in Dehé and Kavalova 2007; Kaltenböck et al. 2011, among others), and to parenthetical clauses in particular (see Schneider 2007; Brinton 2008a). Among the latter, epistemic/evidential parentheticals such as I think, I believe and I guess, which include an epistemic/evidential verb and a first-person subject, have been the object of abundant research (Thompson and Mulac 1991; Brinton 1996a, 2008a; Boye and Harder 2007; Dehé and Wichmann 2010b). Impersonal verbs such as seem, appear, look and sound, which in Present-Day English (PDE) require a third-person dummy subject it, can also occur in epistemic parenthetical constructions (e.g. it seems, it appears). Parentheticals of this kind, which have not yet been studied in detail, are the focus of our ongoing research. Thus far we have analysed what we call like-parentheticals (see (1) below), a relatively recent development in American English (see López-Couso and MéndezWe gratefully acknowledge the financial support of the European Regional Development Fund and the following institutions: the Spanish Ministry of Economy and Competitiveness (grant FFI2011–26693-C02–01) and the Autonomous Government of Galicia (Directorate General for Scientific and Technological Promotion, grant CN2012/012). We are also grateful to an anonymous reviewer and to the editor of this volume, Marianne Hundt, for helpful comments on an earlier version of the chapter. 1
‘Anchor’ is usually understood as the element to which a parenthetical is semantically related (Huddleston and Pullum et al. 2002: 1351), while ‘host’ is used for the clause in which a parenthetical is contained (see e.g. Kaltenböck 2007: 39 fn5). In their definition of parentheticals, Huddleston and Pullum et al. (2002: 895–96) use the term ‘anchor (clause)’ for what other authors call the ‘host clause’.
291
292 López-Couso & Méndez-Naya
Naya 2014a), and the early history of clausal parentheticals with seem and with impersonal think (OED s.v. think v.1 ‘seem, appear’), the default verb of ‘seeming’ in Old and Early Middle English prior to the introduction of seem in the thirteenth century (OED s.v. seem v.2; MED s.v. se¯men v.2) (see LópezCouso and Méndez-Naya 2014b), as shown in (2) and (3). (1) She has, it looks like, a mixture of people who have given her bad
advice, (COCA, 2007, SPOK, CNN_Grace) (2) þan þe kyng as it semed bytook recheslesliche þe governaunce of his kingdom to his chaunceller, bishop of Hely, (HC, QM3_NN_HIST_TREVISA) (3) I se on the firmament, Me thynk, the seven starnes. (HC, QM4_XX_MYST_TOWN) The present chapter is conceived of as a continuation of López-Couso and Méndez-Naya (2014b) and aims to complete the picture of the diachronic development of seem-parentheticals, extending the analysis into the Late Modern English (LModE) period. The evidential verb seem started out as a perception verb (a verb of appearance, see OED seem v.2 II). For Aijmer (2009: 72), for instance, seem indicates that ‘there is a phenomenon or stimulus (visual, aural, tactile) which gives rise to a reaction (e.g. a belief or opinion that something may be the case)’, although it differs from other perception verbs like appear, look and sound in being indeterminate as regards its evidential source. As Lampert and Lampert (2010: 319) note, seem ‘just implies that a source of evidence exists’. This lack of specification has been interpreted by some as inference or as hearsay (OED s.v. seem v.2 II.7d; see also Chafe 1986: 267–68; Plungian 2001; Gisborne and Holmes 2007: 7–8). Evidentials have been shown to acquire ‘epistemic extensions’ cross-linguistically (Aikhenvald 2004: 187). This is also the case with seem, which typically expresses a medium degree of speaker commitment (Huddleston and Pullum 2002: 175–77). As an epistemic verb, seem could be classified within the group of ‘propositional attitude predicates’, which convey ‘an attitude regarding the truth of the proposition expressed as their complement’ (Noonan 1985: 113). Since seem combines the notions of epistemicity and evidentiality, it can be characterised as an ‘epistemic/evidential predicate’.2 Interestingly, epistemic/ evidential verbs are commonly used parenthetically in many languages (see Noonan 1985: 114; Boye and Harder 2007: 583). This chapter is organised as follows: in Section 15.2 we describe the most important characteristics of parenthetical clauses identified in the relevant literature. In the following sections, and using data from the ARCHER 3.1 corpus, we establish a taxonomy of the different construction types in 2
Brinton (2008a: ch. 10) uses the term ‘epistemic/evidential’ to refer to comment clauses like I gather and I find, which combine both notions. See also Boye and Harder (2007: 583).
Epistemic parentheticals with seem 293
which the verb seem appears in our LModE material (Section 15.3) and offer a detailed description of the parenthetical types with seem recorded in the corpus (Section 15.4), paying attention to their morphosyntactic and semantico-pragmatic characteristics. 15.2 Parenthetical clauses As mentioned in Section 15.1, one of the most salient features of parenthetical clauses is that they are independent from their anchor clause in several respects. Thus, for example, they are syntactically independent: while they ‘seem to be linearly integrated’ in their anchor clause (Dehé and Kavalova 2007: 4), they do not form a single grammatical construction with it (see e.g. Burton-Roberts 2006: 179).3 Despite their syntactic independence, parentheticals and their anchor clauses can be interrelated by cross-referencing (Ackema and Neeleman 2004; Dehé and Kavalova 2007: 4; Kaltenböck et al. 2011: 854), that is, by means of some sort of anaphoric element. Consider in this connection (4), in which the pronoun it of the parenthetical everybody is aware of it is co-referential with the proposition our chairman is a liar. (4) Our chairman is, everybody is aware of it (…), a liar. (from
Kaltenböck et al. 2011: 854)
As stated in Huddleston and Pullum’s definition in Section 15.1 above, parentheticals usually have a non-parenthetical counterpart which takes the form of a complement structure, as illustrated in (5a–b). There is, however, a critical difference between the two patterns: the parenthetical I think in (5a) realises the function of sentence adverbial; by contrast, in its complementation counterpart (5b), I think serves as the matrix clause of a following complement clause (it is folly). Note that in (5a) such a relationship does not hold,4 and therefore the verb think occurs without a complement. This has led some scholars to state that parentheticals are syntactically defective (see Quirk et al. 1985: 1114; Huddleston and Pullum et al. 2002: 896; Kaltenböck et al. 2011: 872). (5)
3
a. ‘I say I will not be carried and you say this is folly!’ ‘So it is.’ ‘‘No,
it isn’t!’ ‘It is folly, I think. At any rate, the origin of it all is.’ (ARCHER, 1873hard.f6b) b. ‘I think (that) it is folly.’
Many scholars have argued for the syntactic independence of parentheticals (e.g. Espinal 1991; Haegeman 1991), since, unlike constituents of the anchor clause, they are transparent to VP ellipsis, cannot be the focus of a cleft sentence or be questioned. However, there are also proponents of the opposing view (e.g. Ross 1973a; McCawley 1982; Potts 2002; de Vries 2007). See Dehé and Kavalova (2007b: 4–7) and Kavalova (2007: 157–66) for valuable discussions of the two views. 4 For Huddleston and Pullum et al. (2002: 1350) the relation between I think and its anchor clause is one of ‘supplementation’, which is different from both subordination and co-ordination.
294 López-Couso & Méndez-Naya
The syntactic independence of parentheticals is reflected in their positional mobility. They may be juxtaposed (occurring before or, especially, after their anchor)5 or interpolated in their anchor clause. As we will discuss in Section 15.4.4 below, clauses in initial position are ambiguous between a matrix clause and a parenthetical interpretation, especially when the clause that follows lacks an overt complementiser.6 Therefore, unambiguous parentheticals occur in noninitial position (see e.g. Thompson and Mulac 1991: 313).7 (5a) above illustrates final position, while (6) shows the parenthetical it seems in medial position. (6)
He dresses very gaily too. He is at the head of the fashion, as, it seems, he thinks; (ARCHER, 1753rich.f4b)
Tightly linked with the syntactic independence of parentheticals is the observation that many of them are also prosodically independent. Clausal parentheticals are typically set off from their anchor clause by pauses in intonation and by commas in writing (see e.g. Burton-Roberts 2006: 180, who regards this as a central feature of all parentheticals). It must be noted, however, that prosodic independence is not necessarily found with short parentheticals (see e.g. Crystal 1969: 258), in particular with formulae such as please (see Wichmann 2004; Dehé and Wichmann 2010a; Kaltenböck et al.’s 2011: 876 ‘formulaic theticals’), or when parentheticals appear in ‘phases of disfluency and hesitation’ (Dehé and Wichmann 2010b: 24). Furthermore, parentheticals also show semantic independence from their anchor clause. While their anchor conveys the most important information (Huddleston and Pullum 2002: 896), parentheticals have a ‘secondary status’ (Boye and Harder 2007: 584), normally introducing an evaluation or a comment by the speaker on the information expressed in their anchor clause (hence the common label ‘comment clauses’, as in Quirk et al. 1985: 1112–18; Brinton 2008a). It is precisely because they do not carry the most important information that parentheticals cannot be addressed by really or by a tag question (see Boye and Harder’s 2007: 584 ‘addressability’), as shown in (7). In this example, while *does it really? and *doesn’t it? are ungrammatical, was she really? and wasn’t she? are felicitous since they refer to the statement Leonora was a Woman Beautiful, and … rather than to the parenthetical it seems. 5
Precisely because these elements can appear juxtaposed to their anchor (i.e. not interpolated), Kaltenböck et al. (2011: 856) propose the term ‘thetical’ instead of ‘parenthetical’. In the literature on parentheticals, initial zero clauses are variously regarded as parentheticals, matrix clauses, or structures ambiguous between the two readings. For a summary of the different views, see Kaltenböck 2007: 45 and Dehé and Wichmann 2010a. The presence of the complementiser that is usually regarded as an indication of non-parenthetical status (e.g. Thompson and Mulac 1991). But see Kärkkäinen (2003); Dehé and Wichmann (2010a); Kaltenböck (2011) for whom that-clauses can sometimes be parenthetical. 7 An anonymous reviewer notes, however, that in the early literature on parentheticals some scholars (e.g. Ross 1973a) argue for a main clause analysis even for non-initial parentheticals. See Dehé (2009) for a recent overview of the discussion. 6
Epistemic parentheticals with seem 295
(7)
a. For Leonora it seems was a Woman Beautiful, and otherwise
of an excellent Disposition; but in the Bottom a very Woman. (ARCHER, 1692cong.f2b) b. Was she really? Wasn’t she? *Does it really? *Doesn’t it?
The semantic independence of parentheticals is also reflected in the fact that they can easily be omitted without affecting the acceptability of the rest of the utterance (see Biber et al. 1999: 1067; Kearns 2007), as seen in (8a–b) below. In (8a) I mean can be left out because it has procedural rather than propositional meaning (it is used by the speaker to repair an earlier mistake). This contrasts with (9), where the omission of the matrix I mean (that) yields an ungrammatical sequence in (9b). This is so because in (9a) I mean conveys the propositional meaning ‘to intend a particular meaning when I say something’. (8)
a. {=F FANNY.} Oh, Martin, be not blind, – deaf, I mean, to our
entreaties. (ARCHER, 1813pool.d5b)
b. Oh, Martin, be not blind, – deaf to our entreaties. (9) a. phlogiston: that air is impure (by which I mean that it is unfit
for respiration, and for the purpose of supporting flame) in proportion as it contains more of that principle; (ARCHER, 1775prie.s4b) b. *(...) (by which it is unfit for respiration, and for the purpose of supporting flame) (...)
The non-propositional status of parentheticals is evidenced in their ‘metacommunicative or metatextual’ functions (Kaltenböck et al. 2011: 865). They are used to organise discourse, to show the source of information, to indicate the speaker’s claim to truth, and to express politeness, among other functions (Urmson 1952; Schneider 2007: 2–3, 55; Kaltenböck et al. 2011: 865–68, among many others). Note that pragmatic functions of this kind are characteristic of processes such as subjectification and intersubjectification (see Traugott 1995, 2003; López-Couso 2010). In addition to the characteristics mentioned so far, parentheticals typically show morphosyntactic fixation, a feature which can be related to increasing grammaticalisation: they do not usually allow alteration in tense, aspect and mood (TAM variation)8 or adverbial modification (see Brinton 2008a: 131 in connection with I mean), being typically affirmative, with a verb in the present tense (see Brinton 2008a: 2), as in I believe, it seems, you know, etc.
8
But cf. Van Bogaert (2010), who shows that high frequency sometimes correlates with formal variability. This is the case of the meso-construction I think, which can occur in a number of variations (I was thinking, I’m thinking, I thought, I do think, I think, I should think, I would think, etc.). On the notion of meso-construction, see Trousdale (2008a).
296 López-Couso & Méndez-Naya
Finally, parentheticals have been claimed to be a typical feature of speech, as opposed to writing (Thompson and Mulac 1991; Biber et al. 1999: 197, 862–65, 983).9 15.3 Syntactic patterns with seem in ARCHER This section describes the various syntactic patterns in which the verb seem is attested in ARCHER. We carried out an exhaustive search for all the occurrences of this predicate in the corpus, which yielded a total of 1,382 hits. The data showed that the verb seem is involved in four major construction types: (i) predicative structures; (ii) raising constructions; (iii) constructions with extraposed subject clauses; and (iv) parentheticals. 15.3.1 Predicative constructions The verb seem is frequently found in the corpus in two subtypes of predicative construction, depending on the nature of their subject, which may be phrasal, as in (10), or clausal, as in (11). Clausal subjects can be either finite (11a) or non-finite (11b). The predicative complement in these constructions can be realised by various types of phrases: an NP, as in (10a) and (11a), an AdjP, as in (10b) and (11b), but also by an AdvP, as in (10c), and a PP, as in (10d).10 The predicative complement is sometimes an epistemic adjective, which qualifies the degree of speaker’s commitment expressed by seem, as plausible in (11c), or an evidential adjective, such as evident in (11d).11 (10) a. It hardly seems a week; it almost seems a dream. (ARCHER,
1867robe.d6b) b. and when I last saw her, a month since her severe attack, she
seemed grateful for what had been done, (ARCHER, 1820tath. m5b) c. He had distributed funds sent him and kept many from perishing, and the worst seemed over. (ARCHER, 1880haml.j6b) d. George, dear, you seem out of spirits. (ARCHER, 1867robe.d6b)
9
An anonymous reviewer points out, however, that certain types of parentheticals (e.g. appositive relative clauses) are frequent in written language and in formal spoken language. This pattern corresponds to Seem 1 in Gisborne and Holmes (2007: 9–10). ARCHER also contains a few instances with the dummy there as subject (see Seem 7 in Gisborne and Holmes (2007: 13): ‘there extraposition of a nonclausal element’). An example is given in (i):
10
(i) As he grew older there seemed no hope of a child of his very own. (ARCHER, 1951brod.h8b) 11
The epistemic adjectives combining with seem in this pattern typically indicate certainty (e.g. likely, plausible, possible, probable), but on some occasions the adjective is located at the uncertainty end of the scale (e.g. unlikely).
Epistemic parentheticals with seem 297
(11) a. It seems a good consequence, that (…) the same coldness and want of motion after death might as easily make the same product in the Vessels (…). (ARCHER, 1684wgmb.m2b) b. At first sight it seems difficult to understand how our Lord could have used the simile in the text (…); (ARCHER, 18xxlidd. h6b) c. It seems plausible that increased overlap will increase the covalent character of the ‘ionic bond’, and consequently enhance its diamagnetism. (ARCHER, 1975duru.s8b) d. It seems quite evident from all accounts, that the French army, (…) is about to return home, (ARCHER, 1833tim1.n5b) 15.3.2 Raising constructions Seem also occurs in ARCHER in raising constructions (see Gisborne and Holmes’s Seem 2 (2007: 10–11)), which are related to the predicative structures just described. In these catenative constructions, illustrated in (12), an NP (here the scoundrel) is syntactically the subject of seem, but is semantically related to the subordinate non-finite clause (here it conveys the agent of the verb pursue) (see Quirk et al. 1985: 146–47; 1188; Huddleston and Pullum 2002: 1194ff). In other words, the scoundrel is a raised subject (see Huddleston and Pullum 2002: 1194).12 (12) The scoundrel seemed to pursue his journey quietly enough,
(ARCHER, 1751cove.f4b)
15.3.3 Extraposed subject clauses A very common structure is what Gisborne and Holmes (2007: 12) term Seem 3 or the ‘existential’ use of seem, in which the verb combines with a dummy subject it and an extraposed finite complement clause in subject function.13 The complement clauses in this pattern allow for some 12
The verb seem can also occur in the so-called ‘pseudo-raising’ (Rooryck 2000: 55–56) or ‘copy-raising’ construction (Asudeh and Toivonen 2005, 2006). This pattern, which is not recorded in ARCHER, appears to be a case of raising out of a finite clause, with co-referentiality between the subject of the matrix clause and that of the subordinate clause, as in (ii): (ii) Each instant he seemed as though he would open his eyes and speak; (NCF, Stoker, 1897, Dracula. Ch. XXI, 284).
In PDE, pseudo-raising is only available with comparative complementisers (see LópezCouso and Méndez-Naya 2012a); in earlier English it was occasionally found with that/ zero complement clauses as well (see OED s.v. seem v.2 II.4.f). 13 The analysis of such extraposed clauses is controversial; while some grammars regard them as subjects (see Quirk et al. 1985: 1183), with obligatory extraposition (Quirk et al. 1985: 1392, note a), others, such as Huddleston and Pullum et al. (2002: 960–62), reject the subject analysis on account of the clauses’ inability to occupy the preverbal position, and therefore regard the pattern as an ‘impersonal construction with it as subject’ (2002: 960).
298 López-Couso & Méndez-Naya
variation as regards complementiser choice. Along with the zero complementiser, as in (13), we may find cases of overt connectives, as in (14). Note that in addition to the major complement clause link that (see (14a)), ARCHER also contains a handful of examples of the use of so-called ‘comparative complementisers’ which we have examined in detail elsewhere (López-Couso and Méndez-Naya 2012a, 2012b), namely as if (14b), as though (14c) and like (14d). On one occasion (14e) the complement clause is introduced by a combination of two of these comparative complementisers, like as if (see Kortmann 1997: 371, note 159). (13) It seems ø I am the golden prize for which you plead; (ARCHER,
1753foot.d4b) (14) a. It seems that the waves have passed the theoretical stability point before the vortices are properly cast off from the crests. (ARCHER, 1975crap.s8b) b. It seemed as if all our moral vigor was dying down, and as if nothing could restore it. (ARCHER, 18xxdale.h6b) c. To outside eyes, it seems as though the most populous nation in the world is on the point of disintegration; (ARCHER, 1967stm1. n8b) d. ‘Seems like he tried to take a crap in it.’ (ARCHER, 1976coov. f8a) e. ‘Sometimes it seems like as if he was, and then again more like as if he wasn’t.’ (ARCHER, 18783eggl.f6a) As we have argued in López-Couso and Méndez-Naya (2012a, 2012b), complementiser selection with seem may reflect the degree of the speaker’s endorsement of the embedded proposition, which is higher in the case of that. 15.3.4 Parenthetical constructions The predicate seem is also found in ARCHER in parenthetical constructions. The material provides evidence of the existence of two major parenthetical types, which differ as regards the presence or absence of an adverbial element as or so referring anaphorically to the anchor. Adverbial parentheticals are illustrated in (15a) with as and (15b) with adverbial so, the only example of so-parentheticals with seem recorded in ARCHER. (16), in turn, is an example of the bare parenthetical type, which lacks the anaphoric element. (15) a. Mere accident, as it should seem, directed her into the direct
road to Heart, (ARCHER, 1797butl.f4a)
b. But a certain cruel clearness of vision, or so at least it seemed
to her, made all reflections on this theme but an aggravation of her despair. (ARCHER, 1880bell.f6a) (16) Your father’s debtor dies, it seems, on his arrival at Jamaica. (ARCHER, 1799brow.f4a)
Epistemic parentheticals with seem 299
Note that seem also occurs in ARCHER in structures bearing a strong resemblance to the parentheticals just discussed but occurring independently as responses. As with parentheticals, the verb seem lacks here its clausal complement, which is resumed by an anaphoric element, either the pro-form so, as in (17), or the pronoun it, as in (18) (see Quirk et al. 1985: 1115, note c).14 (17) ‘I suppose what it amounts to, Jeeves, is that, when young Bingo really
takes his coat off and starts in, there is no power of God or man that can prevent him making a chump of himself.’ ‘It would seem so, sir,’ said Jeeves. (ARCHER, 1923wode.f7b) (18) ‘Herndon knocked him down?’ ‘Seems like it. Herndon had the fire poker in his hand –’ (ARCHER, 1953wrig.f8a) 15.3.5 Optional elements
The various syntactic patterns with seem described in the preceding paragraphs may contain a number of optional elements which are closely related to our earlier characterisation of the verb seem as an epistemic/ evidential predicate. These elements, which occur only infrequently judging from the data in ARCHER, convey the experiencer and the source of evidence for seem. Experiencer NPs were only present in about 7 per cent of all occurrences of seem in the corpus. When the experiencer is left unexpressed, it is (by default) associated with the speaker (see Gisborne and Holmes 2007: 7), and the interpretation of seem is closer to hearsay. In turn, overt experiencers typically take the form of a to-PP. The most common pattern involves a first-person singular pronominal experiencer (me) occupying postverbal position in the parenthetical, as in (19). Note that here, rather than hearsay, it seems to me clearly conveys the speaker’s opinion, and thus comes close to I think. More unusual experiencer phrases are shown in (20a–b), which contain a compound NP experiencer and a pronominal experiencer in preverbal position, respectively. (19) A good lady, but too lively, too gay – it seems to me, like two queen
bees in one beehive; (ARCHER, 1799dunl.d4a) (20) a. This history of a forlorn, seemed stranger to the young lady and her father than even the account of my journey through the inside of a mountain, (ARCHER, 1756amor, f4b) b. ‘To us,’ says he, ‘it seems that a steady and uniform march, according to the provisions of the Charter, is almost all that is 14
Interestingly, examples such as (17) and (18) count as theticals for Kaltenböck et al. (2011). According to them, theticals can occur interpolated, adjacent to their anchor clause, or forming an independent utterance (2011: 861, 863).
300 López-Couso & Méndez-Naya
necessary to secure and consolidate the repose and happiness of France.’ (ARCHER, 1819mor1.n5b) On only one occasion is the experiencer realised by an object personal pronoun rather than a PP. This is (21), which is strongly reminiscent of the old impersonal construction me seemeth, hym seemeth, etc. (see OED s.v. seem v.2 8). (21) (…) to say from the heart, and in every trying hour, ‘It is the Lord: let
him do what seemeth him good.’ (ARCHER, 17xxwesl.h4b)
In addition to an overt experiencer, the four major syntactic patterns in which seem is attested in the corpus may occasionally contain an explicit indication of the source of evidence, which mostly takes the form of a PP with by (see Gisborne and Holmes 2007: 8) or, especially, with from. Examples of this kind are given in (22) for the preposition by, with a raising construction, and in (23) for from, illustrating the predicative type. (22) You seem by your habit, Sir, to be a Physician, or something that
way. (ARCHER, 1693powe.d2b) (23) It seems quite evident from all accounts, that the French army, having accomplished the immediate object of its mission, (…), is about to return home, (ARCHER, 1833tim1.n5b) 15.4 Seem-parentheticals in Late Modern English: quantitative evidence from ARCHER As seen in Section 15.3 above, the verb seem may occur in parenthetical structures. This section presents a detailed analysis of the seem-parentheticals attested in ARCHER, paying attention to issues such as the parenthetical type, the expression of the experiencer and of the evidential source, tense, aspect and mood variability, the position the parenthetical constructions occupy in the clause, and the semantico-pragmatic functions they realise. 15.4.1 Parenthetical types Figure 15.1 provides the distribution of the two major parenthetical types identified in Section 15.3 for the verb seem, bare parentheticals and adverbial parentheticals, in the British English material in ARCHER. In order to give a full picture of the development of seem-parentheticals over time, the figure also contains data from the Helsinki Corpus (HC) for the Middle and Early Modern English periods (see López-Couso and Méndez-Naya 2014b). As regards their overall frequency in the two corpora analysed, seemparentheticals appear to be a low-frequency construction throughout. The
Epistemic parentheticals with seem 301
19 50 –
bare adverbial
13 50 –1 42 0 14 20 –1 50 0 15 00 –1 57 0 15 70 –1 64 0 16 40 –1 71 0 17 00 –1 74 9 17 49 –1 79 9 18 00 –1 84 9 18 50 –1 89 9 19 00 –1 94 9
7 6 5 4 3 2 1 0
Figure 15.1 Adverbial and bare seem-parentheticals across time
material does not yield any clear tendencies, with the exception of a changeover in the preferred parenthetical type. As Figure 15.1 shows, the earliest instances of unambiguous parentheticals with seem in the HC are recorded in subperiod M3 (1350–1420) (7 examples), one century after the introduction of this Old Norse loanword in the thirteenth century (OED s.v. seem v.2; MED s.v. se¯men v.2). These instances correspond to the adverbial parentheticals with as, which remain the most common parenthetical type until the early seventeenth century. Example (24) illustrates the earliest instance of any seem-parenthetical in the HC, while (25) provides a LModE example of the adverbial type. (24) = (2) þan þe kyng as it semed bytook recheslesliche þe
g overnaunce of his kingdom to his chaunceller, bishop of Hely, (HC, QM3_NN_HIST_TREVISA) (25) he was now doubly terrified at its result, fearing, and as it seemed, not without cause, that the unfortunate puritan was actually murdered. (ARCHER, 1847lefa.f5b) As shown in Figure 15.1, bare parentheticals emerge later, over the course of the fifteenth century. The earliest examples of this type in our data correspond to the impersonal me seems (see López-Couso and Méndez-Naya 2014b), as in (26), which is modelled on its equivalent methinks.15 (26) And this me semeth shuld be sufficient instruction for the husbande
to kepe measure. (HC, QE1_IS_HANDO_FITZH)
15
The impersonal construction me seems appears to have undergone a process of fossilisation similar to that identified in the literature for methinks (see López-Couso 1996; PalanderCollin 1999; Wischer 2000). See also Elmer (1981: 134), who offers a similar account for me seems to that of methinks, and López-Couso (1996: 163, fn 15). For examples, see OED s.vv. seem v.2 8 and meseems impers.v.
302 López-Couso & Méndez-Naya Table 15.1 Distribution of parenthetical types with seem in the complete ARCHER
1650–1699 1700–1749 1750–1799 1800–1849 1850–1899 1900–1949 1950– TOTAL
Bare parentheticals
Adverbial parentheticals
6 8 4 (BrE) 5 1 (BrE) 3 6 (BrE)
1 – 1 (BrE) 4 – 1 –
8 (AmE) 4 (AmE) 1 (AmE) 46
3 (AmE) 2 (AmE) – 12
The ‘modern’ bare parenthetical it seems is not recorded in the material until the late seventeenth century, when bare parentheticals become the default parenthetical type with seem.16 Interestingly, bare seem-parentheticals show a peak at the beginning of the eighteenth century. Whether this increase can be linked to the rise in the frequency of use of first-person epistemic parentheticals reported by Wierzbicka (2006: 207), which she attributes to the impact of British Empiricism and its emphasis on ‘acknowledging the limitations of one’s understanding’ (2006: 242), remains to be confirmed through the analysis of a larger amount of data, and scientific texts in particular. Table 15.1 gives the distribution of seem-parentheticals in the ARCHER material according to type. For the second half centuries, the figures for the American English component of the corpus are also provided.17 15.4.2 Expression of the experiencer NP and of the evidential source Only eight of the fifty-eight cases of seem-parentheticals in ARCHER show an overt experiencer. In six of them, the experiencer is the speaker (to me) (see (27a)), while the other two cases entail a third person (to her, to them) (see (27b)). (27) a. ‘…You are wonderfully important all of a sudden, it seems to me.
You talk like an artist.’ (ARCHER, 1897cran.f6a)
b. They obeyed her in loving reverence; the spirit of awe not a little
augmented among the group by the sudden, and, as it seemed to them, supernatural appearance of the Englishman upon the scene. (ARCHER, 1860cumm.f6a)
As regards the presence of a PP expressing the source of evidence, this is only found in one of our examples, namely: 16
Our study of epistemic/evidential parentheticals in COCA (López-Couso and MéndezNaya 2014a) shows the clear prevalence of the bare parenthetical pattern with seem in contemporary American English data. 17 Note that ARCHER 3.1 only contains American English material for the second half centuries. On the structure of ARCHER, see Yáñez-Bouza (2011) and www.alc.manchester. ac.uk/subjects/lel/research/projects/archer/archer-versions/archer3_1/.
Epistemic parentheticals with seem 303 Table 15.2 Distribution of verb forms according to parenthetical type in the complete ARCHER Form
Bare PARs
Adverbial PARs
Total
seems seemed may seem should seem would seem
39 2 – 1 4
3 6 1 2 –
42 (72.4%) 8 (13.8%) 1 (1.7%) 3 (5.2%) 4 (6.9%)
(28) the door of a small tavern (…) was suddenly thrown open, and forth
issued two roystering blades, members of the university, it would seem from their scholastic caps and gards, who had evidently been indulging in copious libations, and were now, in all probability, retreating to their place of rest for the night. (ARCHER, 1837ains.f5b)
15.4.3 Tense-aspect-mood variability As seen in Section 15.2 above, as they grammaticalise, parentheticals show a tendency towards morphosyntactic fixation: they are typically affirmative and contain a present-tense verb (Brinton 2008a). This tendency towards fixation is followed by the seem-parentheticals in ARCHER. As the data in Table 15.2 show, ARCHER yields no examples of negative parentheticals, and the present verb form seems is the prevalent one, accounting for almost three-quarters of all examples. It should be noted that the tense of the verb in the parenthetical construction is not necessarily dependent on the verb in the anchor clause (see Howe and Heller 2010). Thus, a present-tense parenthetical can apply to a preterite anchor, as in (29). (29) (…) and cries out with a Fearful Oath that it was his cousin; who had
it seems been a Highway Man and but lately caught and hanged; (ARCHER, 1752knyv.j4b)
If we plot the form of the verb against parenthetical type, an interesting picture emerges. As can be seen in Table 15.2, morphological fixation is particularly relevant for bare parentheticals, which show a present verb form in almost 85 per cent of cases, while adverbial parentheticals still leave room for some variation in verb morphology. This seems to indicate that adverbial parentheticals exhibit a lower degree of grammaticalisation than their bare counterparts.18 The older adverbial type with so/as also shows additional indication of greater variability: as opposed to the bare parenthetical 18
As we argue in López-Couso and Méndez-Naya (2014b), seem-parentheticals would have followed a developmental path similar to that proposed by Brinton (1996, 2008a) for I think, which has evolved, in her view, from structures of the type as/so I think, after the loss of as/so.
304 López-Couso & Méndez-Naya
pattern, which allows for variation only as regards the presence/absence of the experiencer NP, adverbial parentheticals occasionally show adverbial modification: e.g. or so at least it seemed to her (see example (15b) above). 15.4.4 Position of seem-parentheticals As mentioned in Section 15.2, once they reach parenthetical status, the structures under analysis take on a new syntactic function as disjuncts. As such they may occur in the same variable positions as adverbs, that is, initially, medially and finally. When occurring in initial position (see (30)), it seems may have a double reading either as a matrix clause taking a zero complement or as a parenthetical construction (see Thompson and Mulac 1991; Diessel and Tomasello 2001: 107–08; Boye and Harder 2007: 586; Fischer 2007: 298; Kaltenböck 2007: 45; Brinton 2008a: 12). In such cases, prosodic information turns out to be crucial for disambiguation in contemporary spoken English (see Quirk et al. 1985: 1113; Hirschberg 2004: 524; Dehé and Wichmann 2010a). In historical texts, by contrast, disambiguation is more problematic, since punctuation was far less regulated than nowadays (see, for example, Denison and Hundt 2013 for punctuation in Modern English relative clauses), and therefore cannot be taken as a reliable indication of intonation. (31) below, for instance, shows how earlier English made use of punctuation marks such as commas in sentences containing an overt complementiser. (30) It seems ø he is a Relation of mine, (ARCHER, 1739swft.x3b)
(31) It seems, that after the former action he found means to assemble some thousands of the fugitives, (ARCHER, 1774lon1. n4b) In the absence of unequivocal evidence in favour of the parenthetical reading, potentially ambiguous instances such as (30) have therefore been considered as involving a matrix clause followed by a zero complement, rather than as illustrating the parenthetical use of it seems. Table 15.3 shows the distribution of the different parenthetical types with the verb seem in ARCHER in positions other than initial. The table shows that medial is the most common position for both bare and adverbial seem-parentheticals in ARCHER, accounting for 74.1 per cent of all cases. In (32), for instance, the bare parenthetical it seems occurs medially, absolute initial position in the sentence being occupied by the adjunct near the first, whose scope is the clause are the limits of two Lordships (notice here the inverted word order VS triggered by the initial adverbial). (32) We proceeded to another elevation a little farther on; but in 5 Minutes
found that it was only an old heap of Earth thrown up, when they were digging for Turf or making Peat land, there being not the least indication of burning, and the natural soil soon appearing. Near the first it seems are the limits of two Lordships, and it might have served as the Boundary mark. (ARCHER, 1809fent.j5b)
Epistemic parentheticals with seem 305 Table 15.3 Medial vs. final position of seem-parentheticals in the complete ARCHER corpus Type
Medial position
Final position
Bare PARs Adverbial PARs
32 11
14 1
The most frequent slot for medial seem-parentheticals in the ARCHER material (12 instances) is that between the subject and the verb, as in (33), one of the ‘weak spots’ identified by Kaltenböck (2007: 43) in which parentheticals can easily be inserted. t
(33) This Article, it seems, is interpreted as an insinuation y they had not
a Convoy in due time, (ARCHER, 1706stel.x3b)
The material also contains examples of positions in which parentheticals tend to be avoided in PDE (see Brinton 2008a: 8 and references therein; Kaltenböck 2007: 42ff.). Disfavoured positions include that between the verb and its complements, as in (34), within a premodifier string, as in (35), within an NP between the head and its modifier, as in (36), and between an auxiliary and a main verb, as in (37). (34) These were, it seems, Charles XII. of Sweden, and Alexander of
Macedon. (ARCHER, 1743fiel.f3b) (35) = (27b) They obeyed her in loving reverence; the spirit of awe not a little augmented among the group by the sudden, and, as it seemed to them, supernatural appearance of the Englishman upon the scene. (ARCHER, 1860cumm.f6a) (36) (...) she got into a state which she called ‘chrysalis’ a state it seems of partial unconsciousness. (ARCHER, 1912yeat.x7b) (37) The English merchants, (…) have, it seems, devised a plan of continuing it, by fitting out and freighting their vessels under our flag. (ARCHER, 1802joh2.n5b). 15.4.5 Pragmatic functions of seem-parentheticals Two of the pragmatic functions of parentheticals discussed by Kaltenböck et al. (2011) in their study of (paren)thetical grammar prove relevant for our seem-constructions here, namely source of information and attitude of the speaker.19
19
Other components associated by Kaltenböck et al. (2011) with the functions of parentheticals are speaker–hearer interaction, text organisation, discourse setting and world knowledge.
306 López-Couso & Méndez-Naya
As regards source of information, seem is indeterminate with respect to the source of evidence, though it has also been regarded as inferential or as indicating hearsay (see Section 15.1). In (28) above, repeated here as (38) for convenience, the context makes it clear that the evidence is visual; the speaker knows that the two gallants leaving the tavern are members of the university because he can see their scholastic caps and gards. (38) = (28) the door of a small tavern (…) was suddenly thrown open,
and forth issued two roystering blades, members of the university, it would seem from their scholastic caps and gards, who had evidently been indulging in copious libations, and were now, in all probability, retreating to their place of rest for the night. (ARCHER, 1837ains.f5b)
In (39), in turn, the parenthetical it seemed indicates inference; it can be inferred that nobody is at home because no one has opened the door. (39) I waited; one minute, two, three. The door didn’t open. There was
nobody at home, it seemed. (ARCHER, 1935ishe.f7b)
By contrast, in example (40), it seems indicates hearsay, as is evident from the comment Such was this man’s narrative… (40) He proceeded: ‘This man is a brother of that. Their resemblance in
figure occasioned your mistake. Your father’s debtor dies, it seems, on his arrival at Jamaica. There he met with this brother, and bequeathed to him his property and papers. Some of these papers are in my possession. They are letters from Constantia Dudley, and are parts of an intrigue which, considering the character of the man, was not much to her honor. Such was this man’s narrative, told to me some time before your meeting with him at this house. (…)’ (ARCHER, 1799brow.f4a)
In any case, seem-parentheticals usually indicate a distancing on the part of the speaker from the truth of the proposition in the anchor clause (see Kaltenböck et al.’s 2011 attitudes of the speaker). This is exploited in fiction by the narrator to detach him/herself from what is being narrated, and is particularly clear in examples where the NP experiencer is in the third person, as in (35) above, where the narrator does not seem to align him/herself with the idea that the appearance was supernatural. When the parenthetical contains a first-person NP experiencer, as in (27a), repeated here as (41), by contrast, the speaker’s commitment is increased, it seems to me coming close to I think. (41) = (27a) ‘…You are wonderfully important all of a sudden, it seems to
me. You talk like an artist.’ (ARCHER, 1897cran.f6a)
Epistemic parentheticals with seem 307
15.5 Summary Drawing on data from ARCHER, this chapter has described the different construction types in which the epistemic/evidential verb seem is attested in the LModE period. In addition to predicative structures, raising constructions and constructions with extraposed subject clauses, seem is used parenthetically. It is precisely this latter use that has been the main focus of this study. The verb seem was recorded in ARCHER in two parenthetical types: bare (it seems) and adverbial (as/so it seems) parentheticals. Judging from the corpus evidence, bare parentheticals constitute the default pattern with this verb in LModE, thus continuing a tendency established in the early seventeenth century, when they replaced adverbial parentheticals as the most frequent parenthetical type (see López-Couso and Méndez-Naya 2014b). Of the two seem-parenthetical patterns, bare parentheticals are the more grammaticalised, as evidenced by their marked tendency towards morphosyntactic fixation (they are always affirmative and show a present-tense verb form in almost 85 per cent of cases). Adverbial parentheticals, by contrast, allow for greater variation in the VP (e.g. they may contain preterite and modally marked VPs) and for adverbial insertion. As regards position, unambiguous parentheticals may occur medially and finally. Seem-parentheticals in ARCHER show a marked preference for the medial position. They usually intervene between clausal constituents (their preferred slot being between S and V), although the material also yields interesting examples of parentheticals occurring in less favoured positions (e.g. within an NP). As far as their semantico-pragmatic features are concerned, and in line with the epistemic/evidential nature of their predicate, seem-parentheticals are used to qualify the speaker’s commitment towards the truth of the proposition expressed in their anchor clause, which is greater when a first-person NP experiencer is overtly expressed. Seem-parentheticals are usually indeterminate as regards their evidential source, though this is sometimes made explicit by means of a by- or a from-PP. The scenario described in this chapter for bare and adverbial parentheticals with seem, however, needs to be confirmed using a larger amount of data. We believe, nevertheless, that a small but balanced corpus like ARCHER provides valuable indications of the behaviour of seem-parentheticals in LModE. The exploration of further data from the period, together with the analysis of the parenthetical patterns with other epistemic/evidential verbs, such as appear, look and sound, must be left for further research.
Part VI Text-type related change
16
Syntactic stability and change in nineteenth-century newspaper language Er ik Smitterberg
16.1 Introduction 16.1.1 Aim and scope Studies of late twentieth-century English such as Mair (1997) and Hundt and Mair (1999) have identified an important type of ongoing stylistic shift in the distribution of linguistic features in some written genres. Features characteristic of informal spoken interaction, such as contractions (e.g. don’t instead of do not) and phrasal verbs (e.g. put off instead of postpone), have become increasingly frequent in some written genres, e.g. newspaper language, while a genre like academic writing is less affected by such tendencies. Hundt and Mair (1999: 235–36) account for this phenomenon, known as ‘colloquialisation’, partly in terms of production circumstances: newspapers are subject to market forces in order to maintain and increase circulation, whereas academic writing ‘is less affected by economic pressures to win bigger audiences’ (Hundt and Mair 1999: 236). However, as Hundt and Mair (1999: 222) note, several of the changes attested are likely to be ‘part of a more general and long-term development’; Mair (2006: 187) also points out that colloquialisation is present ‘to a greater or lesser extent … at different times diachronically’. As Westin (2002) shows, this change in newspaper English has been in progress for much of the twentieth century. However, less is known about the English of nineteenthcentury newspapers. The aim of this chapter is to contribute to filling this gap in our knowledge by examining the extent to which nineteenth-century British newspaper English shows signs of colloquialisation. I shall focus on four features whose distribution has been shown to differ between speech This article was written as part of a research fellowship funded by the Royal Swedish Academy of Letters, History and Antiquities, supported by a grant from the Knut and Alice Wallenberg Foundation. I am grateful to the participants in the workshop Late Modern English Syntax (Zurich, 19–20 August 2012) for their comments on the paper I presented there, which forms the basis for the present study. I am also indebted to Christer Geisler for writing Perl scripts that facilitated the retrieval of linguistic features from a tagged version of the Corpus of Nineteenth-century Newspaper English. The editor of the volume, Marianne Hundt, and an anonymous reviewer provided valuable comments on draft versions of this study.
311
312 Smitterberg
and writing: the progressive, phrasal verbs, not-contraction, and the different units co-ordinated by and. Several of my results will be compared with the development of the same features in other genres. The present study is structured as follows. In Section 16.1.2, I briefly describe the process of colloquialisation. Section 16.2 addresses the primary material used for this study. My results are presented in Section 16.3 and Section 16.4 contains a discussion of the results against the background of previous research on colloquialisation, language change in Late Modern English (LModE), nineteenth-century British society, and the development of the newspaper industry. 16.1.2 The process of colloquialisation Mair (2006: 187) describes colloquialisation as a stylistic shift that causes written norms to become less linguistically different from spoken usage and more tolerant of informality and anti-formality. Thus, ‘literate’ features like the passive tend to decrease in frequency in genres that undergo colloquialisation, while ‘oral’ features such as contracted forms instead become more frequent. The result is increased linguistic differences among written genres, as some but not all types of writing are affected by the change. As shown in Biber and Finegan’s (1997) factor score analysis, LModE genres exhibit increasing linguistic differentiation in diachrony: stereotypically oral and literate genres become more clearly separated in terms of their linguistic make-up. As colloquialisation would cause the norms for oral genres to become more similar to those for contemporary speech, this stylistic shift may have contributed to the increase in linguistic genre differentiation. Interestingly, Biber and Finegan (1997: 266–69) demonstrate that news reportage first develops towards a literate style, but then displays a reversal towards more oral characteristics on two of the three dimensions that distinguish oral from literate genres. As this reversal sets in between the periods 1800–99 and 1900–90, it is of interest to examine whether any trends towards more oral language can be discerned within the nineteenth century. As shown by Mair (1997, 2006), colloquialisation as a linguistic process correlates with specific trends in late twentieth-century western society; for instance, there has been a democratisation of discourse conventions, which has led to sectors of society that would previously have been silenced acquiring a public forum (Mair 2006: 186).1 In order to support a hypothesis that 1
Farrelly and Seoane (2012) divide democratisation into democratisation proper, which ‘refers to speakers’ tendency to avoid unequal and face threatening modes of interaction’ (p. 393), colloquialisation, which ‘is the tendency for written language to incorporate features of the spoken language’ (p. 394), and informalisation, which is ‘the process whereby the distance between addresser and addressee is shortened’ (p. 395). As Farrelly and Seoane (2012: 395) argue that informalisation but not colloquialisation affects ‘expository writings
Syntactic stability & change in 19th-c. newspapers 313
colloquialisation is taking place in nineteenth-century English, it is thus necessary to consider not only the linguistic evidence, but also the society in which the texts were produced. I shall return to this issue when the results of the present study are discussed in section 16.4. 16.2 Material 16.2.1 The Corpus of Nineteenth-Century Newspaper English The main source of data for the present study is the Corpus of NineteenthCentury Newspaper English (CNNE). There are several reasons why examining nineteenth-century newspapers is of considerable importance in accounting for the development of Late Modern British English. First, as Lee (1976: 18) argues that the press ‘was the most important single medium of the communication of ideas’ during the 1800s, usage current in newspaper language can be expected to have influenced other forms of popular writing, such as fiction. Second, the newspaper industry itself went through extensive change during this time; Brown (1985: 1) even claims that ‘“[t]he news” as we understand it is a nineteenth-century creation’. The readership of newspapers grew dramatically, as a result of changes such as falling newspaper prices and the development of mass literacy. As one illustration of this growth, Brown (1985: 38) shows that there were a total of 4,416 newsagents and newsroom keepers in England and Wales in 1871, and this number had grown to 9,708 in 1891. CNNE covers metropolitan and provincial newspapers published in England. The texts are grouped into two periods to enable diachronic research: 1830–50 (period 1) and 1875–95 (period 2). The periodisation was influenced by two main factors. First, the sampling frame was designed so that the two periods would be separated by a larger time span than that covered by each period. Second, the period spans were chosen so that the periods would be separated by extralinguistic events that may have had an impact on the newspaper genre. In this regard, I focused on two events singled out by Brown (1985: 4) as especially important: the abolition of the stamp duty on newspapers in 1855, which made it possible to lower the price of newspapers and thus reach a larger proportion of the population; and the formation of the Press Association in 1868, which was of central importance as regards, for instance, securing cheap access to news for provincial papers. Ten articles from each newspaper and period were included in the corpus. The newspapers sampled are given in Table 16.1, which also provides such as the language of the academy and press’, the features investigated in this chapter would arguably encompass both colloquialisation and informalisation as defined by Farrelly and Seoane.
314 Smitterberg Table 16.1 Newspapers included in CNNE as of December 2012 Paper
Code
Period
Birmingham Daily Post Daily News Examiner Leeds Mercury Liverpool Mercury Lloyd’s Weekly London Newspaper Manchester Guardian Morning Chronicle Northern Echo Northern Star and Leeds General Advertiser Pall Mall Gazette Poor Man’s Guardian Reynolds’s Weekly Newspaper The Times
BDP DN EX LM LIM LL MG MC NE NS PMG PG RW TT
2 1 + 2 1 1 + 2 2 1 + 2 1 + 2 1 2 1 2 1 1 + 2 1 + 2
the code used to identify each paper in tables and in references to corpus examples. As shown in Table 16.1, some newspapers were sampled for both periods and thus contribute twenty articles to CNNE. The newspapers were selected from three digital newspaper archives: ProQuest Historical Newspapers (the Manchester Guardian), The Times Digital Archive (The Times), and Century British Library Newspapers (all others). Articles from these archives were downloaded as PDF files and converted to text files using optical character recognition software complemented by manual proofreading. The articles sampled belong to the subgenres of editorials and reportage. The current size of CNNE is 323,658 words (159,881 from period 1, 163,777 from period 2), which is small by present-day standards;2 however, as the size of each period sample is close to that of the press sections in members of the Brown family of corpora, it is hoped that CNNE will still yield statistically robust results as regards the mid-frequency and high-frequency features investigated in this chapter. 16.2.2 A Corpus of Nineteenth-Century English To be able to relate the trends attested in CNNE to nineteenth-century English as a whole, I will also report on previous research on A Corpus of Nineteenth-Century English (CONCE).3 CONCE contains approximately 2
As the scope and compilation of CNNE have not yet been finalised, word counts given in the present study should not be taken as referring to the final version of the corpus. The word counts used in this study are those returned by Notepad++ for corpus files from which headers, headings, etc. had been removed. 3 For more detailed descriptions of CONCE, including word counts, see Kytö et al. (2000), Smitterberg (2005: 17–24, 271–75), and Kytö et al. (2006).
Syntactic stability & change in 19th-c. newspapers 315 Table 16.2 Descriptions of the genres in CONCE (based on Kytö et al. 2000: 88) Genre
Characteristics
Debates Drama Fiction History Letters Science Trials
Recorded debates from the Houses of Parliament Prose comedies or farces Novels Historical monographs Private, personal letters between relatives or close friends Monographs pertaining to the natural or social sciences Recorded trial proceedings (chiefly in dialogue format)
one million words. The texts in the corpus are divided into three periods (1800–30, 1850–70, and 1870–1900) and seven genres, brief descriptions of which are given in Table 16.2. As shown in Table 16.2, CONCE includes informal and/or speechrelated genres such as Letters and Drama as well as formal, literate genres like Science. It is thus a suitable corpus for exploring colloquialisation in English, as such a change should be visible as increased linguistic differentiation between the oral genres that take part in the change and the literate genres which resist it. In previous research such as Geisler’s (2002) factor score analysis of the corpus, the difference between the expository genres (Debates, History, and Science) and the non-expository genres (Drama, Fiction, Letters, and Trials) has been shown to be of particular importance: non-expository genres are consistently closer than are expository genres to the oral ends of Biber’s (1988) dimensions of linguistic variation. In Sections 16.3.1 and 16.3.2, I will compare results based on CNNE with findings for parts of CONCE from previous research. Such comparisons are not wholly reliable, as there are differences in periodisation between CNNE and CONCE. Nevertheless, it is hoped that comparing results based on the two corpora will shed light on the extent to which the patterns attested in CNNE are likely to reflect colloquialisation. 16.3 Results The token frequency of a linguistic feature can be measured in different ways. As Smith and Rayson (2007: 134) point out, the main decision concerns whether to present the frequency relative to a certain text length (i.e. a normalised frequency) or to the frequency of one or several other linguistic features which, together with the feature analysed, make up a variant field. The incidence of progressives and phrasal verbs is typically expressed in terms of normalised frequencies rather than percentages of occurrence within variant fields; as both features are also part of the verb phrase, I treat them together in Section 16.3.1. Separate sections are then devoted to not-contraction (16.3.2) and the units co-ordinated by and (16.3.3), which can be analysed
316 Smitterberg
from a variationist perspective (although, as I show in Section 16.3.2, the distribution of the data makes such an analysis unnecessary regarding not-contraction). The Concordancer program in the WordSmith Tools 5 software suite was used for all automatic searches for linguistic features. 16.3.1 Progressives and phrasal verbs The progressive (e.g. was going in example (1)) has been increasing in frequency more or less throughout the Modern English period (see, for instance, Hundt 2004a: 115; Kranich 2010: 95). (1)
Fortunately the train was going very slowly at the time, and no serious consequences are reported. (RW0020, 1875–95)
Phrasal verbs (e.g. give up in example (2)) display a less clear frequency development in Modern English. (2)
Unpopular, as he is we apprehend he is too bold and haughty to give up his office in a mere panic, or until he finds that he can keep it no longer. (LM0007, 1830–50)
Results presented in Spasov (1966: 21) and Claridge (2000: 177–79) indicate that their frequency may have decreased in late seventeenth-century and eighteenth-century English, whereas the 1800s appear to have witnessed rising frequencies in many genres (Smitterberg 2008). This increase continues in twentieth-century English: both the progressive and phrasal verbs take part in the colloquialisation of newspaper English reported on in Hundt and Mair (1999). But despite increasing in frequency in some written genres, these two features have not wholly lost their connotations of orality; as Biber’s (2003a) factor analysis shows, they are still oral features in present-day academic English, and Biber et al. (1999: 409) show that phrasal verbs are more than twice as common in present-day conversation and fiction as in academic writing. Charting the frequency of the progressive and of phrasal verbs in CNNE would thus give a valuable indication of whether nineteenth-century newspaper English underwent colloquialisation. Progressives were retrieved from CNNE by means of searches for combinations of forms of be and –ing (including spelling variants), followed by manual scrutiny. A form of be had to occur no more than six words to the left of the word in –ing to be included in the searches. The manual postprocessing of the concordancer output excluded cases where the –ing-form was adjectival, gerundial, part of a be going to future construction, etc. (see Smitterberg 2005: 25–37 for a detailed account of the selection of data). The retrieval of phrasal verbs was based on a version of CNNE that had been tagged using Conexor’s Machinese Phrase Tagger, from which potential phrasal-verb particles tagged as adverbs were retrieved for manual
Syntactic stability & change in 19th-c. newspapers 317 Table 16.3 Progressives and phrasal verbs by period in CNNE (raw frequencies and frequencies per 1,000 words) Period
1 2 Total
Progressives
Phrasal verbs
#
#/1,000 words
#
#/1,000 words
256 386 642
1.6 2.4 2.0
731 792 1,523
4.6 4.8 4.7
ost-processing.4 Raw and normalised frequencies of the two features in p CNNE are given in Table 16.3. The normalised frequencies of the two features in CNNE as a whole can be compared with their incidence figures in CONCE, reported on in Smitterberg (2008). This comparison is shown in Figure 16.1. As mentioned in Section 16.2.2, comparisons between CNNE and CONCE like the one presented in Figure 16.1 are not exact: whereas period 3 in CONCE (1870– 1900) matches period 2 in CNNE (1875–95) closely, period 1 in CONCE (1800–30) pre-dates period 1 in CNNE (1830–50).5 These caveats notwithstanding, Figure 16.1 may provide a rough indication of where CNNE is situated relative to the genres in CONCE. Interestingly, CNNE takes up an intermediate position with regard to both linguistic features investigated. In both cases, the frequency in CNNE falls between the highest frequency reported for an expository genre (History) and the lowest frequency found in a non-expository genre (Fiction for progressives, Drama for phrasal verbs). These results tally well with Biber et al.’s (1999) findings for PDE. They demonstrate that, with regard to the frequency of both features, News is situated between Conversation and Fiction, on the one hand, and Academic Writing, on the other (Biber et al. 1999: 409, 461–62).6 4
As in Smitterberg (2008), I included both literal and idiomatic phrasal verbs in the counts. Literal phrasal verbs are those where the meaning of the phrasal verb can be seen as a combination of those of the verb and the adverbial particle, e.g. go back ‘return’. By contrast, the meaning of an idiomatic phrasal verb such as give up in example (2) is not simply a combination of the semantics of the verb and the adverbial particle. 5 In addition, while phrasal verbs were retrieved from both CONCE and CNNE through searches for potential particles tagged as adverbs in tagged versions of the corpora, CONCE was tagged with Conexor’s Eng-CG2 tagger whereas, as mentioned above, Conexor’s Machinese Phrase Tagger was applied to CNNE. This may have caused slight differences in retrieval. A manual check of five randomly chosen CNNE texts indicated that 97.3% (36 of 37) of the phrasal verbs in the material had been retrieved. I similarly chose five texts at random from CONCE and selected two random pages from each of those five texts for manual checking; the corresponding result for CONCE was 100% (23 of 23). This was considered a sufficient indication that using the tagged versions of the corpora produced reliable and comparable results. 6 Biber et al. (1999) appear to have used a narrower definition of phrasal verbs than that employed here, as their frequency counts for PDE are in fact lower than those reported in
318 Smitterberg Progressives
Phrasal verbs
Occurrences per 1,000 words
8 7 6 5 4 3 2 1 0 Debates Drama
Fiction
History
Letters
Science
Trials
CNNE
Figure 16.1 Progressives and phrasal verbs per 1,000 words by genre in CONCE (periods 1 and 3 only) and in CNNE
Progressives and phrasal verbs exhibit conflicting frequency developments in CNNE, as seen in Table 16.3. Whereas the frequency of the progressive increases by as much as 47 per cent between the two periods, the increase in the frequency of phrasal verbs is a mere 6 per cent, which may be best interpreted as stability in diachrony. While the progressive would thus contribute to a hypothesised process of colloquialisation in CNNE, there are no such indications with regard to phrasal verbs. In CONCE, the two non-expository genres, Drama and Letters, were those which displayed clear increases in the frequency of both features (Smitterberg 2008: 276–77), which lends some support to a colloquialisation hypothesis for nineteenthcentury English in general. Further detail can be added to the picture by considering developments in individual newspapers.7 As Table 16.1 in Section 16.2.1 made clear, six newspapers in CNNE were sampled for both periods covered by the corpus; it is thus possible to establish tentative paper-specific trajectories of linguistic development, although some raw frequencies are too low to be reliable when these newspapers are considered in isolation. The results are presented in Tables 16.4–16.5. Figure 16.1. Nevertheless, the similarity of the genre patterns is indicative of the intermediate position of News with respect to oral and literate language. The assumption of sameness when one newspaper is compared across time is not unproblematic. If, for instance, a newspaper changes owners and/or editors, this may have a considerable impact on its characteristics. Nevertheless, considering trajectories for individual papers may help to determine how widespread a quantitative change is across the papers sampled.
7
Syntactic stability & change in 19th-c. newspapers 319 Table 16.4 Progressives by period and paper in CNNE (raw frequencies and frequencies per 1,000 words) Paper
DN LM LL MG RW TT
Period 1
Period 2
#
Words
#/1,000 words
#
Words
#/1,000 words
29 14 31 41 23 18
18,481 18,361 15,004 17,710 15,649 15,633
1.6 0.8 2.1 2.3 1.5 1.2
28 53 46 44 33 18
15,644 22,697 17,058 15,865 13,346 14,521
1.8 2.3 2.7 2.8 2.5 1.2
Table 16.5 Phrasal verbs by period and paper in CNNE (raw frequencies and frequencies per 1,000 words) Paper
DN LM LL MG RW TT
Period 1
Period 2
#
Words
#/1,000 words
#
Words
#/1,000 words
104 52 84 44 87 67
18,481 18,361 15,004 17,710 15,649 15,633
5.6 2.8 5.6 2.5 5.6 4.3
62 110 99 67 71 46
15,644 22,697 17,058 15,865 13,346 14,521
4.0 4.8 5.8 4.2 5.3 3.2
There is considerable consistency in how the six newspapers pattern relative to one another with regard to the frequency of these two features. The two papers that exhibit a clear decrease (> –25%) in the frequency of phrasal verbs, the Daily News and The Times, also display stability (± < 25%) with regard to the frequency of the progressive. Conversely, the Leeds Mercury exhibits the biggest frequency increase with regard to both the progressive (+ 206%) and phrasal verbs (+ 71%). Of the remaining papers, both Lloyd’s and Reynolds’s are stable with regard to the frequency of phrasal verbs but display increases in the frequency of the progressive, which might be expected given the overall pattern attested for CNNE. The Manchester Guardian stands out from the others by exhibiting the opposite pattern: phrasal verbs become more frequent over time while the incidence of the progressive is stable. But with that exception, no paper displays a greater increase in the frequency of phrasal verbs than in that of progressives. It thus seems clear that the latter feature is more likely to contribute to a possible colloquialisation of nineteenth-century newspaper writing.8 8
Some of the patterns attested for individual papers may be due in part to the inclusion of texts with unexpectedly high or low frequencies of progressives and/or phrasal verbs.
320 Smitterberg
16.3.2 Not-contraction As shown in Biber’s (2003a) factor analysis of academic English, not-contraction (e.g. don’t in example (3)) is a feature of spoken production in PDE. (3)
‘Silence all of you,’ he cried. ‘You are like the wind, which says nothing when he speaks. Don’t you see what my father means?’ (DN0011, 1875–95)
Despite the fact that not-contraction is characteristic of speech, Hundt and Mair (1999: 227–30) show that the proportion of not-contracted forms rises in late twentieth-century newspaper English, and López-Couso (2007: 305) demonstrates that negative forms of the verbs be and have display increasing proportions of contraction in LModE as well.9 As I have shown elsewhere (Smitterberg 2012), two of the non-expository genres in CONCE, Drama and Fiction, display clear increases in the percentage of not-contracted forms, while the proportion is low and stable in Trials; moreover, there are no occurrences of not-contracted variants in the expository genres.10 Thus, not-contraction makes inroads into written non-expository English during the 1800s, and linguistic differences among written genres increase in diachrony because expository writing is resistant to this change; both developments tally with an interpretation in terms of colloquialisation. Against this background, it is of obvious interest to see whether any tendencies towards increasing proportions of not-contraction can be attested in nineteenth-century newspaper English. Potential not-contracted forms in CNNE were retrieved through searches for words ending in –n’t and –nt; in addition, a number of nonstandard spellings identified by Brainerd (1989 [1993]) and López-Couso (2007) were included in the searches. Not-contractions were then selected from the concordances through manual screening. The results show that not-contraction is very rare in CNNE: only twenty-four instances were found, of which nineteen occurred in period 1 Four of the Leeds Mercury texts from period 1 do not contain any progressives, which contributes to the dramatic increase between periods 1 and 2 for that paper; in period 2, four texts from The Times contain no progressives, which may lie behind some of the stability in the frequency of the construction as far as that paper is concerned. Conversely, as regards the frequency of phrasal verbs, the stability in the Daily News, the increase in the Leeds Mercury, and the decrease in Reynolds’s Weekly Newspaper are partly due to two texts per paper with very high frequencies (> 10) in either the period 1 sample (the Daily News, Reynolds’s Weekly Newspaper) or the period 2 sample (the Leeds Mercury). Nevertheless, the similarity with which the newspapers pattern with regard to the frequency of the two features examined is noteworthy. 9 López-Couso’s (2007) results for contracted forms include proclitic contraction (e.g. ’tis not) and verb contraction (e.g. it’s not) as well as not-contraction. 10 The non-expository Letters genre was excluded from the analysis in Smitterberg (2012) owing to the risk that uncontracted forms had been substituted for not-contractions (or vice versa) before the edited collections of letters on which the corpus texts are based were published.
Syntactic stability & change in 19th-c. newspapers 321
and five in period 2. There is thus an apparent decrease in frequency across the period covered by CNNE; however, the low raw frequencies make results unreliable. Given the dearth of not-contractions in the material, no attempt was made to retrieve their uncontracted counterparts in order to calculate a ratio of not-contraction. However, a qualitative analysis of the few contracted instances that do occur was performed in order to see whether they share any textual or contextual features. To begin with, not-contraction clearly favours spoken contexts: twelve of the twenty-four instances are from real or imaginary quoted speech, as in example (3) above. Contracted forms also appear in other contexts that are characterised by orality or informality; for instance, they occur in an editorial text where the newspaper is referred to by means of a first-person plural pronoun, as in (4), and in an informal report on travelling in Italy (written by Charles Dickens), as in (5): (4) Mr. DUFFY may take our word for it – this swindle won’t go down
with the democrats of either country. (RW0002, 1830–50) (5) But the apothecary’s has its group of loungers, who sit back among the bottles, with their hands folded over the tops of their sticks. So still and quiet, that either you don’t see them in the darkened shop, or mistake them – as I did one ghostly man in bottle-green, the other day, with a hat like a stopper – for Horse Medicine. (DN0003, 1830–50) In both (4) and (5), there are other linguistic indications of informality in the near context: won’t in (4) is part of a verb phrase that also includes the phrasal-prepositional verb go down with, and in (5), the contraction occurs in a sentence fragment. Outside such contexts, not-contraction is very rare in CNNE, and does thus not contribute to colloquialisation in this genre. 16.3.3 Conjoins of and As a co-ordinator, and can link units on different levels of syntactic structure. Following Quirk et al. (1985), I refer to these units as ‘conjoins’. In previous work, a basic distinction is often made between clausal co-ordination, as in (6), where the conjoins are main clauses, and phrasal co-ordination, as in (7), where two noun phrases make up the conjoins.11 (I use bold face for the relevant conjoins in numbered corpus examples.) (6)
11
Some strong language was addressed to one of the attendants, named Christopher Richards, a retired police-officer, now in the service of Messrs. Gatti (the present lessees of the theatre),
Not all scholars agree on what conjoins to include in which category, however. For instance, Huddleston and Pullum et al. (2002: 1,280), whose main categories are labelled ‘clausal’ and ‘subclausal’, include only full main clauses in their clausal category, while Culpeper and Kytö (2010: 161–64) even consider instances of two co-ordinated verb phrases as
322 Smitterberg
and his lordship endeavoured to force his way into the box. (BDP0001, 1875–95) (7) Of late, however, it has been noticed that the man at times was in a very low and desponding state of mind, owing, it is said, to the business not being in a very good condition, and he being in financial difficulties, and he having a dread that he and his wife would come to want. (RW0019, 1875–95) In what follows, I will use a slightly modified categorisation. As Biber et al. (1999: 181) show, speech and writing exhibit considerable differences with regard to the distribution of phrasal and clausal co-ordination in PDE: while clausal co-ordination predominates in conversation, co-ordination of phrases is very frequent in academic writing. (News takes up an intermediate position in this respect.) Moreover, Biber (2003a: 55–61) demonstrates that clausal and phrasal co-ordination characterises spoken and written academic discourse, respectively, in PDE.12 Against this background, it makes sense to assume that a hypothesised colloquialisation of nineteenth-century newspaper English would result in a decrease in the proportion of phrasal co-ordination, as the texts examined would then be moving away from a written towards a spoken norm. To investigate this matter, I retrieved all instances of and in CNNE and classified their conjoins on this parameter. Three categories are recognised in my classification. The first category is labelled ‘super-phrasal’. For an instance of and to be included in this category, both conjoins have to contain (i) more material than can be included in one syntactic phrase and (ii) at least part of a verb phrase, such as the main clauses in (6) above. Conjoins that comprise, for instance, main verbs together with their complements, finite or non-finite subordinate clauses, and larger units such as sentences also belong to this category. For the second category, I use the term ‘phrasal’. Instances of and that belong to this category have conjoins that are on or below the level of a syntactic phrase, e.g. the noun phrases in example (7) above or the two adjectives in (8). (8) The result was that the jury found him guilty of murder, without rec-
ommendation to mercy, and the judge in sentencing him gave him no hope, as he said he had scarcely ever met with a more cruel and deliberate murder. (BDP0010, 1875–95)
In addition to these two categories, I also recognised a third, ‘intermediate’ category. To this category belong, for instance, cases where one conjoin is phrasal and the other super-phrasal, as in (9), instances where each conjoin clausal co-ordination if the phrases denote separate actions. See below for an account of the classification used in the present study. As Biber (2003a: 61) notes, the dimension on which clausal and phrasal co-ordination loads displays ‘a fundamental opposition between the spoken and written modes … regardless of purpose, interactiveness, or other pre-planning considerations’.
12
Syntactic stability & change in 19th-c. newspapers 323 Table 16.6 Conjoins of and in CNNE by period Period
1 2 Total
Phrasal
Intermediate
Super-phrasal
#
%
#
%
#
%
2,375 1,984 4,359
46.9 43.3 45.2
304 224 528
6.0 4.9 5.5
2,382 2,373 4,755
47.1 51.8 49.3
Total
5,061 4,581 9,642
contains more material than one phrase but no part of a verb phrase, as in (10), and cases where the conjoins are verb phrases but contain no other linguistic material, as in (11). (9)
After the prisoner’s acquittal she returned to the village, and, from what has since transpired, there appears to have been much gossip about poison, and how bad husbands could be got rid of. (RW0007, 1830–50) (10) It would produce a double advantage, a multiplied relief to the inhabitants of Great Britain, both lowering the cost of production and extending the demand for their goods, blessing the capitalist with improved profits, and the working class with surer employment, better wages, and cheaper bread. (LM0011, 1830–50) (11) Mr. W. E. Forster, M.P., who attended, strongly urged that as large an endeavour should be made as possible to keep the present agencies at work, which had done and were doing so much good. (BDP0003, 1875–95) Of the 10,033 instances of and in CNNE, 391 (3.9%) were excluded because they were part of proper names or set expressions, or because they resisted classification, leaving 9,642 instances to be analysed. The results are presented in Table 16.6. The results show that super-phrasal co-ordination accounts for a larger share of all occurrences of and across time, at the expense of intermediate and phrasal co-ordination. This difference is statistically significant (d.f. = 2; χ2 = 23.4; p < 0.001). There is thus evidence to suggest that the distribution of different conjoins of and contributes to colloquialisation in nineteenthcentury newspaper English. Looking at paper-specific trajectories can help to determine whether the trend attested in the material as a whole is due to developments for individual newspapers or whether there is a more universal tendency to use and more in super-phrasal contexts. Developments for the six newspapers in CNNE sampled for both periods are therefore given in Figure 16.2. In the interest of clarity, Figure 16.2 includes only the percentage of super-phrasal co-ordination; for raw frequencies of all three categories and the results of significance tests based on those frequencies, see Appendix. Newspapers
324 Smitterberg 60 58 56 54
DN
%
52
LM
50
LL
48
MG
46
RW
44
TT
42 40
1
2 Period
Figure 16.2 The proportion of super-phrasal co-ordination by paper and period in CNNE
that do not change significantly across time (p > 0.05) are represented by dashed lines in Figure 16.2; note also that the value axis crosses the category axis at the 40 per cent mark. As can be seen in Figure 16.2, the Daily News, the Leeds Mercury, the Manchester Guardian and The Times undergo a statistically significant increase in the proportion of super-phrasal co-ordination across the two periods covered by CNNE, while Lloyd’s and Reynolds’s newspapers display stability in this regard. As four of the six newspapers sampled for both periods exhibit a significant shift towards super-phrasal co-ordination, and as there are no significant developments in the other direction, the trend towards more super-phrasal co-ordination seems to be a relatively widespread development in nineteenth-century newspaper English. 16.4 Discussion The present study has reached somewhat conflicting results as regards the occurrence of colloquialisation in nineteenth-century newspaper English. For two of the features of oral discourse examined, namely the progressive and conjoins of and, the findings lend themselves to an interpretation in terms of colloquialisation. By contrast, neither not-contraction nor phrasal verbs become more prominent in CNNE across time, which sets the nineteenth-century newspaper data apart from Hundt and Mair’s (1999) results for late twentieth-century English as well as from results for
Syntactic stability & change in 19th-c. newspapers 325
nineteenth-century plays (see Smitterberg 2008, 2012). The question arises whether it is possible to come up with a consistent interpretation that can account for these findings. I suggest that one explanation for the linguistic discrepancies between nineteenth-century and twentieth-century news texts lies in differences in the constraints operating on the newspaper industry in the two centuries. As Hundt and Mair (1999: 235–36) point out, newspapers are subject to market forces in order to maintain and increase circulation. However, those market forces can be expected to have encouraged slightly different stylistic ideals in the nineteenth and twentieth centuries. The tendency towards informalised discourse norms can be assumed to have been particularly pronounced during the late 1900s, when it coincides with ‘an unprecedented mobilization of formerly relatively stable class-based hierarchies in large parts of the English-speaking world’ (Mair 2006: 185). As Fairclough (1992: 204), writing near the end of the twentieth century, notes, ‘[t]he shifts of speech towards writing may have had their heyday; contemporary cultural values place a high valuation on informality, and the predominant shift is towards speech-like forms in writing’. The nineteenth century, in contrast, presents a more complex picture, in which shifts in both directions may have occurred. On the one hand, several changes in nineteenth-century British society can be expected to have encouraged increased use of oral features in writing. Between periods 1 and 2 of CNNE, the Taxes on Knowledge (the stamp duty on newspapers, and the customs and excise duties on paper) were repealed, which made it possible to lower the price of newspapers considerably; in addition, literacy rates increased (see Altick 1957: 169–72). Changes such as these made newspapers accessible to a larger proportion of the population than previously and would at the same time have diversified the readership of many newspapers in socio-economic terms. Such diversification may well have promoted the occurrence of oral linguistic features, as it would have been important for newspapers to appeal to new groups of readers.13 The Reform Acts (1832, 1867, 1884–85), which enlarged the British electorate considerably, may also have contributed to increased orality in nineteenth-century newspapers, as it became important for newspapers with a political agenda to reach new groups of voters. Towards the end of period 2, the so-called New Journalism, which was characterised by, for 13
The distribution of linguistic features in a newspaper is, of course, not a simple function of readership; for instance, the unstamped Poor Man’s Guardian was explicitly aimed at disfranchised sections of early nineteenth-century society, yet still has one of the lowest frequencies of phrasal verbs of all paper/period samples in CNNE: 3.4 occurrences per 1,000 words. (Other speech-related and informal linguistic features, such as exclamations, occur with considerable frequency in the paper, however.) Nevertheless, it makes sense to assume that the establishment of new target groups for a newspaper may lead to conscious and/or subconscious linguistic adaptation on the part of the newspaper to the usage current among those target groups.
326 Smitterberg
instance, comparatively short news items, human-interest stories, and interviews, ‘triggered discussions among the contemporaries’ (Bös 2012: 100); among other things, it was criticised for promoting ‘a general acceptance of an inferior standard of language use’ (Lee 1976: 130). To the extent that this criticism represented a reaction to real developments, it may provide an indication that the style of at least some newspapers changed in the direction of orality during this time. On the other hand, several forces in nineteenth-century British society can be expected to have counteracted tendencies towards a more speech-like written norm. To begin with, as Görlach (1999: 27) states, the written language often served as a model for speech in the 1800s (see Fairclough 1992: 204, quoted above). Rather than encouraging writing and speech to grow more similar by writing approaching speech, as suggested by Fairclough for late twentieth-century English, such an ideal may have reduced the linguistic distance between the two media through changes in the spoken mode. Although it is impossible to gauge the extent to which this happened without access to spoken language, Görlach (1999: 27) claims that writing and speech became more similar than ever before during this time.14 Moreover, the opinion that speech should approach writing is an instance of a general prescriptive attitude towards language use, which may have affected written usage. Bailey (1996: 215) claims that, during the nineteenth century, ‘[a]ttitudes toward grammar … hardened into ideology’. Arguments such as these imply that the prescriptive evaluation of an oral linguistic feature may have been an important constraint on its frequency in newspapers: proscribed features would have been less likely to be used by journalists and allowed by editors and sub-editors. It may thus be helpful to consider the contemporary stylistic evaluation of the features examined in the present study. Clausal co-ordination with and is not likely to have met with widespread negative evaluation. Although a hypotactic stylistic ideal with extensive use of subordination rather than co-ordination characterises much of the Modern English period (see, for instance, Coates 2004: 17–19), and as a super-phrasal co-ordinator is too widespread and indispensable a feature to be excised from news texts. The only use of super-phrasal and that is likely to have been censored is its occurrence in sentence-initial position, but this happens only rarely in CNNE: there are a mere forty-three instances of And with an initial upper-case letter in the corpus.15 14
Biber and Finegan’s (1997) factor analysis rather indicates that different types of writing grew further apart linguistically, as popular genres became more oral while the style of specialist, expository genres instead grew more literate. Such an increase in linguistic diversity within the written mode obviously complicates any comparisons between speech and writing where the two modes are seen as homogeneous entities. 15 Moreover, of these forty-three instances, thirty occur in period 1 and thirteen in period 2. There is thus no increase in the incidence of this overtly proscribed feature. Instead, like not-contraction, sentence-initial and appears to become less frequent in diachrony, although the low raw frequencies make results unreliable.
Syntactic stability & change in 19th-c. newspapers 327
The progressive is also an unlikely target of prescriptive condemnation. Anderwald (2012: 36) even notes that the progressive was praised in early nineteenth-century grammars (which often plagiarised eighteenth-century sources). The progressive passive, as in (12), is of course the notable exception to this positive evaluation; it met with near universal condemnation during much of the nineteenth century (see, for instance, Visser 1973: 2,427–29; Anderwald 2012: 36–39). (12) An old bridge was being demolished by dynamite. (BDP0007, 1875–1895)
Interestingly, the frequency of the progressive passive skyrockets in CNNE, with only three instances in period 1 as compared with forty in period 2. Nevertheless, the progressive passive is responsible for only a minor part of the overall increase in the frequency of progressive constructions in CNNE (and in other genres; see Hundt 2004a and Smitterberg 2005): with progressive passives removed, the frequency of the progressive per 1,000 words is 1.6 in period 1 and 2.1 in period 2. The bulk of the increase in the frequency of the progressive thus takes place in parts of the paradigm that were not proscribed. Moreover, Visser (1973: 2,428) argues that most prescriptive opposition to the progressive passive ended in the early 1870s, which may have paved the way for its frequent occurrence in newspaper text from the period 1875–95. Anderwald (2012: 38) suggests that there is a slow, gradual acceptance of the construction between the 1820s and the 1860s but locates ‘a “grudging admittance” of the progressive passive’ to the period following the 1860s. Phrasal verbs present a more complex picture. However, Claridge (2000: 178–79) speculates that a possible decrease in their frequency in the late seventeenth century and in the eighteenth century may be due in part to negative prescriptive evaluation. Because phrasal verbs were sufficiently frequent to be noticed and perceived as undesirable, and because it was sometimes possible to avoid using a phrasal verb either by omitting its particle or by substituting a more formal – often Romance – word (e.g. postpone for put off), Claridge argues that they would be likely to be singled out for prescriptive criticism. While Claridge (2000: 212–13) finds no evidence that phrasal verbs in general were proscribed in the eighteenth century, she provides a selection of negative comments on individual instances such as set out and open up; most of her examples are taken from Sundby et al. (1991). However, as Thim (2012: 227, 245) notes, at least in some cases, negative prescriptive reaction may be connected to a general preference for Romance over Germanic lexis rather than to an aversion to phrasal verbs as such. Thim (2012: 215) argues that phrasal verbs were probably ‘perceived as “colloquial” only from the 19th century onwards’; if so, however, the usage of nineteenth-century journalists and editors may still have been affected by such perceptions.16 16
Thim (2012: 217) finds it ‘hard to accept’ that ‘the remarkably uneven distribution of phrasal verbs’ in the genres of CONCE presented in Smitterberg (2008) may be connected
328 Smitterberg
The fourth feature considered in this chapter, not-contraction, is clearly the most oral and proscribed of the features examined. Leonard (1929: 170–72) includes not-contractions like dont and a’nt in his examples of contracted forms that were proscribed in what he refers to as ‘the war upon contractions’ in eighteenth-century English. Stylistic objections to their occurrence in writing continued in the nineteenth century (see, for instance, Phillipps 1984: 68–69). There are probably several reasons for their status as conspicuously oral forms. To begin with, the high frequency of notnegation means that their occurrence potential is considerable, especially in conversation, where not-negation is clearly favoured over no-negation, at least in PDE (see Biber et al. 1999: 170–71). Moreover, their written form represents a clear adaptation to what is likely to have been the predominant pronunciation. 16.5 Conclusion When these differences in the likely stylistic evaluation of the four linguistic features are taken into account, a pattern emerges. Oral features that were not overtly proscribed, such as super-phrasal co-ordination and the progressive, do become more predominant in nineteenth-century newspaper texts; in contrast, (potentially) proscribed features either appear to exhibit stability across time (phrasal verbs) or display very low and possibly decreasing frequencies (e.g. not-contraction and sentence-initial and). The only apparent exception to this pattern is the progressive passive, which increases dramatically between the periods covered by CNNE, despite being one of the most stigmatised features of nineteenth-century English. However, as prescriptive condemnation of the progressive passive became less forceful towards the end of the 1800s (Visser 1973: 2,428), its increasing frequency in CNNE may be due in part to a decrease in stigmatisation. As the high frequency of the progressive passive in media language has also been attested outside CNNE (see Hundt 2004a: 109 for diachronic data from ARCHER and Smith and Rayson 2007 for late twentieth-century English), this matter deserves special investigation that falls outside the scope of the present study. Against this background, my tentative conclusion is that colloquialisation does take place in nineteenth-century newspaper English, but also that this process was more limited and more constrained by contemporary to colloquialisation. However, Thim (2012: 215) himself agrees that phrasal verbs were seen as colloquial during the 1800s, and their incidence in CONCE reveals a clear difference between non-expository and expository genres. Since Geisler (2002) clearly demonstrates that these two groups of genres also pattern differently on the oral/literate dimensions identified in Biber (1988), the association of phrasal verbs with orality in nineteenth-century English seems fairly clear. The fact that phrasal verbs also increase in frequency in Drama and Letters, arguably the two most informal genres in CONCE, lends further support to this association.
Syntactic stability & change in 19th-c. newspapers 329
stylistic ideals than is the case for twentieth-century English. The interaction of potentially conflicting forces such as a widening of the readership and prescriptive perspectives on usage creates a complex pattern where the linguistic features involved must be examined individually as regards their distribution and evaluation. The present study has not exhausted the topic of linguistic change in the direction of orality in nineteenth-century news texts. Several linguistic features remain to be examined, such as the choice between on and upon (see Hundt and Mair 1999). In addition, more detail can be added to the picture presented here; for instance, the distribution of phrasal and super-phrasal co-ordination in CNNE should be compared with the relative frequency of these categories in other genres, such as those represented in CONCE. It is hoped that future studies will continue to open up what seems to be yet another field of complex linguistic variation and change in LModE. Appendix Table 16.7 Conjoins of and by paper and period in CNNE Paper Period 1 Phrasal
DN LM LL MG RW TT
Period 2 Intermediate Super-phrasal Phrasal
Intermediate Super-phrasal
#
%
#
%
#
%
#
%
#
%
#
%
285 330 181 254 234 248
51.3 54.8 40.9 46.4 45.1 50.8
38 28 25 30 39 31
6.8 4.7 5.6 5.5 7.5 6.4
233 244 237 263 246 209
41.9 40.5 53.5 48.1 47.4 42.8
185 277 215 156 185 153
44.5 46.8 43.4 40.2 40.9 39.8
18 22 28 14 30 24
4.3 3.7 5.7 3.6 6.6 6.3
213 293 252 218 237 207
51.2 49.5 50.9 56.2 52.4 53.9
For the Daily News: d.f. = 2; χ2 = 9.35; p = 0.009. For the Leeds Mercury: d.f. = 2; χ2 = 9.74; p = 0.008. For Lloyd’s: d.f. = 2; χ2 = 0.668; p = 0.716. For the Manchester Guardian: d.f. = 2; χ2 = 6.61; p = 0.037. For Reynolds’s: d.f. = 2; χ2 = 2.46; p = 0.292. For The Times: d.f. = 2; χ2 = 11.2; p = 0.004.
Part VII Language contact and syntactic change
17
‘[W]ell are you not got over thinking about going to Ireland yet’: the be-perfect in eighteenth- and nineteenth-century Irish English Kevin McCaffe rty
17.1 All is changed? The be -perfect with intransitive mutative process verbs like change and verbs of motion like go dates back to Old English (Visser 1973: 2054–84; Denison 1993: 359) but is no longer widespread in Present-Day Englishes (PDEs). By 1900, it was essentially obsolete in mainstream standard varieties (Rydén and Brorström 1987; Rydén 1991; Denison 1993: 344; Kytö 1994, 1997; Görlach 2001: 120–21), surviving today only with a few verbs, where it seems to be lexicalised in ‘archaic constructions with specialized, largely adjectival meaning such as He is gone, She is finished’ (Brinton and Traugott 2005: 78). It has been suggested that the influence of normative grammarians – who condemned the use of be without criticising the innovation – might have tipped the balance towards hav e in the late eighteenth century (McFadden 2007a; cited in Tieken-Boon van Ostade 2009: 97). Certainly, there was still disagreement among eighteenth- and nineteenthcentury grammarians, though the balance was already heavily in favour of h av e : far more grammarians condemned the use of auxiliary b e with these verbs and recommended hav e than vice versa (Sundby et al. 1991: 180–81; Anderwald this volume). As Kytö (1997) shows on the basis of data from the Helsinki Corpus and ARCHER, the be -perfect was robust during the period of major British settlement in Ireland, from the mid sixteenth to the early eighteenth century (Fitzgerald and Lambkin 2008; Bardon 2011), and was one of the English/Scots features that went into the feature pool from which Irish English (henceforth IrE) emerged. Summarising a recent survey of PDE grammar in Britain and Ireland, Kortmann notes that the b e -perfect persists in certain regions and reports it as ‘pervasive’ in IrE (2008a: 491, Table 1).1 The use of auxiliary b e with The author acknowledges the support of the University of Bergen’s Meltzer Foundation (Grant No. 9334, 2008–09) and the Research Council of Norway (Grant No. 213245, 2012–15). 1 This survey used They’re not left school yet to exemplify ‘be as perfect auxiliary’ (Kortmann 2008b: xxvi).
333
334 McCafferty
intransitives is included in most surveys as a feature of both Northern and Southern IrE (e.g. Bliss 1984; Harris 1984a, 1993; Filppula 1999, 2008; Hickey 2007a, 2007b; McCafferty 2007; Amador-Moreno 2010; Kallen 2013). There has, however, been little empirical study of this construction in IrE and no diachronic survey, while a reading of the literature raises issues that might usefully be addressed diachronically using CORIECOR, the Corpus of Irish English Correspondence (McCafferty and Amador-Moreno Forthcoming).2 First, as Filppula (2008: 331) remarks, of all the IrE perfectives,3 the be -perfect in particular is recessive. Yet empirical studies (Harris 1984b; Filppula 1999; Ronan 2005) show it remains one of a set of roughly equally robust alternatives to the hav e - perfect, the latter being a minority construction overall in at least some parts of Ireland. When did the b e -perfect begin to recede in IrE, and to what extent did this parallel developments in other varieties? Second, as in other Englishes, the b e -perfect is reportedly restricted in IrE to a limited number of verbs (Kallen 2013: 102–103), go in particular (Filppula 1999: 117; Ronan 2005: 254). Might its survival in IrE be due mainly to retention with such frequent verbs? Third, there is the issue of Irish substrate influence. Filppula suggests the b e -perfect’s survival was the result of convergence between English and Irish – which has a parallel construction, used with transitive and intransitive verbs alike – during prolonged language contact and shift (1999: 122). On the other hand, Hickey acknowledges that the substrate may have supported retention (2007a: 177), but attributes use of auxiliary b e to English input only (2007a: 196, 282, Table 4.39). However, the fact that Irish does not have a verb have but instead uses a construction formed with the substantive verb tá ‘be’ + a form of ag ‘at’ (see 17.2.3, below), both to express possession and in its closest equivalent of the be -perfect, may well have contributed to retention of the be -perfect in IrE. The fact that many intransitives in English can take an object-like complement may have contributed further to convergence and retention of the b e -perfect, as well as facilitating its extension to transitive uses of verbs like finish, a verb with which the Irish construction is also used. 2
CORIECOR is currently under development. At the time of writing, the corpus consists of approximately 5,500 documents (3.6m words), of which some 4,900 are personal letters (3m words) written to and from Irish emigrants from the 1670s onwards. Coverage is good from the 1760s to the 1940s (minimum 55,000 words per twenty-year subperiod). Most of the texts have been downloaded from the Irish Emigration Database, hosted by the Mellon Centre for Migration Studies, which is part of Queen’s University Belfast. There is a bias in this database towards the northern province of Ulster and the east coast region of Leinster, especially in the earlier subperiods, though this bias is in part a reflection of the fact that these were the regions where English was widely spoken before 1800. Several new collections have recently been added, which provide data from previously under-represented subperiods (like the early eighteenth century) and improve the geographical spread of the material. 3 For detailed presentations of IrE perfects, see the surveys cited in the preceding paragraph.
The B E -perfect in 18th- & 19th-c. Irish English 335 100 90 80 70 60 50 40 30 20 10 0
(1 35 0– od (14 142 20 0) EM E 1 – od (15 150 0 00 E EM 2 –1 ) ( 1 57 od 5 AR E 3 70– 0) 1 C 64 H (16 0) AR ER 40 – ( 1 C AR HE 165 71 0– 0) R C H 17 ER 2a 0 (1 AR 2b 70 0) 0 C ( AR HE 175 –50 ) R 0– C H 18 ER 3a 0 ( AR 3b 180 0) 0– (1 C H 5 8 AR ER 50– 0) 19 4a C H ER (19 00) 4a 00– 5 (1 95 0) 0– 90 ) 4
3
E
EM
M
E M
be have
Period
Figure 17.1 B e /have variation with intransitives for subperiods 1350–19904
Diachronic study of regional differentiation in CORIECOR may help clarify this issue further: we might expect auxiliary b e to be more robust in regions where IrE spread through language shift and remained in contact with Irish longer, as opposed to the north-eastern and south-eastern regions centred on Belfast and Dublin, respectively, that became English-speaking early and largely through settlement from Britain. In the meantime, since the transitivity of verbs used with this construction in the two languages is a striking structural difference between Irish and English, we might ask whether the presence of an object-like complement influences the selection of be rather than hav e . Given the history of the transitive verbs used with auxiliary be in examples cited in the literature on IrE, we might ask whether this is due to a transfer effect in the emergent IrE of the eighteenth and nineteenth centuries. This diachronic survey of the b e -perfect in IrE attempts to address these questions. 17.2
Be -
and have-perfects in the history of Englishes
With intransitive verbs, be and hav e were already in competition in Old English (Traugott 1992:191; Hogg 2002:79). H av e gained ground from the 4
This graph is intended to show the general historical decline of the be -perfect. Note that ARCHER includes both BrE and AmE from subperiod 2a onwards. Kytö’s (1997) results from the Century of Prose Corpus, showing an aggregate 67% have use from 1680 to 1780, are excluded because they obscure the diachronic pattern, and COPC overlaps anyway with the HC’s EModE 3 and ARCHER subperiods 1, 2a and 2b.
336 McCafferty
early fourteenth century onwards, but corpus-based diachronic studies show be still dominating with intransitives into Late Modern English (LModE) (Rydén and Brorström 1987; Kytö 1994). In the most detailed study of a multi-genre corpus (Kytö 1997), hav e is shown to have become the majority form in the late eighteenth century and was categorical (minimum 86% use with intransitive verbs) a century later (Figure 17.1).5 This means the be -perfect was the dominant form with intransitives in the English/Scots varieties brought to Ireland in the Plantation era, represented by the first six points from left to right in Figure 17.1. By the upper end of the sharp cline that begins in 1700 and slows after 1900, h ave was categorical with this verb type. The period of rapid change thus coincided with the rise of prescriptivism as the dominant linguistic ideology in the English-speaking world; for detailed analysis of this relationship, see Anderwald (this volume). Crucially, it was also during this period that Irish speakers began shifting to English in increasing numbers. And the period also saw the onset of mass emigration from Ireland, first to North America, and later also to Australia, New Zealand and Great Britain, a movement for which a command of English was widely regarded as essential. Kytö also points to signs of an ‘innovative tendency’ in late eighteenthcentury AmE (1997: 39), which led BrE in replacing b e with h av e . As we have seen, however, the trend towards hav e was already well established by then, so that, as Hundt (2009:17–18, 32) observes, AmE did not initiate the change, but had merely gone further than contemporary BrE, which soon caught up with American usage in the next century: both varieties show categorical hav e use with intransitives after 1850. For about the last 150 years, then, mainstream Englishes on both sides of the Atlantic have had only a residual b e -perfect. Commentators often observe that auxiliary be is retained in PDE only with certain verbs, go in particular, while this and a number of other verbs are also still found in the be -perfect construction in IrE. This implies that, to the extent that the b e perfect is maintained in Ireland after 1750 (when h av e became dominant in AmE, with BrE soon catching up), IrE was conservative in this respect. This is apparent from Figure 17.2, which traces the development of the b e perfect in eighteenth- and nineteenth-century IrE and permits a broad initial comparison with BrE and AmE. From a slight majority (53%) of be -forms in eighteenth-century IrE, use of h ave increases steadily to the 1830s and rises further by the 1880s. However, this is a gradual increase, not the rapid swing reported by Kytö (1994, 1997) and Rydén and Brorström (1987). IrE shifts from 47 per cent h ave to 70 per cent over this century, while Kytö’s data showed higher levels of hav e use over the same period (rising from 61% to 88%). IrE, 5
Rydén and Brorström (1987: 200) corroborate this dating (see also Rydén 1991).
The B E -perfect in 18th- & 19th-c. Irish English 337 80 70 60 50 be have
40 30 20 10 0 18th-c. IrE
IrE 1830s
IrE 1880s
Period
Figure 17.2 B e /have-perfect with intransitive verbs in IrE (CORIECOR) 80 70 60 50 40
be have
30 20 10 0 ARCHER 2 1700–1800
CORIECOR 1701–1800
ARCHER 3 1700–1800
CORIECOR CORIECOR 1830s 1880s
Figure 17.3 B e /have with intransitives in letters only, ARCHER and CORIECOR compared
then, appears to have followed the general development in English, but at a delay of about fifty years. As we will see, the treatment of certain transitive verbs in IrE, which actually increased with auxiliary b e up to 1890, may explain some of the apparent delay in adopting h av e in this variety.6
6
Note that, e.g. finish was not included in Kytö’s study because it can be used transitively. This discrepancy between studies of IrE and the varieties treated by Kytö is unavoidable; finish is often cited as occurring with the be -perfect in IrE, and must be included here.
338 McCafferty
Kytö’s general results are based on multi-genre corpora. However, she also reports results from a small subset of data from letters (Kytö 1997: 42–44), which are more directly comparable to the IrE data. Figure 17.3 compares Kytö’s findings for the eighteenth and nineteenth centuries with CORIECOR data from three subperiods.7 IrE lagged behind in the late eighteenth, but made up considerable ground by the late nineteenth century. In other words, the development seems broadly parallel in both corpora, with IrE trending in the same direction as that found in ARCHER, but at a slower pace. 17.2.1
B e -perfect
in regional varieties of English
Kortmann’s (2008a) grammatical survey of present-day Britain and Ireland suggests a more robust be -perfect in peripheral varieties. Uniquely, Shetland and Orkney dialects have generalised b e to all verbs in the perfect, transitives included (Melchers 2008: 291), which may be a Norn (Scandinavian) substrate influence (Pavlenko 1997).8 Otherwise, as noted above, Kortmann reports the be -perfect as ‘pervasive’ in IrE and ‘attested, but not frequently used’ in south-west England, the north of England and Scotland (2008a: 491, Table 1). However, comparing twentieth-century IrE with dialects from south-west England, Yorkshire and the West Midlands, Filppula (1999: 49–50) reported the b e -perfect as present, but infrequent, in conservative BrE dialects; it was also restricted to use exclusively with go (Filppula 1999: 118). Among regional Englishes, then, leaving aside the unique case of Shetland/Orkney, IrE stands out as a variety that we would expect to have retained a robust b e -perfect throughout the eighteenth and nineteenth centuries. 17.2.2
B e -perfect
in Scots
Since most British settlers in Ireland originated in Scotland (though some spoke Scottish Gaelic rather than Scots/English), it is unfortunate that the historical evidence for the b e -perfect in Scots and Scottish English is patchy. Apart from Shetland/Orkney, brief mentions in surveys are all we have about this feature in Scots. Thus, Trudgill and Hannah simply list transitive I’m finished it as ‘specifically Scottish’ (2008 [1982]: 101), with no further discussion or examples, and many surveys have nothing to say 7
Kytö’s dataset is small compared to CORIECOR, just 79 tokens distributed across ARCHER subperiods: 2a – 11 tokens; 2b – 30; 3a – 13; 3b – 25). Kytö’s analysis by 50-year periods (1997: 44) actually shows a pendulum movement, away from mainly be -forms towards h av e , then back to be . 8 Smith and Durham (2012: 62–63) chart the recent decline of the be -perfect among the younger generations in Shetland.
The B E -perfect in 18th- & 19th-c. Irish English 339
about the be -perfect at all. When it is included, little detail is offered, as in Macafee’s survey of Scots grammar, which merely notes the generalised b e -perfect in the northern isles and then tantalisingly adds that: Be is also the regular auxiliary in Scots generally with a small group of verbs including start and come: when I was just started school in the babies class, … (‘Our Merry’, op. cit, p.15) and when the dog wis come tae its pup – … (‘The princess and the pups’, op. cit., p.260) (Macafee 2011: np.)
Commenting on the paucity of documentation and interest, Melchers observes that: there is hardly any evidence of the construction being a general feature of Scots. It is not mentioned in the standard handbooks and dictionaries and no attempt was made at eliciting it in the – admittedly very restricted – grammatical section of the Linguistic Survey of Scotland. Nor is there any evidence in the recent translation of the New Testament into Scots. (Melchers 1992: 603)
How widespread the be -perfect might be today in Scotland beyond the northern isles remains, therefore, a matter for future study. Documentation in earlier Scots is, if anything, even poorer. There are no empirical diachronic studies, though Moessner notes that auxiliary b e was used variably with verbs of motion in Older Scots (1997: 113), and Görlach (2002: 105) repeats this. Apart from the recent work on Shetland dialect, then, the be -perfect does not appear to have been studied empirically in either Scots or Scottish English at any stage. While we would ideally like to know more about the situation in Scots, especially diachronically, for a study of IrE, at the minute we can only note that, the northern isles aside,9 the b e -perfect was used in Older Scots and is still found elsewhere, with some verbs at least. 17.2.3 ‘B e -perfect’ in Irish In relation to tense and aspect in IrE, it is always necessary to consider possible Irish substrate influence. The potential Irish input in this case is a construction using the substantive verb tá ‘be’ with what is variously termed either the ‘verbal adjective’ or ‘past participle’ (Bliss 1979: 294; Stenson 1981: 148–50; Ó Siadhail 1989: 299–300; Ó Sé 1992: 39; Hickey 2012). Some view this as a passive construction (e.g. Hickey 2012), others 9
Scots settling in Ireland generally came from south-west and central Scotland. There is little evidence of migration from Shetland/Orkney to Ulster or any other part of Ireland (Fitzgerald and Lambkin 2008; Bardon 2011), so that northern isles influence can safely be discounted.
340 McCafferty
as a passive perfective (e.g. Ó Siadhail 1989), but Irish examples are usually translated by the hav e -perfect or one of the IrE alternatives. For Ó Siadhail (1989: 299), it is the ‘passive perfective aspect’, which is closely related to the IrE resultative perfective. The Irish structure uses the substantive verb with the ‘verbal adjective’ (equivalent to the English past participle) and a form of the preposition ag ‘at’, as in (1). (1)
Tá [Is
an leabhar léite the book read
agam at-me]
‘I have the book read’ (after Ó Siadhail 1989: 299) The immediate substrate influence on the IrE b e -perfect is the ‘agent-free passive perfective’ (2), which lacks the prepositional element but is otherwise identical to the construction in (1). Significantly, this has a wider scope than that usually attributed to the English b e -perfect, being used with both intransitive (2) and transitive verbs (3) (Ó Siadhail 1989: 299–300); the latter provides an obvious overlap with the passive. (2)
(3)
Tá sé imithe [Is he gone] ‘He is gone off ’ Tá an leabhar léite [Is the book read] ‘The book is/has been read’ (after Ó Siadhail 1989: 299–300)
A further option in Irish is to use certain transitive verbs, like ‘eat’ and ‘cast (a vote)’, without an object in the ‘agent-free’ construction exemplified in (4) and (5), where the logical objects (e.g. a chuid ‘his share’, vót ‘vote’) are missing. (4)
Tá sé ite [Is he eaten] ‘He has eaten’
(5)
Tá sé caite [Is he voted, lit. ‘spent, thrown, cast’] ‘He has voted’ (after Ó Siadhail 1989: 300)
This latter possibility appears not to have transferred into IrE; consequently, it is seldom remarked on in the literature (but see Filppula 1999: 121), and there are no tokens in the CORIECOR sample used here. Yet the fact that IrE permits the b e -perfect with transitive uses of, for example, finish may prove to be attributable to Irish influence, as Irish has this option.10 10
Mustanoja (1960: 500–01) points to Middle English occurrences of the be -perfect with transitive verbs. If this continued into the Early Modern period, it would weaken the case
The B E -perfect in 18th- & 19th-c. Irish English 341
Finally, it might be important to take account of dialect differences in Irish. Ó Sé (1992: 41) reports that the Irish equivalent of the b e -perfect is a feature of Connacht and Munster but not Ulster dialects. We might, therefore, expect the converging influence of Irish to be stronger in Southern than in Northern IrE, where the be -perfect might be less likely to be affected by Irish substrate influence (Amador-Moreno 2006: 110). It would be useful, therefore, to investigate the geographical distribution of the b e perfect in IrE, although this is beyond the scope of the present study, since CORIECOR is not yet searchable by regional origin. 17.2.4 The be -perfect in Irish English 17.2.4.1 Present-day evidence Kortmann (2008a: 491) characterises the be -perfect as ‘pervasive’ in IrE, and surveys usually list this as a resultative perfect found with intransitive verbs, noting that it is more frequent in IrE than other varieties (e.g. Harris 1984b: 322–23, 1993: 160; Kallen 1989: 19, 2013: 103; Filppula 1999: 116–22, 2003: 166–67). While the three empirical studies that quantify all the IrE perfect constructions in their data sets show the b e -perfect to be a fairly robust alternative to the hav e -perfect in present-day IrE, it is also reported as largely restricted to the verb go (Harris 1984b; Filppula 1999; Ronan 2005), as in other Englishes, standard varieties included. Harris studied the perfect in IrE using data from the Tape-Recorded Survey of Hiberno-English Speech, reporting that the be -perfect is ‘restricted to mutative verbs such as leave, change, die, go’ (1984b: 308). Just over half of all perfects in Harris’s largely Northern IrE sample (n = 119/217, 55 per cent) were non-standard, while 45 per cent were of the standard English h av e perfect. After the preterite, the b e -perfect was the second most frequent non-standard option, accounting for 11 per cent of the data (n = 23/217). Filppula (1999) studied the b e -perfect and other perfects in Southern IrE data that was overall considerably more non-standard than Harris’s: still, only 14 per cent of Filppula’s perfectives were h av e -perfects. Here, too, the preterite (57 per cent of all perfects) was the main non-standard option, while the be -perfect ranked third, with 9 per cent of perfectives in this study (Filppula 1999: 95–126). Finally, Ronan’s (2005) study of usage in Dublin found a rather different picture. Her data were less non-standard generally (41 per cent overall), less than half the Dublin rate reported by Filppula. Also, Ronan’s data were more evenly spread between the non-standard/IrE constructions: none exceeded 12 per cent use, though here, too, the be -perfect occurred at 9 per cent. for Irish influence and further strengthen convergence explanations. As we have noted, auxiliary b e with transitive finish in Scots might also further a convergence argument, if such usage could be documented historically.
342 McCafferty
These studies suggest that the be -perfect is used at similar rates in Northern and Southern IrE, offering little support to the hypothesis that the presence or absence of a parallel in different substrate dialects of Irish might have affected the be -perfect’s regional distribution in IrE.11 Further evidence of the lack of regional differentiation comes from the Survey of Irish English Usage: acceptance rates for the test sentence They’ re finished the work now exceeded 85 per cent in counties as far apart as Derry and Monaghan in Ulster, Offaly in the Midlands, and Kerry in Munster (Hickey 2007a: 178). Elsewhere, Hickey also reports that 80 per cent of Dublin respondents found this test sentence ‘no problem’ (2005:130). The be -perfect appears to be still fairly robust overall in non-standard IrE, where a minority of all perfects are of the standard English h av e perfect type, while the remainder are divided among five other different constructions (for surveys, see Harris 1984b; Kallen 1989; Filppula 1999, 2008; Ronan 2005). Its use contributes to a substantial IrE deviation from mainstream standard Englishes and other varieties in this area of grammar. A recent study based on data from various components of the International Corpus of English (Seoane and Suárez-Gómez 2013: 9, Table 1) reports that be -perfects account for a mean of only 1.6 per cent of all perfects for Hong Kong, Singapore, Indian and Philippines English and 1.4 per cent in BrE. In contrast, the IrE studies cited above consistently show around 10 per cent of all perfects are b e -perfects, so that this perfect may be six to seven times more frequent in IrE than in BrE or East and South-east Asian varieties. For some speakers of IrE, the hav e -perfect remains a minority variant into the late twentieth and early twenty-first century. However, it is a qualified robustness that the be -perfect enjoys, as it appears nowadays to be restricted to use with a small number of verbs in IrE, too. It is also worth noting that Hickey’s test sentence contained the transitive verb finish and an object the work; as we will see, in our period, this verb actually reversed the general trend towards greater use of the hav e -perfect. Hickey also cites further examples with go and change from various data sets (2007a:178, 196): The amusements are gone quite expensive. (TRS-D, M42, M) They’re certainly changed for the better. (TRS-D, U39, M) They’re finished with the repairs now. (DER, M70+) The kids are gone to the strand today. (WER, F55+) They’re finished the school exams now. (RL, F55+) Harris lists the motion and process verbs leave, change, die and go as occurring with auxiliary b e and his example is I’m not too long left (1984b: 308). Ronan (2005: 254) notes that be -perfects occur particularly with go and 11
However Harris’s (1984b) results suggest urban–rural differentiation: standard have -perfects were nearly twice as frequent among urban speakers (65% vs. 35%), while the beperfect was 3.5 times more common in rural areas (14% vs. 4%). The question of regional differences is also worth investigating.
The B E -perfect in 18th- & 19th-c. Irish English 343
cites examples with this and finish (2005: 256). The most detailed IrE study of the be -perfect to date (Filppula 1999: 118) mentions the following: go, leave, finish, change, come, vanish, wear, wither, fade, dry, break up, die, happen, and belong. And Kallen (2013: 102–103) adds pass, build, break down and promise. This amounts to quite a number of fairly frequent everyday verbs, but we should also bear in mind Filppula’s remark that the vast majority of b e -perfects in late twentieth-century IrE involved go (1999: 120). We might, then, conclude this literature survey by noting that the b e perfect is retained in IrE today and is widely regarded as an acceptable usage, but that it is restricted to use with a small number of intransitive mutative verbs of process and motion, especially go. It is also used with a smaller number of transitive verbs, of which finish and do (= ‘finish’) might be the most frequent. Nevertheless, as Kirk and Kallen remark in their discussion of the perfect in standard IrE, while the distinctively Irish realisations of the perfect individually are relatively infrequent compared to the h av e - perfect, even small proportions of the other constructions distinguish IrE from other varieties, providing links between Irish standard English on the one hand and vernacular IrE – with much higher rates of the b e -perfect and other IrE perfective constructions – and the Irish language on the other (2006: 103). That the alternatives to the hav e -perfect may actually combine to constitute an aggregate majority of perfective use underscores the deviation of IrE from other varieties. 17.2.4.2 Historical evidence There has to date been no long-term diachronic study of the b e -perfect in IrE, though its existence in Ireland is documented at various times, and incidental historical evidence for IrE usage is provided in studies concerned with the English language in general, or with other aspects of the perfectives. This section surveys such evidence. Bliss (1979: 294) documents the b e -perfect with verbs of motion in representations of IrE from 1600–1740: di Lady is runne away from dee (v 10) [Thomas Dekker, 1605] I am come a great vay of miles (vi 23) [Ben Jonson, 1613/16] Duke Scomberg and his Army is come (xii 19) [Richard Head, 1663] ’tis come burying you are de corp (xviii 6) [John Dunton, 1698] he’s dead and buried these ten years (xxvii 115) [Thomas Sheridan, 1740] the wind is turn’d (xxvii 152) [Thomas Sheridan, 1740] the wind was turn’d (xxvii 148, 154) [Thomas Sheridan, 1740] (after Bliss 1979:294) Bliss’s texts span the period from the onset of large-scale British settlement to the mid eighteenth century, when the flow of British settlers was reduced to a trickle. Bliss shows at least that the construction continued
344 McCafferty
in use throughout this period in Ireland, when b e was still dominant with intransitives in English generally. Further evidence of continuity comes from Hickey’s survey of Dublin English, which cites examples of the b e -perfect from eighteenth- and nineteenth-century letters, occurring with come, go, arrive and begin (2005: 161–66): I am just come (Cecily Byrne, 1763) They’re gone to town (Thomas Henry Edwards, 1843) I am arrived safe at the close of another week (Thomas Henry Edwards, 1843) Winter is just begun (Owens Letters, 1899, 2 in Hickey 2005: 162–66) But this feature is not listed as used in plays by Dubliners Dion Boucicault and Seán O’Casey, writing in the nineteenth and early twentieth centuries, respectively (Hickey 2005:167–77; also Hickey 2007a: 197). There are some Irish writers represented in the texts used by Rydén and Brorström (1987). The findings for these seem to broadly reflect the general development from be to have in English and the delay in adopting have in IrE. In the eighteenth century, Jonathan Swift (1667–1745) almost categorically preferred be with intransitives, using it in 85 per cent of tokens in letters (Rydén and Brorström 1987: 201, 232, Table I); this is well in excess of the rates reported by Kytö (1997) for this period (see Figure 17.1). Other Irish playwrights of the eighteenth century continued to use the be -perfect more than British and American contemporaries. Hugh Kelly (1739–1777) had be perfects with 63 per cent of intransitives (Rydén and Brorström 1987: 232, Table I), while Thomas Sheridan (1719–1788) and Oliver Goldsmith (1730– 1774) likewise exceeded the rates reported by Kytö (1997) for the eighteenth century: Sheridan used approximately 65 per cent be , Goldsmith 60 per cent (Rydén and Brorström 1987: 21–22). Compared to Swift, then, Kelly, Sheridan and Goldsmith show a decline in be use. In stark contrast with these writers, a century later, Dion Boucicault (1820–1890) was a categorical user of 12 have in his plays (89 per cent) (Rydén and Brorström 1987: 233, Table II). The same applies to Oscar Wilde (1854–1900) – 96 per cent have in letters, 92 per cent in plays – and George Bernard Shaw (1856–1950), whose plays showed 90 per cent have use (Rydén and Brorström 1987: 233, Table II). There are few detailed studies of be / hav e variation in works by twentieth-century Irish authors, but the evidence available is consistent with the decline of the be -perfect and its restriction to particular verbs. In early twentieth-century works by the bilingual Donegal novelist Patrick MacGill, Amador-Moreno found just three be -perfects: one each with go, come and 12
Hickey (2005: 167–77) does not report use of the be -perfect in his samples of either Boucicault or O’Casey. For Boucicault, Hickey and Rydén and Brorström used different samples: Hickey’s Corpus of Irish English contains only The colleen bawn, whereas Rydén and Brorström included The colleen bawn and London assurance.
The B E -perfect in 18th- & 19th-c. Irish English 345
change (2006: 106–08). As Amador-Moreno notes, the existence of this perfect in Ulster, where the Irish substrate lacks the potential Irish source construction, might suggest retention from English rather than substrate influence. Her findings also point to the recessive nature of the b e -perfect in IrE and lexical restriction. Recent work on the use of IrE in twentieth-century films and television sitcoms appears to support this account, too. Walshe (2009: 54–55) found b e -perfects in only seven of the fifty films he investigated, citing just seven tokens: five with go, one each with change and finish (Walshe 2009: Appendix 3, Table 6). Also, Walshe’s study of IrE in the sitcom Father Ted found the b e -perfect just twice; the only example cited uses finish (2011: 132, 136). This historical survey suggests that IrE in most respects participated in the general decline of the be -perfect in both standard and non-standard varieties of the English language since the eighteenth century. However, the shift to h ave instead of auxiliary b e in texts by nineteenth-century Irish literary writers is considerably more rapid than the slow attrition of the b e -perfect in the CORIECOR sample; literary writers from the nineteenth century onwards may have been more in tune with wider mainstream usage than CORIECOR letter-writers. The historical curve suggested by this survey appears to follow a path similar to that found by Rydén and Brorström (1987) and Kytö (1994, 1997) and also supports the view that the change was delayed in IrE by half a century or so. After 1900, the be -perfect also appears, on the evidence collated here from studies based on literary sources and corpora, to have become circumscribed to use with a small number of frequent intransitive mutative verbs of motion and process verbs, as well as some transitive verbs, including finish and do (= ‘finish’).13 While the be -perfect has declined in IrE in the last century or so, it nonetheless remains one of five alternative perfective constructions that occur in different contexts but are still collectively used in preference to the have -perfect. Together, these contribute to a considerable divergence between IrE and other varieties in this area of the grammar. 17.3
Be -perfect
in CORIECOR
Data for the present study was extracted from CORIECOR for the entire eighteenth century, the 1830s, and the 1880s. Since there is relatively little material from the early eighteenth century, this effectively covers a period of about 140 years (1751–1890), during which Ireland was transformed from a 13
Melchers (1992: 602–03) notes be -perfects, including transitives, from Norfolk, Bedfordshire, Northamptonshire, Hertfordshire and Rutland in Wright’s English Dialect Dictionary (1898–1905), and in Cambridgeshire texts in the Helsinki dialect corpus. The issue of whether b e was more widely used with transitives in regional Englishes might therefore be worth studying more widely.
346 McCafferty Table 17.1 Verbs included in CORIECOR searches for be perfects Rydén and Brorström (1987) go come get ‘reach, become’ arrive become
return fall grow pass recover
Kytö (1997) depart
enter
From literature on IrE leave die finish vanish promise
wear wither fade dry do ‘finish’
improve turn ‘change, motion’ (a)rise change run
break up/down happen belong build
predominantly Irish-speaking country to a virtually monolingual Englishspeaking territory. Like Kytö (1997), I searched lexically for the past participles of verbs known from the literature to have occurred with the b e -perfect. The items included (Table 17.1) were: (a) the fifteen verbs that together accounted for 87 per cent of the data in Rydén and Brorström’s (1987: 31) study; (b) the two additional verbs among the twelve that produced 63 per cent of Kytö’s data (1997: 31–32) but which were not among Rydén and Brorström’s; and (c) fourteen further verbs that were not so frequent in these previous studies, but are often mentioned in the literature on IrE as occurring with auxiliary be . This ensured that, if the b e -perfect has indeed been used across a broad range of verbs and verb types, then the full range would be revealed using these mostly fairly frequent verbs, and that any IrE peculiarities might also be captured by inclusion of the verbs most often cited as used with the b e -perfect in this variety. As it happens, five of the verbs in Table 17.1 (belong, build, fade, promise, and wither) did not occur in the perfect in the data anyway, leaving twenty-six verbs. Several of these verbs can be used in either process or motion senses. Of these, go is found only in the sense of motion; turn is only used in the process sense (‘change’) in the first and last subperiods, while it is also a motion verb in the 1830s data; and get occurs with both process (‘change’) and motion (‘reach, arrive’) meanings in all three subperiods. Searches were conducted using Wordsmith 5 (Scott 2009) to retrieve past participles of individual verbs, including all variant participle forms, e.g. got/gotten, fallen/fell, gone/went. Only unambiguous instances of either b e or have are included as data for this study, which means that a number of inherently ambiguous tokens like the following are excluded: (6)
with respect to what I wrote you I do not know whats become of Taylors Daughter (late eighteenth century)
The B E -perfect in 18th- & 19th-c. Irish English 347
(7) I was fully determined to agone home this fall (late eighteenth
century)
(8) I expect he gone to Alabama (1830s) (9) The Minerva Capt Eccles should have been arrived when you wrote
(late eighteenth century)
Tokens excluded from consideration included cases where it is impossible to determine whether an elided form (’s in standard present-day orthography) represents is or has, as in (6). The form agone in (7) might be either have gone or a-going, and may thus be a present rather than a past participle. In (8), the ambiguity is that the auxiliary has been dropped altogether,14 and in (9), both auxiliaries appear, leaving it unclear what the intention is. The results are summarised in Table 17.2. In eighteenth-century IrE, the overall rate of be use with these verbs was 53 per cent (n = 156/292), falling to 39 per cent by the 1830s and 30 per cent in the 1880s. In the eighteenth century, only eight verbs occurred with a majority of b e -forms, though these produced just over half the total data: go, get ‘change’ and become used 57 per cent be -perfects, compared to be -scores upwards of 85 per cent for grow, improve, arrive, return and recover. The single token of vanish in this century also occurred with b e .15 Six of these are process verbs, while only three of those occurring most frequently with be are verbs of motion: arrive, return and go. Note that finish and do ‘finish’ occur exclusively with h av e in this subperiod (though in small numbers), and that the motion verbs most likely to occur with auxiliary be are arrive (86 per cent, n = 36) and return (86 per cent, n = 7). By the 1830s, go was the only motion verb still occurring mainly with b e , and this had just 55 per cent be -perfects (n = 100). The other three verbs with a majority of be -forms denote processes: improve, change, recover, with scores in the range of 64–81 per cent be . In the period to the 1830s, the use of b e declined with thirteen of the eighteen verbs for which we have data. Over the same period, use of b e increased with change, turn ‘change’ and do ‘finish’, and remained stable with go. Finish is not found with b e at all in the 1830s either. In the 1880s, only a handful of verbs occur with auxiliary b e at all, and there are some striking shifts, too, in relation to individual verbs. The motion verbs arrive and return, both categorically used with b e in the eighteenth century, are now found exclusively with hav e , and the same has happened
14
Alternatively, agone might be a representation of a-going in a context where auxiliary be has been deleted, a phenomenon that has recently proven to be variably present in the usage of many CORIECOR letter-writers during the eighteenth and nineteenth centuries, and quite frequent in some of them. 15 Verbs with fewer than five tokens are listed in the tables and contribute to the overall results, but percentages are not calculated for these verbs.
348 McCafferty Table 17.2 Use of b e - perfect in IrE, 1701–1890 (n = 1,110) Verb
1701–1800 be/Total
vanish grow improve arrive return recover get ‘change’ go become run change come get ‘motion’ pass leave turn ‘change’ fall enter happen finish die do ‘finish’ (a)rise wear break dry turn ‘motion’ depart
1/1 7/7 8/9 31/36 6/7 17/20 4/7 34/60 8/14 2/4 2/4 20/41 5/14 1/3 5/16 3/13 1/12 1/8 0/9 0/4 0/2 0/1 – – – – – –
TOTAL
156/292
(n)
1830s be
%
be/Total
1880s (n)
be
%
– 100 89 86 86 85 57 57 57 – (50) 48 36 (33) 31 23 18 12 0 (0) – (0) – – – – – –
0/1 6/13 17/21 7/24 7/19 23/36 15/45 55/100 2/24 0/1 9/13 12/51 0/8 0/6 0/8 3/7 1/15 0/1 1/4 0/4 1/6 1/4 1/1 1/1 1/3 0/1 3/6 –
– 46 81 29 37 64 33 55 8 – 69 24 0 0 0 43 7 – (25) (0) 17 (25) – – – – 50 –
53
166/423
39
be/Total
– 1/11 9/20 0/6 0/11 5/16 8/43 52/97 0/16 1/3 4/15 1/29 4/21 0/16 7/31 0/2 0/7 0/1 0/5 8/17 0/5 17/18 0/1 1/2 0/1 – – 0/1 118/395
(n)
be
%
– 9 45 0 0 31 19 54 0 – 27 3 19 0 23 (0) 0 – 0 47 0 94 – – – – – – 30
to grow, while become, which in the eighteenth century was used almost half the time with b e , is also found exclusively with h av e . There are also notable shifts in relation to the transitive verbs do ‘finish’ and finish itself, which actually suggest the reverse development over the nineteenth century. Do ‘finish’ is categorically found with be in the 1880s, while finish has 47 per cent be . It is also striking that, with the strong exception of go, and weaker exceptions in the cases of run, leave and get ‘reach, arrive’, it is process verbs that maintain the use of be most strongly towards the end of the nineteenth century, while motion verbs cluster towards the bottom of the scale. With some verbs it may sometimes be possible to read b e + past participle as a copula construction with the participle functioning as an adjective, and this possibility may well have facilitated retention of auxiliary b e . But usage varied between be and hav e in the 1830s and continued to do so in the 1880s (10)–(21), even sometimes in the same letter (10) and (11). (10) Dr. Boyd has gone to Kirkcubbin to live. (1830s)
The B E -perfect in 18th- & 19th-c. Irish English 349 100 come become pass arrive return grow turn 'change' fall happen
90 80 70 60 50 40 30 20 10 0
18th-c. IrE
IrE 1830s
IrE 1880s
Figure 17.4 Percentage use of b e -perfect in IrE, 1761–1890
(11) Hugh Boyd has been put out of his house, he is gone to Ann St. to
live. (1830s)
(12) but now this feeling is changed for sadness, and fear (1830s) (13) December here was as warm weather as it Ø with you in the summer-
time but it has now changed to cold (1830s)
(14) I presume to state that your respectable son John is greatly improved
since he came here (1830s)
(15) and everything even in Ireland has improved very much since then
(1830s)
(16) Alicia Sloan is gone back to Liverpool (1880s) (17) Margaret Wilson of Thornhill has gone out to America (1880s) (18) Florence has greatly changed since you were here (1880s) (19) Your old friend Mrs John Moore’s family, is of course very much
changed. (1880s)
(20) I sincerely trust that in the interval Aunt Rose has improved.
(1880s)
(21) any person that seen her the way she landed would scarcely know her
now she is so much improved (1880s)
The presence of object-like complements makes a copula reading unlikely, and continuing variation between b e and hav e justifies inclusion here. Figures 17.4, 17.5 and 17.6 summarise the findings for eighteen verbs in terms of how they evolve over the period studied. Figure 17.4 charts verbs that end up being used categorically with hav e by the late nineteenth century. Eight of these nine verbs, evenly divided between motion and process verbs, use substantial proportions of b e in the late eighteenth century (from a minimum of 18 per cent for fall to a maximum of 100 per cent for grow).
350 McCafferty 100 90 80
go
70
change leave
60
improve
50
recover
40
get 'change'
30
get 'motion'
20 10 0 18th-c. IrE
IrE 1830s
IrE 1880s
Figure 17.5 Percentage use of b e -perfect in IrE, 1761–1890 100 90 80 70 60
do 'finish' finish
50 40 30 20 10 0 18th-c. IrE
IrE 1830s
IrE 1880s
Figure 17.6 Percentage use of b e -perfect in IrE, 1761–1890
A century later, none of these exceeds 10 per cent b e use. Figure 17.5 charts those verbs that maintain substantial be use throughout the period. They range from 30–90 per cent b e in the eighteenth century and are still found with rates from 19 per cent to 54 per cent b e in the 1880s. Here, too, both motion and process verbs are represented. Finally, Figure 17.6 shows the change that results in the transitive verbs do ‘finish’ and finish shifting from non-use of b e in the eighteenth to substantial or categorical use of b e in the late nineteenth century.16 16
Only transitive tokens of do ‘finish’ and finish are included, which means that the tokens counted have active subjects and complements: neigbours are nearly all done ploughing now
The B E -perfect in 18th- & 19th-c. Irish English 351
17.4 Conclusions This survey of the use of the be -perfect in IrE shows that users of this variety participated in the general shift in English away from use of auxiliary b e with intransitive mutative verbs of motion and process. In IrE these verb types still occurred with a slight majority of auxiliary b e in the late eighteenth century. By the early nineteenth century, h av e was already the preferred perfective auxiliary in IrE, and this auxiliary strengthened its position further in the latter half of the century. In this development, IrE usage appears to have been largely following the general development mapped by Kytö (1997) for BrE and AmE, but at a slower pace. The IrE delay in adopting have with all the verbs included here is due in part to the maintenance of the be -perfect until at least the end of the nineteenth century with a broader set of verbs than in other varieties of the language. As noted above, go is widely observed as preserving auxiliary b e in other varieties, while my results show a number of other verbs also retaining b e in the perfect (Figure 17.5). Also, the fact that transitive uses of verbs like finish and do ‘finish’ with auxiliary b e seem to emerge in IrE over the period studied here contributes to the retention of the b e -perfect. It is tempting to speculate that this latter development, though not unknown in earlier stages of English and other varieties of the language, distinguishes IrE from other varieties by the late nineteenth century, but firm conclusions on this issue will have to await empirical historical comparisons with other varieties, especially regional ones. Similarly, the possibility that the uses of b e with these transitive verbs might either have arisen or been strengthened as a result of Irish substrate influence is something that must remain speculation until such time as we are in a position to examine the regional distribution of b e -perfect use in Ireland during the period of language shift. This study represents but a first step in these directions. (1880s) and they’re done with the potatoes are included while adjectival you know as well as I that Ireland is don (1880s) and passive but then their work is done much earlier (1880s) are not.
References
Aarts, Bas, Lopez-Couso, Maria José, and Méndez-Naya, Belén. 2012. ‘Late Modern English Syntax’. In: Bergs, Alexander and Brinton, Laurel J. (eds.). English Historical Linguistics. Vol. i, 869–87. Berlin: de Gruyter. Achard, Michel. 1998. Representation of Cognitive Structures: Syntax and Semantics of French Sentential Complements. Berlin: de Gruyter. Ackema, Peter and Neeleman, Ad. 2004. Beyond Morphology. Interface Conditions on Word Formation. Oxford University Press. Ackema, Peter and Schoorlemmer, Maaike. 1994. ‘The Middle Construction and the Syntax–Semantics Interface’. Lingua 93(1), 59–90. Adamson, Sylvia. 2000. ‘A Lovely Little Example: Word-order Options and Category Shift in the Premodifying String’. In: Fischer, Olga, Rosenbach, Anette, and Stein, Dieter (eds.). Pathways of Change: Grammaticalization in English, 39–66. Amsterdam: Benjamins. Aijmer, Karin. 2009. ‘Seem and Evidentiality’. Functions of Language 16(1), 63–88. Aikhenvald, Alexandra Y. 2004. Evidentiality. Oxford University Press. Alexiadou, Artemis and Doron, Edit. 2011. ‘The Syntactic Construction of Two Non-active Voices: Passive and Middle’. Journal of Linguistics 48(1), 1–34. Algeo, John. 1995. ‘Having a Look at the Expanded Predicate’. In: Aarts, Bas and Meyer, Charles F. (eds.). The Verb in Contemporary English. Theory and Description, 203–17. Cambridge University Press. 2006. British or American English? A Handbook of Word and Grammar Patterns. Cambridge University Press. Algeo, John and Pyles, Thomas. 2005. The Origins and Developments of the English Language. Boston: Wadsworth. Allen, Alexander and Cornwell, James. 1841. A New English Grammar, with Very Copious Exercises, and a Systematic View of the Formation and Derivation of Words. London: Simpkin, Marshall, & Co. Allen, William. 1824. An English Grammar; with Exercises, Notes, and Questions. London: G. and W.B. Whittaker [First published 1813]. Allerton, David J. 2009. ‘Tag Questions’. In: Rohdenburg, Günter and Schlüter, Julia (eds.). One Language, Two Grammars? Differences between British and American English, 306–23. Cambridge University Press. Altick, Richard D. 1957. The English Common Reader: A Social History of the Mass Reading Public 1800–1900. University of Chicago Press. 352
References 353 Amador-Moreno, Carolina P. 2006. An Analysis of Hiberno-English in the Early Novels of Patrick MacGill. Bilingualism and Language Shift from Irish to English in County Donegal. Lewiston, NY: Edwin Mellen Press. 2010. An Introduction to Irish English. London: Equinox. Andersen, Henning. 2001. ‘Actualization and the (Uni)directionality of Change’. In: Andersen, Henning (ed.). Actualization: Linguistics Change in Progress, 225–48. Amsterdam: Benjamins. Anderwald, Lieselotte. 2012. ‘Clumsy, Awkward or Having a Peculiar Propriety? Prescriptive Judgements and Language Change in the 19th Century’. Language Sciences 34(1), 28–53. Submitted. Language Between Description and Prescription: Verb Categories in Nineteenth-Century Grammars of English. Oxford University Press. Appadurai, Arjun. 1996. Modernity at Large: Cultural Dimensions of Globalization. Minneapolis, MN: University of Minnesota Press. Arnaud, René. 1983. ‘On the Progress of the Progressive in the Private Correspondence of Famous British People (1800–1880)’. In: Jacobson, Sven (ed.). Papers from the Second Scandinavian Symposium on Syntactic Variation, Stockholm, May 15–16, 1982, 83–94. Stockholm: Almqvist and Wiksell. 1998. ‘The Development of the Progressive in 19th Century English: A Quantitative Survey’. Language Variation and Change 10(2), 123–52. Arnold, Thomas Kerchever. 1838. An English Grammar, for Classical Schools. Etymology. Parts I. and II. London: Rivington and Parker. Asudeh, Ash and Toivonen, Ida. 2005. ‘Copy Raising and its Consequences for Perceptual Reports’. In: Grimshaw, Jane, Maling, Joan, Manning, Chris, Simpson, Jane, and Zaenen, Annie (eds.). Architectures, Rules, and Preferences. A Festschrift for Joan Bresnan, 1–19. Stanford, CA: CSLI. 2006. ‘Expletives and the Syntax and Semantics of Copy Raising’. In: Butt, Miriam and King, Tracy Holloway (eds.). Proceedings of the LFG06 Conference. Stanford, CA: CSLI. http://csli-publications.stanford.edu/. Auer, Anita. 2006. ‘Precept and Practice: The Influence of Prescriptivism on the English Subjunctive’. In: Dalton-Puffer, Christiane, Kastovsky, Dieter, Ritt, Nikolaus, and Schendl, Herbert (eds.). Syntax, Style and Grammatical Norms: English from 1500–2000. Linguistic Insights, 33–53. Frankfurt: Lang. 2012. ‘Late Modern English Standardization’. In: Brinton, Laurel and Bergs, Alexander (eds.). Historical Linguistics of English: An International Handbook, 939–51. New York: de Gruyter. Baayen, R. Harald. 2008. Analyzing Linguistic Data: A Practical Introduction to Statistics Using R. Cambridge University Press. Bache, Carl. 1978. The Order of Premodifying Adjectives in Present-Day English. Odense University Press. Bailey, Richard W. 1996. Nineteenth-Century English. Ann Arbor, MI: The University of Michigan Press. Barber, Charles, Beal, Joan C., and Shaw, Philip A. 2009. The English Language: A Historical Introduction. Cambridge University Press. Bardon, Jonathan. 2011. The Plantation of Ulster: The British Colonisation of the North of Ireland in the Seventeenth Century. Dublin: Gill and Macmillan.
354 References Beal, Joan. 1999. English Pronunciation in the Eighteenth Century: Thomas Spence’s ‘Grand Repository of the English Language’. Oxford: Clarendon Press. 2004. English in Modern Times. 1700–1945. London: Arnold. Behaghel, Otto. 1909. ‘Beziehungen zwischen Umfang und Reihenfolge von Satzgliedern’. Indogermanische Forschungen 25, 110–42. Bell, Alan. 1984. ‘Language Style as Audience Design’. Language in Society 13(2), 145–204. Berg, Thomas. 2012. ‘The Cohesiveness of English and German Compounds’. The Mental Lexicon 7(1), 1–33. Berglund, Ylva. 2005. Expressions of Future in Present-Day English: A Corpus-based Approach. Uppsala University. Bergs, Alexander. 2012. ‘New Perspectives, Theories and Methods: Construction Grammar’. In: Bergs, Alexander and Brinton, Laurel J. (eds.). English Historical Linguistics: An International Handbook, 1631–46. Berlin: de Gruyter. Berlage, Eva. 2009. ‘Prepositions and Postpositions’. In: Rohdenburg, Günter and Schlüter, Julia (eds.). One Language, Two Grammars? Differences between British and American English, 130–48. Cambridge University Press. 2012. ‘At the Interface of Grammaticalisation and Lexicalisation: The Case of Take Prisoner’. English Language and Linguistics 16(1), 35–55. Biber, Douglas. 1988. Variation Across Speech and Writing. Cambridge University Press. 2003a. ‘Variation among University Spoken and Written Registers: A New Multi-dimensional Analysis’. In: Leistyna, Pepi and Meyer, Charles F. (eds.). Corpus Analysis: Language Structure and Language Use, 47–70 (Language and Computers: Studies in Practical Linguistics 46). Amsterdam: Rodopi. 2003b. ‘Compressed Noun Phrase Structures in Newspaper Discourse: The Competing Demands of Popularization vs. Economy’. In: Aitchison, Jean and Lewis, Diana M. (eds.). New Media Language, 169–81. London: Routledge. Biber, Douglas and Finegan, Edward. 1997. ‘Diachronic Relations among Speechbased and Written Registers in English’. In: Nevalainen, Terttu and KahlasTarkka, Leena (eds.). To Explain the Present: Studies in the Changing English Language in Honour of Matti Rissanen, 253–75 (Mémoires de la Société Néophilologique de Helsinki 52). Helsinki: Société Néophilologique. Biber, Douglas and Gray, Bethany. 2011. ‘Grammatical Change in the Noun Phrase: The Influence of Written Language Use’. English Language and Linguistics 15(2), 223–50. Biber, Douglas, Finegan, Edward, and Atkinson, David. 1994. ‘ARCHER and Its Challenges: Compiling and Exploring a Representative Corpus of Historical English Registers’. In: Fries, Udo, Tottie, Gunnel, and Schneider, Peter (eds.). Creating and Using English Language Corpora, 1–14. Amsterdam: Rodopi. Biber, Douglas, Grieve, Jack, and Iberri-Shea, Gina. 2009. ‘Noun Phrase Modification’. In: Rohdenburg, Günter and Schlüter, Julia (eds.). One Language, Two Grammars? Differences between British and American English, 182–93. Cambridge University Press. Biber, Douglas, Johansson, Stig, Leech, Geoffrey, Conrad, Susan, and Finegan, Edward. 1999. Longman Grammar of Spoken and Written English. London and New York: Longman.
References 355 Bliss, Alan J. 1979. Spoken English in Ireland 1600–1740. Dublin: The Dolmen Press. 1984. ‘English in the South of Ireland’. In: Trudgill, Peter (ed.). Language in the British Isles, 135–51. Cambridge University Press. Bod, Rens, Hay, Jennifer, and Jannedy, Stefanie (eds.). 2003. Probabilistic Linguistics. Cambridge, MA: MIT Press. Bolinger, Dwight. 1967. ‘Adjectives in English: Attribution and Predication’. Lingua 18, 1–34. 1972. Degree Words. The Hague and Paris: Mouton. 1977. ‘Transitivity and Spatiality: The Passive of Prepositional Verbs’. In: Makkai, Adam Becker Makkai, Valerie, and Heilmann, Luigi (eds.). Linguistics at the Crossroads, 57–78. Lake Bluff, IL: Jupiter Press. Bös, Birte. 2012. ‘From 1760 to 1960: Diversification and Popularization’. In: Facchinetti, Roberta, Brownlees, Nicholas, Bös, Birte, and Fries, Udo (eds.). News as Changing Texts: Corpora, Methodologies and Analysis, 91–143. Newcastle upon Tyne: Cambridge Scholars Publishing. Boye, Kasper and Harder, Peter. 2007. ‘Complement-taking Predicates: Usage and Linguistic Structure’. Studies in Language 31(3), 569–606. Braeken, Niels. 2011. ‘The Use of the a-Prefix with -ing-Clauses: A Corpus Analysis of its Distribution in the Suffolk English Dialect.’ Masters thesis, University of Leuven. Brainerd, Barron. 1989 [1993]. ‘The Contractions of Not: A Historical Note’. Journal of English Linguistics 22(2), 176–96. Breban, Tine. 2008. ‘Grammaticalization, Subjectification, and Leftward Movement of Adjectives of Difference in the Noun Phrase’. Folia Linguistica 42(3/4), 259–306. 2009. ‘Structural Persistence: A Case Based on the Grammaticalization of English Adjectives of Difference’. English Language and Linguistics 13(1), 77–96. 2010. English Adjectives of Comparison: Lexical and Grammaticalized Uses. Berlin: de Gruyter. Breban, Tine and Davidse, Kristin. 2003. ‘Adjectives of Comparison: The Grammaticalization of Their Attribute Uses into Postdeterminer and Classifier Uses’. Folia Linguistica 37(3/4), 269–317. Breivik, Leiv Egil and Martínez-Insua, Ana E. 2008. ‘Grammaticalization, Subjectification and Non-concord in English Existential Sentences’. English Studies 89(3), 351–62. Brems, Lieselotte. 2003. ‘Measure Noun Constructions: An Instance of Semantically-driven Grammaticalization’. International Journal of Corpus Linguistics 8(2), 283–312. 2007. ‘The Grammaticalisation of Small Size Nouns: Reconsidering Frequency and Analogy’. Journal of English Linguistics 35(4), 293–324. 2011. Layering of Size and Type Noun Constructions in English (Topics in English Linguistics 74.) Berlin: de Gruyter. 2012. ‘The Establishment of Quantifier Constructions for Size Nouns: A Diachronic Case Study of heap(s) and lot(s)’. Journal of Historical Pragmatics 13(2), 202–31. Bresnan, Joan. 1970. ‘On Complementizers: Toward a Syntactic Theory of Complement Types’. Foundations of Language 6(3), 297–321. 1979. Theory of Complementation in English Syntax. New York: Garland.
356 References Bresnan, Joan and Ford, Marilyn. 2010. ‘Predicting Syntax: Processing Dative Constructions in American and Australian Varieties of English’. Language 86(1), 186–213. Bresnan, Joan and Hay, Jennifer. 2008. ‘Gradient Grammar: An Effect of Animacy on the Syntax of Give in New Zealand and American English’. Lingua 118(2), 245–59. Brinton, Laurel J. 1996a. Pragmatic Markers in English: Grammaticalization and Discourse Functions. Berlin: de Gruyter. 1996b. ‘Attitudes Towards Increasing Segmentalization: Complex and Phrasal Verbs in English’. Journal of English Linguistics 24(3), 186–205. 2008a. The Comment Clause in English: Syntactic Origins and Pragmatic Development. (Studies in English Language). Cambridge University Press. 2008b. ‘Where Grammar and Lexis Meet: Composite Predicates in English’. In: Seoane, Elena and López-Couso, María José (eds.). Theoretical and Empirical Issues in Grammaticalization, 33–53. Amsterdam: Benjamins. Brinton, Laurel J. and Arnovick, Leslie K. 2011. The English Language: A Linguistic History. Don Mills, Ont.: Oxford University Press. Brinton, Laurel J. and Traugott, Elizabeth Closs. 2005. Lexicalization and Language Change. Cambridge University Press. Brown, Lucy. 1985. Victorian News and Newspapers. Oxford: Clarendon Press. Burton-Roberts, Noel. 2006. ‘Parentheticals’. In: Brown, Keith (ed.). Encyclopaedia of Language and Linguistics. 2nd edn. Vol. ix, 179–82. Oxford: Elsevier Science. Bybee, Joan L. 2004. ‘Mechanisms of Change in Grammaticization: The Role of Frequency’. In: Joseph, Brian D. and Janda, Richard D. (eds.). The Handbook of Historical Linguistics, 602–23. Oxford: Blackwell. 2007. Frequency of Use and the Organization of Language. Oxford University Press. 2010. Language, Usage and Cognition. Cambridge University Press. Bybee, Joan L. and Hopper, Paul J. (eds.). 2001. Frequency and the Emergence of Linguistic Structure. Amsterdam: Benjamins. Bybee, Joan L., Perkins, Revere D., and Pagliuca, William. 1994. The Evolution of Grammar: Tense, Aspect, and Modality in the Languages of the World. University of Chicago Press. Cambridge International Dictionary of English. 1995. Procter, Paul (ed.). Cambridge University Press. Canadian Oxford Dictionary. 2004. 2nd edn. Barber, Katherine (ed.). Toronto: Oxford University Press. Chafe, Wallace. 1986. ‘Evidentiality in English Conversation and Academic Writing’. In: Chafe, Wallace and Nichols, Johanna (eds.). Evidentiality: The Linguistic Coding of Epistemology, 261–72. Norwood, NY: Ablex. Channel, Joanna. 1994. Vague Language. Oxford University Press. Chomsky, Noam. 1986. Barriers. Cambridge, MA: MIT Press. Claridge, Claudia. 2000. Multi-word Verbs in Early Modern English: A Corpusbased Study. (Language and Computers: Studies in Practical Linguistics 32). Amsterdam: Rodopi. 2013. ‘The Evolution of Three Pragmatic Markers: As it were, so to speak/say and if you like’. Journal of Historical Pragmatics 14(2), 161–84.
References 357 Claridge, Claudia and Kytö, Merja. 2014. ‘I had lost sight of them then for a bit, but I went on pretty fast: Two Degree Modifiers in the Old Bailey Corpus’. In: Taavitsainen, Irma, Jucker, Andreas, and Tuominen, Jukka (eds.). Diachronic Corpus Pragmatics. Amsterdam: Benjamins. Coates, Jennifer. 1983. The Semantics of the Modal Auxiliaries. London and Canberra: Croom Helm. 2004. Women, Men and Language: A Sociolinguistic Account of Gender Differences in Language. 3rd edn. Harlow: Pearson. Cobbin, Ingram. 1864. Elements of English Grammar: Expressly Designed for the Juvenile Student, either at Home or in Preparatory Schools. Illustrated. London: William Tegg [First published 1828]. Collins, Peter. 2006. ‘Clause Types’. In: Aarts, Bas and McMahon, April (eds.). The Handbook of English Linguistics, 180–97. Oxford and Malden, MA: Blackwell. 2009. Modals and Quasi-modals in English. Amsterdam: Rodopi. Collins, Peter and Yao, Xinyue. 2012. ‘Modals and Quasi-modals in New Englishes’. In: Hundt, Marianne and Gut, Ulrike (eds.). Mapping Unity and Diversity World-wide: Corpus-based Studies of New Englishes, 35–53. Amsterdam: Benjamins. Coupland, Nikolas. 2009. ‘Response: The Mediated Performance of Vernaculars’. Journal of English Linguistics 37(3), 284–300. Croft, William. 2000. Explaining Language Change. London: Longman. 2001. Radical Construction Grammar: Syntactic Theory in Typological Perspective. Oxford University Press. 2004. ‘Logical and Typological Arguments for Radical Construction Grammar’. In: Fried, Mirjam and Östman, Jan-Ola (eds.). Construction Grammar(s): Cognitive and Cross-language Dimensions, 273–314. Amsterdam: Benjamins. Croft, William and Cruse, D. Alan. 2004. Cognitive Linguistics (Cambridge Textbooks in Linguistics). Cambridge University Press. Crystal, David. 1969. Prosodic Systems and Intonation in English. Cambridge University Press. Culpeper, Jonathan and Kytö, Merja. 2010. Early Modern English Dialogues: Spoken Interaction as Writing. Cambridge University Press. Curme, George O. 1913. ‘Has English a Future Tense?’ The Journal of English and Germanic Philology 12(4), 515–39. Dancygier, Barbara and Sweetser, Eve. 2005. Mental Spaces in Grammar: Conditional Constructions. (Cambridge Studies in Linguistics 108). Cambridge University Press. DARE = Cassidy, Frederic G., and Hall, Joan Houston (comps.). 1985–2002. Dictionary of American Regional English. 4 vols. Cambridge, MA: Belknap Press. Davidse, Kristin. 1992. ‘Transitivity/Ergativity: The Janus-headed Grammar of Actions and Events’. In: Davies, Martin and Ravelli, Louise (eds.). Advances in Systemic Linguistics, 105–35. London: Pinter. Davidse, Kristin and Heyvaert, Liesbet. 2003. ‘On the Middle Construction in English and Dutch’. In: Granger, Sylviane, Lerot, Jacques, and PetchTyson, Stephanie (eds.). Corpus-based Approaches to Contrastive Linguistics and Translation Studies, 57–73. Amsterdam: Rodopi.
358 References Davies, Eirlys. 1986. The English Imperative. London: Croom Helm. Davies, Mark. 2012. ‘The 400 Million Word Corpus of Historical American English (1810–2009)’. In: Hegedűs, Irén and Fodor, Alexandra (eds.). English Historical Linguistics 2010, 219–49. Amsterdam: Benjamins. De Clerck, Bernard. 2004. ‘On the Pragmatic Functions of let’s Utterances’. In: Aijmer, Karin and Altenberg, Bengt (eds.). Advances in Corpus Linguistics. Papers from the 23rd International Conference on English Language Research on Computerized Corpora (ICAME 23), 213–37. Amsterdam: Rodopi. Declerck, Renaat. 1981. ‘On the Role of the Progressive Aspect in Nonfinite Perception Verb Complements’. Glossa 15, 83–113. 1982. ‘The Triple Origin of Participial Perception Verb Complements’. Linguistic Analysis 10, 1–26. 1991. A Comprehensive Descriptive Grammar of English. Tokyo: Kaitakusha. Dehé, Nicole. 2009. ‘Clausal Parentheticals, Intonational Phrasing, and Prosodic Theory’. Journal of Linguistics 45(3), 569–615. Dehé, Nicole and Kavalova, Yordanka (eds.). 2007. Parentheticals. Amsterdam: Benjamins. Dehé, Nicole and Wichmann, Anne. 2010a. ‘Sentence-initial I think (that) and I believe (that): Prosodic Evidence for Use as Main Clause, Comment Clause and Discourse Marker’. Studies in Language 34(1), 36–74. 2010b. ‘The Multifunctionality of Epistemic Parentheticals in Discourse: Prosodic Cues to the Semantic–Pragmatic Boundary’. Functions of Language 17(1), 1–28. Denison, David. 1993. English Historical Syntax: Verbal Constructions. London: Longman. 1998. ‘Syntax’. In: Romaine, Suzanne (ed.). The Cambridge History of the English Language Vol. iv: 1776–1997, 92–329. Cambridge University Press. 2006. ‘Category Change and Gradience in the Determiner System’. In: van Kemenade, Ans and Los, Bettelou (eds.). The Handbook of the History of English, 279–304. Oxford: Blackwell. 2010. ‘Category Change in English with and without Structural Change’. In: Traugott, Elizabeth Closs and Trousdale, Graeme (eds.). Gradience, Gradualness, and Grammaticalization, 105–28. Amsterdam: Benjamins. Denison, David and Hundt, Marianne. 2013. ‘Defining Relatives’. Journal of English Linguistics 41(2), 135–67. De Smet, Hendrik. 2008. ‘Diffusional Change in the English System of Complementation.’ PhD dissertation, Department of Linguistics, University of Leuven. 2010. ‘English –ing-Clauses and Their Problems: The Structure of Grammatical Categories’. Linguistics 48(6), 1153–93. 2012a. ‘The Course of Actualization’. Language 88(3), 601–33. 2012b. ‘Review of Yoko Iyeiri. Verbs of Implicit Negation and Their Complements in the History of English’. ICAME Journal 35, 138–43. 2013a. Spreading Patterns: Diffusional Change in the English System of Complementation. New York and Oxford University Press. 2013b. ‘Change through Recombination. Blending and Analogy’. Language Sciences 40, 80–94.
References 359 De Smet, Hendrik and Heyvaert, Liesbet. 2011. ‘The Meaning of the English Present Participle’. English Language and Linguistics 15(3), 473–98. Diessel, Holger and Tomasello, Michael. 2001. ‘The Acquisition of Finite Complement Clauses in English: A Corpus-based Analysis’. Cognitive Linguistics 12(2), 97–141. Diewald, Gabriele. 2002. ‘A Model for Relevant Types of Contexts in Grammaticalization’. In: Wischer, Ilse and Diewald, Gabriele (eds.). New Reflections on Grammaticalization, 103–20. Amsterdam: Benjamins. 2011. ‘Pragmaticalization (Defined) as Grammaticalization of Discourse Functions’. Linguistics 49(2), 365–90. Dixon, Robert M.W. 1991. A New Approach to English Grammar on Semantic Principles. Oxford University Press. 2005. A Semantic Approach to English Grammar. Oxford University Press. Duffley, Patrick. 1992. The English Infinitive. London: Longman. 1999. ‘The Use of the Infinitive and the –ing after Verbs Denoting the Beginning, Middle and End of an Event’. Folia Linguistica 33(3/4), 295–331. Elmer, Willy. 1981. Diachronic Grammar: The History of Old and Middle English Subjectless Constructions. Tübingen: Niemeyer. Erades, P.A. 1950. ‘Points of Modern English Syntax’ xii. English Studies 31(1–6), 153–57. Espinal, M. Teresa. 1991. ‘The Representation of Disjunct Constituents’. Language 67(4), 726–62. Evans, Nicholas. 2007. ‘Insubordination and Its Uses’. In: Nikolaeva, Irena (ed.). Finiteness: Theoretical and Empirical Foundations, 366–431. Oxford University Press. Fagan, Sarah M.B. 1988. ‘The English Middle’. Linguistic Inquiry 19(2), 181–203. 1992. The Syntax and Semantics of Middle Constructions: A Study with Special Reference to German. Cambridge University Press. Fairclough, Norman. 1992. Discourse and Social Change. Cambridge: Polity Press. Faltz, Leonard M. 1985. Reflexivization: A Study in Universal Syntax. New York and London: Garland. Fanego, Teresa. 1996. ‘The Development of Gerunds as Objects of Subject-control Verbs in English (1400–1700). Diachronica 13(1), 29–62. 1998. ‘Developments in Argument Linking in Early Modern English Gerund Phrases’. English Language and Linguistics 2(1), 87–119. 2004. ‘On Reanalysis and Actualization in Syntactic Change: The Rise and Development of English Verbal Gerunds’. Diachronica 21(1), 5–55. Farrelly, Michael and Seoane, Elena. 2012. ‘Democratization’. In: Nevalainen, Terttu and Traugott, Elizabeth Closs (eds.). The Oxford Handbook of the History of English, 392–401. New York: Oxford University Press. Fawcett, Robin P. 1980. Cognitive Linguistics and Social Interaction: Towards an Integrated Model of a Systemic Functional Grammar and the Other Components of a Communicating Mind. Heidelberg: Julius Groos Verlag. Felser, Claudia. 1999. Verbal Complement Clauses. Amsterdam: Benjamins. Fens-de Zeeuw, Lyda. 2011. Lindley Murray (1745–1826), Quaker and Grammarian. Utrecht: LOT Publications. Filppula, Markku. 1999. The Grammar of Irish English: Language in Hibernian Style. London: Routledge.
360 References 2003. ‘The Quest for the Most “Parsimonious” Explanations: Endogeny vs. Contact Revisited’. In: Hickey, Raymond (ed.). Motives for Language Change, 161–73. Cambridge University Press. 2008. ‘Irish English: Morphology and Syntax’. In: Kortmann, Bernd and Upton, Clive (eds.). Varieties of English. Vol. i: The British Isles, 328–59. Berlin: de Gruyter. Fischer, Olga. 1995. ‘The Distinction between to and Bare Infinitival Complements in Late Middle English’. Diachronica 12(1), 1–30. 2007. Morphosyntactic Change: Functional and Formal Perspectives. Oxford University Press. Fisk, Allen. 1822. Murray’s English Grammar Simplified; Designed to Facilitate the Study of the English Language; Comprehending the Principles and Rules of English Grammar, Illustrated by Appropriate Exercises; to which is Added a Series of Questions for Examination. Abridged for the Use of Schools. Troy, NY: Z. Clark. [First published 1821]. Fitzgerald, Patrick and Lambkin, Brian. 2008. Migration in Irish History, 1607– 2007. Basingstoke: Palgrave Macmillan. Fries, Charles Carpenter. 1925. ‘The Periphrastic Future with shall and will in Modern English’. PMLA 40(4), 963–1024. Garcia de la Maza, Casilda. 2012. Gradient Acceptability at the Grammar–Pragmatics Interface: A Study of the Constraints on Middle Formation in English. Munich: Lincom. Geeraerts, Dirk. 1997. Diachronic Prototype Semantics: A Contribution to Historical Lexicology. Oxford: Clarendon Press. Geisler, Christer. 2002. ‘Investigating Register Variation in NineteenthCentury English: A Multi-dimensional Comparison’. In: Reppen, Randi, Fitzmaurice, Susan M., and Biber, Douglas (eds.). Using Corpora to Explore Linguistic Variation, 249–71 (Studies in Corpus Linguistics 9). Amsterdam: Benjamins. Ghesquière, Lobke. 2010. ‘On the Subjectification and Intersubjectification Paths Followed by the Adjectives of Completeness’. In: Davidse, Kristin,Vandelanotte, Lieven, and Cuyckens, Hubert (eds.). Subjectification, Intersubjectification and Grammaticalisation, 277–314. Berlin: de Gruyter. Gisborne, Nikolas. 2010. The Event Structure of Perception Verbs. Oxford University Press. Gisborne, Nikolas and Holmes, Jasper. 2007. ‘A History of English Evidential Verbs of Appearance’. English Language and Linguistics 11(1), 1–29. Gisborne, Nikolas and Patten, Amanda. 2011. ‘Grammaticalization and Construction Grammar’. In: Narrog, Heiko and Heine, Bernd (eds.). The Oxford Handbook of Grammaticalization. Oxford University Press. Givón, Talmy. 1979. On Understanding Grammar. New York: Academic Press. 1980. ‘The Binding Hierarchy and the Typology of Complements’. Studies in Language 4(3), 333–77. Goldberg, Adele. 1995. Constructions: A Construction Grammar Approach to Argument Structure. University of Chicago Press. Goldberg, Adele and Ackerman, Farrell. 2001. ‘The Pragmatics of Obligatory Adjuncts’. Language 77(4), 798–814.
References 361 Görlach, Manfred. 1998. An Annotated Bibliography of 19th-Century Grammars of English. Amsterdam: Benjamins. 1999. English in Nineteenth-Century England: An Introduction. Cambridge University Press. 2001. Eighteenth-Century English. Heidelberg: Universitätsverlag C. Winter. 2002. A Textual History of Scots. Heidelberg: Universitätsverlag C. Winter. Gotti, Maurizio. 2003. ‘Shall and Will in Contemporary English: A Comparison with Past Uses’. In: Facchinetti, Roberta, Krug, Manfred, and Palmer, Frank (eds.). Modality in Contemporary English, 267–300. Berlin: de Gruyter. Grafmiller, Jason. Forthcoming. ‘Variation in English Genitives Across Modality and Genre’. English Language and Linguistics. Gries, Stefan T. 2002. ‘Evidence in Linguistics: Three Approaches to Genitives in English’. In: Brend, Ruth M., Sullivan, William J., and Lommel, Arle R. LACUS Forum XXVIII: What Constitutes Evidence in Linguistics, 17–31. Fullerton, CA: LACUS. 2012. ‘Statistische Modellierung’. Zeitschrift für germanistische Linguistik 40(1), 38–67. Gries, Stefan T. and Hilpert, Martin. 2010. ‘Modeling Diachronic Change in the Third Person Singular: A Multi-factorial, Verb- and Author-specific Exploratory Approach’. English Language and Linguistics 14(3), 293–320. Haegeman, Liliane. 1991. ‘Parenthetical Adverbials: The Radical Orphanage Approach’. In: Chiba, Shuki, Ogawa, Akira, Fuiwara, Yasuaki, Yamade, Norio, Koma, Osamu, and Yagi, Takao (eds.). Aspects of Modern English Linguistics, 232–54. Tokyo: Kaitakushi. Hansen, Maj-Britt Mosegaard. 1998. The Function of Discourse Particles: A Study with Special Reference to Spoken Standard French (Pragmatics and Beyond, New Series 53). Amsterdam: Benjamins. Harris, John. 1984a. ‘English in the North of Ireland’. In: Trudgill, Peter (ed.). Language in the British Isles, 115–134. Cambridge University Press. 1984b. ‘Syntactic Variation and Dialect Divergence’. Journal of Linguistics 20(2), 303–27. 1993. ‘The Grammar of Irish English’. In: Milroy, James and Milroy, Lesley (eds.). Real English. The Grammar of English Dialects in the British Isles, 139–86. London: Longman. Harvey, Thomas W. 1868. A Practical Grammar of the English Language, for the Use of Schools of Every Grade. Cincinnati and New York: van Antwerp, Bragg & Co. Haspelmath, Martin. 1989. ‘From Purposive to Infinitive’. Folia Linguistica Historica 10(1/2), 281–310. Heine, Bernd. 1997. Possession: Cognitive Sources, Forces, and Grammaticalization. Cambridge University Press. 2002. ‘On the Role of Context in Grammaticalization’. In: Wischer, Ilse and Diewald, Gabriele (eds.). New Reflections on Grammaticalization, 83–101. Amsterdam: Benjamins. 2012a. ‘Between Sentence Grammar and Discourse Grammar: On Insubordination’. Unpublished paper. 2012b. ‘On Discourse Markers: Grammaticalisation, Pragmaticalization, or Something Else?’ Unpublished paper.
362 References Heyvaert, Liesbet. 2008. ‘On the Constructional Semantics of Gerundive Nominalizations’. Folia Linguistica 42(1), 39–82. Hickey, Raymond. 2005. Dublin English: Evolution and Change. Amsterdam: Benjamins. 2007a. Irish English: History and Present-Day Forms. Cambridge University Press. 2007b. ‘Southern Irish English’. In: Britain, David (ed.). Language in the British Isles, 135–51. Cambridge University Press. 2012. Discover Irish. Why Study Irish? www.uni-due.de/DI/Syntax.htm (accessed 19 February 2014). Hilpert, Martin. 2013. Constructional Change in English. Developments in Allomorphy, Word Formation and Syntax. Cambridge University Press. Himmelmann, Nikolaus P. 2004. ‘Lexicalization and Grammaticalization: Opposite or Orthogonal?’ In: Bisang, Walter, Himmelmann, Nikolaus P., and Wiemer, Björn (eds.). What Makes Grammaticalization? A Look from its Fringes and its Components, 21–42 (Trends in Linguistics, Studies and Monographs 158). Berlin: de Gruyter. Hinrichs, Lars and Szmrecsanyi, Benedikt. 2007. ‘Recent Changes in the Function and Frequency of Standard English Genitive Constructions: A Multivariate Analysis of Tagged Corpora’. English Language and Linguistics 11(3), 437–74. Hirschberg, Julia. 2004. ‘Pragmatics and Intonation’. In: Horn, Laurence R. and Ward, Gregory (eds.). The Handbook of Pragmatics, 515–37. Oxford: Blackwell. Hoffmann, Sebastian. 2004. ‘Are Low-frequency Complex Prepositions Grammaticalized?’ In: Lindquist, Hans and Mair, Christian (eds.). Corpus Approaches to Grammaticalization in English, 171–210. Amsterdam: Benjamins. Hoffmann, Sebastian, Hundt, Marianne, and Mukherjee, Joybrato. 2011. ‘Indian English – an Emerging Epicentre? A Pilot Study on Light Verbs in Webderived Corpora of South Asian Englishes’. Anglia 129 (3–4), 258–80. Hoffmann, Thomas and Trousdale, Graeme. 2011. ‘Variation, Change, and Constructions in English’. Cognitive Linguistics 22(1),1–23. doi 10.1515/ cogl.2011.01. Hogg, Richard M. 2002. An Introduction to Old English. Edinburgh University Press. Holtz, Louis. 1981. Donat et la tradition de l’enseignement grammatical: étude sur l’Ars Donati et sa diffusion (IVe-IXe siècle) et édition critique. Paris: Centre National de la Recherche Scientifique. Hopper, Paul J. 1991. ‘On Some Principles of Grammaticization’. In: Traugott, Elizabeth Closs and Heine, Bernd (eds.). Approaches to Grammaticalization. Vol. i: Focus on Theoretical and Methodological Issues, 17–35. Amsterdam: Benjamins. Hopper, Paul J. and Thompson, Sandra A. 1980. ‘Transitivity in Grammar and Discourse’. Language 56(2), 251–99. Hopper, Paul J. and Traugott, Elizabeth Closs. 2003. Grammaticalization. 2nd edn. Cambridge University Press. Houston, Ann. 1985. ‘Continuity and Change in English Morphology. The Variable (ING).’ PhD dissertation, University of Pennsylvania.
References 363 Howe, Chad and Heller, Brooke. 2010. ‘Turns Out They Weren’t Much of a Stretch: Variable Patterns of Structural Persistence’. Paper presented at 39th New Ways of Analyzing Variation. University of Texas at San Antonio, November 2010. Huber, Magnus. 2007. ‘The Old Bailey Proceedings, 1674–1834: Evaluating and Annotating a Corpus of 18th- and 19th-century Spoken English’. In: Meurman-Solin, Anneli and Nurmi, Arja (eds.). Annotating Variation and Change (Studies in Variation, Contacts and Change in English 1). n.p. Available at www.helsinki.fi/varieng/series/volumes/01/huber/. Huddleston, Rodney and Pullum, Geoffrey K., et al. 2002. The Cambridge Grammar of the English Language. Cambridge University Press. Hudson, Rodney. 1976. ‘Some Theoretical Issues in the Description of the English Verb’. Lingua 40(4), 331–83. Hundt, Marianne. 2001. ‘What Corpora Can Tell Us about the Grammaticalisation of Voice in Get-constructions’. Studies in Language 25(1), 49–88. 2004a. ‘The Passival and the Progressive Passive: A Case Study of Layering in the English Aspect and Voice Systems’. In: Lindquist, Hans and Mair, Christian (eds.). Corpus Approaches to Grammaticalization in English, 79–120. Amsterdam: Benjamins. 2004b. ‘Animacy, Agentivity, and the Spread of the Progressive in Modern English’. English Language and Linguistics 8(1), 47–69. 2006. ‘“Curtains like these are selling right in the city of Chicago for $1.50” – The Mediopassives in American 20th-century Advertising Language’. In: Renouf, Antoinette and Kehoe, Andrew (eds.). The Changing Face of Corpus Linguistics. Papers from the Twenty-Fourth International Conference on English Language Research on Computerized Corpora (ICAME 24),135–51. Amsterdam: Rodopi. 2007. English Mediopassive Constructions: A Cognitive, Corpus-Based Study of their Origin, Spread and Current Status. Amsterdam: Rodopi. 2009. ‘Colonial Lag, Colonial Innovation or Simply Language Change?’ In: Rohdenburg, Günter and Schlüter, Julia (eds.). One Language, Two Grammars? Differences between British and American English, 13–37. Cambridge University Press. Forthcoming A. ‘The Demise of the Being to V Construction’. To appear in Transactions of the Philological Society. Forthcoming B. ‘Global Spread of English: Processes of Change’. To appear in Kytö, Merja and Pahta, Päivi (eds.). The Cambridge Handbook of English Historical Linguistics. Cambridge University Press. Hundt, Marianne and Mair, Christian. 1999. ‘“Agile” and “Uptight” Genres: The Corpus-based Approach to Language Change in Progress’. International Journal of Corpus Linguistics 4(2), 221–42. Hundt, Marianne and Szmrecsanyi, Benedikt. 2012. ‘Animacy in Early New Zealand English’. English World-Wide 33(3), 241–63. doi:10.1075/eww.33.3.01hun. Hundt, Marianne, Denison, David, and Schneider, Gerold. 2012. ‘Relative Complexity in Scientific Discourse’. English Language and Linguistics 16(2), 209–40. Hurst, Dorothy Ann. 1951. ‘Influence of Subject and Connotation of Force’. Hispania 34(1), 74–78.
364 References Ingersoll, Charles M. 1824. Conversations on English Grammar; Explaining the Principles and Rules of the Language. Illustrated by Appropriate Exercises; Abridged and Adapted to the Use of Schools. 4th edn. Portland: Charles Green. [First published 1821]. Iyeiri, Yoko 2010. Verbs of Implicit Negation and their Complements in the History of English. Amsterdam: Benjamins. Jack, George. 1988. ‘The Origins of the English Gerund’. Nowele 12, 15–75. Jankowski, Bridget. 2004. ‘A Transatlantic Perspective of Variation and Change in English Deontic Modality’. Toronto Working Papers in Linguistics 23, 85–113. 2009. ‘Grammatical and Register Variation and Change: A Multi-corpora Perspective on the English Genitive’. Paper Presented at the American Association for Corpus Linguistics (AACL) 2009. Jankowski, Bridget. 2013. ‘A Variationist Approach to Cross-Register Language Variation and Change’. PhD thesis, University of Toronto. https://tspace. library.utoronto.ca/handle/1807/43604. Jespersen, Otto. 1909–49. A Modern English Grammar on Historical Principles. 7 vols. London: Allen & Unwin. 1940. A Modern English Grammar on Historical Principles. Vol v: Syntax. London: Allen & Unwin. 1942. A Modern English Grammar on Historical Principles. Vol. vi: Morphology. Copenhagen: Ejnar Munksgaard. Jones, Charles. 1989. A History of English Phonology. London: Longman. Joos, Martin. 1968. The English Verb: Form and Meanings. Madison, WI: University of Wisconsin Press. Joseph, Brian. 1992. ‘Diachronic Explanation: Putting Speakers Back into the Picture’. In: Davis, Garry W. and Iverson, Gregory K. (eds.) Explanation in Historical Linguistics, 123–44. Amsterdam: Benjamins. Kallen, Jeffrey L. 1989. ‘Tense and Aspect Categories in Irish English’. English World-Wide 10(1), 1–39. 2013. Irish English. Vol. ii: Republic of Ireland. Berlin: de Gruyter. Kaltenböck, Gunther. 2007. ‘Spoken Parenthetical Clauses in English’. In: Dehé, Nicole and Kavalova, Yordanka (eds.). Parentheticals, 25–52. (Linguistik Aktuell/Linguistics Today 106). Amsterdam: Benjamins. 2011. ‘Explaining Diverging Evidence: The Case of Clause-initial I think’. In: Schönefeld, Doris (ed.). Converging Evidence. Methodological and Theoretical Issues for Linguistic Research, 81–112. Amsterdam: Benjamins. Kaltenböck, Gunther, Heine, Bernd, and Kuteva, Tania. 2011. ‘On Thetical Grammar’. Studies in Language 35(4), 848–93. doi: 10.1075/sl.35.4.03kal. Kärkkäinen, Elise. 2003. Epistemic Stance in English Conversation: A Description of Interactional Functions with a Focus on I Think. Amsterdam: Benjamins. Kavalova, Yordanka. 2007. ‘And-parenthetical Clauses’. In: Dehé, Nicole and Kavalova, Yordanka (eds.). Parentheticals, 145–72 (Linguistik Aktuell/ Linguistics Today 106). Amsterdam: Benjamins. Kearns, Kate. 2007. ‘Epistemic Verbs and Zero Complementizer’. English Language and Linguistics 11(3), 475–505. Kenny, William David. 1858. An English Grammar; Adapted to the Comprehension of Young Persons, with Syntactical Observations for more Advanced Students,
References 365 Orthographical Exercises, Lessons on Parsing, Exercises, and Questions for Examination. London: Routledge. Kirk, John M. and Kallen, Jeffrey L. 2006. ‘Irish Standard English: How Celticised? How Standardised?’ In: Tristram, Hildegard L.C. (ed.). The Celtic Englishes IV, 88–113. Potsdam University Press. König, Ekkehard and Kortmann, Bernd. 1991. ‘On the Reanalysis of Verbs as Prepositions’. In: Rauh, Gisa (ed.). Approaches to Prepositions, 109–25.Tübingen: Narr. Kortmann, Bernd. 1997. Adverbial Subordination: A Typology and History of Adverbial Subordinators Based on European Languages. Berlin: de Gruyter. 2008a. ‘Synopsis: Morphological and Syntactic Variation in the British Isles’. In: Kortmann, Bernd and Upton, Clive (eds.). Varieties of English. Vol. i: The British Isles, 478–95. Berlin: de Gruyter. 2008b. ‘List of Features: Morphology and Syntax’. In: Kortmann, Bernd and Upton, Clive (eds.). Varieties of English. Vol. i: The British Isles, xxv–xxix. Berlin: de Gruyter. Kranich, Svenja. 2010. The Progressive in Modern English: A Corpus-based Study of Grammaticalization and Related Changes (Language and Computers: Studies in Practical Linguistics 72). Amsterdam: Rodopi. Krapp, George Philip. 1909. Modern English: Its Growth and Present Use. New York: Charles Scribner’s Sons. Krug, Manfred. 2000. Emerging English Modals: A Corpus-based Study of Grammaticalization. Berlin: de Gruyter. 2008. ‘Modality and the History of English Adhortatives’. In: Salkie, Raphael, Bussuttil, Pierre, and van der Auwera, Johan (eds.). Modality in English: Theory and Description, 319–48. Berlin: de Gruyter. Kytö, Merja. 1994. ‘BE vs. HAVE with Intransitives in Early Modern English’. In: Fernández, Francisco, Fuster, Miguel, and Calvo, Juan José (eds.). English Historical Linguistics 1992, 179–90. Amsterdam: Benjamins. 1997. ‘Be/Have + Past Participle: The Choice of the Auxiliary with Intransitives from Late Middle to Modern English’. In: Rissanen, Matti, Kytö, Merja, and Heikkonen, Kirsi (eds.). English in Transition: Corpus-based Studies in Linguistic Variation and Genre Styles, 16–85. Berlin: de Gruyter. 1999. ‘Collocational and Idiomatic Aspects of Verbs in Early Modern English: A Corpus-based Study of MAKE, HAVE, GIVE, TAKE, and DO’. In: Brinton, Laurel J. and Akimoto, Minoji (eds.). Collocational and Idiomatic Aspects of Composite Predicates in the History of English,167–206. Amsterdam: Benjamins. Kytö, Merja and Walker, Terry. 2003. ‘The Linguistic Study of Early Modern English Speech-related Texts: How ‘Bad’ Can ‘Bad’ Data Be?’ Journal of English Linguistics 31(3), 221–48. Kytö, Merja, Rudanko, Juhani, and Smitterberg, Erik. 2000. ‘Building a Bridge between the Present and the Past: A Corpus of 19th-Century English’. ICAME Journal 24, 85–97. Kytö, Merja, Rydén, Mats, and Smitterberg, Erik. 2006. ‘Introduction: Exploring Nineteenth-Century English – Past and Present Perspectives’. In: Kytö, Merja, Rydén, Mats, and Smitterberg, Erik (eds.). Nineteenth-Century English: Stability and Change, 1–16. Cambridge University Press.
366 References Labov, William. 1972. ‘Some Principles of Linguistic Methodology’. Language in Society 1(1), 97–120. 1989. ‘The Child as Linguistic Historian’. Language Variation and Change 1, 85–94. Lampert, Günther and Lampert, Martina. 2010. ‘Where Does Evidentiality Reside? Notes on (Alleged) Limiting Cases: Seem and Be Like’. STUF – Language Typology and Universals 63(4), 308–21. Langacker, Ronald W. 1991. Foundations of Cognitive Grammar. Vol. ii: Descriptive Application. Stanford University Press. 2008. Cognitive Grammar: A Basic Introduction. Oxford University Press. Lee, Alan J. 1976. The Origins of the Popular Press in England: 1855–1914. London: Croom Helm. Leech, Geoffrey. 2003. ‘Modality on the Move: The English Modal Auxiliaries 1961–1992’. In: Facchinetti, Roberta, Krug, Manfred, and Palmer, Frank R. (eds.). Modality in Contemporary English, 223–40. Berlin: de Gruyter. Leech, Geoffrey and Mair, Christian. 2006. ‘Current Changes’. In: Aarts, Bas and McMahon, April (eds.). The Handbook of English Linguistics, 318–42. Oxford: Blackwell. Leech, Geoffrey, Hundt, Marianne, Mair, Christian, and Smith, Nicholas. 2009. Change in Contemporary English: A Grammatical Study. Cambridge University Press. Legenhausen, Lienhard. 1998. ‘Mediopassives: Fuzziness and Speaker Evaluation’. In: Kühlwein, Wolfgang (ed.). Language as Structure and Language as Process: In Honour of Gerhard Nickel on the Occasion of His 70th Birthday, 47–62. Trier: Wissenschaftlicher Verlag. Lehmann, Christian. 1982. Thoughts on Grammaticalization: A Programmatic Sketch. Arbeiten des Kölner Universalien-Projektes. University of Cologne, Institut für Sprachwissenschaft. [In 1995 published as Thoughts on Grammaticalization. Munich: Lincom.]. 1985. ‘Grammaticalization: Synchronic Variation and Diachronic Change’. Lingua e Stile 20, 303–18. 1988. ‘Towards a Typology of Clause Linkage’. In: Haiman, John and Thompson, Sandra (eds.). Clause Combining in Discourse and Grammar, 181–225. Amsterdam: Benjamins. 1995. Thoughts on Grammaticalisation. Munich: Lincom. Leitner, Gerhard (ed.). 1986. The English Reference Grammar: Language and Linguistics, Writers and Readers. Tübingen: Niemeyer. (ed.). 1991. English Traditional Grammars: An International Perspective. Amsterdam: Benjamins. Lemmens, Maarten. 1997. ‘The Transitive–Ergative Interplay and the Conception of the World: A Case Study’. In: Verspoor, Marjolijn, Lee, Kee Dong, and Sweetser, Eve (eds.). Lexical and Syntactical Constructions and the Construction of Meaning, 363–82. Amsterdam: Benjamins. 1998a. Lexical Perspectives on Transitivity and Ergativity: Causative Constructions in English. Amsterdam: Benjamins. 1998b. ‘Lexical Constraints on Constructional Flexibility: English ‘Middable’ Verbs’. Paper presented at the Fifth ICLC, Amsterdam. Available at http://stl. recherche.univ-lille3.fr/sitespersonnels/lemmens/docspdf/middable.PDF (accessed 19 February 2014).
References 367 Leonard, Sterling Andrus. 1929. The Doctrine of Correctness in English Usage: 1700– 1800. New York: Russell & Russell. Levin, Beth. 1993. English Verb Classes and Alternations: A Preliminary Investigation. University of Chicago Press. Lewis, Michael. 2002. The English Verb: An Exploration of Structure and Meaning. Boston: Heinle and Thomson. Liberman, Mark. 2003. ‘Like Is, Like not Really Like If You Will’. http://itre.cis.upenn. edu/~myl/languagelog/archives/2003_11.html (accessed 19 February 2014). Live, Anna H. 1973. ‘The take-have Phrasal in English’. Linguistics 95, 31–50. López-Couso, María José. 1996. ‘On the History of Methinks: From Impersonal Construction to Fossilized Expression’. Folia Linguistica Historica 17(1/2), 153–69. 2007. ‘Auxiliary and Negative Cliticisation in Late Modern English’. In: PérezGuerra, Javier, González-Álvarez, Dolores, Bueno-Alonso, Jorge L., and Rama-Martínez, Esperanza (eds.). ‘Of Varying Language and Opposing Creed’: New Insights into Late Modern English, 301–23 (Linguistic Insights 28). Bern: Lang. 2010. ‘Subjectification and Intersubjectification’. In: Jucker, Andreas H. and Taavitsainen, Irma (eds.). Historical Pragmatics, 127–63. (Handbook of Pragmatics, Vol. viii). Berlin: de Gruyter. López-Couso, María José and Méndez-Naya, Belén. 2012a. ‘On the Use of as if, as though and like in Present-Day English Complementation Structures’. Journal of English Linguistics 40(2), 172–95. 2012b. ‘On Comparative Complementizers in English: Evidence from Historical Corpora’. In: Vázquez, Nila (ed.). Creation and Use of Historical English Corpora in Spain, 311–33. Newcastle upon Tyne: Cambridge Scholars Publishing. 2014 a. ‘From Clause to Pragmatic Marker: A Study of the Development of Likeparentheticals in American English’. Journal of Historical Pragmatics 15(1), 66–91. 2014 b. ‘On the Origin of Clausal Parenthetical Constructions: Epistemic/Evidential Parentheticals with Seem and Impersonal Think’. In: Taavitsainen, Irma, Jucker, Andreas, and Tuominen, Jukka (eds.). Diachronic Corpus Pragmatics, 189–212 (Pragmatics & Beyond New Series 243.) Amsterdam: Benjamins. Lorenz, David. 2012. ‘Semi-modal Constructions in English: Emancipation through Frequency.’ PhD thesis, University of Freiburg. Los, Bettelou. 2005. The Rise of the To-infinitive. Oxford University Press. Lowth, Robert. 1762. A Short Introduction to English Grammar. London: Hughs. [1967 Menston: Scolar Press (facsimile)]. Lyons, John. 1977. Semantics. Vol. i. Cambridge University Press. Macafee, Caroline. 2011. ‘Characteristics of Nonstandard Grammar in Scotland’. http://archive.is/apy4m (accessed 19 February 2014). McArthur, Tom (ed.). 1992. The Oxford Companion to the English Language. Oxford University Press. McCafferty, Kevin 2007. ‘Northern Irish English’. In: Britain, David (ed.). Language in the British Isles, 122–34. Cambridge University Press. McCawley, John. 1982. ‘Parentheticals and discontinuous constituent structure’. Linguistic Inquiry 13(1), 91-106.
368 References McFadden, Thomas. 2007a. ‘Auxiliary “Selection” and Restrictions on Perfect Semantics: An Early English/Modern Scandinavian Parallel’. Paper presented at the Sixth York-Newcastle-Holland Symposium on the History of English Syntax, Leiden, 2007. 2007b. ‘Auxiliary Selection’. Language and Linguistics Compass 1(6), 674–708. Mair, Christian. 1997. ‘Parallel Corpora: A Real-time Approach to the Study of Language Change in Progress’. In: Ljung, Magnus (ed.). Corpus-based Studies in English: Papers from the Seventeenth International Conference on English Language Research on Computerized Corpora (ICAME 17), Stockholm, 15–19 May, 1996, 195–209 (Language and Computers: Studies in Practical Linguistics 20). Amsterdam: Rodopi. 2002. ‘Three Changing Patterns of Verb Complementation in Late Modern English’. English Language and Linguistics 6(1), 105–31. 2003. ‘Gerundial Complements after Begin and Start: Grammatical and Sociolinguistic Factors, and How They Work Against Each Other’. In: Rohdenburg, Günter and Mondorf, Britta (eds.). Determinants of Grammatical Variation in English, 329–45. Berlin: de Gruyter. 2005. ‘The Eighteenth and Nineteenth Centuries as a Critical Period in the Formation of the Modern English System of Non-finite Complement Clauses’. In: Houswitschka, Christoph, Knappe, Gabriele, and Müller, Anja (eds.). Anglistentag 2005 Bamberg, 531–42. Trier: Wissenschaftlicher Verlag. 2006. Twentieth-century English: History, Variation, and Standardization. Cambridge University Press. 2009. ‘Corpus Linguistics Meets Sociolinguistics: The Role of Corpus Evidence in the Study of Sociolinguistic Variation and Change’. In: Renouf, Antoinette and Kehoe, Andrew (eds.). Corpus Linguistics: Refinements and Reassessments. Proceedings of the 2007 ICAME Conference – Stratford-upon-Avon, 1–26. Amsterdam: Rodopi. 2012. ‘Using the World-Wide Web to Study Language Change in Progress: From Opportunistic to Systematic Use’. In: Nevalainen, Terttu and Traugott, Elizabeth Closs (eds.). The Oxford Handbook of the History of English, 245–55. New York: Oxford University Press. Mair, Christian and Garley, Matthew. Forthcoming. ‘Support for Do-support with Got in Informal Mainstream Usage? – Changing Speaker Attitudes to a Stigmatised Construction in North American English’. Mason, Charles Peter. 1858. English Grammar; Including the Principles of Grammatical Analysis. London: Walton and Maberly. Mathews, Harriet. 1892. Outlines of English Grammar with Continuous Selections for Practice. Boston, MA: D.C. Heath and Co. Melchers, Gunnel 1992. ‘‘Du’s no heard da last o’ dis’ – On the Use of Be as a Perfective Auxiliary in Shetland Dialect’. In: Rissanen, Matti, Ihalainen, Ossi, Nevalainen, Terttu, and Taavitsainen, Irma (eds.). History of Englishes. New Methods and Interpretations in Historical Linguistics, 602–10. Berlin: de Gruyter. 2008. ‘English Spoken in Orkney and Shetland: Morphology, Syntax, and Lexicon’. In: Kortmann, Bernd and Upton, Clive (eds.). Varieties of English Vol. i: The British Isles, 285–98. Berlin: de Gruyter.
References 369 Melchers, Gunnel and Sundkvist, Peter. 2010. ‘Orkney and Shetland’. In: Schreier, Daniel, Trudgill, Peter, Schneider, Edgar W., and Williams, Jeffrey P. (eds.). The Lesser-Known Varieties of English, 17–34. Cambridge University Press. Méndez-Naya, Belén. 2008. ‘Special Issue on English Intensifiers: Introduction’. English Language and Linguistics 12(2), 213–19. Miller, D. Gary. 2001. Nonfinite Structures in Theory and Change. Oxford University Press. Miller, Jim. 2004. ‘Perfect and Resultative Constructions in Spoken and Nonspoken English’. In: Fischer, Olga, Norde, Muriel, and Perridon, Harry (eds.). Up and Down the Cline: the Nature of Grammaticalization, 229–46. Amsterdam: Benjamins. Mindt, Dieter. 1992. Zeitbezug im Englischen: eine didaktische Grammatik des englischen Futurs. Tübingen: Narr. 2000. An Empirical Grammar of the English Verb. Berlin: Cornelsen. Moessner, Lilo. 1997. ‘The Syntax of Older Scots’. In: Jones, Charles (ed.). The Edinburgh History of the Scots Language, 112–55. Edinburgh University Press. Mondorf, Britta. 2010. ‘Causative Verbs in British and American English’. Paper presented at the ICAME conference, University of Giessen, Germany, 26–30 May 2010. Mulligan, John. 1868. Exposition of the Grammatical Structure of the English Language; Being an Attempt to Furnish an Improved Method of Teaching Grammar. For the Use of Schools and Colleges. New York: D. Appleton and Company [First published 1858]. Murray, Lindley. 1795. English Grammar, Adapted to the Different Classes of Learners. With an Appendix, Containing Rules and Observations for Promoting Perspicuity in Speaking and Writing. York: Wilson, Spence, and Mawman. Murray, Thomas E. and Simon, Beth Lee. 1999. ‘Want + Past Participle in American English’. American Speech 74(2), 140–64. Mustanoja, Tauno F. 1960. A Middle English Syntax. Helsinki: Société Néophilologique. Næss, Åshild. 2007. Prototypical Transitivity. Amsterdam: Benjamins. Navest, Karlijn. 2011. John Ash and the Rise of Children’s Grammar. Utrecht: LOT Publications. Nesselhauf, Nadja. 2007. ‘The Spread of the Progressive and Its “Future” Use’. English Language and Linguistics 11(1), 193–209. 2010. ‘The Development of Future Time Expressions in Late Modern English: Redistribution of Forms or Change in Discourse?’ In: Depraetere, Ilse and Williams, Christopher (eds.). Future Time Reference in English, Special Issue of English Language and Linguistics 14(2), 163–86. 2012. ‘Mechanisms of Language Change in a Functional System: The Recent Semantic Evolution of English Future Time Expressions’. Journal of Historical Linguistics 2(1), 83–132. 2013. ‘The Emergence and Early Development of Present-Day ’ll’. Studia Neophilologica 85(2), 157–173. Nickel, Gerhard. 1966. Die Expanded Form im Altenglischen. Vorkommen, Funktion und Herkunft der Umschreibung beon/wesan + Partizip Präsens. Neumünster: Karl Wachholtz.
370 References 1968. ‘Complex Verbal Structures in English’. International Review of Applied Linguistics in Language Teaching 6(1–4), 1–21. Noël, Dirk. 2003. ‘Is There Semantics in all Syntax? The Case of Accusative and Infinitive Constructions vs. That-clauses’. In: Rohdenburg, Günter and Mondorf, Britta (eds.). Determinants of Grammatical Variation in English, 347– 77. Berlin: de Gruyter. 2007. ‘Diachronic Construction Grammar and Grammaticalization Theory’. Functions of Language 14(2), 177–202. Noonan, Michael. 1985. ‘Complementation’. In: Shopen, Timothy (ed.). Language Typology and Syntactic Description. Vol. ii, 42–140. Cambridge University Press. Nunberg, Geoffrey, Sag, Ivan A., and Wasow, Thomas. 1994. ‘Idioms’. Language 70(3), 491–538. Ó Sé, Diarmaid. 1992. ‘The Perfect in Irish’. Ériu 43, 39–67. Ó Siadhail, Mícheál. 1989. Modern Irish: Grammatical Structure and Dialectal Variation. Cambridge University Press. Övergaard, Gerd. 1995. The Mandative Subjunctive in British and American English in the 20th Century. Uppsala: Acta Universitatis Upsaliensis and Stockholm: Almquist and Wiksell. Palander-Collin, Minna. 1999. Grammaticalization and Social Embedding: I THINK and METHINKS in Middle and Early Modern English (Mémoires de la Société Néophilologique de Helsinki 55). Helsinki: Société Néophilologique. Palmer, Frank Robert. 1988. The English Verb. 2nd edn. London: Longman. Paradis, Carita. 2008. ‘Configurations, Construals and Change: Expressions of degree’. English Language and Linguistics 12(2), 317–43. Parker, Richard G. and Fox, Charles. 1835. Progressive Exercises in English Grammar. Part ii. Containing the Principles of the Synthesis or Construction of the English Language. Boston: Crocker & Brewster; and New York: Leavitt, Lord & Co. [First published 1834]. Partington, Alan. 1996. Patterns and Meanings: Using Corpora for English Language Research and Teaching. Amsterdam: Benjamins. Pavlenko, Alexander 1997. ‘The Origin of the be-perfect with Transitives in the Shetland Dialect’. Scottish Language 16, 88–96. Peitsara, Kirsti. 2004. ‘Variants of Contraction: The Case of it’s and ’tis’. ICAME Journal 28, 77–94. Perek, Florent and Lemmens, Maarten. 2010. ‘Getting at the Meaning of the English at-Construction: The Case of a Constructional Split’. CogniTextes 5. http://cognitextes.revues.org/331 (accessed 19 February 2014). Peters, Hans. 1993. Die englischen Gradadverbien der Kategorie booster. Tübingen: Narr. Phillipps, K.C. 1984. Language and Class in Victorian England. Oxford: Blackwell. Pinheiro, José C. and Bates, Douglas M. 2000. Mixed-Effects Models in S and S-PLUS. New York: Springer. Plungian, Vladimir A. 2001. ‘The Place of Evidentiality within the Universal Grammatical Space’. Journal of Pragmatics 33(3), 349–57. Potts, Chris. 2002. ‘The Syntax and Semantics of as-parentheticals’. Natural Language and Linguistic Theory 20(3), 623–89. Poutsma, Hendrik 1926. A Grammar of Late Modern English. Part ii: The Parts of Speech. Section ii: The Verb and the Particles. Groningen: P. Noordhoff.
References 371 Priestley, Joseph. 1768. The Rudiments of Grammar, Adapted to the Use of Schools; With Notes and Observations, for the Use of Those Who have Made some Proficiency in the Language. 2nd edn. London: printed for T. Becket and P.A. De Hondt, and J. Johnson. [First published 1761]. Prince, Ellen. 1972. ‘A Note on Aspect in English: The take a walk Construction’. In: Ploetz, Senta (ed.). Transformationelle Analyse, 409–19. Frankfurt: Athenum. Pullum, Geoffrey K. 1997. ‘The Morpholexical Nature of To-contraction’. Language 73(1), 79–102. 2003. ‘It’s like, so unfair’. http://itre.cis.upenn.edu/%7Emyl/languagelog/ archives/000138.html (accessed 19 February 2014). Quackenbos, George P. 1868. An English Grammar. New York: D. Appleton and Company [First published 1862]. Quirk, Randolph, Greenbaum, Sidney, Leech, Geoffrey, and Svartvik, Jan. 1972. A Grammar of Contemporary English. London: Longman. 1985. A Comprehensive Grammar of the English Language. London: Longman. R Development Core Team. 2011. R: A Language and Environment for Statistical Computing. Vienna, Austria. www.R-project.org/ (accessed 19 February 2014). Radford, Andrew. 1997. Syntactic Theory and the Structure of English. Cambridge University Press. Rampton, Ben. 1995. Crossing: Language and Ethnicity Among Adolescents. London: Longman. Renouf, Antoinette and Sinclair, John. 1991. ‘Collocational Frameworks in English’. In: Aijmer, Karin and Altenberg, Bengt (eds.). English Corpus Linguistics: Studies in Honour of Jan Svartvik, 128–43. London: Longman. Rissanen, Matti. 1999. ‘Syntax’. In: Lass, Roger (ed.). The Cambridge History of the English Language. Vol. iii 1476–1776, 187–331. Cambridge University Press. Rizzi, Luigi. 1990. Relativized Minimality. Cambridge, MA: MIT Press. Rohdenburg, Günter. 1995. ‘On the Replacement of Finite Complement Clauses by Infinitives in English’. English Studies 76(4), 367–88. 1996. ‘Cognitive Complexity and Increased Grammatical Explicitness in English’. Cognitive Linguistics 7(2), 149–82. 2006a. ‘The Role of Functional Constraints in the Evolution of the English Complementation System’. In: Dalton-Puffer, Christiane, Kastovsky, Dieter, Ritt, Nikolaus, and Schendl, Herbert (eds.). Syntax, Style and Grammatical Norms: English from 1500–2000, 143–66. Bern: Lang. 2006b. ‘Discrepancies Between the Rule Formulations Advanced by Famous Linguists and Their Own Written Usage’. In: Kötter, Markus, Traxel, Oliver, and Gabel, Stephan (eds.). Investigating and Facilitating Language Learning. Papers in Honour of Lienhard Legenhausen, 47–63. Trier: Wissenschaftlicher Verlag. 2008. ‘On the History and Present Behaviour of Subordinating Conjunctions in English’. In: Seoane, Elena and Lopéz-Couso, María José (eds.). Theoretical and Empirical Issues in Grammaticalization, 315–31. Amsterdam: Benjamins. 2009. ‘Grammatical Divergence between British and American English in the Nineteenth and Early Twentieth Centuries’. In: Tieken-Boon van Ostade, Ingrid, and van der Wurff, Wim (eds.). Current Issues in Late Modern English, 301–29. Bern: Lang.
372 References 2013. ‘Using the OED Quotations Database as a Diachronic Corpus’. In: Krug, Manfred and Schlüter, Julia (eds.). Research Methods in Language Variation and Change, 136–157. Cambridge University Press. Rohdenburg, Günter and Schlüter, Julia. 2009. ‘New Departures’. In: Rohdenburg, Günter and Schlüter, Julia (eds.). One Language, Two Grammars? Differences between British and American English, 364–423. Cambridge University Press. Ronan, Patricia. 2005. ‘The After-perfect in Irish English’. In: Filppula, Markku, Klemola, Juhani, and Palander, Marjatta (eds.). Dialects Across Borders. Selected Papers from the 11th International Conference on Methods in Dialectology (Methods XI), Joensuu, August 2002, 253–70. Amsterdam: Benjamins. Rooryck, Johan. 2000. Configurations of Sentential Complementation: Perspectives from Romance Languages. London: Routledge. Rosenbach, Anette. 2002. Genitive Variation in English: Conceptual Factors in Synchronic and Diachronic Studies. Berlin: de Gruyter. 2003. ‘Aspects of Iconicity and Economy in the Choice Between the S-genitive and the Of-genitive in English’. In: Rohdenburg, Günter and Mondorf, Britta (eds.). Determinants Of Grammatical Variation in English, 379–412. Berlin: de Gruyter. 2004. ‘The English S-genitive: A Case of Degrammaticalization?’ In: Fischer, Olga, Perridon, Harry, and Norde, Muriel (eds.). Up and Down the Cline: The Nature of Grammaticalization, 73–96. Amsterdam: Benjamins. 2005. ‘Animacy Versus Weight as Determinants of Grammatical Variation in English’. Language 81(3), 613–44. 2008. ‘Animacy and Grammatical Variation: Findings from English Genitive Variation’. Lingua 118(2), 151–71. Rosenbaum, Peter. 1967. The Grammar of English Predicate Complementation. Cambridge, MA: MIT Press. Ross, John Robert. 1973a. ‘Slifting’. In: Gross, Maurice, Halle, Morris, and Schützenberger, Marcel-Paul (eds.). The Formal Analysis of Natural Languages. Proceedings of the First International Conference, 133–69. The Hague and Paris: Mouton. 1973b. ‘A Fake NP Squish’. In: Bailey, Charles-James N. and Shuy, Roger (eds.). New Ways of Analyzing Variation in English, 96–140. Washington, DC: Georgetown University Press. 1995. ‘Defective Noun Phrases’. In: Dainora, Audra, Hemphill, Rachel, Luka, Barbara, Need Barbara, and Pargman, Sheri (eds.). Proceedings of the ThirtyFirst Regional Meeting of the Chicago Linguistic Society, 398–440. University of Chicago Press. Rosta, Andrew. 1995. ‘How Does This Sentence Interpret? The Semantics of English Mediopassives’. In: Aarts, Bas and Meyer, Charles F. (eds.). The Verb in Contemporary English: Theory and Description, 123–44. Cambridge University Press. Rouchota, Villy. 1998. ‘Procedural Meaning and Parenthetical Discourse Marker’. In: Jucker, Andreas H. and Ziv, Yael (eds.). Discourse Markers: Descriptions and Theory, 97–126 (Pragmatics & Beyond, New Series 57). Amsterdam: Benjamins. Rudanko, Juhani 1996. Prepositions and Complements: A Syntactic and Semantic Study of Verbs Governing Prepositions and Complement Clauses in Present-Day English. Albany, NY: State University of New York Press.
References 373 1998. Change and Continuity in the English Language: Studies on Complementation over the Past Three Hundred Years. Lanham, MD: University Press of America. 1999. Diachronic Studies of English Complementation Patterns: Eighteenth Century Evidence in Tracing the Development of Verbs and Adjectives Selecting Prepositions and Complement Clauses. Lanham, MD: University Press of America. 2000. Corpora and Complementation. Lanham, MD: University Press of America. 2006. ‘Watching English Grammar Change’. English Language and Linguistics 10(1), 31–48. 2010. ‘Explaining Grammatical Variation and Change: A Case Study of Complementation in American English over Three Decades’. Journal of English Linguistics 38(1), 4–24. 2012. ‘Exploring Aspects of the Great Complement Shift, with Evidence from the TIME Corpus and COCA’. In: Nevalainen, Terttu and Traugott, Elizabeth Closs (eds.). The Oxford Handbook of the History of English, 222–32. New York: Oxford University Press. Rydén, Mats. 1984. ‘The Study of Eighteenth Century English Syntax’. In: Fisiak, Jacek (ed.). Historical Syntax, 509–20. Berlin: de Gruyter. 1991. ‘The be/have Variation with Intransitives in its Crucial Phases’. In: Kastovsky, Dieter (ed.). Historical English Syntax, 343–54. Berlin: de Gruyter. Rydén, Mats and Brorström, Sverker. 1987. The be/have Variation with Intransitives in English, with Special Reference to the Late Modern Period. Stockholm: Almqvist and Wiksell. Sahkai, Heete. 2009. ‘Implications of a Construction for the Theory of Grammaticalization: Estonian Adhesive Manner and Cause Adverbials’. TRAMES 13(4), 374–400. Schelfhout, Carla, Coppen, Peter-Arno, and Oostdijk, Nelleke. 2004. ‘Finite Comment Clauses in Dutch: A Corpus-based Approach’. Journal of Germanic Linguistics 16(4), 331–49. doi: http://dx.doi.org/10.1017/ S1470542704040437. Schlüter, Julia. 2009. ‘The Conditional Subjunctive’. In: Rohdenburg, Günter and Schlüter, Julia (eds.). One Language, Two Grammars? Differences between British and American English, 277–305. Cambridge University Press. Schneider, Gerold. 2012. ‘Adapting a Parser to Historical English’. In: Tyrkkö, Jukka, Kilpiö, Matti, Nevalainen, Terttu, and Rissanen, Matti (eds.). Outposts of Historical Corpus Linguistics: From the Helsinki Corpus to a Proliferation of Resources (Studies in Variation, Contact and Change in English, Vol. x). Helsinki University. www.helsinki.fi/varieng/series/volumes/10/schneider/ (accessed 19 February 2014). Schneider, Stefan. 2007. Reduced Parenthetical Clauses as Mitigators: A Corpus Study of Spoken French, Italian and Spanish. (Studies in Corpus Linguistics 27). Amsterdam: Benjamins. Schönberger, Axel. 2009. Die Ars maior des Aelius Donatus: lateinischer Text und kommentierte deutsche Übersetzung einer antiken Lateingrammatik des 4. Jahrhunderts for den fortgeschrittenen Anfängerunterricht. Frankfurt am Main: Valentia. Scott, Mike. 2009. Wordsmith Tools 5.0. Oxford University Press. Searle, John R. 1976. ‘A Classification of Illocutionary Acts’. Language in Society 5(1), 1–23.
374 References 1979. Expression and Meaning: Studies in the Theory of Speech Acts. Cambridge University Press. Seggewiß, Friederike. 2012. ‘Current Changes in the English Modals: A Corpusbased Analysis of Present-Day Spoken English’. PhD dissertation, University of Freiburg. Seiler, Hans-Jakob. 1982. Possession as an Operational Dimension of Language. Tübingen: Narr. Seoane, Elena and Suárez-Gómez, Cristina. 2013. ‘The Expression of the Perfect in East and South-East Asian Englishes’. English World-Wide 34(1), 1–25. Seppänen, Aimo. 1977. ‘The Position of let in the English Auxiliary System’. English Studies 58(6), 515–29. Shih, Stephanie, Grafmiller, Jason, Futrell, Richard, and Bresnan, Joan. Forthcoming. ‘Rhythm’s Role in Genitive Construction Choice in Spoken English’. In: Vogel, Ralf and van de Vijver, Ruben (eds.). Rhythm in Phonetics, Grammar and Cognition, Berlin: de Gruyter. Smith, Jennifer and Durham, Mercedes. 2012. ‘Bidialectalism or Dialect Death? Explaining Generational Change in the Shetland Islands, Scotland’. American Speech 87(1), 59–88. Smith, Michael B and Escobedo Joyce. 2002. ‘The Semantics of to-Infinitival vs. –ing Verb Complement Constructions in English’. In: Andronis, Mary, Ball, Christopher, Elston, Heidi, and Neuvel, Sylvain (eds.). The Proceedings from the Main Session of the Chicago Linguistic Society’s Thirty-seventh Meeting, 549– 65. Chicago Linguistic Society. Smith, Nicholas and Rayson, Paul. 2007. ‘Recent Change and Variation in the British English Use of the Progressive Passive’. ICAME Journal 31, 129–60. Smith, Roswell Chamberlain. 1841. English Grammar on the Productive System: A Method of Instruction Recently Adopted in Germany and Switzerland. Designed for Schools and Academies. Stereotype edition. Hartford: John Paine [First published 1831]. Smitterberg, Erik. 2005. The Progressive in 19th-century English: A Process of Integration. (Language and Computers: Studies in Practical Linguistics 54). Amsterdam: Rodopi. 2008. ‘The Progressive and Phrasal Verbs: Evidence of Colloquialization in Nineteenth-Century English?’ In: Nevalainen, Terttu, Taavitsainen, Irma, Pahta, Päivi, and Korhonen, Minna (eds.). The Dynamics of Linguistic Variation: Corpus Evidence on English Past and Present, 269–89 (Studies in Language Variation 2). Amsterdam: Benjamins. 2012. ‘Colloquialization and Not-Contraction in Nineteenth-Century English’. In: Markus, Manfred, Iyeiri, Yoko, Heuberger, Reinhard, and Chamson, Emil (eds.). Middle and Modern English Corpus Linguistics: A Multi-dimensional Approach, 191–206 (Studies in Corpus Linguistics 50). Amsterdam: Benjamins. Spasov, Dimiter. 1966. English Phrasal Verbs. Sofia: Naouka i Izkoustvo. Stein, Gabriele. 1991. ‘The Phrasal Verb Type “to have a look” in Modern English’. International Review of Applied Linguistics in Language Teaching 19(1–4), 1–29. Stenson, Nancy. 1981. Studies in Irish Syntax. Tübingen: Narr.
References 375 Straaijer, Robin. 2009. ‘Deontic and Epistemic Modals as Indicators of Prescriptive and Descriptive Language in the Grammars by Joseph Priestley and Robert Lowth’. In: Tieken-Boon van Ostade, Ingrid and van der Wurff, Wim (eds.). Current Issues in Late Modern English, 57–87. Bern: Lang. 2011. Joseph Priestley, Grammarian: Late Modern English Normativism and Usage in a Sociohistorical Context. Utrecht: LOT Publications. Stuurman, Frits. 1990. Two Grammatical Models of Modern English. The Old and the New from A to Z. London: Routledge. Sundby, Bertil, Bjørge, Anne Kari, and Haugland, Kari E. 1991. A Dictionary of English Normative Grammar: 1700–1800 (Studies in the History of the Language Sciences 63). Amsterdam: Benjamins. Swan, Toril. 2003. ‘Present Participles in the History of English and Norwegian’. Neuphilologische Mitteilungen 104, 179–95. Szmrecsanyi, Benedikt. 2006. Morphosyntactic Persistence in Spoken English: A Corpus Study at the Intersection of Variationist Sociolinguistics, Psycholinguistics, and Discourse Analysis. Berlin: de Gruyter. 2013. ‘The Great Regression: Genitive Variability in Late Modern English News Texts’. In: Börjars, Kersti, Denison, David, and Scott, Alan (eds.). Morphosyntactic Categories and the Expression of Possession, 59–88. Amsterdam: Benjamins. Szmrecsanyi, Benedikt and Hinrichs, Lars. 2008. ‘Probabilistic Determinants of Genitive Variation in Spoken and Written English: A Multivariate Comparison Across Time, Space, and Genres’. In: Nevalainen, Terttu, Taavitsainen, Irma, Pahta, Päivi, and Korhonen, Minna (eds.). The Dynamics of Linguistic Variation: Corpus Evidence on English Past and Present, 291–309. Amsterdam: Benjamins. Tagliamonte, Sali A. and D’Arcy, Alexandra. 2007. ‘The Modals of Obligation/ Necessity in Canadian Perspective’. English World-Wide 28(1), 47–87. Tagliamonte, Sali A., D’Arcy, Alexandra, and Jankowski, Bridget. 2010. ‘Social Work and Linguistic Systems: Marking Possession in Canadian English’. Language Variation and Change 22(1), 149–73. Taglicht, Josef. 1970. ‘The Genesis of the Conventional Rules for the Use of Shall and Will’. English Studies 51(3), 193–213. Thim, Stefan. 2012. Phrasal Verbs: The English Verb-particle Combination and Its History (Topics in English Linguistics 78). Berlin: de Gruyter. Thomas, Russel. 1931. ‘Syntactical Processes Involved in the Development of the Adnominal Periphrastic Genitive in the English Language’. PhD thesis, University of Michigan. Thompson, Sandra A. and Mulac, Anthony. 1991. ‘A Quantitative Perspective on the Grammaticalization of Epistemic Parentheticals in English’. In: Traugott, Elizabeth Closs and Heine, Bernd (eds.). Approaches to Grammaticalization. Vol. ii, 313–29. Amsterdam: Benjamins. Tieken-Boon van Ostade, Ingrid (ed.). 2008. Grammars, Grammarians and GrammarWriting in 18th Century England. Berlin: de Gruyter. 2009. An Introduction to Late Modern English. Edinburgh University Press. 2011. The Bishop’s Grammar: Robert Lowth and the Rise of Prescriptivism in English. Oxford University Press.
376 References Timberlake, Alan. 1977. ‘Reanalysis and Actualisation in Syntactic Change’. In: Li, Charles N. (ed.). Mechanisms of Syntactic Change, 141–77. Austin, TX: University of Texas Press. Torres Cacoullos, Rena and Walker, James A. 2009. ‘The Present of the English Future: Grammatical Variation and Collocations in Discourse’. Language 85 (2), 321–54. Traugott, Elizabeth Closs. 1992. ‘Syntax’. In: Hogg, Richard M. (ed.). The Cambridge History of the English Language. Vol. i: The Beginnings to 1066, 168–289. Cambridge University Press. 1995. ‘Subjectification in Grammaticalisation’. In: Stein, Dieter and Wright, Susan (eds.). Subjectivity and Subjectivisation: Linguistic Perspectives, 31–54. Cambridge University Press. 2003. ‘From Subjectification to Intersubjectification’. In: Hickey, Raymond (ed.). Motives for Language Change, 124–39. Cambridge University Press. 2006. ‘The Semantic Development of Scalar Focus Modifiers’. In: van Kemenade, Ans and Los, Bettelou (eds.). The Handbook of the History of English, 335–59. Oxford: Blackwell. 2007. ‘The Concepts of Constructional Mismatch and Type-shifting From the Perspective of Grammaticalization’. Cognitive Linguistics 18(4), 523–57. 2008a. ‘The Grammaticalisation of NP of NP Patterns’. In: Bergs, Alexander and Diewald, Gabriele (eds.). Constructions and Language Change, 23–45. Berlin: de Gruyter. 2008b. ‘Grammaticalization, Constructions and the Incremental Development of Language: Suggestions from the Development of Degree Modifiers in English’. In: Eckardt, Regine, Jäger, Gerhard, and Veenstra, Tonjes (eds.). Variation, Selection, Development: Probing the Evolutionary Model of Language Change, 219–50. Berlin: de Gruyter. 2010. ‘(Inter)subjectivity and (Inter)subjectification: A Reassessment’. In: Davidse, Kristin, Vandelanotte, Lieven, and Cuyckens, Hubert (eds.). Subjectification, Intersubjectification and Grammaticalization, 29–71. (Topics in English Linguistics 66). Berlin: de Gruyter. 2012a. ‘Constructional Changes and Constructionalization’. In collaboration with Graeme Trousdale. Lecture presented at the University of Santiago de Compostela. 16 Oct. www.stanford.edu/~traugott/resources/ TraugottSantiago1.pptx. 2012b. ‘Grammatical Constructionalization: Rethinking Grammaticalisation in the Light of Constructionalization’. In collaboration with Graeme Trousdale. Lecture presented at the University of Santiago de Compostela. 17 Oct. www. stanford.edu/~traugott/resources/TraugottSantiago2.pptx. 2012c. ‘The Status of Onset Contexts in Analysis of Micro-changes’. In: Kytö, Merja (ed.). English Corpus Linguistics: Crossing Paths, 221–55. Amsterdam: Rodopi. Traugott, Elizabeth Closs and Dasher, Richard B. 2002. Regularity in Semantic Change (Cambridge Studies in Linguistics 97). Cambridge University Press. Traugott, Elizabeth Closs and König, Ekkehard. 1991. ‘The Semantics–Pragmatics of Grammaticalization Revisited’. In: Traugott, Elizabeth Closs and Heine,
References 377 Bernd (eds.). Approaches to Grammaticalization. Vol. i: Focus on Theoretical and Methodological Issues, 189–218. Amsterdam: Benjamins. Traugott, Elizabeth Closs and Trousdale, Graeme. 2010. ‘Gradience, Gradualness and Grammaticalization: How Do They Intersect?’. In: Traugott, Elizabeth Closs and Trousdale, Graeme (eds.). Gradience, Gradualness, and Grammaticalization, 19–44. Amsterdam: Benjamins. 2013. Constructionalization and Constructional Changes. (Oxford Studies in Diachronic and Historical Linguistics 6). Oxford University Press. Trousdale, Graeme. 2008a. ‘Constructions in Grammaticalization and Lexicalization: Evidence from the History of a Composite Predicate Construction in English’. In: Trousdale, Graeme and Gisborne, Nikolas (eds.). Constructional Approaches to English Grammar, 33–67. (Topics in English Linguistics). Berlin: de Gruyter. 2008b. ‘A Constructional Approach to Lexicalization Processes in the History of English: Evidence from Possessive Constructions’. Word Structure 1, 156–77. 2010. ‘Issues in Constructional Approaches to Grammaticalization in English’. In: Stathi, Katarine, Gehweiler, Elke, and König, Ekkehard (eds.). Grammaticalization: Current Views and Issues, 51–72. (Studies in Language Companion Series 119). Amsterdam: Benjamins. 2012. ‘Grammaticalization, Constructions, and the Grammaticalization of Constructions’. In: Davidse, Kristin, Breban, Tine, Brems, Lieselotte, and Mortelmans, Tanja (eds.). Grammaticalization and Language Change: New Reflections, 167–98. (Studies in Language Companion Series 130). Amsterdam: Benjamins. 2013. ‘Multiple Inheritance and Constructional Change’. In: De Smet, Hendrik, Ghesquière, Lobke, and Van de Velde, Freek (eds.). Multiple Source Constructions in Language Change. Special issue of Studies in Language 37(3), 491–514. Trudgill, Peter. 2002. Sociolinguistic Variation and Change. Washington DC: Georgetown University Press. Trudgill, Peter and Hannah, Jean 2008 [1982]. International English. A Guide to the Varieties of Standard English. 5th edn. London: Hodder Education. Trudgill, Peter, Nevalainen, Terttu, and Wischer, Ilse. 2002. ‘Dynamic Have in North American and British Isles English’. English Language and Linguistics 6(1), 1–15. Urmson, James O. 1952. ‘Parenthetical Verbs’. Mind 61, 480–96. Van Bogaert, Julie. 2010. ‘A Constructional Taxonomy of I think and Related Expressions: Accounting for the Variability of Complement-taking Mental Predicates’. English Language and Linguistics 14(3), 399–427. Van der Auwera, Johan, Dobrushina, Nina, and Goussev, Valentin. 2005. ‘ImperativeHortative Systems’. In: Haspelmath, Martin, Dryer, Matthew S., Gil, David, and Comrie, Bernard (eds.). The World Atlas of Language Structures, 294–95. Oxford University Press. Van de Velde, Freek. 2011. ‘Left-peripheral Expansion of the English NP’. English Language and Linguistics 15(2), 387–415. Van Valin, Robert D. 1999. ‘Functional Relations’. In: Brown, Keith and Miller, Jim (eds.). Concise Encyclopedia of Grammatical Categories, 150–62. Amsterdam: Elsevier.
378 References Visser, Frederikus Theodorus. 1963–73. An Historical Syntax of the English Language. 3 vols. Leiden: Brill. 1973. An Historical Syntax of the English Language. Part iii, 2nd Half: Syntactical Units with Two and with More Verbs. Leiden: Brill. Vogt, Helmut. 2002. ‘Semantische Unterschiede zwischen komplexen Verbalfügungen vom Typ give/have/take a look und ihren einfachen Verbentsprechungen’. In: Rapp, Reinhard (ed.). Sprachwissenschaft auf dem Weg in das Dritte Jahrtausend. Akten des 34. Linguistischen Kolloquiums in Germersheim 1999, 237–44. Bern: Lang. Vosberg, Uwe. 2003. ‘The Role of Extractions and Horror Aequi in the Evolution of –ing Complements in Modern English’. In: Rohdenburg, Günter and Mondorf, Britta (eds.). Determinants of Grammatical Variation in English, 329–45. Berlin: de Gruyter. 2006. Die große Komplementverschiebung. Außersemantische Einflüsse auf die Entwicklung satzwertiger Ergänzungen im Neuenglischen. Tübingen: Narr. Vries, Mark de. 2007. ‘Invisible Constituents? Parentheses as B-merged Adverbial Phrases’. In: Dehé, Nicole and Kavalova, Yordanka (eds.). Parentheticals, 203– 34. Amsterdam: Benjamins. Walshe, Shane. 2009. Irish English as Represented in Film. Frankfurt am Main: Lang. 2011. ‘‘Normal people like us don’t use that type of language. Remember this is the real world’. The Language of Father Ted: Representations of Irish English in a Fictional World’. In: Amador-Moreno, Carolina P. and McCafferty, Kevin (eds.). Fictionalising Orality. Special issue of Sociolinguistic Studies 5(1),127–48. Warner, Anthony. 1982. Complementation in Middle English and the Methodology of Historical Syntax. Pennsylvania State University Press. Wekker, Herman Ch. 1976. The Expression of Future Time in Contemporary British English: An Investigation into the Syntax and Semantics of Five Verbal Constructions Expressing Futurity. Amsterdam: North-Holland. Wells, William H. 1847. A Grammar of the English Language; For the Use of Schools. Andover: Allen, Morrill, and Wardwell; and Boston: John P. Jewett & Co. [First published 1846]. Westergren Axelsson, Margareta. 1998. Contraction in British Newspapers in the Late 20th Century. Uppsala University Press. Westin, Ingrid. 2002. Language Change in English Newspaper Editorials (Language and Computers: Studies in Practical Linguistics 44.) Amsterdam: Rodopi. Wichmann, Anne. 2004. ‘The Intonation of Please-requests: A Corpus-based Study’. Journal of Pragmatics 36(9), 1521–49. Wierzbicka, Anna. 1988. The Semantics of Grammar. Amsterdam: Benjamins. 2006. English: Meaning and Culture. Oxford University Press. Wischer, Ilse. 2000. ‘Grammaticalization Versus Lexicalization: Methinks There is Some Confusion’. In: Fischer, Olga, Rosenbach, Anette, and Stein, Dieter (eds.). Pathways of Change. Grammaticalization in English, 355–70. Amsterdam: Benjamins. Wolk, Christoph, Bresnan, Joan, Rosenbach, Anette, and Szmrecsanyi, Benedikt. 2013. ‘Dative and Genitive Variability in Late Modern English: Exploring Cross-constructional Variation and Change’. Diachronica 30(3): 382–419.
References 379 Yáñez-Bouza, Nuria. 2011. ‘ARCHER Past and Present (1990–2010)’. ICAME Journal 35, 205–36. Zaenen, Annie, Carlette, Jean, Garretson, Gregory, Bresnan, Joan, KoontzGarboden, Andrew, Nikitina, Tatiana, O’Connor, M. Catherine, and Wasow, Tom. 2004. ‘Animacy Encoding in English: Why and How’. In: Byron, Donna and Webber, Bonnie (eds.). Proceedings of the 2004 ACL Workshop on Discourse Annotation, Barcelona, July 2004, 118–25. Zwicky, Arnold M. 1987. ‘Suppressing the Zs’. Journal of Linguistics 23(1), 133–48.
Author index
Aarts, Bas, 13 Ackema, Peter, 92, 293 Algeo, John, 39, 42, 46, 209 Amador-Moreno, Carolina P., 334, 341, 344 Andersen, Henning, 113 Anderwald, Lieselotte, 327 Arnaud, René, 13 Auer, Anita, 2, 52 Bailey, Richard W., 326 Beal, Joan, 1, 13, 52 Behaghel, Otto, 138 Berlage, Eva, 208, 209, 223 Biber, Douglas, 16, 40, 57, 78, 137, 173, 222, 224, 243, 295, 296, 312, 315, 316, 317, 320, 322, 328 Bolinger, Dwight, 119, 239, 251 Breban, Tine, 114, 115, 117, 119, 120, 122, 127, 129, 131 Brems, Lieselotte, 114, 239, 240, 241, 242, 244, 254 Bresnan, Joan, 133, 134, 183 Brinton, Laurel, 1, 207, 209, 220, 221, 271, 274, 275, 283, 287, 291, 294, 295, 303, 304, 305, 333 Brorström, Sverker, 16, 333, 336, 344, 345, 346 Bybee, Joan, 83, 85, 88, 89, 132, 284 Chafe, Wallace, 292 Chomsky, Noam, 183 Claridge, Claudia, 209, 240, 244, 252, 259, 272, 273, 274, 275, 278, 316, 327 Coates, Jennifer, 79, 326 Collins, Peter, 40, 56, 57 Croft, William, 77, 89, 184, 231 Crystal, David, 294 D’Arcy, Alexandra, 73 Davidse, Kristin, 100, 119 De Smet, Hendrik, 113, 128, 130, 131, 132, 183, 184, 185, 187, 201, 225, 227, 228, 230, 232, 235
380
Declerck, Renaat, 182, 224, 231, 236 Dehé, Nicole, 275, 291, 293, 294, 304 Denison, David, 1, 13, 14, 46, 50, 76, 114, 129, 185, 202, 203, 227, 304, 333 Diewald, Gabriele, 127 Dixon, Robert M.W., 90, 92, 183, 231 Duffley, Patrick, 183 Fagan, Sarah M.B., 92 Fanego, Teresa, 184, 226 Felser, Claudia, 183 Filppula, Markku, 334, 338, 340, 341, 342, 343 Finegan, Edward, 16, 312 Fischer, Olga, 184, 200, 201, 304 Fries, Charles Carpenter, 78 Gisborne, Nikolas, 231, 292, 297, 299, 300 Goldberg, Adele, 77, 92, 107 Görlach, Manfred, 18, 326, 333, 339 Gray, Bethany, 222 Gries, Stefan T., 132, 134 Harris, John, 334, 341, 342 Heine, Bernd, 127, 148, 274, 275, 285, 286 Heyvaert, Liesbet, 100, 230, 235 Hickey, Raymond, 334, 339, 342, 344 Hoffmann, Sebastian, 132, 209 Hoffmann, Thomas, 288, 289 Hopper, Paul J., 40, 41, 88, 122, 129, 132, 155, 157, 200, 241, 248, 284 Hundt, Marianne, 9, 13, 92, 93, 94, 95, 96, 97, 107, 134, 137, 150, 173, 209, 241, 245, 304, 311, 316, 320, 324, 325, 327, 328, 329, 336 Jankowski, Bridget, 73, 137 Jespersen, Otto, 78, 221, 224 Kallen, Jeffrey L., 334, 341, 342, 343 Kaltenböck, Gunther, 274, 275, 291, 293, 294, 295, 304, 305, 306 König, Ekkehard, 127, 132, 207 Kortmann, Bernd, 207, 298, 333, 338, 341
Author index 381 Krug, Manfred, 40, 41, 43, 44, 56 Kytö, Merja, 1, 2, 13, 14, 16, 17, 186, 221, 240, 243, 244, 252, 259, 333, 336, 338, 344, 345, 346, 351 Labov, William, 135, 229 Langacker, Ronald W., 122, 183, 185 Leech, Geoffrey, 56, 171, 222 Lehmann, Christian, 113, 130, 199, 200 Leitner, Gerhard, 13 Lemmens, Maarten, 93, 94, 101, 106 Levin, Beth, 92, 94, 107 López-Couso, María José, 13, 291, 292, 295, 298, 300, 301, 307, 320 Mair, Christian, 9, 56, 57, 59, 60, 75, 179, 184, 311, 312, 316, 320, 324, 325, 329 McFadden, Thomas, 14, 333 Melchers, Gunnel, 338, 339 Méndez-Naya, Belén, 13, 243, 292, 298, 300, 301, 307 Miller, Jim, 74, 184 Nesselhauf, Nadja, 77, 79, 81, 82, 83, 84, 85, 87, 88, 89 Noël, Dirk, 88, 184 Palmer, Frank R., 78, 91 Pullum, Geoffrey K., 40, 41, 42, 61, 78, 79, 85, 131, 209, 217, 221, 224, 228, 272, 274, 275, 284, 291, 292, 293, 294, 297 Rohdenburg, Günter, 155, 157, 169, 173, 175, 178, 184, 185, 198, 199, 202, 203
Rosenbach, Anette, 133, 137, 138, 144, 146, 147, 148, 149, 150 Ross, John R., 210, 211, 215 Rudanko, Juhani, 157, 184 Rydén, Mats, 2, 13, 16, 25, 333, 336, 344, 345, 346 Schlüter, Julia, 173 Searle, John R., 156, 163 Seoane, Elena, 342 Smitterberg, Erik, 13, 316, 317, 318, 320, 325, 327 Straaijer, Robin, 13, 34 Szmrecsanyi, Benedikt, 133, 134, 135, 137, 144, 145 Tagliamonte, Sali A., 73 Tieken-Boon van Ostade, Ingrid, 1, 13, 52, 333 Traugott, Elizabeth C., 40, 41, 42, 88, 114, 120, 127, 129, 130, 132, 200, 207, 209, 220, 225, 238, 239, 241, 242, 244, 249, 253, 258, 259, 283, 288, 289, 295, 333, 335 Trousdale, Graeme, 114, 130, 219, 220, 221, 225, 238, 288, 289 Trudgill, Peter, 57, 62, 338 Visser, Frederikus T., 14, 46, 78, 79, 207, 226, 228, 232, 327, 328, 333 Vosberg, Uwe, 6, 155, 157, 184 Walshe, Shane, 345 Warner, Anthony, 183 Wierzbicka, Anna, 183, 302
Subject index
acceptability judgements, 92, 95 active – passive – neuter, 20, 21, 24, 25, 29, 30, 35 actualisation, 113, 127–32 adhortative let’s construction, see construction American English (AmE), 4, 6, 10, 16, 17, 32, 35, 39, 41, 42, 43, 44, 46, 47, 48, 49, 54, 55, 57, 59, 61–73, 74, 75, 133, 134, 136, 141, 144, 149, 152, 159, 160, 163, 170, 171, 172, 173, 174, 175, 176, 178, 179, 235, 280, 291, 302, 336, 351 analogy, 84, 130, 132, 258, 259 anchor clause, 9, 291, 293, 294, 303, 306, 307 animacy, 6, 134, 135, 137, 140, 144, 145, 146, 147, 148, 149, 150, 151, 152, 198 apodosis, 273, 275, 281, 282, 284, 285–86, 287, 288, 289, 290 then-clause, 285–86 ARCHER, 17, 43, 44, 45, 46, 47, 80, 82, 134, 135, 138, 139, 140, 144, 150, 151, 292, 296–306, 307, 333, 338 attach, 96, 102–3, 109 binominal, 8, 114, 120, 121, 242, 252, 255, 256, 257, 258 BNC, 73, 86, 92, 95, 97, 102, 131, 155, 159, 160, 168, 172, 212, 214 British English (BrE), 4, 10, 39, 42, 43, 44, 46, 47, 48, 49, 54, 55, 57, 62, 63, 73, 74, 135, 141, 144, 152, 160, 163, 170, 171, 172, 173, 174, 175, 176, 178, 235, 280, 300, 313, 336, 338, 342, 351 Brown family corpora, 171, 314 Cambridge Grammar of the English Language, 291 CB, 228, 236
382
CED, 239 CEMET, 233, 234 CEN, 44, 47, 50 clamp on, 93, 96, 103–4, 109 CLMET, 2, 5, 6, 7, 44, 46, 124, 190, 280 CNG, 19, 23, 25, 34 CNNE, 313–14, 315, 316, 317, 318, 319, 320, 321, 322, 323, 324, 325, 326, 327, 328, 329 COCA, 57, 61, 62, 63–72, 91, 95, 96, 103, 104, 105, 155, 160, 271, 280 cognitive complexity, 7, 185, 198 COHA, 2, 16, 17, 32, 44, 48, 49, 50, 52, 54, 55, 60, 61, 62, 91, 95, 96, 97, 101, 102, 103, 104, 106, 280 cohesiveness, 7, 207, 208, 209–11, 212, 214, 218, 219, 220 collocation, 71, 93, 103, 121, 132, 212, 220 collocational, 5, 113, 114, 115, 116, 120, 122, 128, 129, 130, 241, 242 collocational distribution, 129, 130, 132 colloquialisation, 311, 312–13, 315, 316, 318, 319, 320, 321, 322, 323, 324, 328 competing gerunds, see gerund: rivalling gerunds complementation, 6–7, 17, 155, 179, 183–84, 191–98, 199, 200, 202, 203, 209, 226, 229, 293 complement, 8, 48, 50, 52, 163, 169–73, 182, 184, 185, 189, 200, 201, 202, 207, 208, 209, 210, 211, 212, 213, 214, 215–19, 220, 223, 244, 272, 273, 284, 286, 287, 290, 291, 296, 299, 304, 305, 334, 335, 349 complement clause (CC), 6, 182, 183, 184, 185, 186, 187, 188, 191–98, 199, 200, 201, 202, 203, 237, 293, 297, 298 complement construction, 210, 225, 229–37, 238 complement structure, 238, 285, 288, 290, 293 that complementation, 6, 155, 156, 157, 158, 159, 160, 162, 163, 164,
Subject index 383 165, 166, 167, 168, 169, 170, 171, 172, 175, 176, 177, 178, 179, 183, 184, 192, 199, 201, 202 composite predicates, 7, 207, 208, 209, 220, 222, 223 CONCE, 314–15, 317, 318, 320, 329 conceptual change, 145, 147–48, 152 conditional meaning, 273, 274, 275, 283 conjunctional force, 178 conjunctions complex, 6, 173 emergent, 173, 178 established, 173, 178 construal, 97, 102, 103, 104, 106, 119, 123, 127, 129, 131 construction, 77 adhortative let’s, 40–43, 48 complement, see complementation gerundial, see gerund mediopassive, 5, 90, 91, 92–95, 96, 97, 98, 99, 100, 101, 102, 103, 104–8, 109 middle, 92 NP of NP, see NP of NP construction Construction Grammar, 4, 5, 7, 88, 91, 95, 107, 108, 241, 288–90 constructionalisation, 5, 8, 84, 285, 286, 288, 289, 290 context contextual, 5, 113, 114, 115, 116, 117, 118, 119, 120, 122, 128, 129, 130, 155 contextual distribution, 128, 129, 130, 132 contraction, 52, 54, 77, 78, 84, 87, 88, 179, 311, 328 ‘ll-contraction, 77, 78, 79, 80, 81, 82, 84, 87, 88 see also ‘ll not-contraction, 320–21, 324, 328 co-ordination, 321, 322, 323, 324, 326, 328, 329 CORIECOR, 334, 335, 338, 340, 341, 345–50 decategorialisation, 200, 219, 220, 259 degree modifier, see modifier delexicalisation, 7, 208, 219–20, 221, 222, 223 deny, 183, 184, 185, 188, 189, 191, 192, 193, 201, 202 detransitivisation, 90, 155, 157, 158, 166, 168, 169 Dictionary of American Regional English (DARE), 58, 59 directives, see verbs distribution, 119, 128, 130, 148, 191–98, 311, 322, 323, 329 distributional, 127, 129, 134, 184, 227, 231
distributional behaviour, 128, 130, 131, 132 don’t let’s, 39, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 54, 55 do-support, 4, 57, 58, 59–61, 62, 63–72, 73, 74, 75, 98 EAF, 159, 171, 175 ECCO, 44, 52 ECF, 169, 211, 212, 214, 280, 284 economy of expression, 208, 223 ED, 280, 281, 284, 285, 286 EEPF, 160, 167, 168, 176 eighteenth century, 2, 8, 83, 102, 159, 160, 166, 170, 172, 176, 179, 213, 214, 232, 233, 242, 258, 281, 302, 327, 333, 336, 343, 344, 345, 347, 348, 349, 350, 351 EPD, 180 epistemic, 8, 9, 291, 292, 296, 299, 302, 307 epistemicity, 292 ergative, 90, 97, 103, 108, 109 ETC, 180 ethnic variety, 4, 61, 62, 75 evaluation, 252, 254, 294, 326, 327, 328, 329 evidential, 9, 291, 292, 296, 299, 300, 302–3, 307 evidentiality, 292 expressives, see verbs fossilisation, 220 FRED, 229 frequency, 6, 17, 39, 65, 93, 96, 109, 131, 135, 144, 151, 236, 244, 245, 248, 272, 280, 284, 289, 300, 315, 316, 317, 318, 319, 326, 327 change in, 24, 61, 74, 85, 94, 103, 104, 147, 151, 166, 168, 169, 178, 245, 302, 312, 316, 318, 319, 321, 327, 328 discourse, 8, 149, 150 distribution, 185, 194, 198, 201, 203 high, 85, 89, 284, 295, 314, 317, 328 low, 4, 8, 55, 93, 96, 104, 242, 259, 289, 300, 317 mid, 314 relative, 17, 73, 84, 96, 97, 98, 125, 128, 329 text, 95, 100, 109, 168 future marker, 87 genitive, 133, 134, 137–40, 141, 144, 146, 147, 150, 151, 226 genitive alternation, 5, 133, 134, 135–36, 137–40, 145, 149, 152 of-genitive, 133, 135, 137, 138, 139 s-genitive, 6, 133, 135, 137, 138, 139, 141, 144, 145, 146, 147, 148, 149, 150, 151
384 Subject index genitive (cont.) genre, 3, 5, 9, 44, 45, 52, 82, 88, 93, 96, 104, 107, 109, 186, 187, 190, 198, 214, 260, 311, 312, 313, 315, 316, 317, 318, 320, 321, 326, 328, 329 gerund, 7, 156, 158, 162, 178, 179, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 235, 236, 237, 238 gerundial complement, 229, 230, 231, 233, 234, 235, 237 gerundial construction, 156, 159, 179, 235 prepositional gerunds, 157, 173–76 rivalling gerunds, 156, 157, 158, 160, 173, 176 grammar writing, 13, 15, 18, 21, 25, 29, 31, 33, 35 grammaticalisation, 3, 4, 5, 6, 7, 8, 14, 40–43, 59, 71, 74, 75, 85, 88, 89, 113, 114, 122, 129, 130, 132, 147, 169, 185, 199, 200, 201, 202, 207, 239, 240, 241, 242, 245, 246, 248, 252, 258, 259, 288, 295, 303, 307 HC, 227, 239, 292, 300, 301, 333 hedging, 20, 240, 246, 249, 252, 254, 255, 259, 272, 278 hedge, 8, 239, 242, 247, 252, 253, 272, 276, 278 if-elliptical clause, 8, 273, 274, 275–90 inanimate subjects, 167, 168, 169, 177, 178 indirect conditional, 274, 275–90 inheritance, 95, 107 intransitive, see transitivity Irish, 9, 10, 334, 335, 339–41, 351 Irish English (IrE), 9, 10, 333, 334, 335, 336, 337, 338, 339, 340, 341–45, 346, 347, 351 language contact, 3, 9, 10, 60, 334 LC, 44, 239 length, 138, 141, 144 let’s don’t, 38, 39, 42, 44, 46, 48, 49, 50, 52, 54, 55 letters, 16, 82, 134, 150, 151, 315, 318, 338, 344 lexicalisation, 7, 207, 208, 219–20, 222 linguistic change, 145, 146, 147, 150, 152, 329 LION, 44, 46, 47, 49, 54, 55 literacy, 313, 325 ‘ll, 77, 79, 80–88, 89 shall, 77, 78, 80, 82, 85, 87 will, 77, 79, 80, 83, 84, 85, 87, 88 LNC, 171, 180 mediopassive, see construction metalinguistic function, 8, 271, 272, 275, 276–79, 280, 281, 282, 283, 284, 285, 286, 287, 288, 289, 290
middle, see construction modality, 56, 57, 75, 100, 109, 202, 289 modifiability, 210, 211, 215 modification constraint, 5, 91, 92–95, 104, 107–8 modifier, 100, 117, 118, 119, 129, 192, 197, 211, 215, 226, 241, 244, 245, 250–52, 257, 305 degree, 243, 244, 259 NCF, 167, 169, 175, 180 newspaper language, 222, 311, 313, 316, 319, 320, 322, 323, 324, 328 nineteenth century, 13–15, 17, 18, 21, 27, 31, 33, 35, 60, 75, 88, 94, 100, 107, 144, 148, 158, 160, 165, 166, 168, 169, 171, 172, 173, 174, 176, 178, 179, 245, 251, 258, 281, 287, 312, 325, 326, 327, 328, 338, 345, 348, 349, 350, 351 non-finite clause, 175, 192, 200, 202, 224, 226, 238, 297 non-standard English, 58, 74, 345 NP of NP construction, 240, 244, 245, 246–50, 257, 258 NP status, 209, 210, 211, 212, 214, 215, 216, 217, 218 OBC, 6, 8, 45, 46, 47, 184, 185, 186, 190, 239, 240, 242, 244, 249, 250, 253, 258 OBP, 45, 280, 281, 285 OED, 41, 42, 43, 44, 78, 83, 85, 102, 105, 106, 114, 124, 155, 157, 165, 175, 208, 239, 240, 271, 272, 280 paradigmaticisation, 113, 127–32 parenthetical adverbial parenthetical, 298, 302, 303, 304, 307 bare parenthetical, 9, 298, 300–2, 303, 304, 307 partitive, 129, 246, 247, 252, 253, 255, 257, 258, 259 passive, 19, 20, 24–25, 29, 31, 33, 34, 35, 163, 164, 165, 166, 167, 169, 170, 171, 172, 173, 177, 178, 210, 211, 212–15, 220, 227, 232, 312, 339, 340 perfect, 14, 15, 17, 25, 33–35, 169, 341, 342, 343 be-perfect, 13, 14, 15–18, 35, 333, 334, 335–50, 351 have-perfect, 10, 14, 15, 17, 27, 32, 34, 334, 335–45 persistence, 122, 131, 248, 259 phrasal verbs, see verbs politeness function, 272, 276–79, 283, 287, 290
Subject index 385 PPCEME, 227 PPCMBE, 2, 5, 114, 115, 116–27, 129 prescription, 13, 39, 52, 325, 326, 328, 329, 336 present participle, 7, 224, 225, 226, 227, 228, 229, 230, 231, 232, 233, 234, 237, 238 probabilistic analysis, 191–98 grammar, 137, 150 variation, 182, 184, 202 progressive, 57, 91, 150, 229, 231, 315, 316–19, 324, 327, 328 proscribed feature, see prescription psychological event verbs, see verbs quantification, 256, 259 quantifier, 5, 8, 113, 114–15, 116, 117, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 131, 132, 241, 242, 245, 246, 253, 255, 256, 258 quantifying, 240, 252, 253, 254, 255, 258 raising constructions, 169, 170, 171, 172, 173, 296, 300, 307 structures, 156, 169–73, 178 subject-to-object, 156 reanalysis, 5, 73, 74, 84, 89, 113, 120, 127, 128, 130, 131, 132, 258 reflexive, 157, 158, 163, 177, 233 objects, 158, 168, 169, 178 pronoun, 90 pronouns, 90, 158, 167, 169, 178 verbs, 102, 160 regression, 135, 138, 140–45, 146, 152, 185, 194, 195, 198 logistic regression, 134, 140, 194, 202 regret, 182, 183, 184, 185, 188, 191, 192, 193, 201, 202 remember, 182, 183, 184, 185, 187, 191, 192, 193, 201, 202 representatives, see verbs Scots, see Scottish English (ScotE) Scottish English (ScotE), 333, 336, 338–39 Sears & Roebuck corpus, 93, 94, 95, 97, 98, 100
seem, 9, 291, 292, 293, 296–306, 307 sell, 90, 92, 93, 96, 97, 98, 99, 101–2, 108 semi-auxiliary, 57 Strathy, 273, 275, 276, 277, 278, 279, 280 stylistic ideals, see prescription subjectification, 242, 287, 295 subjectivity, 242, 246, 252, 256, 283 subjunctive, 41, 169, 170 mandative subjunctive, 6, 173, 178 syntactic change, 1, 2, 9, 13, 113, 128, 129, 130, 219, 222, 288 syntactic merger, 225 text type, see genre that-clause, see that complementation TIME, 105, 106 to-infinitive, 169, 182, 183, 184, 186, 191, 199, 200, 201, 284 transitivity, 6, 94, 102, 155, 157, 177, 335 see also detransitivisation features, 155, 158, 160, 167, 169, 177, 178 high, 178 intransitive, 14, 19, 20, 21, 22, 25, 29, 30, 31, 33, 35, 108, 109, 155, 165, 166, 167, 168, 334, 344 low, 158, 160, 167, 169, 177, 178 mutative intransitive verbs, 9, 14, 35, 333, 343, 345, 351 transitive, 10, 14, 19, 21, 23, 25, 29, 30, 31, 35, 57, 90, 155, 156, 168, 275, 334, 335, 343, 351 twentieth century, 1, 4, 6, 57, 75, 79, 82, 83, 87, 94, 98, 100, 102, 103, 104, 106, 107, 159, 165, 170, 171, 172, 174, 175, 176, 213, 214, 281, 311, 325 verbs directive, 6, 155, 156, 163, 169–73, 176, 178, 179 expressive, 6, 155, 156–62, 176, 179 mutative intransitive, see transitivity phrasal, 311, 315, 316–19, 324, 327, 328 psychological event, 94, 95, 96, 97, 104–7, 108 reflexive, see reflexive representative, 6, 155, 162–69, 176, 178, 179
