The Syntax of Nonsententials
Linguistik Aktuell/Linguistics Today Linguistik Aktuell/Linguistics Today (LA) provide...
50 downloads
1072 Views
3MB Size
Report
This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form
The Syntax of Nonsententials
Linguistik Aktuell/Linguistics Today Linguistik Aktuell/Linguistics Today (LA) provides a platform for original monograph studies into synchronic and diachronic linguistics. Studies in LA confront empirical and theoretical problems as these are currently discussed in syntax, semantics, morphology, phonology, and systematic pragmatics with the aim to establish robust empirical generalizations within a universalistic perspective.
Series Editors Werner Abraham
Elly van Gelderen
University of Vienna
Arizona State University
Advisory Editorial Board Cedric Boeckx
Ian Roberts
Harvard University
Cambridge University
Guglielmo Cinque
Ken Safir
University of Venice
Rutgers University, New Brunswick NJ
Günther Grewendorf
Lisa deMena Travis
J.W. Goethe-University, Frankfurt
McGill University
Liliane Haegeman
Sten Vikner
University of Lille, France
University of Aarhus
Hubert Haider
C. Jan-Wouter Zwart
University of Salzburg
University of Groningen
Christer Platzack University of Lund
Volume 93 The Syntax of Nonsententials: Multidisciplinary perspectives Edited by Ljiljana Progovac, Kate Paesani, Eugenia Casielles and Ellen Barton
The Syntax of Nonsententials Multidisciplinary perspectives
Edited by
Ljiljana Progovac Kate Paesani Eugenia Casielles Ellen Barton Wayne State University
John Benjamins Publishing Company Amsterdam/Philadelphia
8
TM
The paper used in this publication meets the minimum requirements of American National Standard for Information Sciences – Permanence of Paper for Printed Library Materials, ansi z39.48-1984.
Library of Congress Cataloging-in-Publication Data The syntax of nonsententials : multidisciplinary perspectives / edited by Ljiljana Progovac, ... [et al.]. p. cm. (Linguistik Aktuell/Linguistics Today, issn 0166–0829 ; v. 93) Includes bibliographical references and index. 1. Grammar, Comparative and general--Syntax. 2. Language acquisition. 3. Pidgin languages. 4. Creole dialects. I. Progovac, Ljiljana. P291.S9577 2006 417´.22--dc22 isbn 90 272 3357 8 (Hb; alk. paper)
2006042931
© 2006 – John Benjamins B.V. No part of this book may be reproduced in any form, by print, photoprint, microfilm, or any other means, without written permission from the publisher. John Benjamins Publishing Co. · P.O. Box 36224 · 1020 me Amsterdam · The Netherlands John Benjamins North America · P.O. Box 27519 · Philadelphia pa 19118-0519 · usa
JB[v.20020404] Prn:16/08/2006; 12:42
F: LA93CO.tex / p.1 (48-105)
Table of contents
Preface
ix
Introduction Ljiljana Progovac, Kate Paesani, Eugenia Casielles, and Ellen Barton 1. Background: Defining and describing nonsententials 1 2. Goals of the current volume 4 3. Volume overview 6 chapter 1 Toward a nonsentential analysis in generative grammar Ellen Barton 1. Introduction 11 2. Early arguments for an ellipsis analysis 12 3. Early arguments for a nonsentential analysis 17 4. Mixed analyses 21 5. Data and sentential/nonsentential analyses in Minimalism 6. Conclusion 27
11
25
chapter 2 The syntax of nonsententials: Small clauses and phrases at the root Ljiljana Progovac 1. Introduction 33 2. Outline of the proposed analysis 37 3. Sentence and Tense 42 4. Noun Phrase, Determiners, and Case 45 5. Small clauses in current syntactic theory 52 6. Tense, Time, assertion, and the role of context 54 7. Conclusion 61 chapter 3 “Small structures”: A sententialist perspective Jason Merchant 1. Introduction 73 2. Fragment answers 74
1
33
73
JB[v.20020404] Prn:16/08/2006; 12:42
F: LA93CO.tex / p.2 (105-146)
Table of contents
3. Discourse-initial (?) fragments 4. Conclusion 89
83
chapter 4 Neither fragments nor ellipsis Robert J. Stainton 1. Introduction 93 2. Merchant’s proposal and its advantages 95 3. Empirical objections 103 4. A methodological objection: “Fragments” and simplicity 113 5. Concluding remarks 114 chapter 5 Big questions, small answers Eugenia Casielles 1. Introduction 117 2. The elliptical view 119 3. The nonelliptical view 122 4. The sluicing view 125 5. Toward a phrasal analysis of small answers 6. Conclusions and further issues 134
93
117
130
chapter 6 Extending the nonsentential analysis: The case of special registers Kate Paesani 1. Introduction 147 2. An overview of special registers 149 3. The (non)sentential nature of special registers 157 4. The syntax of nonsententials in special registers 159 5. Expressing time in special registers 167 6. Conclusion 172 chapter 7 The narrowing acquisition path: From expressive small clauses to declaratives Christopher Potts and Tom Roeper 1. Expressive small clauses 183 2. Small clauses in adult English 185 3. Small clauses in child English 186 4. A note on expressive content 187 5. When speaking expressively, we’re all children 188 6. Self-directed disapprobation cross-linguistically 189 7. The acquisition of expressive small clauses 193 8. Analysis of self-disapprobation small clauses 194
147
183
JB[v.20020404] Prn:16/08/2006; 12:42
F: LA93CO.tex / p.3 (146-186)
Table of contents
9. Analysis of incredulity small clauses 198 10. Summary and conclusions 199 chapter 8 Nonsententials and second language acquisition Nicola Work 1. Introduction 203 2. Background 204 3. Early L2 acquisition data: The Basic Variety 206 4. Learner language and native adult nonsententials 211 5. Nonsentential grammar and theories of L2 acquisition 219 6. Teacher talk 221 7. Conclusion 223 chapter 9 How language adapts to the brain: An analysis of agrammatic aphasia Herman Kolk 1. Overview 229 2. Variability of symptoms 230 3. Theories of agrammatism as a variable phenomenon 232 4. Evidence for the temporal-window hypothesis 234 5. Adaptation 237 6. Conclusion 251 chapter 10 Nonsententials and agrammatism Patricia Siple 1. Introduction 259 2. Syntactic deficits associated with agrammatic production 260 3. Theoretical accounts of agrammatic production 263 4. Extending Kolk: Syntactic structure and adaptation theory 268 5. Conclusion 277 chapter 11 Reduced syntax in (prototypical) pidgins Donald Winford 1. Introduction 283 2. Structural characteristics of pidgins 286 3. Processes of pidginization 288 4. Pidgins and second language acquisition 292 5. Stages of development in pidginization and early IL 295 6. Conclusion 304
203
229
259
283
JB[v.20020404] Prn:16/08/2006; 12:42
F: LA93CO.tex / p.4 (186-212)
Table of contents
chapter 12 Copula variation in Guyanese Creole and AAVE: Implications for nonsentential grammar Walter Edwards 1. Introduction 309 2. The copula in Rural Guyanese Creole 309 3. Urban Guyanese Creole and the present-tense copula 311 4. Copula variation in Urban Guyanese Creole 311 5. Guyanese Creole and the nonsentential analysis 314 6. Copula variation in AAVE 315 7. AAVE, Guyanese Creole, and the nonsentential analysis 318 8. Discussion 319
309
Epilogue: Wherefrom and whereto? Ljiljana Progovac, Kate Paesani, Eugenia Casielles, and Ellen Barton 1. Nonsententials, individual variation, and multiple grammars 323 2. Why nonsententials: A historical perspective 328 3. Elliptical versus nonsentential approaches 333 4. Some remaining issues and future prospects 343
323
Index of languages
355
Index of names
357
Index of subjects
361
JB[v.20020404] Prn:19/07/2006; 14:17
F: LA93PR.tex / p.1 (46-117)
Preface
The Editors of this volume would like to express their gratitude to everyone involved in this project. It has been a most rewarding intellectual journey. This volume is the culmination of 4 years of collaborative effort on the part of the Working Group on Nonsententials, funded by the Wayne State University Humanities Center. We would especially like to thank Walter Edwards, Director of the Humanities Center, for this generous support. We are also grateful to several programs and departments at Wayne State University for additional support, including the Linguistics Program, the Department of Romance Languages and Literatures, and the Department of English. The Working Group on Nonsententials brought together a number of Wayne State University faculty and graduate students, listed here in alphabetical order: Ellen Barton, Eugenia Casielles, Walter Edwards, Kate Paesani, Ljiljana Progovac, Patricia Siple, and Nicola Work. The Working Group held two workshops in 2003 (Small Structures: Sentential and Nonsentential Analyses) and 2004 (Small Utterances: Views on Syntax, Acquisition, Aphasia, and Pidgins), which featured presentations from members of the group as well as invited speakers. We would like to thank the following colleagues for their contributions to the workshops: Herman Kolk, University of Nijmegen; Jason Merchant, University of Chicago; Thomas Roeper, University of Massachusetts; Robert Stainton, University of Western Ontario; and Donald Winford, The Ohio State University. This volume benefited from an extensive review process. Michel DeGraff reviewed the entire volume, and his suggestions were invaluable in sharpening the analyses presented here. We thank him here for what was a monumental effort. We also thank additional reviewers, including Alan Beretta, Rita Berndt, Anthony Grant, Lisa Green, Liliane Haegeman, Barbara Partee, Julia Herschensohn, Caroline Heycock, Ruth Kempson, Eric Potsdam, and Tim Stowell. The Series Editors of Linguistics Today, Elly van Gelderen and Werner Abraham, as well as Kees Vaes at John Benjamins were most helpful in bringing this volume to press. Thanks also go to Rachel Anderson for her work on copyediting and formatting. Finally, we would like to thank all of the authors in the volume for working with us in this exploration of the syntax of nonsententials.
JB[v.20020404] Prn:19/07/2006; 14:24
F: LA93IN.tex / p.1 (46-114)
Introduction Ljiljana Progovac, Kate Paesani, Eugenia Casielles, and Ellen Barton
.
Background: Defining and describing nonsententials
This volume brings together experts from syntax, semantics, pragmatics, philosophy of language, language acquisition, aphasia, and pidgin and creole languages to examine the structure and interpretation of utterances that are smaller than a sentence. These utterances are considered first from the perspective of core adult data from conversational speech and question-answer pairs. The syntactic analysis developed to account for these data, couched within the theoretical framework of Minimalism (Chomsky 1995), is then extended to address data from a wide variety of contexts, including special registers, child language, nonnative varieties, impaired language, pidgins, and creoles. The traditional term for an utterance smaller than a complete sentence has been fragment (Sweet 1900; Quirk, Greenbaum, Leech, & Svartvick 1972). This term, however, implies a particular analysis, namely that structures smaller than a sentence are derived via ellipsis from a full-sentence source, an analysis that we will call here a sentential analysis. A competing term is nonsentential, used more recently in the literature (Barton 1990, 1991; Barton & Progovac 2005). The term nonsentential also implies a particular analysis, namely that structures smaller than a sentence are derived via base generation, an analysis that we will call here a nonsentential analysis (see Barton, Chap. 1, this volume for a detailed discussion of these competing analyses; see also Note 1 for a brief discussion of terminology). As the title of this volume suggests, the chapters herein examine nonsentential utterances from multidisciplinary perspectives and many of these chapters pursue a nonsentential analysis.1 Grammarians and linguists have not always considered the structure of nonsentential utterances an important question. In the traditional grammar literature, for example, Follett (1966: 326) remarks that such structures are “only a part masquerading as a whole.” Some generative linguists have been equally dismissive. Matthews (1981: 14), for instance, claims that structures smaller than a sentence “are of no concern to syntax, except as a source of confusion in our data.” In a recent article in the journal Language, Newmeyer
JB[v.20020404] Prn:19/07/2006; 14:24
F: LA93IN.tex / p.2 (114-188)
Ljiljana Progovac et al.
(2003: 690) also defends a similar view, stating that “speakers mentally represent the full grammatical structure, even if they utter only fragments.” But other grammarians and linguists have long been intrigued by the question of the structure underlying nonsentential utterances, as they appear in adult discourse such as casual conversation, (1) and (2), and question-answer pairs, (3) and (4). (1) Problem solved. (2) Me first. (3) What does John do? Play baseball. (4) Who ate the cake? Him. Historically and traditionally, utterances like (1) through (4) have been called fragments and have been presumed to derive via ellipsis from full-sentence sources. In this approach, the answer to the question in (5) would be analyzed as deriving from a full sentential source such as (6). (5) Which movie did you see? Casablanca. (6) I saw Casablanca. A problem that the ellipsis analysis faces, however, is the fact that none of the examples in (1) through (4) can be straightforwardly derived by ellipsis, as shown in (7) through (10): (7) *Problem is solved. (8) *Me am first. (9) *John play baseball. (10) *Him ate the cake. However, this volume includes a recent and innovative elliptical analysis that may be able to account for some of these data. Merchant (2004, Chap. 3, this volume) proposes a move-and-sluice analysis that insures that deletion happens only once and that it affects only single constituents. Two authors in the volume respond to Merchant’s elliptical analysis: Casielles, regarding his analysis of question-answer pairs; and Stainton, regarding his analysis of so-called limited ellipsis. In addition, this issue is revisited in the Epilogue. Yet the rest of the volume is not set up as a response to Merchant’s approach, or to other elliptical approaches. Instead, it is devoted to building a coherent nonelliptical analysis of nonsententials. Indeed, an alternative to the ellipsis analysis is to consider the utterances in (3) and (4) as base-generated phrases and those in (1) and (2) as base-generated small clauses. Under this analysis, NP, VP, PP, AdjP, and other phrases and small clauses are built from the bottom up, and do not contain a syntactic Tense node. Thus, utterances
JB[v.20020404] Prn:19/07/2006; 14:24
F: LA93IN.tex / p.3 (188-250)
Introduction
that appear to be smaller than a sentence are in fact smaller than a sentence and do not involve ellipsis from a full-sentence source. Moreover, this nonelliptical approach can be extended to account for similar data in a variety of contexts and registers. Previous research on nonsentential utterances – both elliptical and nonelliptical – has looked at this type of adult English data in isolation. Yet this volume shows that similar utterances are regularly attested in a variety of other contexts, including special registers such as headlines (11), first language acquisition (12), second language acquisition (13), agrammatism (14), pidgin languages (15), and creoles (16). (11) Proof in the pudding. (12) Me big. (13) Girl go. (14) Me forget. (15) As handrid piptin dala ei dei. us hundred fifteen dollar a day ‘We [made] a hundred and fifteen dollars a day.’ (16) Country people more nice. Moreover, such nonsententials are attested cross-linguistically – in adult speech, and in the same contexts illustrated in (11) through (16). (17) Spanish question-answer pair: a. ¿Qué hace Juan? ‘What does John do?’ b. Jugar al béisbol. play-inf part-det baseball ‘Play baseball.’ (18) French headline: Risque d’avalanche maximal ce weekend. ‘Avalanche risk maximal this weekend.’ (19) French child language: Pas manger la poupée. ‘Not eat the doll.’ (20) Dutch second language: Meisje weg. ‘Girl gone.’ (21) Dutch aphasia: Hersenen niet goed. brain no good ‘My brain is not good.’
JB[v.20020404] Prn:19/07/2006; 14:24
F: LA93IN.tex / p.4 (250-325)
Ljiljana Progovac et al.
(22) Russernorsk pidgin: Russman bra mann. Russian-man good man ‘The Russian is a good man.’ The grammatical structure underlying nonsentential utterances, then, may be a more important question than it would first appear.
. Goals of the current volume This volume explores an empirically grounded and theoretically unified account of the type of nonsentential data presented earlier by bringing together multidisciplinary perspectives from syntax, semantics, philosophy of language, first and second language acquisition, aphasia, pidgins, and creoles. This exploration is motivated by two interrelated goals: (1) to develop a syntactic analysis that captures the core properties of adult nonsentential data in a unified and elegant way and (2) to extend the syntactic analysis to account for similar nonsentential data across contexts, registers, and languages. These goals are detailed next, and are revisited in the Epilogue to this volume. The first goal of this volume is to develop a nonelliptical analysis of core adult nonsentential data, of the type in (1) through (4), within the theoretical framework of Minimalism (Chomsky 1995). These data will be shown to exhibit the following properties: (23) a. Tenselessness b. Caselessness Tenselessness is evidenced through the use of verbless/copulaless clauses, as in (1) and (2), or through the use of tenseless/nonfinite verb forms, as in (3). Caselessness is evidenced through the use of default case forms (accusative in English), as in (2) and (4), or through the omission of determiners on subjects, as in (1). The central claim of the syntactic analysis of these properties is that the grammar base generates nonsentential phrases and root small clauses. Progovac (Chap. 2, this volume) shows that there is a systematic correlation in nonsententials between the omission of tense/agreement on VPs and the omission of nominative case on NPs, which accounts for the data in (1) through (4). Importantly, the correlation accounts not only for the good examples in (1) through (4), but also for the bad examples in (7) through (10). The latter are bad because the correlation does not hold; that is, we have tense but no nominative case, exhibited by the omission of the article or by the accusative form of the subject pronoun, for example. The defining property of nonsententials is thus argued to be the absence of the syntactic Tense node and, with it, all the syntactic phenomena that rely on Tense, including structural nominative case. The theoretical framework of Minimalism provides a syntactic frame in which to capture the nonsentential properties in (23) and the crucial correlation between
JB[v.20020404] Prn:19/07/2006; 14:24
F: LA93IN.tex / p.5 (325-380)
Introduction
Tense and Case, and to consider the base generation of nonsententials. Most research in syntactic theory focuses primarily on the sentence. In Minimalism, it is commonly assumed that what makes a sentence a sentence is the use of a finite verb (Tense), which correlates with the use of the nominative case on the subject. So important is Tense for the formation of the sentence, that current syntactic theory in fact analyzes the sentence as a TP (Tense Phrase). One of the crucial postulations of this framework is the claim that tensed/finite verb forms necessarily engage in a feature-checking/matching relationship with the nominative subject, thus explaining the correlation between Tense/finiteness and nominative case in sentences. Moreover, it is typically assumed that this checking relationship requires a projection of a Determiner Phrase for the subject, as well as a full sentential projection TP to accommodate this feature-checking relationship between the subject and the finite verb in full sentences. However, Progovac (Chap. 2, this volume) maintains that the Minimalist Program also allows another option, an option that does not create a full TP structure as the final or only output of syntax. In this analysis, verbs and subjects are selected from the lexicon with default Tense and default Case features. As such, no Tense/Case feature checking is required, and consequently no sentential (TP) or Determiner Phrase projections are required. The immediate prediction of this analysis is that nonsententials, particularly small clauses, will be characterized by a correlation between the use of nonfinite verbs/predicates (tenselessness) and the absence of structural nominative case assignment to the subject position (caselessness). This correlation is indeed the defining property of nonsententials. The analysis thus brings the nonsentential data, which have often been dismissed as peripheral to human language, straight into the grammar of the linguistic competence of an adult native speaker. The second goal of this volume is to make a strong case for the existence of nonsentential data across contexts, registers, and languages, and to extend the core syntactic analysis of adult nonsentential data to account for them. These data come from special registers (Paesani), child language (Potts and Roeper), second language acquisition (Work), aphasia (Kolk; Siple), pidgins (Winford), and creoles (Edwards), as illustrated in (11) through (22). The aim in this volume is to bring all of these data together under the same analysis, not only to establish the parallelisms among different contexts and languages, but also to provide a theoretical account of such data that applies in a unified fashion to all of the varieties considered. The volume thus approaches the analysis of nonsententials from an innovative perspective, one that is multidisciplinary and multilingual. The implications of this approach to the analysis of nonsententials are far-reaching for generative grammar. In addition to working out an innovative structural analysis of nonsententials, which enhances and sharpens the syntactic theory itself, this research addresses a long-standing puzzle about the possibility of multiple grammars making up linguistic competence in a language (e.g., Roeper 1999). Some linguists working on first language acquisition have proposed that children’s initial grammars are, essentially, nonsentential grammars, based on phrases and root small clauses as initial projections (e.g., Radford 1990, 1996). If this is tenable, then this nonsentential
JB[v.20020404] Prn:19/07/2006; 14:24
F: LA93IN.tex / p.6 (380-437)
Ljiljana Progovac et al.
grammar is not a stage grammar replaced as the sentence grammar emerges. Instead, a nonsentential grammar remains available to both children and adult speakers parallel to the sentence grammar in their linguistic competence. In this view, a nonsentential grammar coexists with the sentence grammar even for normal adult speakers and is the basis for nonsententials in discourse (e.g., question-answer pairs, casual conversation), in special registers (e.g., headlines, recipes), in first and second language acquisition, and in processes of pidginization and creolization. Likewise, this same nonsentential grammar is arguably the grammar to which speakers with aphasia frequently resort, even when they are occasionally capable of producing full sentences.
. Volume overview In this volume, several chapters pursue this analysis of nonsententials and argue for the existence of a nonsentential grammar to account for the structure and interpretation of utterances that are smaller than a sentence. These chapters present an analysis of the syntax and semantics of nonsententials in detail, drawing on a wide variety of cross-linguistic data. Other chapters extend the analysis to describe nonsententials in first and second language acquisition, in agrammatism/aphasia, and in pidgins and creoles. A final Epilogue written by the four editors draws the formal analysis and its extensions together. The first six chapters in the volume focus on the syntactic and semantic analysis of core adult nonsentential data. In Chapter 1, “Toward a Nonsentential Analysis in Generative Grammar,” Barton sets the stage for this analysis by critically reviewing competing approaches (elliptical, nonelliptical, mixed) to the analysis of structures smaller than a sentence, and discussing crucial data for comparing sentential and nonsentential analyses in current research. Chapter 2, “The Syntax of Nonsententials: Small Clauses and Phrases at the Root,” provides the theoretical foundation for the volume. In this chapter, Progovac argues that the defining property of nonsententials is the absence of Tense and any syntactic phenomena that rely on it, including structural nominative case. Due to the absence of Tense, nonsententials are analyzed as root phrases and small clauses. Progovac’s chapter also addresses some semantic and pragmatic implications of this analysis, including frequent nonindicative interpretations among nonsententials, and offers an account that draws on the historical development of the indicative. Merchant provides an alternative to Progovac’s analysis in Chapter 3, “‘Small Structures’: A Sententialist Perspective.” Merchant proposes an ellipsis account for small utterances, arguing that fragments in question-answer pairs result from sentential sources. In this analysis, a constituent is moved to a clause-peripheral position prior to ellipsis of the remaining TP. In Chapter 4, “Neither Fragments nor Ellipsis,” Stainton responds to Merchant’s ellipsis analysis, arguing that ellipsis is neither empirically nor theoretically adequate to account for the variety of nonsentential utterances that can be used as propositional speech acts. Stainton concludes that a nonsentential analysis is theoretically more conservative than an ellipsis analysis in mapping forms
JB[v.20020404] Prn:19/07/2006; 14:24
F: LA93IN.tex / p.7 (437-480)
Introduction
to propositions. Casielles further investigates the import of question-answer pairs in Chapter 5, “Big Questions, Small Answers.” She explores a nonelliptical analysis of question-answer pairs and shows that such an approach does not run into syntaxsemantic mapping problems, and that it can account for the focal nature of short answers without costly operations such as movement and deletion. Finally, in Chapter 6, “Extending the Nonsentential Analysis: The Case of Special Registers,” Paesani explores adult nonsentential data outside of the context of casual conversation and question-answer pairs. She develops an analysis in which certain special registers (e.g., headlines, recipes) are argued to provide evidence for root phrases, small clauses, and root infinitives of the type proposed in Progovac (Chap. 2). This chapter serves as segue to the second part of the volume, which addresses nonsentential data from a wide variety of contexts. Chapters 7 through 12 extend the nonsentential analysis to describe data from first and second language acquisition, agrammatism/aphasia, and pidgins and creoles. In Chapter 7, “The Narrowing Acquisition Path: From Expressive Small Clauses to Declaratives,” Potts and Roeper contrast adult small clauses with small clauses in first language acquisition. Their analysis predicts which root-level small clauses present in child language will survive into adult grammar, and which will be blocked by the acquisition of higher functional projections. In Chapter 8, “Nonsententials in Second Language Acquisition,” Work extends the nonsentential analysis to early second language (L2) learner varieties. She argues that certain L2 structures are nonsententials and that learners employ this nonsentential grammar early in the acquisition process before potentially developing a full-sentence grammar. Chapters 9 and 10 explore nonsententials in the speech of agrammatic aphasiacs, and both point out that the greatest challenge in analyzing aphasia is the variability of data within the same individual, that is, the use of both sentential and nonsentential speech by one and the same speaker. In Chapter 9, “How Language Adapts to the Brain: An Analysis of Agrammatic Aphasia,” Kolk argues that the major symptoms of agrammatic aphasia reflect functional reorganization. Language adapts to the brain condition by an increase in the frequency of elliptical/nonsentential utterances and by an increase in the frequency of covert repair behaviors. Siple then examines theoretical accounts of agrammatism in Chapter 10, “Nonsententials and Agrammatism.” She argues that the nonsentential analysis proposed by Progovac (Chap. 2) may provide a theoretical framework for explaining variability within the same individual. The final two chapters extend the nonsentential analysis to pidgin and creole data. In Chapter 11, “Reduced Syntax in (Prototypical) Pidgins,” Winford describes the characteristics of pidgin languages and suggests that the grammar of pidgins is the result of the processes by which they were formed, processes similar to those that operate in early L2 acquisition. He further suggests that the grammar that produces pidgin structures shares characteristics with the grammar of small clauses for adult speakers. Finally, in Chapter 12, “Copula Variation in Guyanese Creole and AAVE: Implications for Nonsentential Grammar,” Edwards examines copula omission in Rural Guyanese Creole, Urban Guyanese Creole, and African American Vernacular English, and considers what the syntactic behavior of
JB[v.20020404] Prn:19/07/2006; 14:24
F: LA93IN.tex / p.8 (480-550)
Ljiljana Progovac et al.
copulas in these varieties implies for the underlying grammar of small clauses, particularly in nonpast contexts. The volume concludes with the Epilogue, in which the editors (Progovac, Paesani, Casielles, and Barton) extract the common themes, threads, and conclusions of the chapters. Specifically, the Epilogue presents empirical and theoretical implications of the analysis, and offers a historical perspective on the coexistence of nonsentential and sentential grammars in linguistic competence. Moreover, the Epilogue identifies prospects for future research into the syntax of nonsententials.
Notes . Nonetheless, specific authors may use different terminology for the same phenomenon, precisely because each comes from different disciplines that may have their own accepted terminology. Work in language acquisition has used terms like one-word stage, multi-word stage, telegraphic speech, and elliptical speech, again without necessarily committing to a sentential analysis (Brown & Fraser 1963; cf. Radford 1990). Work in aphasia and agrammatism similarly uses the term elliptical speech, even though researchers in this area are often not explicit about whether aphasic speech is derived from full sentences or not (e.g., Kolk, Chap. 9, this volume). Similarly, again, work in second language acquisition uses terms like basic variety without a commitment to a sentential or nonsentential analysis (Klein & Perdue 1997). In yet another example, work on pidgins has used terms like reduced syntax and described utterances in terms of missing or absent elements, again without necessarily committing to a sentential analysis with ellipsis (e.g., Bloomfield 1933; Winford, Chap. 11, this volume). Within this volume, some of the authors working in syntax and semantics (Barton; Progovac; Merchant; Stainton; Casielles; and Paesani) consistently use terms like fragment in the sense of a commitment to a sentential analysis via ellipsis; these authors also consistently use the term nonsentential in the sense of a commitment to a nonsentential analysis via base generation. Other authors working in acquisition, agrammatism, pidgins, and creoles (Potts & Roeper; Work; Kolk; Siple; Winford; and Edwards) use terms like elliptical or reduced or basic without a strict commitment to either a sentential or nonsentential analysis.
References Barton, E. (1990). Nonsentential Constituents: A Theory of Grammatical Structure and Pragmatic Interpretation. Amsterdam: John Benjamins. Barton, E. (1991). Nonsentential constituents and theories of phrase structure. In K. Leffel & D. Bouchard (Eds.), Views on Phrase Structure (pp. 193–214). Dordrecht: Kluwer. Barton, E. & Progovac, L. (2005). Nonsententials in Minimalism. In R. Elugardo & R. J. Stainton (Eds.), Ellipsis and Non-Sentential Speech (pp. 71–93). Dordrecht: Kluwer. Bloomfield, L. (1933). Language. New York, NY: Holt. Brown, R. & Fraser, C. (1963). The acquisition of syntax. In C. Cofer & B. Musgrave (Eds.), Verbal Behaviour and Learning: Problems and Processes (pp. 158–201). New York, NY: McGraw-Hill.
JB[v.20020404] Prn:19/07/2006; 14:24
F: LA93IN.tex / p.9 (550-595)
Introduction
Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: The MIT Press. Follett, W. (1966). Modern American Usage. New York, NY: Hill and Wang. Klein, W. F. & Perdue, C. (1997). The Basic Variety: Or, couldn’t natural languages be much simpler? Second Language Research, 13, 301–347. Matthews, P. (1981). Syntax. Cambridge: CUP. Merchant, J. (2004). Fragments and ellipsis. Linguistics and Philosophy, 27, 661–738. Newmeyer, F. (2003). Grammar is grammar and usage is usage. Language, 79, 682–707. Quirk, R., Greenbaum, S., Leech, G., & Svartvick, J. (1972). A Grammar of Contemporary English. London: Longman. Radford, A. (1990). Syntactic Theory and the Acquisition of English Syntax. Cambridge, MA: Blackwell. Radford, A. (1996). The nature of children’s initial clauses. In M. Aldridge (Ed.), Child Language (pp. 112–148). Clevedon: Multilingual Matters. Roeper, T. (1999). Universal bilingualism. Bilingualism: Language and Cognition, 2(3), 169–186. Sweet, H. (1900). New English Grammar. Oxford: Clarendon Press.
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.1 (47-131)
chapter
Toward a nonsentential analysis in generative grammar* Ellen Barton This chapter reviews the development in generative grammar of a nonsentential analysis for the explanation of the grammatical structure underlying utterances smaller than a sentence, discussing competing approaches suggested in the literature – an ellipsis analysis and a nonsentential analysis. The chapter considers the data and arguments for each approach, and concludes by discussing important data for comparing sentential and nonsentential analyses in current research within a Minimalist framework.
.
Introduction
In generative linguistics, there are two competing approaches to the analysis of structures underlying utterances that are smaller than a sentence like the following: (1) a.
A: What stops the White House staff from visiting the Speaker of the House in his Congressional office? B: An old grudge. b. B: The old grudge.
(2) A: The White House staff doesn’t visit the Speaker of the House in his Congressional office. B: Old grudge. In early generative grammar, working from data like (1), an ellipsis analysis was presented in a classic paper by Morgan (1973). Essentially, an ellipsis analysis presumes that the discourse provides a full sentential source from which syntactically and semantically identifiable material has been deleted in order to produce a fragment. The early arguments for an ellipsis analysis will be presented in Section 2. In contrast, a nonellipsis analysis of structures smaller than a sentence has occasionally been suggested in the generative grammar literature, proposing, more or less strongly, that independent phrases like those in (1) may be base generated (Shopen 1972; Yanofsky 1978; Brame 1979; Napoli 1982). Working from data like (2) as well as (1), Barton (1990, 1991) developed a nonsentential analysis of structures smaller than a sentence. Essentially, a nonsentential analysis allows base generation of any major syntactic category
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.2 (131-167)
Ellen Barton
as the source for independent phrases. The early arguments for a nonsentential analysis will be presented in Section 3. For a time, it seemed that both an ellipsis analysis and a nonsentential analysis might be necessary in order to explain the structure underlying utterances smaller than a sentence (Morgan 1989; Barton 1998); arguments for this mixed analysis will be presented in Section 4. Recently, research in generative grammar, specifically in the Minimalist framework (Chomsky 1995), has once again focused on the data and analysis of structures smaller than a sentence, either within a sentential analysis (Merchant 2004, and Chap. 3, this volume) or a nonsentential analysis (Barton & Progovac 2005; Progovac, Chap. 2, this volume), and there is currently some attention to what the important data might be for developing and testing sentential versus nonsentential approaches; these data will be discussed in Section 5. Overall, this chapter reviews the development of a nonsentential analysis in generative grammar up to the publication of this volume, setting the stage for further analyses of nonsententials that are developed herein.
. Early arguments for an ellipsis analysis The key evidence in arguments for an ellipsis analysis of structures smaller than a sentence, traditionally termed fragments, consists of the apparent dependence, connectivity, or matching effects between fragments and their supposed full-sentence sources, which are presumed to be recoverable in the discourse context. In the early generative grammar literature, the most detailed presentation of this approach is in Morgan (1973). To account for sentence fragments, Morgan proposed a rule of General Ellipsis, which looks to the discourse to find a full-sentence source that is posited to undergo ellipsis to a fragment. General Ellipsis is constrained under a discourse version of the Condition of Recoverability, so the deleted material is predictable syntactically and thereby interpretable semantically from its full-sentence source. It is important to note that Morgan’s data come almost exclusively from question-answer pairs, where the discourse seems to readily provide a possible full-sentence source. It is also important to note with respect to Morgan’s rule and condition that the source for a fragment must be an extant sentence in the discourse, such as a question for an answer, which guards against truly unconstrained deletion in the syntax, as would be the case if the source for a fragment were allowed to be any sentence at all. Within the standard theory of the time, Morgan (1973) argued that a well-formed sentential source necessarily underlies a well-formed fragment structure derived via General Ellipsis. Morgan looked at a wide variety of connectivity effects, such as subcategorization: (3) A: What does John think? B: That Tricia has given birth to a 7-pound chin. (John thinks that Tricia has given birth to a 7-pound chin.)
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.3 (167-248)
Toward a nonsentential analysis
(4) A: What does John think? B: *For Tricia to have given birth to a 7-pound chin. (*John thinks for Tricia to have given birth to a 7-pound chin.) In each of these cases, a well-formed sentential source predicts a well-formed fragment, and an ill-formed sentential source predicts an ill-formed fragment. Three sets of Morgan’s (1973) arguments have remained particularly important in the generative grammar literature.1 One set rested on the analysis of fragments that include binding phenomena such as reflexive pronouns and bound epithets: A: Who does Johni want to shave? B: Himselfi . (Johni wants to shave himselfi .) b. B: *Himi . (*Johni wants to shave himi .)
(5) a.
(6) a.
A: What does Johni think? B: That the bastardj is being spied on. (Johni thinks (that) the bastardj is being spied on.) b. B: *That the bastardi is being spied on. (*Johni thinks (that) the bastardi is being spied on.)
(7) a.
A: When did hei leave? B: When Johnj began to feel sick. (Hei left when Johnj began to feel sick.) b. B: *When Johni began to feel sick. (Hei left when Johni began to feel sick.)
The connectivity effects here fall out from then current analyses of binding: a wellformed sentential source with grammatical binding relations predicts a well-formed fragment, and an ill-formed sentential source with ungrammatical binding relations predicts an ill-formed fragment. Another set of arguments rested on the analysis of a variety of island phenomena (all judgments are Morgan’s (1973)): (8) a.
A: Did John and Bill leave this morning? B: No, John and Harry left this morning. b. B: *No, Harry. (*No, Harry, John and left this morning.)
(9) a.
A: Did the man who arrested Martha leave town? B: No, the man who arrested Thelma left town. b. B: *No, Thelma. (*No, Thelma, the man who arrested left town.)
Morgan argued that fragment NPs like (8b) and (9b) are ungrammatical because they would result from deletion within an island in a full-sentence source. Morgan went on
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.4 (248-305)
Ellen Barton
to note some oddities about island data, however. First, acceptability is improved when a questioned constituent is at the end of the previous sentence: (10) a.
A: John and someone just left. B: *Who? (*Who John and just left.) b. A: Bill saw John and someone. B: ?Who? (*Who Bill saw John and.)
Second, in contrast to the examples in (8) and (9), where the fragments are direct answers to questions, Morgan suggested that acceptability is improved when the fragment is used as an answer to an echo question or as a correction rather than a direct answer: (11) A: John kidnapped Martha and who? B: Thelma. (12) a.
A: Did the man who arrested Martha leave town? B: No, Thelma. b. B: *No, Thelma.
(as a correction) (as a direct answer)
(As noted earlier, these judgments are Morgan’s, and he did not discuss these examples in terms of possible complications of stress and focus with respect to islands, but see such discussion in Merchant, Chap. 3; Stainton, Chap. 4; and Casielles, Chap. 5, in this volume.) A third set of arguments noted that fragments were proper constituents of their full-sentence sources, pointing to the ungrammaticality of NPs without determiners and VPs without complements: (13) a.
A: Does Mary like the soprano? B: No, the tenor. (No, Mary likes the tenor.) b. B: *No, tenor. (*No, Mary likes tenor.)
(14) a.
A: Does John want to kiss Martha? B: No, (to) hit her. (No, John wants to hit her.) b. B: *No, (to) hit. (*No, John wants (to) hit.)
On first analysis, then, it would seem that a grammatical full-sentence source predicts a grammatical fragment in (13a) and (14a), and further captures the generalization that sentence fragments are full grammatical constituents of their sources, thus ruling out (13b) and (14b). Morgan later noted complications for this constituent condition in
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.5 (305-372)
Toward a nonsentential analysis
the case of bare VPs, noting that an ungrammatical sentence appears to be the source for a grammatical fragment: (15) A: What does Martha want to do with him? B: Talk him to death. (*Martha wants to (do) talk him to death with him.) It is tempting to tweak the source a bit, suggesting that a grammatical cleft like What Martha wants to do is talk him to death could delete to the fragment in (15). But this specifically violates Morgan’s Condition on Recoverability constraint by creating a new sentence source and not restricting the source to the sentence already present in the discourse, so a cleft would not be a possible source in Morgan’s (1973) analysis. Interestingly, Morgan (1973) proposed that an ellipsis analysis incorporates one other rule besides General Ellipsis, which is Sluicing, noting the similarity of these as rules that operate to produce sentences as well as fragments, also with the result that a well-formed sluiced sentence produces a well-formed fragment and vice versa (again, the judgments are Morgan’s): (16) a.
John did away with someone, but he didn’t say who he did away with. John did away with someone, but he didn’t say who. b. *John did away with someone, but he didn’t say with whom he did away. *John did away with someone, but he didn’t say with whom. c. *John did away with someone, but he didn’t say who with he did away. *John did away with someone, but he didn’t say who with.
(17) a.
A: B: b. A: B: c. A: B:
John did away with someone. Who? John did away with someone. *With whom? John did away with someone. *Who with?
Sluicing is the centerpiece of the sentential analysis of fragments proposed recently by Merchant (2004, and Chap. 3, this volume). Finally, Morgan (1973) noted some puzzling data for the General Ellipsis rule. First is the appearance of grammatical fragments that do not fit into a grammatical sentential source, including the bare VP in (15), repeated here as (18): (18) A: What does Martha want to do with him? B: Talk him to death. (*Martha wants to (do) talk him to death with him.) (19) a.
A: John has been here. B: When? b. A: Yesterday. (*John has been here yesterday.)
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.6 (372-426)
Ellen Barton
Second is the occurrence of ungrammatical fragments that derive from a grammatical sentence source: (20) A: Who shot Cock Robin? B: *I/*She/*He. (I/She/He shot Cock Robin.) Third is the distribution of negation in fragments, where in contrast to the generalization that fragments must be full constituents within their sentential sources, Morgan (1973: 747) proposes the “not particularly pleasant” idea that not plus NP might be a constituent of some sort: (21) A: Who did Martha talk to? B: Not Kissinger. (*Martha talked to not Kissinger.) Barton (1990, 1991) based many of her arguments for a nonsentential analysis on data with mismatches like these. Morgan (1973) paid considerable attention to the implications of his analysis in the theoretical framework of generative grammar. For example, he noted that his analysis preserves the generalization of a standard theory model that the initial node of a generative grammar is S, but he also noted that sentencehood should not be a necessary condition for the acceptability of utterances; in other words, utterances that are not sentences, specifically fragments, are acceptable even though they are not sentences. He also pointed out that the placement of a General Ellipsis Rule is not perfectly straightforward: in standard theory, surface structure is the most plausible source for the operation of General Ellipsis, but it is not a perfect source because many fragments appear to look at levels either before or after surface structure for the distribution of pronominal person and number and the necessary changes in verb tense as in (19). With respect to the relationship between syntax and semantics, Morgan (1973: 721) stipulated that the relationship is sentential: “I will not argue here for the obvious fact that [fragments] have the semantic import of full sentences,” but he invites future research on this assumption. Perhaps most importantly, Morgan acknowledged that an ellipsis theory violates the autonomy of syntax, because it requires direct input from discourse to the operation of the ellipsis rule, which is in the syntax. The autonomy of syntax was then (and is now) one of the long-standing conditions on generative grammar, and the Condition of Recoverability on Deletion was an important constraint in support of autonomy (Chomsky 1965). Morgan (1973: 731) noted that in his analysis “only the letter and not the spirit of this condition is violated.” More specifically, he extended the Condition of Recoverability very narrowly to just the discourse context of the fragment, which rules out any unconstrained ellipsis analyses that would allow the creation of any kind of source and thus allow any kind of unconstrained deletion to produce fragments. In this way, Morgan’s analysis is relatively tight, and it works particularly well to account for the data of question-answer pairs.
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.7 (426-480)
Toward a nonsentential analysis
In summary, the generative grammar of the time was argued to account for the syntactic and semantic properties of fragments fairly straightforwardly by utilizing the construct of a full-sentence source from the discourse plus relatively constrained ellipsis.
. Early arguments for a nonsentential analysis An ellipsis analysis with intersentential deletion and a nonsentential analysis with base generation both provide a syntactic account for the majority of utterances that are smaller than a sentence. As noted before, the key evidence supporting a sentential analysis is connectivity effects, particularly visible in question-answer pairs. The key evidence in arguments for a nonsentential analysis of structures smaller than a sentence consists of the apparent syntactic and semantic independence of many nonsententials, ones that have properties that are crucially without straightforward dependence, connectivity, or matching effects to full-sentence sources in the discourse context. Yanofsky (1978) first noted that some discourse-initial NP utterances like those in (22) occur without a discourse context to support recoverable syntactic deletion: (22) a. b. c. d.
The time? Thief! Thief! Teamwork. Your move.
[at a bus stop to a stranger] [as a mugger is running away] [after winning tennis doubles] [during a game of Monopoly]
Yanofsky noted the close connection between these initial NPs and their pragmatic contexts, in the spirit of Hankamer and Sag (1976), but she also pointed out that there seems to be no motivation for syntactic deletion without an obvious sentential source. Yanofsky herself did not develop a syntactic analysis of the derivation of these NPs, but some researchers (Brame 1979; Napoli 1982) suggested the possibility of base generating such phrases in generative grammar. Barton (1990) went on to show that independent phrases without grammatical full-sentence sources in their linguistic discourses or ostensible connections with their pragmatic contexts are widespread in English, arguing that independent phrases are not restricted to related question-answer pairs or pragmatically bound utterances. Barton (1990, 1991) offered a systematic presentation of examples of well-formed independent phrases that could not be derived syntactically from full-sentence sources in their discourses and further could not be predictably associated with a single fullsentence source for semantic or pragmatic interpretation. Some examples are NPs like (23), VPs like (24) and (25), and small clauses like (26a) and (26b):2 (23) A: The White House staff doesn’t visit the Speaker of the House in his Congressional office. B: Old grudge.
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.8 (480-559)
Ellen Barton
(24) A: What does John do in the summer? B: Play baseball. (25) A: John doesn’t know what the best defense against ethics charges would be. B: Ask any lawyer. (26) a.
A: Our classmate John is probably making a million a year by now. B: John a successful business man? Don’t make me laugh. b. B: Him a successful business man? Don’t make me laugh.
Barton argued that it is syntactically impossible to derive these structures from discourse sentence sources: (27) a. *The White House staff doesn’t visit the Speaker of the House in his Congressional office because of old grudge. b. *Old grudge stops the White House staff from visiting the Speaker of the House in his Congressional office. (28) *John play baseball in the summer. (29) *John ask any lawyer. (30) a. *Our classmate John is probably making a million a year by now him a successful businessman. b. *Our classmate John is probably making a million a year by now John a successful businessman. It is important to note here that Barton (1990, 1991) argued specifically against Morgan’s (1973) analysis, which restricts sentential sources to the extant discourse context in order to guard against completely unconstrained deletion. The argument, then, does not claim that utterances like old grudge, play baseball, ask any lawyer, or him a successful businessman are not possible constituents in some full sentence. However, if an ellipsis analysis allows any full-sentence source to stand as the source for independent constituents, it would have to rely on completely unconstrained deletion because the source could be any sentence at all. For example, the following sources may seem plausible for the utterances in (23) through (26): (31) A: The White House staff doesn’t visit the Speaker of the House in his Congressional office. B: It’s (the/an) old grudge between them. (32) A: What does John do in the summer? B: What John does in the summer is play baseball. (33) A: John doesn’t know what the best defense against ethics charges would be. B: He should ask any lawyer. Ask any lawyer is what he should do. (34) A: Our classmate John is probably making a million a year by now. B: Do you consider John/him a successful businessman?
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.9 (559-628)
Toward a nonsentential analysis
Plausibility, however, can be misleading. If an analysis allows freely created sentences to serve as sources, it has no way of ruling out any other sources, including wildly implausible ones such as the following for the examples in (23) through (26): (35) B: There’s an old grudge between Michigan and Ohio State. (36) B: No hockey player would like to play baseball all summer. (37) B: Aristotle could ask any lawyer a question about tax evasion. (38) B: The Communist Party considers John/him a successful businessman. Theoretically, unconstrained deletion is highly undesirable, and, as noted earlier, Morgan (1973) did not utilize unconstrained deletion in his ellipsis analysis. Morgan (1973) suggested that the syntax-semantics relationship between fragments and sentential sources is straightforward, but Barton (1990) argued that it is not always possible to establish a semantic interpretation of some features of independent phrases and small clauses by using the construct of a full-sentence source in the discourse. For example, the definiteness of the NP in (23) is not certain, as shown in (39a) and (39b), nor are the semantic roles of the thematic entities, in other words, who holds a grudge against whom, as shown in (40a) and (40b): (39) a.
The old grudge between the White House staff and the Speaker of the House. [A assumes B knows that there is a grudge.] b. An old grudge between the White House staff and the Speaker of the House. [A indicates to B that there is an old grudge.]
(40) a.
The White House staff doesn’t visit the Speaker of the House in his Congressional office because they have an old grudge against him. b. The White House staff doesn’t visit the Speaker of the House in his Congressional office because he has an old grudge against them.
Similarly, it is impossible to establish definitively the entities associated with the agentive role in the VP in (25), in other words, who exactly must, could, or should ask any lawyer: (41) a. b. c. d.
(You will) ask any lawyer. [imperative] John could/should ask any lawyer. You could/should ask any lawyer. Anybody could ask any lawyer.
For this example, it is also impossible to establish the pragmatic speech act of the utterance: it could be an imperative, a suggestion, or, perhaps, an assertion. (See n. 3 of this chapter, as well as Stainton 2004, and Chap. 4, this volume; Progovac, Chap. 2, this volume; and the Epilogue of this volume for more on the semantics and pragmatics of nonsententials.)
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.10 (628-681)
Ellen Barton
Barton (1991) proposed the Xmax Generalization within the framework of government and binding theory to describe base generation of nonsentential constituents: (42) Xmax Generalization: The initial node of a generative grammar is Xmax . This generalization, which was an extension of the principles of X-bar theory then current in generative grammar, argued against the long-standing assumption stipulating that the initial node of a generative grammar is S. Under the Xmax Generalization, a grammar generates not only sentences under the initial node of S but also nonsentential constituents with initial nodes of NP, VP, AdjP, AdvP, and PP, all of which are grammatical, acceptable, and attested nonsententials in a variety of discourses and contexts. Barton (1990) argued that most of the grammatical properties of well-formed versus ill-formed nonsententials fall out from their derivation under a maximal category as an initial node: for example, phrasal heads must occur with their subcategorized complements, and so on. Furthermore, the occurrence and properties of nonsentential NPs without determiners, bare VPs, and independent small clauses are explained as well. A nonsentential NP like old grudge is assumed to be a projection to NP without a determiner, a type of NP the grammar already allows in general (e.g., for noncount nouns, some proper nouns, and pronouns, under the analyses of the time). A bare VP like play baseball is assumed to be a projection to VP before Tense and Agreement features, again a type of constituent the grammar already generates in general. A nonsentential small clause like him/John a successful businessman is assumed to be a projection to the phrasal node of its head, here NP, following Stowell (1983). Concerning Morgan’s (1973) generalization that utterances must be proper constituents of their full-sentence sources, Barton (1991) argued that the Xmax Generalization also predicts the well-formedness and ill-formedness of Morgan’s NP examples like (13), repeated here: (43) a. No, the tenor. b. *No, tenor. With the advent of a functional category analysis of DPs (Abney 1987), a nonsentential analysis can accommodate the fact that this kind of no with comma intonation would only be grammatical for a common noun if adjoined to a functional category projection, like the DP projection in (43a). Because a nonsentential analysis would analyze (43a) as a DP and (43b) as an NP, the contrast in well-formedness would be predicted straightforwardly: (43a) is a functional DP projection, as shown by the presence of a determiner, so no with comma intonation can adjoin; (43b), however, is a projection to NP without a determiner or DP projection, so no with comma intonation cannot adjoin, and the nonsentential is predicted to be ungrammatical. Similarly, a nonsentential analysis predicts Morgan’s VP examples like (14) and (15)/(18), repeated here:
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.11 (681-757)
Toward a nonsentential analysis
(44) a.
A: Does John want to kiss Martha? B: No, to hit her. b. B: *No, hit.
(45) A: What does Martha want to do with him? B: Talk him to death. (*Martha wants to (do) talk him to death with him.) Both a sentential analysis and a nonsentential analysis predict the examples in (44). Within a sentential analysis, the fragment answer in (44a) derives straightforwardly from the question. Within a nonsentential analysis, the adjunction of no with comma intonation would only be grammatical for a VP within some projection of IP; arguably, an infinitive VP with to has some kind of functional projection, which allows the adjunction of no in (44a). Both analyses rule out (44b) because the verb hit requires a complement. The derivation of (45), however, differs in the two analyses: as Morgan noted, a sentential analysis cannot generate this bare VP without Tense and Agreement features, so it stands as a problematic counterexample to his analysis as discussed with respect to (18). But in a nonsentential analysis talk him to death is projected as a bare VP without any further functional projection, which predicts the grammaticality of (45b) straightforwardly. Bare NPs and VPs, then, are explained straightforwardly in a nonsentential analysis. Barton (1990, 1991) argued that a nonsentential analysis could be theoretically desirable for two related reasons: by eliminating the necessary recourse of an ellipsis analysis to discourse-based deletion, the nonsentential analysis strongly respects the autonomy of syntax by base generating constituents directly.3 In summary, a nonsentential analysis was argued to account for the grammatical properties of independent phrases and small clauses straightforwardly by utilizing the Xmax Generalization in combination with theoretical developments in generative grammar.
. Mixed analyses Morgan (1973) and Barton (1990, 1991) both proposed that all structures underlying utterances smaller than a sentence were derived exclusively through ellipsis or base generation respectively. Working further on cross-linguistic data, however, both Morgan (1989) and Barton (1998) proposed a mixed analysis whereby some of these structures were derived through ellipsis and some through base generation. Morgan formulated the problems of the competing analyses as follows: The evidence is not unequivocal. The base generation analysis predicts that fragments should have no sentence-derivative syntactic properties; this prediction is clearly falsified by the data. The deletion analysis predicts that fragments should have such sentence-derivative properties. Unfortunately, the deletion analysis in its simpler form makes the stronger prediction that fragments should have all such properties. . . This prediction is too strong. (Morgan 1989: 232–233)
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.12 (757-812)
Ellen Barton
Working on data from Korean, Morgan found both case-marked and non-casemarked structures were grammatical and acceptable as independent NPs: (46) a.
A: Nu-ka ku chaek-ul sa-ass-ni? ‘Who bought the book?’ B: Yongsu John-no.case ‘John’ b. B: Yongsu-ka John-nom ‘John’
(47) a.
A: Nuku-eke k chaek-ul pnasess-ni? ‘Who did you send the book to?’ B: Yongsu John-no.case ‘John’ b. B: Yongsu-eke John-dat ‘(To) John’
(48) a.
A: Muos-uro phyonci-rule ssoss-ni? ‘With what did you write the letter?’ B: Yonphil pencil-no.case ‘Pencil’ b. B: Yonphil-lo pencil-instr ‘(With a) pencil’
Morgan also noted that some strongly context-dependent NPs must occur without any case marking in Korean: (49) a.
Phyo hang-cang ticket one-no.case ‘One ticket!’ b. *Phyo hang-cang-l ticket one-nom c. *Phyo hang-cang-ul ticket one-acc
Within a mixed analysis, Morgan suggested that the forms without case marking were base-generated NPs, whereas the forms with case marking were NPs from ellipsis; the presence of case, he suggested, forces the analysis of the reconstruction of a full-sentence source. In his conclusion, Morgan noted the empirical need for a mixed analysis:
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.13 (812-860)
Toward a nonsentential analysis
In fact there is some evidence from Korean that suggests that there are two parallel systems. . . [F]ragments have some, but not all, of the properties one would expect them to have if they were sub-constituents of sentences. Taken as a complete account of fragments, neither base generation nor deletion is consistent with this partiality. The base generation analysis falsely predicts that fragments don’t behave in any way as if they were sub-constituents of sentences. The syntactic deletion approach avoids this false prediction, but errs in the other direction, in that in its simplest form it predicts that such fragments behave in every way as if they were sub-constituents of sentences. The data make it clear that this prediction is also false. (Morgan 1989: 237–238)
Working on an analysis of cross-linguistic data from a corpus of telegraphese, Barton (1998) noted that some telegraphic utterances showed evidence of sentential derivation in VPs with overt Tense and Agreement features: (50) English: a. Am ill. b. Was to present a paper. c. Am at border in Newport, Vermont. (51) German: a.
Bin krank be-1sg ill ‘Am ill’ b. Bin verhaftet be-1sg arrested ‘Am arrested’ c. Habe Autopanne have-1sg car-breakdown ‘Have car breakdown’
(52) Dutch:
a.
Ben ziek be-1sg sick ‘Am sick’ b. Heb autopech have-1sg car-breakdown ‘Have car breakdown’
Other telegraphic utterances, however, including initial utterances, are independent phrases straightforwardly derived in a nonsentential analysis: (53) English: a. Sudden flu attack b. Car problem (54) Dutch:
a.
Oponthoud door bommelding delay through bomb-threat ‘Delay through bomb threat’
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.14 (860-921)
Ellen Barton
b. Autopech car-problem ‘Car problem’ (55) German: a.
Plotzlich Krankheit sudden illness ‘Sudden illness’ b. Grenzproblem border-problem ‘Border problem’ c. Drogenschmuggelverdacht drug-smuggling-suspicion ‘Suspicion of drug smuggling’
(56) Swedish: a.
Influenza ‘Influenza’ b. Motorhaveri motor-breakdown ‘Car breakdown’ c. Tullproblem customs-problem ‘Customs problem’ d. Bombhot bomb-threat ‘Bomb threat’
Note that many of these examples are bare NPs without determiners, generated unproblematically within a nonsentential analysis. Barton (1998) thus proposed that a sentential analysis would be necessary for just those telegraphic utterances that show evidence of a full-sentence source, as in the examples in (50) through (52), whereas a nonsentential analysis could account straightforwardly for telegraphic utterances that are independent phrases, like the examples in (53) through (56). Some telegraphic utterances, however, are not as straightforward, particularly examples like the following: (57) English: Car broken down (58) Swedish: Bilen sonder car-def broken-down (59) German: Auto defect car defective At the time, Barton (1998) thought these examples might have been sentential, but Barton and Progovac (2005) reanalyzed them as small clause nonsententials. Small clause nonsententials are discussed in Progovac (Chap. 2, this volume).
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.15 (921-980)
Toward a nonsentential analysis
Similarly to Morgan (1989), then, Barton (1998) concluded that a mixed analysis was necessary to account empirically for the structures underlying telegraphic utterances: I argue that both frameworks [an ellipsis analysis and a nonsentential analysis] are necessary for a full explanation of the corpus [of telegraphese]. . . [A] nonsentential analysis does not eliminate ellipsis entirely from a generative grammar, although it restricts it significantly by limiting ellipsis to instrasentential deletions. . . The combination of an ellipsis analysis and a nonsentential analysis thus allows a set of principled distinctions to be made between sentential structures that have undergone intrasentential deletion and nonsentential structures that have derived from major grammatical categories as initial nodes. (Barton 1998: 39, 44)
Morgan (1989) and Barton (1998) differ in the scope of deletion versus base generation in their mixed analyses. Morgan argued that most structures smaller than a sentence are fragments that should be derived through sentential deletion, with a restricted number of nonsententials derived through base generation, just those with notable pragmatic immediacy. Barton argued that most structures smaller than a sentence are nonsententials that should be derived through base generation, with a restricted number of fragments derived through ellipsis, just those that show overt evidence of sentential derivation.
. Data and sentential/nonsentential analyses in Minimalism Although the literature in a variety of frameworks includes some other discussions of fragments versus nonsententials, usually in the service of an argument about a different topic (van Riemsdijk 1978; Ginzburg & Sag 2000; Culicover & Jackendoff 2005 for a nonsentential approach; Hankamer 1979; Stanley 2000 for a sentential approach), recent work in the Minimalist framework of generative grammar has again taken up the data and analysis of structures smaller than a sentence as an important topic. In particular, Merchant (2004, and Chap. 3, this volume) has developed a new sentential analysis based on sluicing, and Barton and Progovac (2005) and Progovac (Chap. 2, this volume) have continued work on a nonsentential analysis with base generation. In this new wave of work, one important development is that advocates of both sentential and nonsentential approaches seem to be converging on a common identification of the important data that must be accounted for explicitly in an analysis. Specifically, data exhibiting properties of case marking are now seen as crucial data for developing, testing, and choosing between competing analyses. Case has been noted to exhibit properties of both connectivity and nonconnectivity in structures smaller than a sentence. As Morgan (1989) noted, however, the evidence is not unequivocal: some data seem to lend support to a sentential analysis, whereas other data seem to lend support to a nonsentential analysis.
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.16 (980-1038)
Ellen Barton
For example, Merchant (2004, and Chap. 3, this volume) has provided an extensive list of question-answer pairs that seem to exhibit straightforward connectivity effects. Barton (1990), however, has provided examples to indicate that the facts of well-formed independent pronouns do not necessarily align with sentential syntax. For example, nonsentential pronouns typically appear in the accusative case, even when they seem to play an agentive role in the utterances of a discourse: (60) a.
A: B: b. B: c. B:
Someone gave a book to John. Me/Him/Her/?Us/Them. ?*I/*He/*She/*We/*They. I did/He did/She did/We did/They did.
Native speakers differ somewhat in their judgments about these examples: almost all speakers accept the accusative pronouns in (60a) in casual speech, and most prefer the short sentences in (60c) to the independent nominative pronouns in (60b). (Some linguists such as Emonds (1986) have suggested that the examples in (60b), particularly the first-person singular I, may be the residue of prescriptive grammar hypercorrection.) Barton (1991) also noted that the unmarked accusative case appears as the subject of nonsentential small clauses: (61) a. Him a successful businessman? Don’t make me laugh. b. *He a successful businessman? Don’t make me laugh. Morgan (1989) similarly noted this pattern and its variable judgments in (61), also pointing to a similar pattern with negation in (62): (62) a.
A: B: b. B: c. B:
Who can eat another piece of cake? Me/?Us/Him/Her. ?*I/*We/?*He/?*She. I can/We can/He can/She can.
(63) a.
A: Who shot Cock Robin? B: Not me/Not you/Not him/Not her/Not us/Not them. b. B: Not I/?Not he/?Not she/*Not we/*Not they.
Both Morgan and Barton noted that an ellipsis analysis does not easily predict such examples because they would be ungrammatical in their discourse sentence sources: (64) a. *Me/*Him/*Her/*Us/*Them gave a book to John. b. *Me/*Us/*Him/*Her can eat another piece of cake. c. *Not me/*Not you/*Not him/*Not her/*Not us/*Not them shot Cock Robin. Following Radford (1981), Barton (1990, 1991) suggested that Accusative is the unmarked Case for independent pronouns in English, which accounts for the data in (60) through (64) generally, including the negated accusative pronouns. Although this
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.17 (1038-1096)
Toward a nonsentential analysis
proposal was more of a stipulation than an analysis, the point here is that independent pronouns may not match Case in sentential syntax straightforwardly. The explanation of Case thus appears to be crucial in current research, and will be discussed extensively in the chapters in this volume (see especially Merchant, Progovac, and the Epilogue, which includes a discussion of Case cross-linguistically).
. Conclusion With structures smaller than a sentence, it is often tempting to consider these data not crucial and not to pursue the importance and implications of an analysis of these data within syntactic theory. It is sometimes easier to set what appear to be recalcitrant data aside and settle for an intuitively plausible version of a mixed-analysis approach that defaults to a sentential analysis for data like question-answer pairs with seeming connections to an obvious sentential source, and that defaults to a nonsentential analysis for data like pragmatically bound utterances or other data without obvious connections to a sentential source. Theoretically, however, a mixed analysis is less than parsimonious, especially because many utterances with the same structure occur in both contexts: (65) a.
A: Who moves next? [referring to another player during a board game] B: Him? b. [during a board game, asking a question to a second player while handing the dice to a third player] A: Him?
To derive the same NP in separate analyses seems somewhat counterintuitive and undesirable, both empirically and theoretically. Furthermore, current research on both sentential and nonsentential analyses has attempted to widen the scope of data. The current version of a sentential analysis (Merchant 2004, and Chap. 3, this volume) seeks to extend its scope from question-answer pairs like (65a), where the analysis works quite elegantly, to data from a wider variety of discourse and pragmatic contexts: Merchant utilizes the notion of “limited ellipsis” to account for data like (65b) in a principled way (although see Stainton, Chap. 4, this volume, for a critique). Similarly, the current version of a nonsentential analysis seeks to extend its scope from pragmatically salient utterances like (65b), where the analysis works quite well, to explaining data like question-answer pairs like (65a) in a principled way: Progovac (Chap. 2, this volume) utilizes the notion of Default (nonstructural) Case in nonsententials, and Casielles (Chap. 5, this volume) extends this analysis specifically to question-answer pairs. These efforts follow the long-standing practice in the field of linguistics: research in generative grammar has always operated by working to push an analysis as far as it can go, which is a goal of some of the chap-
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.18 (1096-1169)
Ellen Barton
ters in this volume in terms of developing a nonsentential analysis of the grammatical structure underlying utterances that are smaller than a sentence.
Notes * I would like to thank the members of the Wayne State University Humanities Center Working Group on Nonsententials for taking up this topic and for collaborating to the point of this volume. My greatest thanks go to the coeditors of this volume – Ljiljana Progovac, especially; Kate Paesani; and Eugenia Casielles – this has been a remarkable collaborative and collegial experience. . In Section 2, I focus on the data and arguments that continue to play an important role in the generative grammar literature on the structures underlying utterances that are smaller than a sentence. I do not recount arguments that have faded from the literature, such as fragments that derive from transformations of the time (e.g., passive, tough movement, each other formation, and affix hopping) or fragments that derive from specific rule ordering (e.g., adjective preposing and ellipsis to fragments). . Barton (1990, 1991) also presents examples of independent AdjPs, AdvPs, and PPs without obvious sentential sources, but this chapter will focus mainly on NPs, VPs, and small clauses, anticipating later arguments in Progovac (Chap. 2, this volume). . An autonomous nonsentential analysis within the syntax does, however, raise questions about interpretation. Barton (1990) proposed a model for the interpretation of nonsententials based on Gricean pragmatics, but the operations of this model are quite outdated. Nevertheless, reviewers for the volume have asked whether a nonsentential syntactic analysis is forced to rely on adding unmotivated complications to theories of discourse or pragmatic interpretation. Here, I argue briefly that the contextual interpretation of nonsententials does not require unmotivated complications of discourse and pragmatics; in fact, the interpretation of nonsententials fits well into existing mechanisms of theories of interpretation in context. One test case for this argument is the interpretation of nonsententials in question-answer pairs, a case where an independent phrase seems to have close connectivity effects to a sentential source as proposed in an ellipsis analysis. In an ellipsis analysis, the interpretation of an answer to a question is elegantly handled within sentential semantics (Merchant 2004, and Chap. 3, this volume). In a nonsentential analysis, answers to questions are proposed to be focal elements (Casielles, Chap. 5, this volume), but that leaves open the process by which nonsententials are interpreted as playing roles with respect to the previous question, as in the following examples: (i)
A: B:
(ii) A: B:
What does John see? A yellow bird. Who loves John? Mary.
In terms of thematic roles within a sentence, a yellow bird would function as a patient, and Mary as an agent. A similar process is needed for the interpretation of focal nonsentential pronouns, including those in Default Case: (iii) A: B:
Did someone give a book to John? Me/You/Him/Her/? Us/Them.
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.19 (1169-1255)
Toward a nonsentential analysis
Because a nonsentential analysis does not construct a sentential source for answers to questions, it must shift interpretation to a discourse/pragmatics component, and what needs to be explained is the most obvious interpretation of the NPs above in terms of the linguistic context, specifically the preceding question. In the discourse/pragmatics literature, there are a couple of approaches that could explain the interpretation of answers to questions without the need to specify any special mechanisms specifically for nonsententials. In the pragmatics literature, Stainton (2004, Chap. 4, this volume, and elsewhere) argues that nonsententials can perform the entire repertoire of speech acts, including assertion. When a nonsentential performs the speech act of asserting, it uses a phrase in order to pragmatically express a proposition (Stainton argues that semantically a nonsentential is either an argument or a property but pragmatically it expresses a proposition). This proposition is constructed within pragmatics through an enrichment process. Using relevance theory (Sperber & Wilson 1986), specifically their notion of enriching to a proposition, Stainton (2004: 283) observes that information to enrich a nonsentential to a pragmatically derived proposition can come from the multitude of information available to any speaker/hearer, including information from “long-term memory, from short-term memory, from the imagination, from a perceptual module, etc.” It is not unreasonable to suppose that short-term memory contains the previous utterance in a discourse (Chafe 1994); thus, it can be used in the pragmatic process of enrichment to a proposition. From this approach, the nonsentential answers to questions can be interpreted as propositions: (iv) saw (John, a yellow bird) (v) loves (Mary, John) (vi) give (me, a book, to John) It is important to note here that a nonsentential like a yellow bird in (i) does not necessarily have to be interpreted as a proposition that answers the question in terms of the proposition in (iv). (iv) would be the most straightforward and conventional interpretation, perhaps signaled by normal intonation. With marked intonation, however, the interpretation might shift to a version of one of Grice’s (1975) well-known examples of conversational implicature (for this example, assume that there is a yellow bird in the physical context of the exchange): (viii) A: What did John see? B: A yellow bird! (B sees John himself walking up to them and so flouts the maxim of relevance by describing an unrelated object in order to implicate to A that he does not wish to answer the question, perhaps because John is walking up to the speakers and B does not want him to know that they have been talking about him.) The point here is that nonsententials can be interpreted as propositions that answer questions, with a proposition derived from both the linguistic expression and its enrichment through inferences, in this case based on the discourse context that is available to the speaker/hearer through short-term memory. This explanation utilizes independently motivated mechanisms already argued to be part of a pragmatic component (or speakers’/hearers’ pragmatic competence, if that locution is preferred). A second, converging approach to establishing the nonsententials in (i) through (iii) as answers to questions in a discourse comes from another body of research literature in conversation analysis (for a still-current summary of conversation analysis written for researchers in generative linguistics, see Levinson 1983, Chap. 6). Conversation analysis describes the way that interac-
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.20 (1255-1332)
Ellen Barton
tion proceeds in a stepwise, locally managed fashion, with speakers producing and interpreting utterances primarily with respect to the linguistic context, particularly the previous utterance: speakers observably display their orientation to the previous utterance by designing their turns within a variety of sequence organizations. Sequence organizations consist of adjacency pairs; an adjacency pair has a first part with a strongly preferred second part. In conversation analysis, the question-answer pair is the king of adjacency pairs: simply put, when a speaker observably orients to this adjacency pair by asking a question as a first part, the strongly preferred second part is an answer, and any nonpreferred response (i.e., an utterance that is not an answer to the question) is notably marked. Under this view of interaction, the unmarked nonsententials in (i) through (iii) are observably oriented to the preferred second part of a question-answer adjacency pair; in other words, they are turns designed to be answers to the questions. At first, this technical apparatus of turn design and adjacency pairs may seem simply descriptive in an obvious way for describing question-answer pairs. But it similarly accounts for example (viii): a nonsentential with marked intonation would not be interpreted as the preferred second-part member of an answer in response to a question. In summary, the interpretation of nonsententials in question-answer pairs, both in terms of proposition enrichment and in terms of interaction (turn design within adjacency sequences), is explained without the addition of specific mechanisms to existing theories of discourse and pragmatics. Instead, discourse- and pragmatic-based interpretation of nonsententials utilizes the existing and independently motivated constructs in current theories.
References Abney, S. (1987). The English Noun Phrase in its Sentential Aspect. Ph.D. dissertation, Massachusetts Institute of Technology. Barton, E. (1990). Nonsentential Constituents: A Theory of Grammatical Structure and Pragmatic Interpretation. Amsterdam/Philadelphia: John Benjamins. Barton, E. (1991). Nonsentential constituents and theories of phrase structure. In K. Leffel & D. Bouchard (Eds.), Views on Phrase Structure (pp. 193–214). Dordrecht: Kluwer. Barton, E. (1998). The grammar of telegraphic structures: Sentential and nonsentential derivation. Journal of English Linguistics, 26, 37–67. Barton, E. & Progovac, L. (2005). Nonsententials in Minimalism. In R. Stainton & R. Elugardo (Eds.), Ellipsis and Nonsentential Speech (pp. 71–93). Dordrecht: Kluwer. Brame, M. (1979). A note on COMP S grammar vs. sentence grammar. Linguistic Analysis, 5, 383–386. Chafe, W. (1994). Discourse, Consciousness, and Time: The Flow and Displacement of Conscious Experience in Speaking and Writing. Chicago, IL: University of Chicago Press. Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, MA: The MIT Press. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: The MIT Press. Culicover, P. & Jackendoff, R. (2005). Simpler Syntax. New York, NY: OUP. Emonds, J. (1986). Grammatically deviant prestige dialect constructions. In M. Brame, H. Contreras, & F. Newmeyer (Eds.), A Festschrift for Sol Saporta (pp. 93–129). Seattle, WA: Noit Amrofer. Ginzburg, J. & Sag, I. (2000). Interrogative Investigations: The Form, Meaning, and Use of English Interrogatives. Stanford, CA: CSLI.
JB[v.20020404] Prn:19/07/2006; 15:21
F: LA9301.tex / p.21 (1332-1399)
Toward a nonsentential analysis
Grice, H. P. (1975). Logic and conversation. In P. Cole & J. Morgan (Eds.), Syntax and Semantics Volume 3: Speech Acts (pp. 41–58). New York, NY: Academic Press. Hankamer, J. (1979). Deletion in Coordinate Structures. New York, NY: Garland. Hankamer, J. & Sag, I. (1976). Deep and surface anaphora. Linguistic Inquiry, 7, 391–428. Levinson, S. (1983). Pragmatics. Cambridge: CUP. Merchant, J. (2004). Fragments and ellipsis. Linguistics and Philosophy, 26, 661–738. Morgan, J. (1973). Sentence fragments and the notion ‘sentence.’ In B. Kachru, R. Lees, Y. Malkiel, A. Pietrangeli, & S. Saporta (Eds.), Issues in Linguistics: Papers in Honor of Henry and Renée Kahane (pp. 719–751). Urbana, IL: University of Illinois Press. Morgan, J. (1989). Sentence fragments revisited. In B. Music, R. Graczyk, & C. Wiltshire (Eds.), CLS Parasession on Language in Context (pp. 228–241). Chicago, IL: Chicago Linguistics Society. Napoli, D. (1982). Initial material deletion in English. Glossa, 16, 85–111. Radford, A. (1981). Transformational Syntax. Cambridge: CUP. Shopen, T. (1972). A Generative Theory of Ellipsis. Ph.D. dissertation, University of California, Los Angeles. Sperber, D. & Wilson, D. (1986). Relevance: Communication and Cognition. Cambridge, MA: Harvard University Press. Stainton, R. (2004). The pragmatics of non-sentences. In L. Horn & G. Ward (Eds.), The Handbook of Pragmatics (pp. 266–287). Oxford: Blackwell. Stanley, J. (2000). Context and logical form. Linguistics and Philosophy, 23, 391–434. Stowell, T. (1983). Subjects across categories. Linguistic Review, 2/3, 285–312. van Riemsdijk, H. (1978). A Case Study in Syntactic Markedness: The Binding Nature of Prepositional Phrases. Dordrecht: Foris. Yanofsky, N. (1978). NP utterances. In D. Farkas, W. Jacobsen, & K. Todnys (Eds.), CLS 14 (pp. 491–502). Chicago, IL: Chicago Linguistic Society.
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.1 (47-118)
chapter
The syntax of nonsententials Small clauses and phrases at the root* Ljiljana Progovac I argue for the existence of root phrases and small clauses (Me; Me dancing; Class in session) as terminal syntactic objects. Their defining property is the absence of Tense, and, with it, any syntactic phenomena that rely on Tense, including structural nominative case. This predicts, for English: default accusative on subject pronouns; possible absence of otherwise obligatory articles; systematic lack of finiteness. The analysis capitalizes on the role of Economy in Minimalism, as well as on its theoretical assumption that every sentence is underlyingly a small clause. The lack of tense specification is arguably also responsible for the dearth of indicative interpretations among base-generated nonsententials (the phenomenon finds a historical correlate), as well as for their reliance on pragmatic context.
.
Introduction
The proposal in this chapter is an attempt to bring together research on nonsententials and current syntactic theory, in particular, Minimalism (e.g., Chomsky 1995), arguably to the benefit of both. The argument is that what you see in nonsententials is more or less what you get, basically phrasal projections smaller than a sentence. In this respect, Barton and Progovac (2005) point out that analyzing nonsententials as phrases/small clauses is the most economical derivation for such structures, in the spirit of Minimalism. It is an empirical fact that there is a difference between, for example, Me first! and I am first, not only syntactically, but also when it comes to interpretation. Whereas I am first is a full finite sentence, with a nominative subject, interpreted as a truth-evaluable assertion, Me first!, which I analyze here as a tenseless small clause with a default case (accusative) subject, can be interpreted as anything but an assertion, ranging over exclamative, imperative, optative, and other possibilities. Similar considerations of form and interpretation underlie the contrast between, for example, He worries and Him worry?! The goal of this chapter is primarily to characterize the structural differences between the two types by a concrete theoretical analysis, and, secondarily, to use this syntactic analysis to illuminate the semantic differences, as well as to discuss the role
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.2 (118-173)
Ljiljana Progovac
of pragmatic context. This chapter will thus inevitably touch on some basic linguistic questions, such as what exactly is a sentence (in contrast to a nonsentential), as well as what is the role of Tense or Time in creating a sentence or an assertion. Noticing that nonsentential grammars often coexist with full grammars, in both children and adults, Roeper (1999) concludes that this state of affairs constitutes what he calls “Universal Bilingualism,” that is, access of each speaker to two grammar systems, the full grammar, and the small/minimal grammar. For Roeper, the two systems are the product of the tension between the two conflicting requirements on language: the need for structural Economy, on the one hand, and the need for Explicitness, on the other. Even though descriptively speaking one might say that there are two kinds of grammar, one sentential (e.g., I am dancing; I am first), and one nonsentential (e.g., Me dancing; Me first!), my approach in this chapter will characterize the differences by appealing to different feature specifications, or in fact underspecifications, rather than to different grammars. More precisely, I will argue that nonsententials result from selecting lexical items with unspecified/default forms of Tense and Case, rendering the projection of a TP (Tense Phrase) superfluous. On the other hand, when the corresponding lexical items are selected with Tense or nominative Case features, a projection of a full sentence (TP) will be forced. But, essentially, both sentential and nonsentential structures are governed by the same principles of Universal Grammar, and the derivation of both types is naturally accommodated in the current Minimalist theory, as will be argued in this chapter. It has long been noted that even normal adult speakers produce utterances that do not look like (full) sentences at all, as illustrated in (1) through (28). I refer to such utterances as nonsententials, and, anticipating the analysis, I categorize them as specific phrases, that is, as NPs, VPs, APs, PPs, and DPs, which is meant to be a representative, although not exhaustive, list of phrasal categories (for previous phrasal approaches, see Barton, Chap. 1, this volume). The reader will notice that some of these examples include subjects, in which case I refer to them more specifically as small clause nonsententials or as root small clauses. Thus, the term nonsentential will be used as a cover term for both single-phrase (nonclausal) utterances (e.g., Nice lady! (1)) and root small clauses (e.g., Problem solved (6)), the latter including a subject and a predicate. As a working hypothesis, I propose that a nonsentential is any utterance whose structure can be analyzed as smaller than a (full) sentence, that is, smaller than a TP. In other words, I consider the TP layer as a cutoff point between what we perceive as a nonsentential and what we perceive as a full clause/sentence, at least in English and other languages with grammaticalized Tense (in this respect, I do not consider infinitive to clauses as complete TP projections; see also n. 9). Each of the following isolated noun phrases (NPs) surfaces in an unusual form: (1) and (2) surface without articles, which would typically be obligatory if these phrases were integrated in full sentences; the pronouns in (3) surface with the accusative case, even though they correspond to the subjects of the predicate ate the pie.
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.3 (173-269)
Small clauses and phrases at the root
NP: Noun Phrase (1) Nice lady! (2) Old grudge. (3) Me./Him./Them.
(in response to: Who ate the pie?)
The Verb Phrases (VPs) in (4) through (15) are equally unusual. Not only are they headed by a nonfinite form, either a participle as in (4) through (9), or bare infinitive form as in (10) through (15), but many of these nonfinite VPs actually include subjects, as in (6), (7), (9), (13), (14), and (15). And, just as is the case with the isolated NPs in (1) through (3), these subjects also surface in unexpected forms: The subject in (13) surfaces in the Accusative Case form, whereas the subjects in (6), (7), and (9) surface without an article, which would otherwise be required in finite sentential counterparts. VP: Verb Phrase (4) Felled by flu.
(in a telegram)
(5) Arrested US border.
(in a telegram; Barton 1998)
(6) Problem solved. (7) Point well taken. (8) How about staying at home? (9) Washing machine dancing across the floor? (RepairClinic.com, March 2005) (10) Play baseball.
(in response to, e.g.: What did you do?)
(11) Why not take a holiday? (12) Get lawyer./Ask any lawyer.1 (13) (What?) Him worry?!
(Entertainment Weekly, 11/12/2004)
(14) Sarah do a thing like that?! (Never!)
2
(15) John be rude to a friend?! (That’s certainly not like him.)
(Davies 1986) (Davies 1986)
Adjective Phrases (APs) and Prepositional Phrases (PPs) can likewise appear with or without subjects, as illustrated in (16) through (25). When they appear with subjects, these subjects again show the same properties as do isolated NPs as in (1) through (3) and subjects of nonfinite verb phrases above: pronouns surface in accusative case forms (22), whereas nouns typically lack determiners (20), (25). AP: Adjective Phrase (16) Fond of animals?/Not fond of animals? (17) Urgent./Not urgent. (18) Not sweet.
(in response to: How do you like your coffee?)
(19) Very sick. (20) Car dead./Battery dead.
(in a telegram)
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.4 (269-343)
Ljiljana Progovac
(21) John tall?! (I’d never have described him that way.)
(Davies 1986)
(22) Me first! PP: Prepositional Phrase (23) Still at JFK./Not at JFK. (24) Martha Stewart in jail. (25) Class in session. Determiner Phrases (DPs) also occur as nonsententials, either as nonclausal phrases in (26) and (27) or small clauses in (28) (the latter may be analyzable as a Number Phrase instead, as per Ritter’s (1991) and Lobeck’s (1995) proposals). DP: Determiner Phrase (26) That man! (27) Rob’s mom. (28) This a bargain?! (I don’t think so.) Not only is there a preponderance of such nonsentential utterances in everyday (casual) speech, as well as in special registers, such as telegraphese, titles, headlines, and so on (Paesani, Chap. 6, this volume), but this nonsentential grammar option may also characterize much of the speech of language learners (L1 and L2 acquisition), aphasiacs, and pidgin speakers (see the relevant chapters of this volume: Potts and Roeper, Chap. 7, and Work, Chap. 8, for language acquisition; Kolk, Chap. 9, and Siple, Chap. 10, for aphasia; Winford, Chap. 11, for pidgins).3 Although nonsentential types, their interpretation, and their frequency of use may vary from group to group, it is possible that all nonsententials have the same minimal grammar at their core, a grammar that is moreover systematic and rule governed. Theoretically speaking, all other things being equal, the base generation of nonsentential data as phrases is much more economical, and thus much more in the spirit of Minimalism, than any sentential derivation. There is in fact, as I argue, nothing in the theory that would preclude base generation of structures smaller than TP. First, the derivation of each phrase or sentence in the current Minimalist framework proceeds in a bottom-up fashion, and concludes with the last node for whose projection there is linguistic evidence or need. It is the very essence of Minimalism that only phrases for which there is linguistic evidence will be projected. Just observing that a certain utterance pertains to a certain time does not constitute linguistic evidence for the presence of the Tense node. Not only is there no evidence for TP structure in nonsententials, but there is in fact counter-evidence, which comes from the previously observed patterns of default accusative case and article omission. In addition, according to X-bar theory, there is no real boundary between a phrase and a sentence, given that a sentence is but a phrase whose head happens to be Tense. Thus, there is nothing in the theory per se that would privilege the projection of a TP
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.5 (343-401)
Small clauses and phrases at the root
over other projections (but see Section 3 for possible semantic arguments for privileging TP). In this respect, even the distinction between TP and non-TP structures is arbitrary, and I propose it here only as a working assumption, given that the effects of the absence of Tense are much more readily observable, at least in English, than the effects of the absence of other functional projections. Similarly, the term nonsentential is also potentially misleading, because it is, first of all, a relative term in the sense that an utterance will be perceived as a nonsentential only if there are more complex utterances that can be analyzed as sentences. In addition, in order to define this relative term nonsentential, one would first have to have a clear definition of what a sentential, or a sentence, is (and vice versa, perhaps). Although my proposed distinction between sentences and nonsententials, in terms of the presence versus absence of the tense node, may be feasible for English, it will not necessarily work cross-linguistically or diachronically (see Section 6 and the Epilogue, Section 3, for some historical discussion). But, as with so many other fundamental concepts, their precise characterization is usually the result of an investigation, rather than a prerequisite for it; my chapter is written with that kind of investigation in mind. This chapter is organized in the following way. Capitalizing on the observed correlation between Tense and structural nominative case, Section 2 spells out the basic proposal of this chapter: the analysis of nonsententials as tenseless syntactic objects. This section is only an outline of the analysis, which receives its rationale and defense in subsequent sections. For example, Section 3 illuminates the current understanding of the relationship between sentencehood and Tense in syntactic and semantic theories, providing the background for my analysis of nonsententials as tenseless. Of crucial importance in this respect is the current analysis of (full) sentences as Tense Phrases, that is, phrases headed by Tense. Section 4 focuses on the current theoretical take on the interrelationship among Case, Tense, and the use of determiners (and DPs), which will be argued to be consistent with my analysis of nonsententials. Section 5 highlights the similarities between what are typically analyzed as embedded small clauses, and what I analyze here as root small clauses, providing further support for the analysis. Finally, Section 6 suggests that semantic considerations are not inconsistent with the tenseless analysis of nonsententials, especially when the element of context is taken into account, and especially when one recognizes a prevalence of nonindicative (nonassertive) interpretations among nonsententials. Drawing on an analysis of the historical development of Indicative, I argue that these “irrealis” interpretations are favored by nonsententials exactly because they lack Tense specifications.
. Outline of the proposed analysis As illustrated in the previous section, nonsententials exhibit the systematic absence of Tense and Agreement, which correlates with the absence of structural nominative case on subjects, whether in nonclausal or small clause nonsententials. The absence of structural nominative case is most readily observable with subject pronouns in
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.6 (401-455)
Ljiljana Progovac
the (default) accusative case in English or with the possible absence of articles, even with count nouns in argument positions (see relevant subsections of Section 4 for the characterization of Default Case (4.3) and for reasons for article absence (4.2)).4 This correlation is further supported by the examples that follow, which show that when it comes to Tense and nominative case in English, it is both or neither; the intermediate possibilities are at best marginal, and invoke special interpretations:5 (29) a. Battery dead./Problem solved. b. The battery is dead./The problem is solved. c. ?*The battery dead./?*The problem solved. d. ?*Battery is dead./?*Problem is solved.6 (30) a. Me first!/Him worry?! b. I am first./He worries. c. ?*I first./??He worry?! d. *Me is/am first./*Him worries. In the spirit of Barton (1990, 1991, 1998) and Barton and Progovac (2005), my proposal for all the nonsententials of the type illustrated in (29) and (30) is that they involve base-generated phrases, and not full sentences that have undergone deletion or ellipsis. As for the nonsententials in (1) through (5), (8), (10) through (12), (16) through (19), (23), (26), and (27), the claim is that they are base generated NPs, PPs, VPs, and so on. Even though my focus in this chapter is on small clause nonsententials, all the arguments that have to do with Tense and Case are also relevant for these singlephrase (nonclausal) nonsententials. The basic argument is that each of these phrases is a complete syntactic object, with no features remaining to be checked, leaving us with no motivation to project an additional projection. For example, the NP me in (3) is selected without specific Case features, which explains not only why it surfaces in the default accusative form, but also why it does not require a projection of Tense, in contrast to an answer such as I did, in which nominative case requires the tensed element in order to surface. Likewise, the VP answer to a question in (10), Play baseball, can be analyzed as involving the verb form without any Tense or Case features, a nonfinite form, which thus does not require (in fact, does not motivate) a projection of TP, or nominative case checking (see also n. 19 for similar question-answer strategies in Sanskrit and Old Irish). Small clauses with nonfinite verb forms are discussed later.7 I return to the interpretation of single-phrase nonsententials in Section 6. As for the examples in (6), (7), (9), (13) through (15), (20) through (22), (24), (25), and (28), I analyze them as phrases as well, but phrases that happen to have a subject in their specifier position, as the result of predication. I refer to such phrases as small clause nonsententials, or root small clauses, as mentioned earlier. In fact, phrases with subjects (VPs, DPs, APs, and PPs), when embedded in other clauses, are analyzed in the syntactic literature as small clauses, a well-established concept (see Section 5). By analyzing nonsententials with subjects as small clauses, I am not in-
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.7 (455-502)
Small clauses and phrases at the root
troducing a new concept, but only extending the use of the small clause to syntactic isolation/root contexts. Adopting the phrasal small clause structure advocated in Stowell (1981), to be further discussed in Section 5, I analyze small clause nonsententials as illustrated in the diagrams in (31) through (34) (see also Barton & Progovac 2005). This basic small clause analysis has been proposed by Radford (1990) for similar “telegraphic” utterances in first language acquisition, and by Ouhalla (1993) for agrammatic aphasia (see n. 3). VP
(31)
V’
Him
worry?! AP
(32)
A’
John
A tall?!
(33)
PP Class
(34)
P’ P
N
in
session.
DP This
D’ D a
N bargain?!
(I assume that the negative not and (other) adverbial elements, such as the ones illustrated in the examples (11), (16), (18), and (23) are attached by adjunction, but I leave the exact nature of attachment in these cases open for future research.)
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.8 (502-555)
Ljiljana Progovac
Current syntactic theory in fact claims that every sentence starts out as a small clause, where the subject and the predicate stand in a thematic relationship (see, e.g., Stowell 1978, 1981; Burzio 1981; Kitagawa 1986; Koopman & Sportiche 1991; Chomsky 1995; Radford 1997; Hale & Keyser 2002). For example, the sentence with a copular verb such as John is tall derives from the underlying Adjective Phrase (AP) small clause (John tall), where the subject John moves from the subject position of the small clause (AP) to the subject position of the sentence (TP) in order to check nominative case features against the Tense features of the copular verb (35). Case in English, including nominative, is taken to have Uninterpretable features, that is, purely formal features, which need to be checked/eliminated during the derivation. What requires then such a transformation from the small clause to a TP is the selection of lexical items with purely formal features that need to be checked, such as nominative case features: (35) [TP is [AP John [A’ tall]]] → [TP John [T’ is [AP t [A’ tall]]]] Similarly, as illustrated in (36), the sentence He will worry starts out as a small clause VP (He worry), which gets transformed into a full sentence (TP) by the merger of Tense (will), and by the subsequent movement of nominative he to the specifier of TP for case-checking purposes. (36) [TP will [VP he[V’ worry]]] → [TP he [T’ will [VP t [V’worry]]]] Notice that the verb worry itself is selected without any Tense or Case features here, which is the reason why it does not have a relationship with the Tense node. That is exactly what I am claiming for the verbs in nonsententials such as Play baseball (10) and Him worry (13) – that they are selected from the Lexicon with no Tense or Case features, a possibility that is clearly independently available for lexical items. In fact, each full sentence seems to have a nonsentential counterpart, the one in which Tense has failed to merge. Such nontensed counterparts also occur as embedded small clauses, as discussed later, and in Section 5. My proposal is thus that a root small clause results when Tense fails to merge over a small clause. This analysis is consistent with the current Minimalist idea that structure is built bottom up, rather than top down, and that it is built only in the presence of linguistic evidence for further projections. As will be argued throughout the chapter, there is no linguistic evidence for the presence of a TP projection in nonsententials. More specifically, my argument is that if lexical items are selected without formal features that need to be checked, then the creation of a TP becomes unnecessary. So, if the selected lexical items are the default case him, and tenseless worry, one can only create the small clause Him worry, the small clause that is nonetheless a legitimate syntactic object, because it contains no unchecked Uninterpretable features. The small clause Me first (22) would be created in a parallel fashion. Similarly, the small clause Battery dead (20) is created by merging the tenseless predicate dead (Adjective) and the NP battery, arguably also selected without formal Case features, in the default case form. As will be discussed in detail in Section 4, it is the choice of a default case form
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.9 (555-637)
Small clauses and phrases at the root
that makes it possible to omit the article. The analysis directly extends to the PP small clauses such as Class in session (25). This analysis can be extended to passive examples as well, such as Problem solved/Point well taken in (6) and (7). One can generate the subject Problem in the specifier position of VP (as a theme argument), and the participle solved in the head V position of VP, essentially creating a small clause. If the NP is selected without a specific Case feature (which in this case would be signaled by the absence of the article), and if no tensed lexical item is selected, then there will be no Tense/Case features to be checked, and the small clause will be a well-formed syntactic object. Notice that on this analysis, in which themes are merged in the specifier positions of their predicates (see, e.g., Chomsky 1995; Baker 1996), even passive small clauses can be derived by a single merger, and need not involve movement. To appreciate the parallelism between what I analyze here as root small clauses and embedded small clauses, consider some additional examples of root small clauses, immediately followed by their embedded counterparts. (37) Head ache?/I felt [my head ache]. (38) Heart throb?/I heard [her heart throb]. (39) Her give up?! (Never!)/I never saw [her give up]. (40) Them not go to the party?!/I cannot see [them not go to the party]. (41) (What?) Me worry?!/You never see [me worry]. (42) Heart not healthy?/They consider [her heart not healthy]. (43) Car dead./They pronounced [her car dead]. (44) Martha Stewart (still) in jail./They want [Martha Stewart in jail]. (45) Class in session./I saw [the class in session]. (46) This a bargain?!/I consider [this a bargain]. (47) Him (not) a loser?!/They consider [him a loser]. A careful reader will notice, however, that there is one difference between root small clauses and their embedded counterparts: the subject NPs of embedded small clauses cannot drop their determiners (with singular count nouns), the way their root counterparts can (see the examples in (37), (38), (42), (43), and (45)). This difference can be made to follow from independent factors, and should not need to be stipulated. It is commonly held in syntactic theory that embedded small clause subjects check structural accusative case (as opposed to default case) with the preceding verb, through a mechanism of Exceptional Case Marking (ECM). It is possible that once a small clause is merged as a complement of a verb in a full sentence, it becomes subject to sentential rules, including structural Case checking (see also n. 10). If this is correct, then it follows that a root small clause is even “smaller” than an embedded small clause.8
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.10 (637-680)
Ljiljana Progovac
. Sentence and Tense This section provides a brief overview of the current understanding of the relationship between Tense and sentencehood in the theory, both from syntactic and semantic perspectives. It is an attempt to tease apart the defining, essential components of the current theory from its nonessential, unproven assumptions. The basic argument is that the analysis of nonsententials explored in this chapter follows naturally from the theory if only some of the unproven assumptions are abandoned or reevaluated. At the heart of a typical sentence/clause is a predicate/argument relationship – this much seems to be uncontroversial in the linguistic literature. Where theories diverge is with respect to how much more structure a sentence/clause must have. In this respect, syntactic theories have seen a most radical change. A historical perspective helps one recognize that our knowledge of sentential structure is still very much in flux. Only around 25 years ago, syntacticians (e.g., Chomsky 1981) represented sentences/clauses as typically consisting of a subject, an auxiliary slot (referred to as AUX or INFL), and a verb phrase (VP): (48)
S NP AUX VP John will stay
However, with the advent of X-bar theory and the push for binary branching only, as well as with the attempt to provide parallel structures for phrases and sentences (see, e.g., Stowell 1981; Pesetsky 1982; Kayne 1984; Chomsky 1986; Pollock 1989), the structure of the sentence has been revised in such a way that the AUX/INFL node now heads the sentence, and the whole sentence is now typically called Tense Phrase (TP), to accommodate the assumption that Tense is the crucial element of the AUX/INFL node, and, as a result, the head of the sentence (see also Ross 1967). Here, I am presenting only the rudimentary structure of the sentence, reasonably well established and accepted in the Minimalist framework (for additional functional projections postulated over VP, see some discussion in Section 5). TP
(49) NP
T’ VP
Johni T will
ti
V’ stay
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.11 (680-738)
Small clauses and phrases at the root
As explained in Section 2, in this framework every sentence/clause is analyzed as a small clause underlyingly, which is transformed into a TP only after the merger of Tense, and the subsequent movement of the subject from the specifier of VP to the specifier of TP position, as indicated by the co-indexing. The last 25 years or so have seen an enormous advance in characterizing the sentential structure of many languages, thanks to the analysis in (49), and I adopt this basic analysis for typical (full) sentences and clauses in these languages, including English. In order to justify (or just understand) the analysis that renders Tense the head of the sentence, syntacticians often claim that Tense is an indispensable element in a sentence, without which it is not possible to assign the sentence its meaning or truth conditions. For example, the following sentences will have different truth conditions, and thus different meanings, due to the different Tense/Time specifications: (50) John is tall./John will be tall./John was tall. Current semantic theories consider the truth conditions of a (declarative) sentence the most important aspect of its meaning, and in order to establish the truth conditions it is necessary to situate the predication in time, which is normally achieved via Tense (for an overview of the semantic and philosophical approaches to the relationship between meaning and truth, see, e.g., Larson & Segal 1995; Chierchia & McConnell-Ginet 1990). What I question in this chapter, however, is the premise that Time must always be represented via a Tense node. Instead, I allow that, in principle, Time specification can also be achieved by providing a time frame pragmatically, by ostension, or by the use of temporal adverbials (regarding the latter, see, e.g., Winford, Chap. 11, this volume, and Winford 2003, for pidgin languages). For example, the spatio-temporal frame can be provided by a photograph, labeled as Me dancing. Also, as discussed in Paesani (Chap. 6, this volume), the time frame can be provided by a special register, such as headlines (e.g., Athens in crisis. . . ). In each of these cases, the time is not specified by Tense, but rather by a discourse time frame. In this respect, it may be that timeless statements found in proverbs such as Nothing ventured, nothing gained also frequently occur as nonsententials by virtue of not needing a time specification (see also Epilogue, Section 5.3). What is typical of many nonsententials is that their event is situated in the here and now as a default place/time (e.g., Battery dead; Class in session). It may well be that the default time specification is universally present time, or time of utterance, and it is not a coincidence that in a variety of languages, such as Russian, Hebrew, or dialects of English (e.g., African American Vernacular English (AAVE), as in She pretty, see, e.g., Labov 1969; Rickford, Ball, Blake, Jackson, & Martin 1991), one can omit the copular verb when present time is to be expressed, but not otherwise (see also Edwards, Chap. 12, this volume, for AAVE and Guyanese Creole). Thus Russian uses copulaless utterances only to express the present time (see, e.g., Pesetsky 1982; Schein 1995):
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.12 (738-799)
Ljiljana Progovac
(51) Ivan veren. / Ivan byl veren. ‘Ivan (is) faithful.’ ‘Ivan was faithful.’ Similarly, Hebrew can optionally omit the copula in the examples pertaining to the present time, but not otherwise (see Rothstein 1995). (52) Dani (hu) nehmad / rofe / al ha-gag. Dani m.sg nice / doctor / on the-roof ‘Dani (is) nice/a doctor/on the roof.’ Interestingly, in contrast to the examples with the copula, root small clauses without a copula in (52) can only be interpreted as stage level (see Déchaine 1993; Carlson 1977), which basically means that they pertain to the time of the utterance. The present tense in the semantic literature is normally characterized as “coincident with the time of the context in which the sentence is produced” (Chierchia & McConnell-Ginet 1990: 266). In fact, Tense has been analyzed on a par with pronouns and demonstratives as an indexical or deictic category, whose meaning varies with context, and whose interpretation can be determined by ostension (pointing) (Chierchia & McConnell-Ginet; Bar-Hillel 1954). I interpret this to mean that a nonsentential that has no Tense node can nonetheless be situated/anchored in Time by the context of the utterance, or by some other time frame, and thus constitute a truth-evaluable assertion. This is also in keeping with Stainton’s view (e.g., Chap. 4, this volume), as well as with the argument in Avrutin (2004: 74) that “the function of a functional category can be sometimes taken over by the context.” If this discussion is on the right track, then situating an utterance in time need not be done via a Tense node, even though in full sentences (TPs) it must be, at least in languages like English, in which the expression of Tense/Time is grammaticalized. The relevant chapters in this volume explore the possible absence of the Tense node in question-answer pairs (Casielles, Chap. 5), special registers (Paesani, Chap. 6), pidgins (Winford, Chap. 11), Guyanese Creole (Edwards, Chap. 12), language acquisition (Potts & Roeper, Chap. 7; Work, Chap. 8), and aphasia (Kolk, Chap. 9; Siple, Chap. 10). Two reviewers of this volume point out that the projection of Tense in every sentence may be needed in order to satisfy interface conditions, in particular, Logical Form (or the Conceptual-Intentional (C-I) interface), which is often taken to operate only over full sentences. According to this line of reasoning, meaning/interpretation cannot be assigned to anything that is not a sentence, that is, a full TP. Although this assumption is typically present in current syntactic theory, it has not been shown to follow from anything essential in the theory. Instead, it remains at the level of assumption, or stipulation. In fact, Chomsky in his most recent work (e.g., 2005) no longer assumes that only sentences with Tense can interface with the C-I system. He specifically proposes that some special subparts of sentences, which he calls phases in this framework, for example, vPs, and possibly DPs, can directly interface with both C-I and Sensory-Motor (S-M) systems. I believe that the theory leaves the door wide open
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.13 (799-853)
Small clauses and phrases at the root
for interface conditions to interact with nonsentential syntactic objects, and I explore this interaction in more detail in Section 6.
. Noun Phrase, Determiners, and Case This section explores in some theoretical detail the correlation between tenselesness and caselesness, the backbone of the syntactic analysis developed in this chapter. Section 4.1 looks at the theoretical and empirical significance of the DP analysis, especially in its relation to Case and Tense. Section 4.2 clarifies the correlation between the lack of structural Case and the possibility of omitting articles. Section 4.3 establishes a need to appeal to Default Case, not only in the analysis of nonsententials, but more generally, and offers a way to implement the notion of Default Case into the current theory.
. DP as a theoretical construct Parallel to the addition of the TP layer over the verb phrase, the theory of syntax has also seen the addition of a layer of structure over the noun phrase, referred to as the Determiner Phrase (DP). Some 20 years ago (e.g., Chomsky 1981, 1986), it was still generally assumed that the noun heads the noun phrase, and that determiners are premodifiers in the noun phrase: NP
(53) D
N
the/this
book
Due to the influential work of Szabolcsi (1983, 1987) and Abney (1987), it is now commonly accepted in the syntactic literature that determiners, and not nouns, head such nominal phrases, now called Determiner Phrases – not only in English, but cross-linguistically (for arguments for the DP proposal, as well as for the additional functional projections in the DP domain, the reader is referred to the cited references, as well as to Alexiadou & Wilder (1998) for more recent work). (54)
DP D
NP
the/this
book
It is also well established that pronouns normally occupy the D position, rather than the N position (e.g., Postal 1969; Longobardi 1994). Based on cross-linguistic data, Cardinaletti (1993) and Progovac (1998) further argue that pronouns may be gener-
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.14 (853-909)
Ljiljana Progovac
ated in the N position, and then moved to the D position for (Case) feature-checking purposes. In fact, the occurrence of a (default case) pronoun in N is evident in English examples such as The real/new me is emerging, where the presence of a determiner shows that the pronoun is in N, not in D. It may be that default case pronouns do not move to D positions, given that they do not check structural Case, and that examples such as (3), (13) and (22) from Section 1 illustrate pronouns as NPs, not DPs (see also Schütze 2001; Barton & Progovac 2005). Default Case is further discussed in Section 4.3. It has further been proposed that those NPs that serve as arguments in a sentence must have DP structure. In English, arguments require a determiner with count nouns (55), whereas nonargument NPs, such as predicates (56) or vocatives (57), can occur without determiners (e.g., Stowell 1989 and previous work cited there; also Longobardi 1994): (55) *(The) doctor locked the office. (56) She is doctor to the stars in Hollywood. (57) Doctor, can you please examine me? I take the structure in (54) to be well established cross-linguistically, and I adopt it here for English and languages comparable to English in this respect. I also consider it firmly established that an argument in a typical (full) sentence, which involves structural Case, has to be a DP, as per the references mentioned before. However, my claim is that there are instances of nonsententials in which the argument of a predication does not have to be a DP, but rather can be an NP. In addition to the examples with determinerless noun phrases in the subject position (e.g., (6), (7), (9), (20), (25)), or determinerless noun phrases in isolation ((1), (2)), I argue that these include the examples with subject pronouns in the default case, such as (13) and (22), as well as isolated pronouns in examples such as (3). In fact, the proposals of TP and DP are interrelated in current syntactic theory – they reinforce each other. The standard assumption is that an argument is required to be a DP in a typical sentence in order for it to be visible for interpretation, that is, visible at LF (Logical Form). The claim is that a DP is visible as an argument only if it is assigned abstract/structural Case (e.g., nominative for subjects in English), and in order to be assigned structural Case, the argument has to be a DP. In other words, structural Case is the property of DPs, and not NPs. On the other hand, it is standardly assumed that Tense, the head of the TP, is responsible for assigning/checking the nominative case of the subject DP. Thus, only sentences with (finite) Tense can have nominative subjects, but, for example, embedded small clauses or infinitive clauses cannot (even though, of course, they still receive structural accusative case through ECM; see also Section 2):9 (58) She/*Her ate the dinner. (59) I consider [her/*she intelligent].10
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.15 (909-972)
Small clauses and phrases at the root
(60) [For her/*she to eat the dinner] would be unwise. There is thus a clear correlation in English between having a full sentence (TP) and having full, structurally case-marked DP subjects, and I do not question this correlation. On the other hand, I call into question the premise that an argument has to be a DP in order to be interpreted. I will argue that it is possible to have a nonsentential in which neither Tense nor structurally case-marked determiners occur, but that can still express a predicate-argument relationship, as well as serve as a truth-evaluable assertion. It is worth pointing out that some of the claims of current syntactic theory are essential components, whereas others are mainly attempts to justify or understand those essential claims. For example, the postulation of the functional projections such as TP and DP is a central component, as well as the claims that Tense participates in checking the nominative Case feature, and that structural Case checking involves full DPs. On the other hand, that a Tense node must project in every utterance, or that arguments can only be visible for interpretation if they are DPs, are only nonessential assumptions in the theory. What I am reevaluating here are thus not the essential claims of the theory, but rather some of its unproven assumptions. That standard Case assignment to full DPs is not the only way to make an argument visible is argued for extensively in Baker’s (1988, 1996) influential work on incorporation. His main proposal for object incorporation in polysynthetic languages is that the head noun of the object NP (and not a full DP) incorporates into the verb by head adjunction, and that the noun-incorporation process establishes a relationship between the two heads, sufficient for argument visibility. In other words, head-to-head incorporation of the argument into the predicate V is an alternative way to render the argument visible. In addition to incorporation, Baker (1988, 1996) proposes that Agreement is yet another mechanism of making an argument visible. If this analysis of small clause nonsententials is on the right track, then it follows that the argument’s visibility can be achieved in still another way, possibly by the argument remaining in the lexical projection of its predicate, that is, in its theta position (for nonclausal nonsententials, see Section 6). If Case and Agreement are morphological mechanisms for ensuring argument visibility, it may be that they are only needed for arguments that get displaced from their original theta positions, that is, from their predicates. The precise analysis of this matter, however, remains open for future research. The basic conclusion of this section is that nonsententials, as a result of being tenseless, differ from their sentential counterparts in that their (subject) arguments do not have to be DPs, but can rather be NPs, including the default case (NP) pronouns. The following section shows how this analysis can be used to explain the possibility of article omission with nonsentential subjects.
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.16 (972-1032)
Ljiljana Progovac
. Tense, Case, and article omission As pointed out in the previous section, the basic theoretical requirement on structural Case assignment in current syntactic theory is that it requires a DP. Structural Case here contrasts with Default Case, which is argued in this chapter not to require a projection of DP (see also Sections 4.1 and 4.3). Thus, it is assumed in syntactic theory that there is an implication of the following kind: (61) If structural Case is assigned to a category, then that category has to be a DP. This is a one-way implication, however. From this it does not necessarily follow that if something is a DP it must receive structural Case. In fact, there are exclamatives that are clearly DPs, but that are not considered to engage in structural Case checking, as in (62). Likewise, pronouns, which are typically analyzed as Ds, can be used for summoning in vocative functions (63), another position where structural Case is not checked.11 In addition, DPs are attested in predicate positions (64), which also do not seem to involve structural Case (see also the discussion regarding examples (56)–(57)): (62) That man! (63) You, come over here! (64) John has been (the) president of our club for many years. What is often claimed in the syntactic literature is that an argument in a sentence must have structural Case in order to be “visible” for interpretation (this is clearly a rather vague characterization, bound to see more specific characterizations in the future). Given the implication in (61), this in turn means that a typical argument in a full sentence must be a DP. This chapter only modifies this basic proposal for nonsententials, and argues that (subject) arguments within nonsententials do not participate in structural (nominative) Case checking, and that thus they need not be DPs. For the reasons pointed out earlier, this is not to say that the determiner must be absent. But one may wonder why the determiner would be there if it is not required. Would not Economy preclude that? If the determiner’s sole function were formal Case feature checking, then such a determiner would indeed be excluded in nonsententials, for Economy reasons. But, in addition to this purely formal function, determiners also have a referential function (as well as possibly some other functions, such as quantificational). Longobardi (1994: 610) dubs these two functions “expletive” and “substantive,” respectively. Thus, the determiners in (62) and (63) can be considered to be used solely for referential purposes, that is, in their substantive functions. Conversely, determiners, typically articles, can be used for purely formal reasons, that is, for case-checking purposes. Consider again the pattern illustrated in Section 2: (65) Problem solved. (66) ?*The problem solved.
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.17 (1032-1097)
Small clauses and phrases at the root
(67) ?*Problem is solved. (68) The problem is solved. Although it may seem that the definite article in (68) is required for referential reasons, to establish definite reference, its use for that purpose is obviously not necessary, given the possibility of (65), with the same definite interpretation. The definite reference seems to be established in (65) as a given of the context. In fact, if the definite article is used in a small clause such as (66), it results in a marked interpretation, in which the problem is now to be interpreted referentially, almost as a name/label for a group of people who can be identified as “the problem” (see n. 5). I conclude that the article in (68) is there for formal reasons, to fulfill the sentential requirement on structural Case checking, which requires a DP. This dual function of determiners is also evident with the pronoun it, which has exactly these two kinds of functions: expletive (69) and referential/substantive (70): (69) It seems clear that John is ill. (70) It arrived safely. (e.g., a package) When used referentially (70), the pronoun it has a double duty: both to check structural Case and to refer to an object in the real world. In (69), the pronoun is taken to have a purely formal function.12 It follows from this analysis, then, that subjects of nonsententials will occur without determiners whose sole function would be formal feature checking; on the other hand, they can occur with substantive, referential determiners. This explains the optional omission of articles with nonsentential subjects, as established empirically in Section 2.
. What is Default Case? The analysis of nonsententials in this chapter relies heavily on the notion of Default Case. This section points to the need to appeal to Default Case not only in the analysis of nonsententials, but also more generally. Next, it offers a way to implement the notion of Default Case into current theory, as well as some speculations on why Default Case should exist. Default accusative case in (present-day) English is most readily observable when it surfaces on pronouns in predicative positions (71); in noun positions, preceded by adjectives and/or determiners (72) (see also Section 4.1); when it serves as a label (e.g., [+/– Me] feature in Morphology for first-person singular, used, for example, in Spencer (1991: 216); in book titles: Me, e.g., by Delores Minor or by William Saroyan; see also Almighty Me by Robert Bausch); and possibly in coordinated structures (73). (In addition, it is quite possible that left-dislocation structures also involve default accusative case, an issue that is addressed in the Epilogue, Section 4, in the context of Merchant’s elliptical analysis.) These positions are typically considered not to involve structural Case.
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.18 (1097-1170)
Ljiljana Progovac
(71) It is me. (72) The real me is emerging. (73) [Me and my friend]/[My friend and me] wrote this book. As pointed out in Section 4.1, examples such as (72) point to the possibility of analyzing instances of other default case pronouns as occurring in N positions, rather than D positions. Such examples provide especially strong support for Default Case, given that nominative case, required by the tensed auxiliary, must be checked with the article in D, which is reflected in the third-person agreement on the auxiliary (see Schütze 2001). Because the and me refer to a single individual, and constitute a single argument, the theory would not allow them to receive two separate structural cases, leaving us with only one possibility: to analyze the case on me as default case. Thus, default accusative case is needed even independently of the analysis of root small clauses (e.g., Me dancing; Me worry?!) or NP answers to questions (Me). Barton (1990, 1991) and Barton and Progovac (2005) invoke Default Case in their discussion of nonsententials and Roeper (1999) uses it in his characterization of default grammars. Default Case is also often invoked to describe children’s early speech (e.g, Radford 1990; Lebeaux 1990; Vainikka 1990; Schütze & Wexler 1996). Notice that Jespersen (1954: 372–377) offers some English examples of what I analyze here as root small clauses (italicized), but with nominative subjects, suggesting that the English shift from nominative to accusative default case was relatively recent, which is also quite evident in the prescriptive pressures to use nominative case in some of the default situations described earlier. (74) Why, his grandfather was a tradesman! He a gentleman! (75) She a beauty! I should as soon call her mother a wit! (76) He an enemy of our beloved church! He esteemed and honored it. Kiparsky (1968) argues convincingly that nominative was the default case in ProtoIndo-European and in early Indo-European, and many present-day languages have not changed in this regard. For example, it is easy to establish that nominative is the default case in Serbian: it surfaces on isolated citation forms, such as single-NP exclamatives (77) and labels (78), in predicative positions (79), and in root small clauses (80): (77) Kiša! /*Kišu! rain-nom rain-acc ‘Rain!’ (78) Rakija /*Rakiju brandy-nom brandy-acc (79) To je Grozdana/ona/*nju. this is Grozdana-nom/she-nom/her-acc ‘This is Grozdana/her.’
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.19 (1170-1231)
Small clauses and phrases at the root
(80) Ona/*Nju predsednik kluba?! (Vi se šalite.) she-nom/her president club-gen you se kid ‘Her president of the club?! (You must be kidding!)’ Nominative would also be used as an isolated comment to a previous utterance (81), comparable to the English Old grudge comment, given in (2): ˇ (81) A: (Cula sam šta mu se dogodilo.) Stara priˇca! ‘(I heard what happened to him.) Old(-nom) story(-nom)!’ Of course, the general question is why languages would have Default Case at all. As Schütze (2001: 231) puts it: “I am not sure that it makes sense to try to answer that question while having no answer to the question of why Case exists at all.” But in fact, there is a reasonable answer to this question. Syntacticians would have to agree that, for whatever reason, there are syntactic positions in which NPs/DPs occur, but which are not specific Case positions or argument positions (see Section 6.2 for possible diachronic reasons for the existence of Default Case). Such non-case positions include at least predicates, citation forms, and titles, as pointed out previously. An approach that would try to ascribe all or some of these uses of accusative in English to several instances of special structural Case would miss the generalization that all these instances of Default Case are lumped together into a single case form (whatever it is), not only in English, but in other languages as well. In addition, when that default form changes over time, as it did in English, it changes across the board, for all the instances. A reviewer raises the question of how exactly Default Case can be implemented in current syntactic theory. In fact, it is generally assumed that lexical items are selected from the Lexicon with certain features (e.g., Chomsky 1995; see also Radford 1997). Thus, the pronoun he is selected from the Lexicon with the feature Nominative, the pronoun him with the feature Accusative. All that one needs to add is the possibility for the pronoun Him to occur in the Lexicon with the feature Default Case, or No Case, which in English coincides with accusative case. It is likewise not unusual for one phonological form to correspond to multiple lexical items, with distinct grammatical features. For example, given the current theory, there are two lexical items corresponding to You, one checking accusative case, and one checking nominative case. Moreover, Radford specifically points out that determiners, such as the, are listed in the Lexicon twice, once with the Accusative feature, and once with the Nominative feature, and that the appropriate form must be selected for each case-checking situation. Given the proposal developed here, each D element would also need to be listed in the Lexicon with a Default Case feature, including determiners and (accusative) pronouns. Thus, the accommodation of Default Case/No Case does not really add any extra burden to the theory, whether one considers this particular version of the theory, or some other; it only adds some additional lexical items, and the theory has a natural way of accommodating such items in the Lexicon – no better or worse than accommodating other case distinctions. Further discussion of Default Case in other languages, as
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.20 (1231-1303)
Ljiljana Progovac
well as Case cross-linguistically in general, can be found in the Epilogue to this volume, Section 4.
. Small clauses in current syntactic theory It is rather common in current syntactic literature to assume the existence of structures that do not carry Tense, but still express predicate-argument relationships: the so-called small clauses. The very term small clause clearly suggests that these are clauses smaller than typical clauses/sentences in some respect. This section reviews the most influential arguments for the existence of small clause syntactic constituents, reinforcing the main claim in this chapter that many nonsententials can be analyzed as root (unembedded) small clauses. The ground-breaking work on small clauses is Stowell (1981, 1983) (see also Manzini 1983; Contreras 1987; Rothstein 1995; Chomsky 1995; Hale & Keyser 2002), who analyzes various types of small clauses not as full clauses/sentences, but as basegenerated (lexical) phrases, including the bracketed phrases in the following examples: Adjective Phrases (APs) as in (82), Prepositional Phrases (PPs) as in (83), Participle VPs as in (84) and (85), and bare VPs as in (86) and (87): (82) I consider [AP John [A’ very stupid]]. (83) I expect [PP that sailor [P’ off my ship before noon]]. (84) We considered/feared [VP John [V’ killed or captured by the enemy]]. (85) There was [VP an alcoholic [V’ sitting in the room]]. (86) I heard [VP Jack [V’ come into the kitchen]]. (87) Nancy made [VP Scott [V’ take out the garbage]]. It is commonly accepted now that the bracketed strings in (82) through (87) are indeed constituents, and it is difficult to find a recent text on syntax that does not appeal, in some way or other, to small clauses (but see, e.g., Bresnan 1978; Gazdar 1982; Williams 1980, 1983, for early arguments against small clauses as constituents). The arguments for the constituency of small clauses are based on familiar constituency tests, such as coordination, movement, and binding theory (e.g., Stowell 1981, 1983; Contreras 1987; Chung & McCloskey 1987, Rothstein 1995). Chung and McCloskey add another, very powerful argument for the constituency of small clauses in Irish – their ability to occur in syntactic isolation/root contexts, a point to which I return later. In addition, Rothstein argues that Hebrew copulaless sentences are likewise root small clauses, which would also be the case with Russian copulaless sentences, as discussed in Section 3 (see also Edwards, Chap. 12, this volume, for root small clauses in AAVE and Guyanese Creole). Whereas Stowell (1981, 1983) specifically claims that small clauses cannot be sentences, other linguists have argued or assumed that small clauses are sentences, of the
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.21 (1303-1359)
Small clauses and phrases at the root
category S or S’ (e.g., Chomsky 1981, 1986; Kitagawa 1985; Chung & McCloskey 1987; Starke 1995).13 Notice, however, that it was much easier to make this assumption in a framework in which the Sentence node had no semantic/morphological content (see Section 3). Now that sentences have been reanalyzed as Tense Phrases (TPs), it is indeed hard to make the argument that small clauses are TPs, with no evidence at all of Tense. In fact, Chung and McCloskey (their n. 13) struggle with this issue, and actually conclude that Irish root small clauses cannot be projections of infl. Because infl (or Tense) is the head of the sentence in the current framework, this leads to the inevitable conclusion that these small clauses are not sentences, at least not in the TP/IP sense. The infinitive tense node in embedded contexts is often assumed to be anaphoric on (dependent on) the main/matrix clause tense (e.g., Stowell 1981), so that some kind of tense specification is available, and is established by a relationship between the infinitive to and the matrix tense node. But notice that the time of the embedded small clause is similarly dependent on that of the matrix clause, and yet there is no evidence of any tense node inside the small clause to mediate this dependence. I conclude that the time of the small clause can be determined contextually, relative to the time of the matrix clause, even without there being a tense node in the small clause itself. When it comes to root small clauses, my argument is that their time specification can likewise be determined contextually, whether by linguistic or pragmatic context (see also discussion in Sections 3 and 6). Chung and McCloskey (1987) show that small clauses in Irish can be used as adjunct phrases formed with the conjunction agus (‘and’), in constructions with the conjunction ach (‘but’), when selected as complements of certain verbs, such as tarlaigh (‘happen’), and, crucially, they can also be used in syntactic isolation, that is, as root small clauses: (88) a.
Ghaibh criú naomhóige isteach. came crew currach-gen in b. Iad righin fadthruslógach. them tough with-long-loping-stride c. Iad ag guailleáil rompu tríd an slua. them shoulder prog before-them through the crowd ‘The crew of a currach came in. They were tough and walked with a long loping stride. They shouldered their way through the crowd.’
Chung and McCloskey (177) state: “A very common type of discourse structure, in both oral and written narrative, consists of a sequence of clauses only the first of which contains a verb. The other clauses in the sequence are the kinds of predicative NP XP sequences. . .” Significantly, such root small clauses use the (default) accusative case for subjects, as pointed out in Chung and McCloskey (177). The correlation between the lack of Tense and the use of the default case in the subject position holds in Irish as well. In addition, Irish short answers to questions also employ accusative pronouns corresponding to subjects (Schütze 2001: 228).
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.22 (1359-1435)
Ljiljana Progovac
Note, incidentally, that English, in addition to having subcategorized embedded small clauses with verbs such as consider or with perception and causative verbs (82) through (87), also has small clauses in adjunct positions (89), as well as in coordinated positions (90), even though such uses may not be as productive as they are in Irish:14 (89) Susan/Her having gone, Fred left town. (90) I’m not going to have any woman rummaging about my house, and me in bed. Given the analysis of nonsententials proposed here, English would have all the varieties of small clauses attested in Irish: subcategorized embedded small clauses (with certain verbs), adjunct/coordinated small clauses, and root small clauses. Even among the proposals that analyze small clauses as tenseless, there are claims that small clauses necessarily have an Agreement Phrase (AgrP), which may be considered to be a sentential projection. Such proposals are based mainly on the presence of subject-predicate agreement in languages with gender/number agreement (e.g., Italian), as in Cardinaletti and Guasti (1992, 1995), Raposo and Uriagereka (1990), Contreras (1995), and Guéron and Hoekstra (1995).15 In contrast, Moro (1995, also 1988) has argued against an (obligatory) AgrP in Italian small clauses, offering examples such as (91), where the predicate (the cause of the riot) of the small clause does not agree with its subject (these books) in gender or number, suggesting that an agreement relationship between the subject and the predicate is not a necessary condition for building a small clause (which is also obvious in the English translation): (91) Gianni ritiene [questi libri la causa della rivolta.] Gianni believes these books-pl the-sg cause-sg of-the riot ‘Gianni considers these books the cause of the riot.’ My contention for root small clauses coincides with Stowell’s (1995) for embedded small clauses: projecting an AgrP (or another functional projection, such as Aspect Phrase) in small clauses would have to be required by an independent factor (such as strong Agreement features on the predicate adjectives or nouns, or evidence of movement), and is not a necessary or defining property of the small clause itself. Chomsky (1995) also suggests that small clauses should not be treated as AgrPs for theoretical reasons, as an attempt to obviate the need for AgrPs. In fact, when it comes to small clauses, Chomsky (354) suggests going back to “something like the original assumptions of Stowell’s (1978).” The following section returns to root small clauses and other nonsententials, and considers their interpretation.
. Tense, Time, assertion, and the role of context This section explores some semantic and pragmatic consequences of the phrasal/smallclausal analysis of nonsententials. It is rather tentative in nature, and is meant to open
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.23 (1435-1480)
Small clauses and phrases at the root
a discussion, rather than to conclude one. The basic argument is that any syntactic object, sentential or not, if syntactically well formed, that is, if it does not contain unchecked, purely formal (Uninterpretable) features, can be dispatched to the interfaces. Given this approach, the basic challenge is to determine how such tenseless objects can be anchored in time, and how they can express assertions. Section 6.1 focuses on root small clauses and phrases that are used to make assertions, and argues that the context, either linguistic or pragmatic, can provide such nonsententials with the spatio-temporal anchor necessary to yield truth-evaluable propositions and assertions. Section 6.2, on the other hand, points out that many nonsententials in fact are not interpreted as assertions, but instead receive a range of nonindicative, “irrealis” readings, such as incredulity, imperative, optative, and so on. I argue that these irrealis interpretations are favored by nonsententials exactly because they lack Tense specifications. This idea finds a parallel in the historical development of Indicative.
. The role of context If it is indeed true that a predication needs to be situated in time in order to qualify as a truth-evaluable proposition (and basis for an assertion; see Section 3), then, given the approach explored in this chapter, the conclusion has to be that a syntactic Tense node is not the only way to anchor a predication in time. Instead, a time anchor can be provided in a variety of alternative ways, including by the linguistic context in question-answer pairs, and by the pragmatic context, such as the context of the here and now, which is a frequent setting for nonsententials.16 As pointed out in Section 2, there is nothing in the theory to prevent the syntactic component from submitting a phrase or a small clause to the interfaces, as long as that phrase is a well-formed syntactic object. If the small clause Class in session is dispatched by syntax, it can still be assigned some interpretation in LF, in particular, that of a predication in which the predicate in session takes the theme class as its argument. The rest can then be deduced from the pragmatic context of the here and now. Avrutin (2004: 74) suggests that the function of a functional category, such as Tense or Determiner, can sometimes be taken over by the context (see also Avrutin 1999; Bouchard 2002). To be sure, this can also be seen from a different perspective: functional categories themselves may have developed exactly in order to reduce reliance on context, that is, they may have themselves taken over the functions initially performed solely by the context (see also the Epilogue). Notice that linguistic theory cannot isolate itself from pragmatic context in other areas either. For example, pronouns, such as he, she, and so on, are typical examples of so-called variable reference terms (together with Tense; see Section 3). As pointed out in, for example, Larson and Segal (1995: 201–202), pronouns get their semantic value in the context of the utterance, which can involve either a linguistic antecedent, as in (92), or “pointing” (another term used is ostension), as given in (93). (92) John does not think that he is tall.
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.24 (1480-1564)
Ljiljana Progovac
(93) He should sit between her and him. What this really means is that, in order to determine the meaning of the sentence, including its truth conditions, one must take into account an element of context, which can be provided by pointing. It seems that it is not much of a stretch from here to claim that Time can also be provided contextually, or deictically, or ostensively, just as the reference for pronouns can, as argued for example in Partee (1973) and Burge (1974). Partee (609) speaks of the striking parallelism between tenses and pronouns in this respect, and Burge (212) points out that sentences containing demonstratives, including pronouns and Tense, are open sentences, which are neither true nor false apart from the actual use, and which get completed only extra-linguistically. According to Burge, “a gesture saves a thousand words” (205). Recall from Section 5 that the temporal properties of small clauses can likewise be determined by their dependence on linguistic context (matrix clause) when they are embedded, or, if the analysis in this chapter is on the right track, on pragmatic context or ostension, when they are not embedded. In neither case is there a Tense node to mediate such temporal dependence of the small clause. Many single-phrase nonsententials discussed in this chapter seem to be interpreted as predicates, as is the case with the examples repeated in (94). (94) Nice lady!/Felled by flu./Arrested US border./Play baseball./ (Not) Fond of animals./(Not) Urgent./(Not) sweet./Very sick./ Still at JFK./Not at JFK. Because there are no unchecked Uninterpretable features in these phrases, such as Case or Tense features (see Section 2), these phrases are well-formed syntactic objects, even though they stop short of creating a full sentence/TP. If such phrases are taken to be predicates, and if predicates necessarily predicate over something, it will also be the part of their interpretation that there is some (although unidentified/implicit) argument in each case.17 And if such utterances are used felicitously, there will be a salient individual in the context that can be interpreted as specifying the implicit argument of the predicate. Thus, Nice lady can be felicitously uttered in the salient presence of a lady who just arrived with a cake, whereas Very sick can be felicitously uttered in the salient presence of a patient in a hospital bed. Stainton (2005) makes a convincing argument that some nonsententials are not only meaningful assertions, but also legally binding, as would be the case with a participle phrase such as Driven only 10,000 km, posted as a label on a car. Such isolated phrases are sometimes interpreted as arguments, whose predicate is salient in discourse – linguistic or nonlinguistic, as is respectively the case with the examples in (3) and (27), repeated here. (3) Me./Him./Them. (27) Rob’s mom.
(in response to: Who ate the pie?) (while holding a jar of jam)
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.25 (1564-1607)
Small clauses and phrases at the root
Suppose that in the case of (27) syntax simply creates the DP Rob’s mom, where D is selected with a default case feature, as per the discussion in Sections 2 and 4. This DP can now be sent to the interfaces without causing a crash, given that it does not contain any Uninterpretable features. A similar argument can be made for answers to questions, such as (3), except that in this case the predicate is established by the linguistic context – the question. As far as syntax is concerned, these isolated pronouns selected with the default case feature are well-formed objects because they contain no Uninterpretable features. (In contrast, an isolated nominative pronoun, such as I, would not constitute a good answer to a question in English, exactly because it contains unchecked Uninterpretable nominative case features.) These pronouns can now be interpreted as referring to individuals, potentially arguments of some predicate. In case of answers to questions, the predicate is not only salient in the immediate linguistic discourse, but the conversation frame creates an expectation that any answer to such a question will supply the relevant argument (for a detailed nonsentential analysis of question-answer pairs, see Casielles, Chap. 5, this volume; see Merchant, Chap. 3, this volume, for the elliptical view; see also the Epilogue, Section 4, for a comparison between the two views). Nonsententials thus often rely heavily on the context, linguistic or pragmatic. At the same time, reliance on pragmatic context is not always possible, given that normal adults rarely situate their speech in the context of the here and now. Even when they do, such communication is vague and imprecise (see Roeper’s (1999) claim that our speech is shaped by the tension between Economy and Explicitness). It is significant to point out that the degree of precision/explicitness necessarily expressed in full sentences may not always be desirable, in which case a nonsentential might be preferred, both for Economy reasons, and for its expressive force (this may also be the case with poetry and other artistic uses of language). The following section looks more closely at some nonsententials that seem to embrace the inexplicitness.
. When assertion is not available Intriguingly, many nonsententials in fact do not get interpreted as assertions, but rather involve nonindicative, irrealis readings, including interrogative, exclamative, desiderative, imperative, and/or incredulitive (e.g., Sick?!; Him stop smoking!?; Me Dancing?!; Him a successful businessman!?; Me first!). I argue that such irrealis interpretations are typical of nonsententials exactly because they lack Tense/Time specification. This idea finds some support in the historical development of Indicative. If these syntactic structures do not have Tense/Time specifications, a critical ingredient of an assertion, then they will be expected to favor irrealis interpretations. For example, Me first! conveys a combination of desiderative/optative and command/request moods (not unlike in the fossilized forms such as Long live the King!, whose significance will become clear shortly). On the other hand, utterances such as Sick!? and Him a successful businessman!? convey a combination of exclamation, interrogation, incredulity, and possibly more (see also Potts & Roeper, Chap. 7, this volume). There is in fact
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.26 (1607-1681)
Ljiljana Progovac
a great degree of indeterminacy/inexplicitness, as well as overlap, when it comes to the illocutionary force of such nonsententials, unlike in full sentential counterparts, which usually exhibit a specific morpheme, or a modal verb, specializing for only one of such moods. I do not believe that these irrealis interpretations in nonsententials should be associated with any syntactic functional projection(s); instead, they seem to come exactly from the lack of such functional specification, and are subject to extra-syntactic cues, such as intonation. Of course, intonation can also operate on full sentences.18 However, whereas a full sentence (e.g., He is a successful businessman) receives an indicative interpretation with unmarked intonation, many nonsententials can only be used with marked intonation to yield nonindicative interpretations (e.g., Me first!; Him a successful businessman!?). In other words, for full sentences such as He is a successful businessman, the indicative is the unmarked, default interpretation, whereas for nonsententials, such as Him a successful businessman!?, if indicative interpretation is available at all, it will be a marked, special interpretation. This is to be expected: if the structure of full sentences differs from that of nonsententials, then their interpretation is not expected to be identical either. More precisely, if the structure of nonsententials excludes Tense specification, then, all other things being equal, nonsententials are more likely to receive nonindicative/nonassertion interpretations. In fact, in his discussion of Mad Magazine examples, Akmajian (1984) claims that the unavailability of assertion as an illocutionary force is the direct result of the absence of the AUX node, which would be Tense, in the present framework (Zhang (1990) also discusses Mad Magazine data, but concludes that they are CPs). A reviewer rightly raises the question of where these irrealis interpretation possibilities come from. Surely, the context does not provide interrogative/exclamative/ imperative/irrealis possibilities. I believe that a historical perspective on the development of Indicative sheds direct light on the issue. Many Indo-Europeanists have proposed a notion of Default or Zero Tense (e.g., the Prague School (Jakobson 1964)); in most Indo-European (IE) languages the default is present indicative tense, but there are other possibilities. The notion of Default/Zero Tense is defended by, for example, Kiparsky (1968), who argues that this kind of tenseless form was used after conjunctions (Conjunction Reduction contexts) in early IE, but also survives in one way or another in various modern languages. Thus, in examples such as He came and dance, the second conjunct would be left in the tenseless/default form, given that the past time frame is already established by the first conjunct.19 This analysis also extends to the more familiar Historical Present (actually Zero Tense, for Kiparsky), that is, the present tense used in narratives to tell a story, once the setting is established to be the past time. Of note is the fact that these same contexts, conjunction and narratives, are the contexts in which Irish (verbless) small clauses abound (see Section 5 for Irish and English small clauses after conjunctions). This can hardly be a coincidence; instead, both phenomena are characterized by omitting Tense where Time is already established, in which case either the default tense form of the verb will be used or, even more economically, a small clause without a verb.
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.27 (1681-1732)
Small clauses and phrases at the root
Kiparsky (1968) also suggests that examples such as He came and dance are not possible in Modern English, for example, because Tense in Modern English is a grammaticalized feature on the verb, and not an independent constituent (in full sentences, that is), for which he gives independent evidence. This can be interpreted in the present framework to mean that the (finite) verb in English (full sentences) has to check the Tense feature in a functional projection, TP, but that in early IE Tense was a separate/separable constituent, and was interpreted on a par with other adverbials. There are also modern languages in which Tense is adverbial in nature, including Tongan, a Polynesian language, as described in Churchward (1953). Moreover, in Greek and Sanskrit, verbs are commonly put into (what looks like) present tense when modified by adverbs denoting past time (Kiparsky: 47). Brugmann (1883) noted, and this view is widely accepted, that this is a proto-IE feature. Kiparsky argues that this form probably is not present tense, but rather a relic of the proto-IE Injunctive, which combines with adverbs to express time. Crucially, Kiparsky assumes that this form instantiates a Zero Tense, or no Tense, which will become relevant for the following discussion. This gives credence to the claim that Tense/Time can be established independently of a TP node, as well as to the idea that language, in principle, can tolerate default verb forms (unmarked for Tense). In such cases, reference to Time can be established in a variety of alternative ways, including by adverbials, by the first conjunct, by a narrative setting, as well as by other types of (pragmatic) context, including that of the here and now. As pointed out earlier, some IE-ists, including Kiparsky (1968), have argued that the default tense forms have their origin in the ancient proto-IE Injunctive form, which initially performed both indicative and nonindicative mood functions (which “neutralized” mood and tense distinctions), but later specialized for nonindicative moods (e.g., imperative, exclamative, optative, desiderative – some relics in English are: Long live the King!; God forbid!). The basic difference between Indicative and other moods is that Indicative yields assertions, but other moods do not (in this respect, Lyons (1968: 308), e.g., claims that Interrogative is a nonindicative mood). Perhaps the bare form of the verb in nonsententials in English (e.g., Him worry?!) can be compared to the Injunctive, nonspecified/default mood, showing a range of nonindicative interpretations. It is crucial to point out that the Injunctive began to specialize for nonassertive/ irrealis functions only after the Indicative broke away from it; before that, one and the same unmarked form of the verb, Injunctive, was used to express any mood, including Indicative. After the assertive Indicative mood broke away, the Injunctive was an elsewhere, miscellaneous mood for everything but assertion. Notice that a similar path is arguably attested in the change in interpretation of small clauses from childhood to adulthood (see the Epilogue to this volume, Section 3; also Potts & Roeper’s Chap. 7 for a different perspective). The reviewer’s question mentioned earlier now becomes relevant: why are all such cases with irrealis interpretation not analyzed as projecting a functional projection involving a null Injunctive morpheme, especially given that the context itself cannot
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.28 (1732-1783)
Ljiljana Progovac
supply irrealis interpretations (see also Zhang 1990)? Going back to the historical argument, Injunctive was never a specific mood, but rather in fact an unmarked, vague form, used only to name the action, which could range over a variety of different moods. Kiparsky (1968) characterizes Injunctive as a No Tense/No Mood form because it in fact preceded the emergence of such specific categories, and was simply underdetermined with respect to them. Because the irrealis interpretation is not a well-defined concept that can be attributed to a single morpheme, if an Injunctive morpheme were to be posited for such examples, one would have to define its meaning contribution in negative rather than positive terms, that is, as anything-but-indicative. For similar reasons, it would be misleading to characterize Default Case as a special kind of case, rather than just a caseless form, given that Default Case is used in a variety of contexts that in fact have nothing in common other than that they are contexts in which no specific case is available (e.g., predicative positions, labels and titles, left dislocation; see Section 4.3). It is quite possible that Default Case, like Injunctive, is an elsewhere Case, that is, miscellaneous Case, which remained after more specific case forms broke away from one unspecified/unmarked form. Returning to the Injunctive, if such a null injunctive morpheme were to be posited, one would expect to find some linguistic evidence for it, but it is hard to see what that evidence would be (see also n. 18 for the lack of evidence for C in intonation questions). For example, if such a morpheme exists, one would expect it to select a specific kind of complement, the way other functional heads do. Instead, as is evident from the examples of various nonsententials that normally exhibit these nonindicative interpretations, this morpheme would occur even in verbless clauses or phrases, such as Him a businessman!?, Sick!?, and Me first! With verbal small clauses, it would be able to select any nonfinite form: Him worry?, Me Dancing!?, and John fired!? To compound the problem, as pointed out earlier, even full sentences (TPs) exhibit irrealis interpretations, with certain marked intonation, and it is not at all clear where an injunctive morpheme would fit in such TPs. Thus, if such a null morpheme exists, it would be unlike any other morpheme. Of course, it is theoretically possible to argue that all the nonsententials with irrealis interpretation have a full sentential structure, complete with a verb in VP and an injunctive modal/morpheme in TP (or a comparable functional projection), from which the injunctive morpheme and the verb delete. Apart from the obvious question of why such morphemes and the accompanying verb would be able (or forced) to delete in these particular cases, but not in others (e.g., He must/should/can become a businessman cannot be reduced to He a businessman), what I show here is that this complex derivation is not necessary for interpretation purposes. The mere unavailability of Tense/Time specification will preclude an indicative/assertive interpretation, leaving us with a variety of other (nonindicative) interpretations. We syntacticians have gained a great deal by discovering numerous non-obvious generalizations and constructs, such as structural Case, TP, and Determiner Phrase, as discussed in the previous sections. However, we stand to lose just as much by routinely imposing these constructs on all the data, in all languages, no matter what the form or function of
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.29 (1783-1863)
Small clauses and phrases at the root
such data is. Minimalism provides useful guidelines in this respect, demanding that structure be built bottom up, and only in the presence of linguistic evidence.20 In summary, my claim is that irrealis is an elsewhere/miscellaneous interpretation, a fallback interpretation where no Tense/Time specification is available to support indicative/assertive interpretations. In other words, my claim is that the irrealis interpretation in nonsententials is the result of an absence or underspecification, rather than a presence, predicting that nonsententials, smaller structures, will be more likely to result in such interpretations than full sentences. If so, then the semantic considerations having to do with Tense, Time, Assertion, and Mood provide indirect support for a nonelliptical approach to nonsententials, as further discussed in the Epilogue, Section 3. Needless to say, a precise characterization of the range and availability of various nonindicative interpretations, as well as the role of intonation and pragmatic context, remain for future research to sort out. One promising line of research regarding pragmatic context is Avrutin (1999, 2004), who attempts to develop a theory of syntax-discourse interface (see also Epilogue, Section 5.1).
. Conclusion The goal of this chapter was to make a case for the existence of nonsententials (e.g., Me; Me dancing; Battery dead) as terminal syntactic objects. Nonsententials are argued to lack, at the very least, the syntactic Tense node and, with it, any syntactic phenomena that rely on Tense, including structural nominative case. This single claim, couched in the Minimalist framework, predicts the core syntactic properties of nonsententials: accusative default case on subject pronouns, possible absence of otherwise obligatory articles, and systematic lack of finiteness. In fact, there is a clear correlation between the choice of the default tense strategy and the default case strategy. My analysis capitalizes on the assumption in current syntactic theory that every sentence/clause is underlyingly a small clause, which gets transformed into a full (finite) sentence/clause only after the merger of the Tense node. This transformation is only possible if verbal elements are selected with Tense features, and if subjects are selected with nominative case features. Having one without the other will result in ungrammaticality, because structural nominative and Tense features must be checked against each other. However, my argument is that nonsententials result from selecting lexical items with neither Tense nor (nominative) case features. Such syntactic objects, even though not full sentences (TPs), are nonetheless well formed syntactically, because they do not contain any illicit unchecked formal features. The lack of Tense specification is arguably also responsible for the dearth of indicative interpretations among nonsententials, as well as for their frequent reliance on pragmatic context and intonation. My argument is that nonsententials, given that they are tenseless syntactic objects, can either be anchored in Time by means other than Tense (e.g., the context of the situation, typically here and now) or receive nonindicative/nonassertive/irrealis interpretations.
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.30 (1863-1918)
Ljiljana Progovac
Theoretically, all other things being equal, the base generation of such examples as phrases/small clauses is more economical, and thus more in the spirit of Minimalism, than any sentential derivation. It is the very essence of Minimalism that only phrases for which there is linguistic evidence will be projected. Not only is there no evidence for TP structure in nonsententials, but there is in fact counter-evidence, which comes from the patterns of Case, Tense, and article omission, as discussed throughout this chapter. Nonetheless, data like these are only rarely discussed in the current Minimalist literature, possibly due to the common belief that they do not fit into the theory. I have shown here that this belief is in error, and that such data in fact follow precisely from this theory, with only some nonessential claims of the theory needing to be reevaluated, and with some more bridges needing to be built with the semantic/pragmatic components. If these data are real, but are excluded by the theory, the theory will necessarily be incomplete. As pointed out in the Epilogue, especially Section 4, there are a variety of phenomena that are currently taken to belong only to full sentential grammars, such as Case and Island effects. If this is not the case, as is suggested by the nonsentential analysis developed here, then the nonsentential data can provide a rich new database to tease apart sentential from nonsentential aspects of these theoretical constructs, offering a better chance of understanding both.
Notes * This chapter owes a great deal to the participants in the discussions of the Wayne State University Humanities Center Working Group over 4 years, out of which this volume grew: Ellen Barton, Eugenia Casielles, Walter Edwards, Kate Paesani, Pat Siple, and Nicola Work. My special thanks go to the coeditors, Ellen Barton, Eugenia Casielles, and Kate Paesani, for careful reading and comments on this chapter. In addition, thanks to Ellen Barton and her work on nonsententials for the inspiration to work on this topic, and for many examples, especially some novel ones, discussed in Section 2. Thanks to Eugenia Casielles for many insightful comments, as well as discussions and references on small clauses cross-linguistically. Thanks to Kate Paesani for many discussions and examples, especially regarding the Default Case. Thanks to Pat Siple for references and discussions on the nature of language acquisition and aphasia. Aspects of this work were presented at the 2003 and 2004 Wayne State Workshops on Small Utterances (as Progovac 2003 and 2004). For their comments, I am grateful to the visiting workshop participants, including Herman Kolk, Jason Merchant, Alan Munn, Tom Roeper, Donald Winford, and, in particular, Rob Stainton, for the comments on terminology and philosophy of language. For judgments and comments, my thanks also go to Marc Authier, Martha Ratliff, Paweł Rutkowski, and Liu Hayong. My deepest thanks go to Michel DeGraff and Tim Stowell – thanks to their detailed, inspiring comments, this chapter improved by leaps and bounds. . These two examples are categorized as VPs in Barton (1998) and (1991), respectively. Ask any lawyer was uttered in response to the statement: John doesn’t know what the best defense against ethics charges would be, and Get lawyer was taken from a telegram of a person who got arrested. The first question is whether these examples are imperatives, or something else. And even if they are to be categorized as imperatives, do imperatives involve more structure than a VP in English? Starting from the last question, it is not clear that English imperatives have more struc-
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.31 (1918-1972)
Small clauses and phrases at the root
ture than VP (see Akmajian (1984) for the argument that there is no AUX in imperatives; but see Zhang (1990), who argues that imperatives are TPs in spite of that). But it is not even clear that the examples in (12) are imperatives, especially Ask any lawyer. Whereas the imperatives are typically analyzed as having you as their (implied) subject, and command as their illocutionary force, Ask any lawyer seems to be much more indeterminate in its interpretation – it is more a comment than a command, more a generic statement than an order to anybody specific, and it does not seem to be an order given to the addressee (see Barton, Chap. 1, this volume, for further discussion of this example). As pointed out by a reviewer, these examples may but need not be imperatives. Whether or not they are imperatives, they can be analyzed as VPs, and I thus include them with the (other) VP examples. Notice also that there are imperatives that are clearly nonsentential, such as (i) and (ii), even in the analysis of Zhang (1990). In fact, the example with the preposition in (i) can easily be expanded into a small clause with a subject, such as (iii), undermining Zhang’s argument that there is a crucial distinction between imperatives with verbs and those without. The small clause in (iii) can be analyzed as a PP. (i) Out (with you)! (ii) Attention! (iii) Everybody out!
(PP) (NP) (PP)
. A reviewer suggests that examples (14) and (15) can be analyzed as full sentences from which the modal verb would has elided. This, however, would be disconfirmed by the fact that a pronoun in the subject position would carry accusative case, the point to which I also return later in the text (see also n. 8): (i) Her/Me do a thing like that?! (Never!) (ii) Him/Me be rude to a friend?! (That’s certainly not like him/me.) The same argument holds for the reviewer’s example in (iii), whose pronominal counterpart would feature an accusative pronoun (iv): (iii) Sarah say that her sister is unhappy?! Never! (iv) Her say that her sister is unhappy?! Never! . An early argument for parallelism between child language and aphasia can be found in Jakobson (1941). This parallelism is also pursued in Ouhalla (1993), who argues that “telegraphic speech” in aphasia and child language involves bare phrases, and no functional projections. Among others, Radford (1990) develops an elaborate characterization of an early stage in child speech in terms of phrasal grammar, with initially absent, and then gradually emerging, functional projections (for criticism, see e.g., Guasti 2002, and references cited there). More recently, Herman Kolk, in a series of papers, has argued for a three-way comparison of “elliptical” speech in aphasia, child language, and normal (unimpaired) “elliptical” speech (see Kolk, Chap. 9, this volume, and references cited there; see also Avrutin 2004). It is a commonly held view in current syntactic theory that the locus of language variation is in functional projections (see Borer 1984). To the extent that this minimal grammar utilizes a minimum of functional projections, if any at all, it is expected that it will differ minimally from group to group. . In fact, Akmajian (1984) discusses a subset of small clause nonsententials, sometimes referred to as Mad Magazine sentences, such as the examples (13), (14), (15), (21), and (28) in the text, and concludes that they are sentences that lack the AUX (auxiliary) node (see Section 3 for the structure of a sentence that includes an AUX node). He also bases this conclusion on the
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.32 (1972-2042)
Ljiljana Progovac
observation that such examples lack the typical AUX constituents, such as Tense, modal verbs, nominative case, and sentential adverbs (e.g., fortunately). . As pointed out by Rob Stainton (personal communication), example (29c) would be acceptable in a situation where the problem is used in a referential way, such as in a situation in which one points to a group of hungry people, and says, “The problem.” Then, when pizza gets delivered, one can say, “The problem solved.” I return to such examples in Section 4.2, where I argue that the problem in this case is used almost as a linguistic label/name for the said group of people, and that in such uses, the determiner has a referential function, rather than a formal (case-related) function. Some additional examples with similar properties are given here. As with Problem solved in the text, the determiners may be possible in such nonsententials, but under special pragmatic circumstances. (i)
Case closed./Mission accomplished./Permission granted.
. A reviewer points out that the examples in (29d) are not completely ungrammatical, and that equivalent examples occur in headlines (see also Paesani, Chap. 6, this volume). . For further details of this nonsentential approach to question-answer pairs, as well as for previous references, the reader is referred to Casielles, Chapter 5, this volume, and Barton, Chapter 1, this volume. For an elliptical approach to question-answer pairs, the reader is referred to Merchant, Chapter 3, this volume. Some comparisons between the two approaches are given in the Epilogue to this volume, Section 4. . An approach that would posit a phonological deletion of the Tense element in root small clauses cannot explain the appearance of accusative case on subject pronouns, or the absence of articles. For comparison’s sake, let us look at one construction in which the postulation of a null tense node finds independent support. English subjunctive clauses such as (i) are typically analyzed as involving a null T element, with a null counterpart of a modal verb, comparable to should. Such clauses still have their subjects in the nominative case, and they do not sanction article omission (ii): (i) John demanded that they / *them ø be transferred. (ii) John demanded that the boy / *boy ø be transferred. If there were a null (or deleted) tense node in the nonsententials discussed earlier, then one would wrongly predict that they would behave essentially like the subjunctive clauses in (i) and (ii). . Notice that I do not discuss infinitive clauses in this chapter, focusing instead on clear-cut cases of finite clauses versus small clauses (some discussion of infinitives can be found in Paesani, Chap. 6, this volume, as well as in the Epilogue, Section 3). The infinitive clauses would seem to fall somewhere in between, given that, in English at least, they do not assign nominative case, but do involve a piece of functional structure, namely the modal particle to, whose exact analysis is still controversial (some analyze it as a complementizer; others place it in the infl (Tense) projection; still others analyze it as a preposition, e.g., Hyde 2000). The ultimate analysis of the infinitive clause will prove important for the analysis of nonsententials, given that there are root infinitives among nonsententials, such as (i) and (ii) (see also headlines, recipes, and directions, discussed in Paesani, Chap. 6, this volume). Likewise, root infinitives are frequent among children’s utterances cross-linguistically (Guasti 2002, and references cited there), as well as in agrammatism (Kolk, Chap. 9, this volume). (i)
Not to worry.
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.33 (2042-2118)
Small clauses and phrases at the root
(ii) Not to forget. . A reviewer points out that the possibility of Default Case might wrongly predict that the examples such as (i) and (ii) would be grammatical, given that the passivization of the main verb renders the subject position of the bracketed small clause caseless, and given that the subject argument appears to be in its theta position. (i) *It is considered [girl intelligent]. (ii) *It is considered [her intelligent]. As noted in Section 2, embedded small clauses differ from root small clauses in that their subjects necessarily receive structural accusative case from the matrix verb and that, consequently, articles cannot be dropped in such positions. It seems that nonsententials with default case cannot be embedded in sentential structures. The very act of embedding may require some kind of functional relationship between the matrix verb and the embedded clause, whether structural Case or some functional projection, such as AgrP or CP. In this respect, notice that language acquisition studies, aphasic studies, and pidgin studies often report lack or dearth of subordination in their respective data (see the relevant chapters in this volume, especially Winford’s Chap. 11, as well as the Epilogue, Section 5.3). Needless to say, a precise characterization of the nature of this restriction on embedding remains open for future research. . Notice that when nouns are used as vocatives, they necessarily occur without determiners, as in (i) and (ii): (i) Kid, come over here!/*The kid, come over here! (ii) Doctor, what is wrong with me?/*The doctor, what is wrong with me? This may be because the noun is used directly as a label associated with the addressee, and thus no article is needed to facilitate reference (see the discussion in Section 6). In addition to that, the definite article would be incompatible with the second-person feature of the vocative, which is basically the second person, you. . Proper nouns also occur freely in nonsentential subject positions (see examples (14), (15), (21), (24)). Proper nouns, as argued in, for example, Longobardi (1994), are associated with the DP projection, but in different ways in different languages. In English, proper nouns originate in N and move to D only in LF (covertly). On the other hand, in some dialects of Italian, proper nouns raise to D in overt syntax, whereas other proper nouns stay in N and are preceded by an (expletive) article in D. Just as it is suggested for pronouns in the text, it may be that the movement of proper nouns to D is for formal case-checking reasons only, which follows from the implication in (61). If so, then such movement may not take place in (default case-marked) nonsentential positions. This issue, however, is not immediately relevant for the main proposal in this chapter. Suffice it to say that nothing in the theory, or in this proposal, excludes the occurrence of proper nouns in nonsententials. . Stowell’s (1981) argument for a phrasal treatment of small clauses rests on the observation that not all verbs take the same small clause type, which would be surprising if all clauses were sentences/TPs (Stowell’s argument has been criticized in Kitagawa (1985), and then refined in Contreras (1987)). For example, consider the ungrammatical example with a small clause (i), which becomes grammatical once expanded into a fuller (infinitive) clause (ii): (i) *I expect [you an attorney by the end of the year]. (ii) I expect [you to be an attorney by the end of the year].
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.34 (2118-2185)
Ljiljana Progovac . The example in (89) is adapted from Jackendoff (2002), who uses examples of this kind to argue for a possible pre-TP stage in the evolution of human language. As pointed out by Eugenia Casielles (personal communication), comparable small clauses are also used in Latin, and are referred to as Absolute Ablative constructions. The example in (90) is from Jespersen (1954). . Although sentences are typically analyzed as TPs (see Section 3), it is usually held that there are several other functional sentential projections, including AgrPs – see, for example, Ouhalla’s (1988) and Pollock’s (1989) split-infl hypothesis, as well as views according to which an AgrP projection occurs between two VP shells (e.g., Bobaljik 1995; Boškovi´c 1995; Carnie 1995; Harley 1995; Koizumi 1995; Lasnik 1999). Even if small clauses (including root small clauses) are characterized by the absence of Tense (or TP node), this still leaves open the theoretical possibility for the presence of a projection such as AgrP/AspP inside the VP shells. Such a projection may be responsible for the subject-predicate agreement in, for example, Italian small clauses, as well as for some aspectual properties of English nonsententials (e.g., contrasts between Me dance? and Me dancing?). Likewise, it may be necessary to project a functional projection above the VP in the examples that involve Quantifier Float, such as Money all gone, as pointed out by a reviewer (see also Cardinaletti & Guasti 1995 for Quantifier Float in embedded small clauses). However, the precise analysis of either aspect or Quantifier Float in English is far from resolved, and at this point cannot serve as a decisive argument for either the presence or the absence of AgrP/AspP in small clauses (for recent views of Quantifier Float, see, e.g., Bobaljik 2003; Boškovi´c 2004). . It is worth pointing out that child language is often described as referring to the here and now. If children’s early language is characterized by a preponderance of nonsentential utterances, as argued for example in Radford (1990), then this would explain why their speech is about the here and now. Or, to see it from another perspective, if children are talking about the here and now, then they can get away with using nonsententials, because the time frame is provided contextually. See the Epilogue, Section 3, for a possible explanation of why this possibility is highly restricted in adult language; see also Potts and Roeper, Chapter 7, this volume. . This poses a potential problem for the Theta Criterion, which is normally taken to require all the (obligatory) arguments to be syntactically present. However, even in sentential syntax there are many situations where arguments are only optionally supplied by syntax, such as with understood recipient roles of various types of verbs (i), (ii), or with predicates in nominalizations (iii), (iv). Determining the exact domain and limits of the application of the Theta Criterion is an important problem, which is nonetheless more general than the scope of this chapter. (i) He said (to us) that the room is empty. (ii) He sold the house (to someone). (iii) The destruction was complete. (cf. John’s destruction of his house.) (iv) The sickness was all around us. (cf. Many people around us were sick.) . It is typically assumed that intonation questions, that is, questions marked only by rising intonation (i), do not involve a CP layer: they neither trigger inversion, nor license negative polarity items (e.g., ever) (ii), which are arguably licensed by operators in C (iii). In other words, the interrogative mood in (i) does not seem to be associated with a functional projection. (i)
John has traveled to London?
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.35 (2185-2284)
Small clauses and phrases at the root
(ii) ? *John has ever traveled to London? (iii) Has John ever traveled to London? . Intriguingly, Kiparsky (1968: 43) also notes that such neutral/injunctive forms were used in Sanskrit as answers to questions: “Rules 3.2.120-1 of P¯anini’s grammar of Sanskrit contain a strange piece of information . . . a question asked in the past tense is answered in the present tense with certain particles” (e.g., Did you make the mat? I make.). A similar strategy is used in Old Irish, and is called Responsive (Draak 1952). . It is also important to remember how we got here. The main argument in the current theory is that every truth-evaluable utterance must involve a full sentence, or a TP, because the presence of T is required for an indicative, assertive interpretation. But now clearly there are examples where T is visibly missing and that, moreover, have a nonindicative interpretation. If we again ascribe to such examples the same functional structure found in indicative sentences, with a null mood morpheme, then our analysis becomes unfalsifiable – there is no structure or interpretation that can prove it wrong.
References Abney, S. (1987). The English Noun Phrase in its Sentential Aspect. Ph.D. dissertation, Massachusetts Institute of Technology. Akmajian, A. (1984). Sentence types and the form-function fit. Natural Language and Linguistic Theory, 2, 1–23. Alexiadou, A. & Wilder, C. (Eds.). (1998). Possessors, Predicates and Movement in the Determiner Phrase. Amsterdam/Philadelphia: John Benjamins. Avrutin, S. (1999). Development of the Syntax-Discourse Interface. Dordrecht: Kluwer. Avrutin, S. (2004). Optionality in child and aphasic speech. Lingue e Linguaggio, 1, 67–97. Baker, M. (1988). Incorporation: A Theory of Grammatical Function Changing. Chicago, IL: University of Chicago Press. Baker, M. (1996). The Polysynthesis Parameter. Oxford: OUP. Bar-Hillel, Y. (1954). Indexical expressions. Mind, 63, 359–379. Barton, E. (1990). Nonsentential Constituents: A Theory of Grammatical Structure and Pragmatic Interpretation. Amsterdam: John Benjamins. Barton, E. (1991). Nonsentential constituents and theories of phrase structure. In K. Leffel & D. Bouchard (Eds.), Views on Phrase Structure (pp. 193–214). Dordrecht: Kluwer. Barton, E. (1998). The grammar of telegraphic structures: Sentential and non-sentential derivations. Journal of English Linguistics, 26, 37–67. Barton, E. & Progovac, L. (2005). Nonsententials in Minimalism. In R. Elugardo & R. J. Stainton (Eds.), Ellipsis and Non-Sentential Speech (pp. 71–93). Dordrecht: Kluwer. Bobaljik, J. D. (1995). Morphosyntax: The Syntax of Verbal Inflection. Ph.D. dissertation, Massachusetts Institute of Technology. Bobaljik, J. D. (2003). Floating quantifiers: Handle with care. L. Cheng & R. Sybesma (Eds.), In The Second Glot International State-of-the-Article Book: The Latest in Linguistics [Studies in Generative Grammar 61] (pp. 107–148). Berlin: de Gruyter. Borer, H. (1984). Parametric Syntax: Case Studies in Semitic and Romance Languages. Dordrecht: Foris.
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.36 (2284-2392)
Ljiljana Progovac
Boškovi´c, Ž. (1995). Principles of Economy in Nonfinite Complementation. Ph.D. dissertation, University of Connecticut. Boškovi´c, Ž. (2004). Be careful where you float your quantifiers. Natural Language and Linguistic Theory, 22, 681–742. Bouchard, D. (2002). Adjectives, Number and Interfaces: Why Languages Vary. Amsterdam: Elsevier. Bresnan, J. (1978). A realistic transformational grammar. In M. Halle, J. Bresnan, & G. Miller II (Eds.), Linguistic Theory and Psychological Reality (pp. 1–59). Cambridge, MA: The MIT Press. Brugmann, K. (1883). Altind. pur¯a und griech. páros mit dem Indikativ des Präsens. Berichte über die Verhandlungen der königlich sächsischen Gesellschaft der Wissenschaften. Philologisch-Historische Classe, 35, 169–173. Burge, T. (1974). Demonstrative constructions, reference, and truth. Journal of Philosophy, 71, 205–223. Burzio, L. (1981). Intransitive Verbs and Italian Auxiliaries. Ph.D. dissertation, Massachusetts Institute of Technology. Cardinaletti, A. (1993). On the internal structure of pronominal DPs. University of Venice Working Papers in Linguistics, 3(2), 1–20. Cardinaletti, A. & Guasti, M. T. (1992). Epistemic small clauses and null subjects. In G. F. Westphal, B. Ao, & H.-R. Chae (Eds.), Proceedings of VII Eastern States Conference of Linguistics 1991 (pp. 23–33). Columbus, OH: Ohio State University Press. Cardinaletti, A. & Guasti, M. T. (1995). Small clauses: Some controversies and issues of acquisition. In A. Cardinaletti & M. T. Guasti (Eds.), Syntax and Semantics 28: Small Clauses (pp. 1–23). San Diego, CA: Academic Press. Carlson, G. (1977). Reference to kinds in English. Ph.D. dissertation, University of Massachusetts, Amherst. Carnie, A. (1995). Non-Verbal Predication and Head Movement. Ph.D. dissertation, Massachusetts Institute of Technology. Chierchia, G. & McConnell-Ginet, S. (1990). Meaning and Grammar. Cambridge, MA: The MIT Press. Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. (1986). Barriers. Cambridge, MA: The MIT Press. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: The MIT Press. Chomsky, N. (2005). Three factors in language design. Linguistic Inquiry, 36, 1–22. Chung, S. & McCloskey, J. (1987). Government, barriers, and small clauses in Modern Irish. Linguistic Inquiry, 18, 173–237. Churchward, C. M. (1953). Tongan Grammar. London: OUP. Contreras, H. (1987). Small clauses in Spanish and English. Natural Language and Linguistic Theory, 5, 225–243. Contreras, H. (1995). Small clauses and complex predicates. In A. Cardinaletti & M. T. Guasti (Eds.), Syntax and Semantics 28: Small Clauses (pp. 135–152). San Diego, CA: Academic Press. Davies, E. (1986). The English Imperative. London: Croom Helm. Déchaine, R.-M. (1993). Predicates Across Categories: Towards a Category-Neutral Syntax. Ph.D. dissertation, University of Massachusetts, Amherst. Draak, M. (1952). Emotional reflexes. Ériu, 16, 74–78. Gazdar, G. (1982). Phrase Structure Grammar. In P. Jacobson & G. K. Pullum (Eds.), The Nature of Syntactic Representation (pp. 131–186). Dordrecht: Reidel.
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.37 (2392-2505)
Small clauses and phrases at the root
Guasti, M. T. (2002). Language Acquisition: The Growth of Grammar. Cambridge, MA: The MIT Press. Guéron, J. & Hoekstra, T. (1995). The temporal interpretation of predication. In A. Cardinaletti & M. T. Guasti (Eds.), Syntax and Semantics 28: Small Clauses (pp. 77–107). San Diego, CA: Academic Press. Hale, K. & Keyser, S. J. (2002). Prolegomena to a Theory of Argument Structure [Linguistic Inquiry Monograph 39]. Cambridge, MA: The MIT Press. Harley, H. B. (1995). Subjects, Events, and Licensing. Ph.D. dissertation, Massachusetts Institute of Technology. Hyde, B. (2000). The structures of the to-infinitive. Lingua, 110, 27–58. Jackendoff, R. (2002). Foundations of Language: Brain, Meaning, Grammar, Evolution. Oxford: OUP. Jakobson, R. (1941). Child Language, Aphasia and Phonological Universals. (Originally published by Mouton, the Hague, in 1968). Jakobson, R. (1964). Zur Struktur des russischen Verbums. In Charisteria Guilelmo Mathesio quinguagenario oblata. Reprinted in J. Vachek (Ed.), A Prague School Reader in Linguistics. Bloomington, IN: Indiana University Press. Jespersen, O. (1954). A Modern English Grammar. Part III: Syntax. London: Allen & Unwin. Kayne, R. (1984). Connectedness and Binary Branching. Dordrecht: Foris. Kiparsky, P. (1968). Tense and Mood in Indo-European syntax. Foundations of Language, 4, 30– 57. Kitagawa, Y. (1985). Small but clausal. Chicago Linguistic Society, 21, 210–220. Kitagawa, Y. (1986). Subjects in English and Japanese. Ph.D. dissertation, University of Massachusetts, Amherst. Koizumi, M. (1995). Phrase Structure in Minimalist Syntax. Ph.D. dissertation, Massachusetts Institute of Technology. Koopman, H. & Sportiche, D. (1991). The position of subjects. Lingua, 85, 211–258. Labov, W. (1969). Contraction, deletion, and the inherent variability of the English copula. Language, 45, 715–762. Larson, R. & Segal, G. (1995). Knowledge of Meaning. An Introduction to Semantic Theory. Cambridge, MA: The MIT Press. Lasnik, H. (1999). Minimalist Analysis. Oxford: Blackwell. Lebeaux, D. (1990). The grammatical nature of the acquisition process: Adjoin-Alpha and the formation of relative clauses. In L. Frazier & J. de Villiers (Eds.), Language Processing and Language Acquisition (pp. 13–82). Dordrecht: Kluwer. Lobeck, A. (1995). Ellipsis: Functional Heads, Licensing, and Identification. Oxford: OUP. Longobardi, G. (1994). Reference and proper names: A theory of N-Movement in syntax and Logical Form. Linguistic Inquiry, 25, 609–665. Lyons, J. (1968). Introduction to Theoretical Linguistics. Cambridge: CUP. Manzini, M. R. (1983). Restructuring and Reanalysis. Ph.D. dissertation, Massachusetts Institute of Technology. Moro, A. (1988). Per una Teoria Unificata delle Frasi Copulari. Rivista Di Grammatica Generativa, 13, 81–110. Moro, A. (1995). Small clauses with predicative nominals. In A. Cardinaletti & M. T. Guasti (Eds.), Syntax and Semantics 28: Small Clauses (pp. 109–132). San Diego, CA: Academic Press. Ouhalla, J. (1988). The Syntax of Head-Movement. Ph.D. dissertation, University College London.
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.38 (2505-2612)
Ljiljana Progovac
Ouhalla, J. (1993). Functional categories, agrammatism and language acquisition. Linguistische Berichte, 143, 3–36. Partee, B. (1973). Some structural analogies between tenses and pronouns in English. Journal of Philosophy, 70, 601–609. Pesetsky, D. (1982). Paths and Categories. Ph.D. dissertation, Massachusetts Institute of Technology. Pollock, J.-Y. (1989). Verb Movement, Universal Grammar, and the structure of IP. Linguistic Inquiry, 20, 365–424. Postal, P. M. (1969). On the so-called ‘pronouns’ in English. In D. Reibel & S. Schane (Eds.), Modern Studies in English (pp. 201–224). Englewood Cliffs, NJ: Prentice Hall. Progovac, L. (1998). Determiner Phrase in a language without determiners. Journal of Linguistics, 34, 165–179. Progovac, L. (2003). Case and Nonsententials. Paper presented at the First Wayne State Workshop on Small Structures, Detroit. Progovac, L. (2004). Small Utterances: Formal Aspects. Paper presented at the Second Wayne State Workshop on Small Utterances, Detroit. Radford, A. (1990). Syntactic Theory and the Acquisition of English Syntax. Oxford: Blackwell. Radford, A. (1997). Syntactic Theory and the Structure of English: A Minimalist Approach. Cambridge: CUP. Raposo, E. & Uriagereka, J. (1990). Long-distance Case assignment. Linguistic Inquiry, 21, 505– 537. Rickford, J. R., Ball, A., Blake, R., Jackson, R., & Martin, N. (1991). Rappin on the copula coffin: Theoretical and methodological issues in the analysis of copula variation in African American Vernacular English. Language Variation and Change, 3, 103–132. Ritter, E. (1991). Two functional categories in Noun Phrases: Evidence from Modern Hebrew. In Susan Rothstein (Ed.), Perspectives in Phrase Structure: Heads and Licensing [Syntax and Semantics 25] (pp. 37–62). New York, NY: Academic Press. Roeper, T. (1999). Universal bilingualism. Bilingualism: Language and Cognition, 2(3), 169–186. Ross, J. R. (1967). Constraints on Variables in Syntax. Ph.D. dissertation, Massachusetts Institute of Technology. (Later published as Infinite Syntax! Norwood, NJ: Ablex, 1986.) Rothstein, S. (1995). Small clauses and copular constructions. In A. Cardinaletti & M. T. Guasti (Eds.), Syntax and Semantics 28: Small Clauses (pp. 27–48). San Diego, CA: Academic Press. Schein, B. (1995). Small clauses and predication. In A. Cardinaletti & M. T. Guasti (Eds.), Syntax and Semantics 28: Small Clauses (pp. 49–76). San Diego, CA: Academic Press. Schütze, T. C. (2001). On the nature of Default Case. Syntax, 4, 205–238. Schütze, C. & Wexler, K. (1996). Subject Case licensing and English Root Infinitives. In A. Stringfellow, D. Cahana-Amitay, E. Hughes, & A. Zukowski (Eds.), Proceedings of the 20th Annual Boston University Conference on Language Development (pp. 670–682). Somerville, MA: Cascadilla. Spencer, A. (1991). Morphological Theory. Oxford: Blackwell. Stainton, R. J. (2005). In defense of non-sentential assertion. In Z. Szabo (Ed.), Semantics vs. Pragmatics (pp. 383–457). Oxford: OUP. Starke, M. (1995). On the format for small clauses. In A. Cardinaletti & M. T. Guasti (Eds.), Syntax and Semantics 28: Small Clauses (pp. 237–269). San Diego, CA: Academic Press. Stowell, T. (1978). What was there before there was there. In D. Farkas, W. M. Jacobsen, & K. W. Todrys (Eds.), Papers from the Fourteenth Regional Meeting, Chicago Linguistic Society (pp. 458–471). Chicago, IL: The Chicago Linguistic Society.
JB[v.20020404] Prn:25/07/2006; 12:26
F: LA9302.tex / p.39 (2612-2659)
Small clauses and phrases at the root
Stowell, T. (1981). Origins of Phrase Structure. Ph.D. dissertation, Massachusetts Institute of Technology. Stowell, T. (1983). Subjects across categories. Linguistic Review, 2/3, 285–312. Stowell, T. (1989). Subjects, specifiers, and X-bar theory. In M. Baltin & A. Kroch (Eds.), Alternative Conceptions of Phrase Structure (pp. 232–262). Chicago, IL: University of Chicago Press. Stowell, T. (1995). Remarks on clause structure. In A. Cardinaletti & M. T. Guasti (Eds.), Syntax and Semantics 28: Small Clauses (pp. 271–286). San Diego, CA: Academic Press. Szabolcsi, A. (1983). The possessor that ran away from home. Linguistic Review, 3, 89–102. Szabolcsi, A. (1987). Functional categories in the noun phrase. In I. Kenesei (Ed.), Approaches to Hungarian, Volume Two (pp. 167–189). Szeged: Jate. Vainikka, A. (1990). The status of grammatical default systems: Comments on Lebeaux. In L. Frazier & J. De Villiers (Eds.), Language Processing and Language Acquisition (pp. 83–103). Dordrecht: Kluwer. Williams, E. (1980). Predication. Linguistic Inquiry, 11, 203–238. Williams, E. (1983). Against small clauses. Linguistic Inquiry, 14, 287–308. Winford, D. (2003). An Introduction to Contact Linguistics. [Language in Society 33.] Oxford: Blackwell. Zhang, S. (1990). The Status of Imperatives in Theories of Grammar. Ph.D. dissertation, University of Arizona.
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.1 (47-126)
chapter
“Small structures” A sententialist perspective* Jason Merchant Fragment answers and some apparently nonsentential utterances are argued to be syntactically elliptical for fully sentential structures; an ellipsis analysis accounts for a wide range of connectivity and anticonnectivity effects. Fragments furthermore undergo movement to a clause-external position prior to ellipsis, capturing a number of otherwise puzzling asymmetries, but not all “small structures” are amenable to an elliptical analysis.
.
Introduction
The nature of short or “fragment” answers as in (1) and that of “out-of-the-blue” utterances as in (2) have long been of interest to linguists and philosophers. (1) [Abby and Ben are at a party. Abby asks Ben about who their mutual friend Sarah is bringing as a date by uttering: “Who is Sarah bringing?” Ben answers:] Alex. (2) [Abby and Ben are at a party. She sees an unfamiliar man with Sarah, a mutual friend of theirs, and turns to Ben with a puzzled look on her face. Ben says:] Some guy she met at the park. The final utterances in (1) and (2) are generally known as “nonsentential speech” or “fragments” in the linguistic and philosophical literature (see the Introduction of this volume for a fuller discussion of the varieties of nomenclature in use). The primary observation to be made about these fragments is that they can have the same conversational function as (3) and (4) respectively: (3) Sarah is bringing Alex. (4) He’s some guy she met at the park. The question that this observation gives rise to is the following: How can such syntactically nonsentential objects give rise to well-formed assertions with fully sentential propositional content? More specifically, does the propositional content of (1) and
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.2 (126-182)
Jason Merchant
(2) come from (a) a novel method of generating and interpreting such fragments or (b) the usual mechanisms? The consequences for the architecture of the grammar are clear, depending on which option is correct. If option (a) is right, we need to allow nonpropositional semantic objects to be used to make assertions, whereas if option (b) is right, we need to implement what appears to be a kind of ellipsis. Option (a), which I will call the “direct interpretation” approach, has enjoyed considerable success, being pursued in various forms by van Riemsdijk (1978), Yanofsky (1978), Barton (1990), Stainton (1995, 1997, 1998), Ginzburg and Sag (2000), Jackendoff (2002), Culicover and Jackendoff (2005), and several contributions to this volume, among others. Option (b), which I will call the “ellipsis” approach, has had fewer adherents, the primary ones being Hankamer (1979), Morgan (1973, 1989), and Stanley (2000). In this chapter, I present data that support the ellipsis approach for some such “fragmentary” utterances, and argue that this ellipsis is preceded by movement of the fragment to a clause-peripheral position. This analysis is shown to correctly capture the full range of facts in English, Greek, and other languages from case matching, preposition stranding, binding theory connectivity, pronominal restrictions, scope, bound variable anaphora, polarity items, complementizer deletion, island sensitivity, and immobility effects. I wish to stress at the outset that it is clear, especially from the other contributions to this volume, that an ellipsis analysis is not appropriate for all such fragments: there are clearly, to my mind, a large number of phenomena for which a direct interpretation approach is preferable. Finding the boundaries between these sets of phenomena (if indeed they are nonoverlapping, which I doubt) is an important task to be taken up. Here I concentrate solely on those facts that seem to me to be the best candidates for an elliptical analysis, and for which I feel no well-worked-out, convincing nonelliptical alternative has been proposed.
. Fragment answers Fragment answers can be of a variety of categories, such as DPs, PPs, and VPs, as seen in the following examples: (5) A: Who did she see? B: John. (6) A: When did he leave? B: After the movie ended. (7) A: What does Bush want to do to Iraq? B: Take it over. In all these cases, I propose that the fragment has undergone movement to a clauseperipheral position, call it specFP, and that the head of this projection, F, hosts a feature E, which licenses the ellipsis of the host clause TP, out of which the fragment has moved. This is illustrated for (5) in (8) (see Merchant 2001 for details of the E feature).
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.3 (182-230)
“Small structures”: A sententialist perspective
(8)
FP
[DP John]2 F[E]
F’ she saw t2
. Evidence for ellipsis in fragments The first kind of evidence for ellipsis in fragments comes from case connectivity effects.
.. Case-matching connectivity effects In languages with overt case morphology on DPs, the case of a fragment must match the case of the wh-phrase that is its correlate (e.g., pjos in (9), pjon in (11)) in the antecedent question. This is illustrated for Greek in (9) through (12), but also holds for Russian, German, Urdu, Hebrew, Korean, and English. (9) Q: Pjos idhe tin Maria? who-nom saw the Maria ‘Who saw Maria?’ a. A: O Giannis. the Giannis-nom b. A: *Ton Gianni. the Giannis-acc (10) a.
A: O Giannis idhe the Giannis-nom saw ‘Giannis saw Maria.’ b. A: *Ton Gianni idhe the Giannis-acc saw ‘Giannis saw Maria.’
(11) Q: Pjon idhe i Maria? who-acc saw the Maria ‘Who did Maria see?’ a. A: *O Giannis. the Giannis-nom b. A: Ton Gianni. the Giannis-acc
tin Maria. the Maria-acc tin Maria. the Maria-acc
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.4 (230-306)
Jason Merchant
(12) a.
A: *I Maria idhe the Maria-nom saw ‘Maria saw Giannis.’ b. A: I Maria idhe the Maria-nom saw ‘Maria saw Giannis.’
o Giannis. the Giannis-nom ton Gianni. the Giannis-acc
On the ellipsis analysis, the case in the fragment answers comes from the usual mechanisms (verbal or T-related) for assigning cases internal to clauses.
.. Other connectivity effects In general, a fragment shows the same kinds of grammatical dependency effects that we find in fully sentential structures. This is illustrated for the binding principles A, B, and C in English in (13) through (15), and for the more complex anaphoric dependency found with the Greek item o idhios in (16) and (17). (13) Q: Where is he2 staying? a. *In John2 ’s apartment. b. *He2 is staying in John2 ’s apartment. (14) Who did John1 try to shave? a. *Him1 . b. *John1 tried to shave him1 . (15) Who does John like? a. Himself. b. John likes himself. (16) a.
Pjos nomizi o Giannis oti tha pari tin dhoulia? who thinks the Giannis that fut gets the job ‘Who does Giannis think will get the job?’ b. O idhios. the same ‘Him.’ (= Giannis1 thinks that he1 will get the job.) c. O Giannis1 nomizi oti tha pari tin dhoulia o idhios1 . the Giannis thinks that fut gets the job the same ‘Giannis1 thinks that he1 will get the job.’
(17) a.
Pjos nomizi oti tha pari tin dhoulia o Giannis? who thinks that fut gets the job the Giannis ‘Who thinks Giannis will get the job?’ b. *O idhios. the same c. *O idhios1 nomizi oti tha pari tin dhoulia o Giannis1 . the same thinks that fut gets the job the Giannis
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.5 (306-375)
“Small structures”: A sententialist perspective
If binding theory effects are due to syntactic constraints on co-occurence stated over phrase markers using c-command, then the behavior of indexed DPs in fragments can be taken to indicate that such phrase markers are present, if unpronounced.
. Evidence for movement in fragments There is also substantial evidence that the fragment has moved prior to the ellipsis.
.. Preposition stranding The first such evidence comes from the cross-linguistic distribution of prepositionless fragment answers to wh-phrases governed by a preposition. While such answers are grammatical in preposition-stranding languages, such as English and the Scandinavian languages, these kinds of answers are impossible in non-preposition-stranding languages like Greek, German, Yiddish, Czech, Russian, Bulgarian, Hebrew, and others (a reviewer points out that French patterns as expected with non-P-stranding languages); the Hebrew example is from Ginzburg and Sag (2000: 299). (18) English:
a. Who was he talking with? b. Mary.
(19) Swedish:
a.
Vem har Peter talat med? who has Peter talked with b. Mary.
(20) Norwegian: a.
Hvem har Per snakket med? who has Per talked with b. Mary.
(21) Danish:
a.
Hvem har Peter snakket med? who has Peter talked with b. Mary.
(22) Icelandic:
a.
(23) Greek:
a.
(24) German:
a.
Hvern hefur Pétur talað við? whom has Petur talked with b. Mary.
Me pjon milise i Anna? with whom spoke the Anna b. Me ton Kosta. with the Kosta c. *Ton Kosta. Mit wem hat Anna gesprochen? with whom has Anna spoken ?? b. Dem Hans. c. Mit dem Hans. with the Hans
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.6 (375-435)
Jason Merchant
(25) Yiddish:
a.
Mit vemen hot zi geredt? ‘With whom has she spoken?’ b. Mit Moshe. ‘With Moshe.’ c. ??Moshe.
(26) Czech:
a.
(27) Russian:
a.
S kým mluvila Anna? with whom spoke Anna b. S Jindˇrichem. ‘With Jindˇrichem.’ c. *Jindˇrichem. S kem ona govorila? with whom she spoke b. S Ivanom. ‘With Ivan.’ c. *Ivanom.
(28) Bulgarian: a.
S koj e govorila Anna? with who aux spoken Anna b. S Ivan. ‘With Ivan.’ c. *Ivan.
(29) Hebrew:
a.
Le-mi hixmeta? to-who you.flattered b. Le-Moti. ‘To Moti.’ c. *Moti.
This distribution is expected if the fragment undergoes movement, subject to language-particular constraints. It is a major puzzle for nonsententialist analyses, and to my knowledge, the nonsententialist approaches ignore this set of data completely (Culicover & Jackendoff 2005, e.g., fail to appreciate how difficult these data are for their analysis, and do not mention it at all. Casielles (Chap. 5, this volume) also overlooks these important data).
.. Islands Perhaps the most striking confirmation that fragments move comes from the fact that they are sensitive to islands. This can be seen in three contexts. The first is in answers to implicit salient questions, discussed by Morgan (1973) and Hankamer (1979). When no island is present, such fragment answers are as licit as their sentential counterparts. (30) a. Does Abby speak Greek fluently? b. No, Albanian. c. No, she speaks Albanian fluently.
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.7 (435-500)
“Small structures”: A sententialist perspective
(31) a. Did Abby claim she speaks Greek fluently? b. No, Albanian. c. No, she claimed she speaks Albanian fluently. But when the correlate to the fragment is embedded in an island, only the sentential answer is possible: The fragment answer, since it requires movement out of the island, is barred: (32) a. Does Abby speak the same Balkan language that Ben speaks? b. *No, Charlie. c. No, she speaks the same Balkan language that Charlie speaks. (33) a. Did Ben leave the party because Abby wouldn’t dance with him? b. *No, Beth. c. No, he left the party because Beth wouldn’t dance with him. (34) a. Did Abby vote for a Green Party candidate? b. *No, Reform Party. c. No, she voted for a Reform Party candidate. The second context where island sensitivity can be observed is in answers to multiple questions: (35) Who’s more likely to be influencing who? The CIA John Foreman, or John Foreman the CIA? (36) a.
Which committee member wants to hire someone who speaks which language? b. Abby wants to hire someone who speaks Greek and Ben wants to hire someone who speaks Albanian. c. *Abby Greek, and Ben Albanian.
It should be noted that while the island facts reported here have been judged unacceptable by more than a dozen speakers, island sensitivities for fragment answers are in general somewhat variable, due to factors that at present are unclear: Culicover and Jackendoff (2005), Stainton (Chap. 4, this volume), and Casielles (Chap. 5, this volume) all adduce examples that appear to involve island violations but that are nevertheless acceptable.
.. Complementizer deletion Morgan (1973) presents a puzzle from complementizer deletion; he notes that while in situ complementizers can be deleted after verbs like believe, in fragment answers the complementizer must be present: (37) What does no one believe? #(That) I’m taller than I really am. (38) No one believes (that) I’m taller than I really am.
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.8 (500-569)
Jason Merchant
This is exactly the pattern found with movement of CPs, as expected on the present analysis: (39) #(That) I’m taller than I really am, no one believes.
.. Polarity items The distribution of polarity items in fragment answers cross-linguistically also patterns after their moveability in the relevant language, as Giannakidou (2000) shows. In English, a polarity item like anything can neither serve as a fragment answer nor be fronted. (40) a. What didn’t Max read? b. *Anything. (41) a. Max didn’t read anything. b. *Anything, Max didn’t read. In a language like Greek, however, as discussed in Giannakidou, certain polarity items can be fronted; it is exactly these items that can also appear as fragment answers. (42) TIPOTA dhen idha. nothing-emphatic not I.saw ‘I didn’t see anything.’ (43) Q: Ti idhes? what you.saw A: TIPOTA. nothing-emphatic ‘Nothing.’
.. Turkish bare objects versus subjects Hankamer (1979: 395) discusses the distribution of “bare” DPs in Turkish (those occurring without the indefinite article/numeral bir ‘a/one’). While bare objects cannot be fronted, bare subjects can be. Expectedly, it is only bare subjects that can form licit fragment answers: (44) Q: Hasan ne yazıyor? ‘What is Hasan writing?’ A: *(Bir) mektup. ‘(A) letter.’ (45) Q: Hayvanların en aptalıne-dir? ‘Of the animals, the most stupid is what?’ A: Ayı. ‘Bear.’
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.9 (569-619)
“Small structures”: A sententialist perspective
.. Raising versus control infinitivals As Chomsky (1981: 62) points out, raising and control infinitivals differ in their moveability: Control infinitivals can be fronted, as in clefts, while raising infinitivals cannot. Expectedly, control, but not raising, infinitivals can appear as fragment answers. (46) a. *It’s [to procrastinate] that people tend. b. Q: How do people tend to behave? A: *To procrastinate. (cf. the well-formedness of the sentential answer People tend to procrastinate.) (47) a. It’s [to get a job in Europe] that she wants. b. Q: What does she really want? A: To get a job in Europe. (cf. She really wants to get a job in Europe.)
.. Predicate answers Answers that are predicates must include an entire VP, regardless of whether the semantics of the question would be satisfied with merely a V, as pointed out in Hankamer (1979: 239–243), and illustrated here: (48) A: What did he do to the car? B: Totaled *(it). (49) A: What did she do with the spinach? B: Washed *(it). (50) A: What did he do for his sister? B: Funded *(her). This follows on the present account as a consequence of structure preservation: Movement to specFP is phrasal, not head, movement, so the minimal informative constituent that can be used as a fragment answer is the VP, not the V.
.. Pronominal answer restrictions In languages with strong/weak pronoun distinctions, we find again a perfect correlation between those pronouns that can be fronted or occur at the left edge of the clause and those that can be used as fragment answers. Weak (or clitic) pronouns can do neither, while strong (or tonic) pronouns can do both. This is shown for Greek, French, German, and Dutch: (51) Greek:
Pjon idhes? whom did.you.see a. Afton. him-strong
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.10 (619-720)
Jason Merchant
b. *Ton. him-weak (52) French: Il voulait qui? ‘He wanted who?’ a. Moi. me-strong b. *Me. me-weak (53) German: Was wolltest du? what wanted-2sg you a. Das. ‘That.’ b. *Es. ‘It.’ (54) Dutch:
Wie heeft ze gezien? ‘Who has she seen?’ a. Jou. you-strong b. *Je. you-weak
(55) Greek:
{Afton / *Ton}, ton idha. {him-strong/him-weak} him I.saw
(56) French: {Moi / *Me}, il me voulait. {me-strong/me-weak} he me wanted (57) German: {Das / *Es} wollte ich. {that/it} wanted I (58) Dutch:
{Jou /*Je} heeft ze gezien. {you-strong/you-weak} has she seen
A similar distribution is found in English, as discussed by Yanofsky (1978), Napoli (1982), and Barton (1990). Only the “strong,” accusative form of pronouns can ordinarily occur in fragment answers. It is this form of the pronoun that similarly can bear focus on the left periphery of the clause as well. (59) Q: Who watered the plants? a. Me. b. *I. (60) a. *Me watered the plants. b. I watered the plants. (61) a. Me, I watered the plants. b. *I, I watered the plants.
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.11 (720-743)
“Small structures”: A sententialist perspective
But the fact remains that in particular the English facts are somewhat puzzling: as Casielles (Chap. 5, this volume), Progovac (Chap. 2, this volume), and Barton (Chap. 1, this volume) point out, something more needs to be said to account for the properties of the English pronouns. Casielles points out correctly that usually (in nonanswer contexts) the pronoun me in (61a) is topical, not focused as required by the answering context. I acknowledge this difficulty, and am inclined to believe that the accusative case that surfaces in English is in fact due to the left periphery – the pronoun receives nominative case in situ, but on moving to the left, it receives a new case assignment (whether by a default case mechanism as Progovac moots or by a functional head in the left periphery itself is immaterial), which in English overwrites the earlier case; this situation is parallel, I believe, to what is found in Greek (and other) apparently phrasal comparatives (see Merchant 2006; Lechner 2004), and to the case of subject-to-object raising in Japanese and Nieuan, and to other Suffixaufnahme (“case-stacking”) cases, for example as found in some Caucasian and Australian languages and in Korean. Case assignment in English is indeed a difficult matter (especially in copular clauses and other noncanonical argument positions), and it is perhaps not surprising that it is difficult in the case of fragment answers as well. What is perhaps more indicative is the situation found in languages where such complications are absent, such as Greek and German. In such languages, pronouns behave exactly like nonpronominal DPs, appearing in fragment answers in the case appropriate to their syntactic function within the elided clause (by hypothesis), including the nominative. If these languages have any default case at all, it is presumably nominative, as this is the case found in citation forms, titles, street signs, and so on. But in answer to a direct object question such as Who did you see? (Greek Pjon idhes? (who-acc you.saw), the only answers possible are in the accusative, including for pronouns (Esena (you-acc), *Esi (you-nom)). This fact appears to me to be unaccounted for on the nonsententialist approaches just cited. In summary, there is considerable reason to believe that fragment answers are syntactically fully sentential, but that ellipsis has rendered part of the clause unpronounced. This ellipsis is consistent with current views of ellipsis, which countenance only constituent deletion, only if the pronounced fragment undergoes movement to a clause-external position prior to the ellipsis. This theoretical requirement was seen to have substantial empirical support from a wide range of phenomena.
. Discourse-initial (?) fragments . Discourse-initial fragments Having seen that ellipsis is implicated in the derivation of fragment answers, we are now in a position to examine the second kind of fragment that opened this chapter, fragments with no linguistic antecedents at all, such as (62) (repeated from (2)), said in response to an inquiring glance.
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.12 (743-808)
Jason Merchant
(62) Some guy she met at the park. Such fragments form the basis for the radical revisions proposed most recently by Stainton in a series of papers (1995, 1997, 1998), following the important initial work of Barton (1990). The core of Stainton’s argument can be stated as follows. Premise 1: Ellipsis requires linguistic antecedents. Premise 2: Such fragments do not have linguistic antecedents. Conclusion: These fragments do not involve ellipsis (by modus tollens). Stainton’s (1995, 1997, 1998) argument fails to go through because Premise 1 is false, however. While ellipses cannot occur in true discourse-intial contexts (“DInull ”), they can occur without previous utterances in the discourse (“DIlang ”). Such fragments, then, may indeed involve ellipsis, but of a limited kind, licensed by the discourse context in the absence of linguistic material that could serve as a more explicit antecedent. There are a considerable number of examples of DIlang ellipses that have been discussed in the literature; I reproduce some of the most important ones here. (See Hankamer & Sag 1976; Schachter 1977, 1978; Hankamer 1978; Stanley 2000; Pullum 2001 for differing views on what to make of these data.) (63) a.
[Miss Clairol advertisement:] Does she or doesn’t she? Only her hairdresser knows. b. [As a response to an offer of a second piece of chocolate cake:] I really shouldn’t. c. [As an invitation to dance:] Shall we? d. [Seeing someone about to do a shot of Jenever:] If you can, I can, too. e. [Looking at someone about to jump off a bridge:] She won’t. f. [Seeing someone about to light their head on fire:] Don’t!
As claimed by Hankamer (1978) and Pullum (2001), these have something of the flavor of utterance idioms, and seem to some extent to be conventionalized. What I claim is conventionalized, however, is simply the VP [VP do it]. The meaning of this VP is licensed by the discourse relevance of some action; it need not have a determinate propositional content. Thus I share the judgment of Stanley (2000), for example (and pace Hankamer & Sag 1976) about Hankamer and Sag’s (1976) example; there is no difference in grammaticality for the two variants in (64). (A reviewer agrees that do it is a felicitous continuation for all of the examples in (63) except for (63c), which “does not sound like an appropriate way to ask someone to dance.” I concur.) (64) [Harry Houdini, before an audience of thousands, is attempting to escape from a locked safe dangling under a blimp. One spectator says to another:] a. Do you think he’ll be able to do it? b. Do you think he’ll be able to? The fact that the missing VP is do it also accounts for the following fact: While such DIlang VPs are possible in questions, as seen in (65), no extraction from the unpronounced VP is possible, as seen in (66).
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.13 (808-881)
“Small structures”: A sententialist perspective
(65) [Seeing three contestants about to buzz in:] Who do you think will first? (66) [Seeing a contestant about to pick among three choices:] *{Which (one)/What} do you think she will? The same holds, as a reviewer notes, for other examples modeled on (63): (67) a. *Key lime pie, I really shouldn’t. b. *The tango, shall we? c. *This bridge, I won’t. Furthermore, because the VP do it is restricted to nonstatives, such DIlang uses should equally bar stative meanings; the correctness of this prediction is borne out in the following: (68) [Abby has a ten-year-old younger sister, who she discovers one day in front of their mother’s dresser. The younger sister has put on their mother’s clothes, done up her hair like their mother, put on their mother’s jewelry, and in general done everything possible to resemble their mother. She is in the very act of applying their mother’s lipstick when Abby enters the room and observes all this. Abby is horrified and shouts:] Don’t! a. = Don’t do it! (Here possibly = Don’t put on that lipstick!) b. = Don’t resemble our mother! The restriction to nonstatives also explains why, although (69a) is a highly conventional invitation to dance, and if anything would therefore be expected to be found in such contexts with a reduced VP, (69b) is actually impossible in such a meaning, since {like/care} to dance is stative. (69) [As an invitation to dance:] a. Would you {like/care} to dance? b. #Would you? b’. Would you do it? = Would you {like/care} to dance? Resolution of DIlang deictics and/or pronominals is similar: (70) [Responding to a puzzled glance at an unfamiliar person:] {That/He}’s some guy she met at the park. Since such deictics and pronominals are licensed in DIlang contexts, and since the copula is similarly licensed, it is possible that DIlang fragments have the following derivation: (71) [FP some guy she met at the park1 ]
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.14 (881-948)
Jason Merchant
Initial support for the correctness of this analysis comes from case facts in casemarking languages like Greek and German. The case that occurs on such fragments is the nominative, the same case that occurs on predicate nominals with the copula. (72) Greek:
a.
Kapjos pu gnorise sto parko. someone-nom that she.met in.the park b. *Kapjon pu gnorise sto parko. someone-acc that she.met in.the park
(73) German: a.
Ein Typ, den sie im Park kennengelernt hat. a-nom guy that she in.the park met has b. *Einen Typ, den sie im Park kennengelernt hat. a-acc guy that she in.the park met has
(74) Greek:
a.
Aftos ine {kapjos/*kapjon} pu he is {someone-nom/someone-acc} that sto parko. in.the park German: b. Das ist {ein/*einen} Typ, den sie im that is {a-nom/a-acc} guy that she in.the kennengelernt hat. met has
gnorise she.met
Park park
Other differences between sentential antecedents and fragment antecedents to ellipsis discussed in Stainton (1998: 326) follow from the proposed analysis. Since the unpronounced material in the fragment in (76) is equivalent to the deictic and copula pronounced in (77), it comes as no surprise that the predicate ellipsis in Mark’s utterance is equally strange in both cases (since it is resolved as And Betty is the man from Paris, too, or at best, And Betty is from Paris, too – in any case, not as And Betty is at the door, too). (75) Jason: Mark:
The man from Paris is at the door. And Betty is, too.
(76) Jason: Mark:
??And
The man from Paris. Betty is, too.
(77) Jason: Mark:
??And Betty is too.
That’s the man from Paris.
Other predicates, in certain extremely limited and highly conventionalized contexts, may be elided as well. Perhaps the most common such context is one of ordering something from a waiter at a restaurant. In this context, a small range of elements (bring, give, I want, I’d like) can be elided, with the consequence that we observe the relevant case: in Greek accusative, in Russian the genitive (partitive).
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.15 (948-1013)
“Small structures”: A sententialist perspective
(78) Greek:
a.
(Enan) kafe (parakalo)! a coffee-acc please ‘(A) coffee (please)!’ b. Ferte mou (enan) kafe (parakalo)! bring-imp me a coffee-acc please ‘Bring me (a) coffee (please)!’
(79) Russian: a.
Vody (pozhalujsta)! water-gen please ‘(Some) water (please)!’ b. Dajte mne vody (pozhalujsta)! give-imp me water-gen please ‘Give me (some) water (please)!’
The English equivalent in such contexts is underdetermined by its lack of distinguishing case morphology: (80) Water! The moral of all these examples is merely that richer contexts help to provide nonlinguistic antecedents to ellipsis. This being the case, a “limited ellipsis” analysis of fragments in DIlang contexts is feasible, and no adjustments to the usual syntaxsemantics-pragmatics interfaces are needed for such phenomena.
. Nonelliptical varia There are a number of other phenomena that are sometimes discussed in connection with nonsententials or “small structures,” and that can in most cases occur with no antecedents (implicit or overt). These other nonsentential types fall roughly into the classes in (81) through (88) (see Shopen 1972; Sadock 1974; Yanofsky 1978; Klein 1985; Morgan 1989; Barton 1990, 1998; Schlangen 2003; Culicover & Jackendoff 2005; Stainton 2004, with additional classes in Klein; Schwabe 1994; and Schlangen). In such cases, there is no good evidence to my mind for any kind of ellipsis being involved. (81) [Short directives:] Left! Higher! Scalpel! (82) [Special registers: telegrams, headlines, weather reports, diaries, recipes, instructions:] If no paper, turn wheel. (83) [Labels, titles: (cf. Bühler’s dingfest angeheftete Namen (‘names adhering to things’), Bühler 1934: §10):] a. Campbell Soup. b. Starbucks. c. Thief! Thief! [cf. Paul 1919: 1 “Wenn jemand den Angst- und Hilferuf ‘Diebe’ ausstößt, so will er, daß der Allgemeinbegriff ‘Diebe’ mit einer von ihm in dem Augenblick gemachten Wahrnehmung in Beziehung gesetzt werde.” (‘When someone cries
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.16 (1013-1087)
Jason Merchant
out in fear and for help “thieves,” he intends that the general concept “thieves” be connected to what he is perceiving at that moment’).] d. Fire! e. And now: the first act of the night: The Rolling Stones! f. To Kill a Mockingbird g. German: Der Zauberberg (‘The Magic Mountain’) h. Next exit: Chicago. i. E 61st St. (84) [Expressive exclamations:] Wonderful! Nonsense! Fate! For Pete’s sake! (85) Utterance idioms (Klein’s (1985) “elliptische Formeln” (‘elliptical formulas’):] a. Up yours. b. German: Gewitter im Mai – April vorbei (lit. ‘storms in May – April over’) c. German: Wenn schon, denn schon (lit. ‘If already, then already’; roughly, ‘In for a penny, in for a pound’) d. Dutch: Met Jason [(‘with Jason’) as a telephone greeting] (86) [Other nonsentential partially fixed material expressions ((b)–(h) from Culicover & Jackendoff 2005):] a. So much for the light of reason. b. Off with his head! c. A good talker, your friend Bill. d. Books open to page 15! e. How about a cookie? f. What, me worry? g. Hey, Phil! h. Vikings 27, Bears 3 (87) [Some kinds of fragments (e.g., Schlangen (2003)’s “explanation” subtype):] a. Mary: Try it. It’s good for you. b. Peter: Why? c. Mary: Lots of vitamins. (88) [Greetings:] Hello. Good-bye. Roger. Over. Out. It is now possible and indeed critical to ask ourselves what separates these classes from the ones discussed earlier, especially question-answer pairs, for which some kind of elliptical analysis still seems to me to be unavoidable. We have arrived at a state of affairs where we must admit of two possible analyses generated by the grammar for, for example, seemingly bare DPs: one when they are used as answers to questions (a sentential source, subject to ellipsis), and another one when they are used as, for example, titles of books. As theorists, we in general try to avoid such redundancy if possible; in the present case, however, it is not possible. The question then becomes under what circumstances one or the other analysis is actually employed: what factors play a role in determining which strategy will be used to generate a “fragment”? At this
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.17 (1087-1142)
“Small structures”: A sententialist perspective
stage in our understanding, no firm answer seems possible, though the most obvious distinction is the one Hankamer and Sag (1976) made between “deep” and “surface” anaphora: the elliptical forms are “surface anaphora,” in their terms, and generally require a linguistic antecedent (modulo the “limited ellipsis” strategy outlined in the previous subsection), while the nonelliptical ones are “deep anaphora.” The latter do not require a linguistic antecedent. Applying this distinction to the phenomena here, we can claim that when a linguistic antecedent is available, the grammar (user) must make recourse to it, forcing an elliptical analysis when one is possible. (This accounts for the fact that in case-marking languages such as German and Greek, a default case such as the nominative found in titles and signs, is not possible on an NP short answer to a question over an oblique case position. The elliptical form must be used, although the other form is generable.) Whether this is a matter internal to the grammar itself (required comparison of derivations with differing numerations relative to a context) or a grammar-external pragmatic matter is not a question I can resolve here, though I suspect it is the latter (i.e., using the nonmatching, nonelliptical form leads to a stark discourse infelicity registered as unacceptability, not ungrammaticality in a narrow sense). It is exactly the fact that a nonelliptical analysis is available in the grammar, and employed in these other contexts, that makes a uniform treatment such as that envisioned in Culicover and Jackendoff (2005) inadvisable.
. Conclusion In brief, I have attempted to show that certain fragments can be analyzed within a conservative theory of the syntax-semantics interface, by positing ellipsis, or in some cases, a “limited ellipsis” analysis. Ellipsis alone does not account for all the attested properties of even these fragments, however; we have seen that the full range of properties – the various connectivity effects in particular – falls out from the movement component of the proposed derivation. This movement furthermore makes the ellipsis implicated in fragments consistent with our current understanding of ellipsis, one that permits only ellipsis of constituents. Finally, it is clear that we must also countenance mechanisms in the grammar for generating nonelliptical “small structures,” and find a satisfying way to delimit their application, prohibiting them from use in contexts where a sentential, elliptical analysis is possible.
Note * I wish to thank the editors of this volume for giving me the opportunity to present this research at their workshop at Wayne State and for the many fruitful discussions that this workshop afforded, as well as for their patience and encouragement in making sure this chapter was included. The main conclusions and data presented here in brief form can be found in much fuller form in Merchant (2004), to which the reader is referred for details and further exposi-
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.18 (1142-1236)
Jason Merchant
tion; this chapter is essentially a slightly amended summary of Sections 3 and 5 of that paper, and is included in this volume in part to give a better sense of the contributions at the workshop. The other contributions to this volume, in particular those of Barton, Stainton, Progovac, and Casielles, have convinced me that a more moderate, intermediate stance is required to account for the full range of data on fragments. Nevertheless, it remains clear to me that completely eschewing structural ellipsis, as for example proposed by Culicover and Jackendoff (2005), cannot at present account for many of the facts discussed here (especially the P-stranding generalization, the distribution of complementizers, predicate answers, and most of the case-matching effects in languages other than English), nor for others relevant to the larger debate (such as the difference in voice mismatch licensing under VP ellipsis vs. sluicing).
References Barton, E. (1990). Nonsentential Constituents. Amsterdam: John Benjamins. Barton, E. (1998). The grammar of telegraphic structures: Sentential and nonsentential derivation. Journal of English Linguistics, 26, 37–67. Bühler, K. (1934). Sprachtheorie. Die Darstellungsfunktion der Sprache. Jena: Fischer. Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris. Culicover, P. & Jackendoff, R. (2005). Simpler Syntax. Oxford: OUP. Giannakidou, A. (2000). Negative . . . concord? Natural Language and Linguistic Theory, 18, 457– 523. Ginzburg, J. & Sag, I. (2000). Interrogative Investigations: The Form, Meaning, and Use of English Interrogatives. Stanford, CA: CSLI. Hankamer, J. (1978). On the nontransformational derivation of some null VP anaphors. Linguistic Inquiry, 9, 66–74. Hankamer, J. (1979). Deletion in Coordinate Structures. New York, NY: Garland. Hankamer, J. & Sag, I. (1976). Deep and surface anaphora. Linguistic Inquiry, 7, 391–428. Jackendoff, R. (2002). Foundations of Language: Brain, Meaning, Grammar, Evolution. Oxford: OUP. Klein, W. (1985). Ellipse, Fokusgliederung und thematischer Stand. In R. Meyer-Hermann & H. Rieser (Eds.), Ellipsen und fragmentarische Ausdrücke (pp. 1–24). Tübingen: Niemeyer. Lechner, W. (2004). Ellipsis in Comparatives. Berlin: de Gruyter. Merchant, J. (2001). The Syntax of Silence: Sluicing, Islands, and the Theory of Ellipsis. Oxford: OUP. Merchant, J. (2004). Fragments and ellipsis. Linguistics and Philosophy, 27, 661–738. Merchant, J. (2006). Phrasal and clausal comparatives in Greek and the abstractness of syntax. Ms., University of Chicago. Morgan, J. (1973). Sentence fragments and the notion ‘sentence.’ In B. Kachru, R. Lees, Y. Malkiel, A. Pietrangeli, & S. Saporta (Eds.), Issues in Linguistics (pp. 719–751). Urbana, IL: University of Illinois Press. Morgan, J. (1989). Sentence fragments revisited. In B. Music, R. Graczyk, & C. Wiltshire (Eds.), CLS 25, Parasession on Language in Context (pp. 228–241). Chicago, IL: Chicago Linguistic Society. Napoli, D. J. (1982). Initial material deletion in English. Glossa, 16, 85–111. Paul, H. (1919). Deutsche Grammatik (Vol. III). Halle: Niemeyer.
JB[v.20020404] Prn:20/07/2006; 10:20
F: LA9303.tex / p.19 (1236-1299)
“Small structures”: A sententialist perspective
Pullum, G. (2001). Hankamer was! In S. Chung, J. McCloskey, & N. Sanders (Eds.), The Jorge Hankamer WebFest. University of California, Santa Cruz: http://ling.ucsc.edu/Jorge/ pullum.html Sadock, J. (1974). Read at your own risk: Syntactic and semantic horrors you can find in your own medicine chest. In M. W. LaGaly, R. A. Fox, & A. Bruck (Eds.), Papers from the 10th Regional Meeting of the Chicago Linguistic Society (pp. 599–607). Chicago, IL: Chicago Linguistic Society. Schachter, P. (1977). Does she or doesn’t she? Linguistic Inquiry, 8, 763–767. Schachter, P. (1978). English propredicates. Linguistic Analysis, 4, 187–224. Schlangen, D. (2003). A Coherence-Based Approach to the Interpretation of Non-Sentential Utterances in Dialogue. Ph.D. dissertation, University of Edinburgh. Schwabe, K. (1994). Syntax und Semantik situativer Ellipse. Tübingen: Narr. Shopen, T. (1972). A Generative Theory of Ellipsis: A Consideration of the Linguistic Use of Silence. Bloomington, IN: Indiana University Linguistics Club. Stainton, R. (1995). Non-sentential assertions and semantic ellipsis. Linguistics and Philosophy, 18, 281–296. Stainton, R. (1997). Utterance meaning and syntactic ellipsis. Pragmatics and Cognition, 5, 51– 78. Stainton, R. (1998). Quantifier phrases, meaningfulness ‘in isolation,’ and ellipsis. Linguistics and Philosophy, 21, 311–340. Stainton, R. (2004). In defense of non-sentential assertion. In Z. Szabo (Ed.), Semantics vs. Pragmatics (pp. 383–457). Oxford: OUP. Stanley, J. (2000). Context and Logical Form. Linguistics and Philosophy, 23, 391–434. van Riemsdijk, H. (1978). A Case Study in Syntactic Markedness: The Binding Nature of Prepositional Phrases. Dordrecht: Foris. Yanofsky, N. (1978). NP utterances. In D. Farkas, W. Jacobsen, & K. Todrys (Eds.), Papers from the 14th Regional Meeting of the Chicago Linguistic Society (pp. 491–502). Chicago, IL: Chicago Linguistic Society.
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.1 (47-122)
chapter
Neither fragments nor ellipsis* Robert J. Stainton Jason Merchant (2004, and Chap. 3, this volume) proposes to account for all speech acts performed with “fragments,” whether in discourse-initial position or otherwise, by appealing to syntactic ellipsis. Though his proposal is insightful, I offer empirical and methodological considerations against it. Empirical problems include: (a) His alleged “elliptical sentences” do not embed the way they should; (b) in some cases where Merchant requires fronting to take place, it is blocked – either by an island (e.g., in English) or because nonsubject fronting is not allowed in the language in question (e.g., in Malagasy); and (c) his “limited ellipsis” strategy, allowing do it and this is __ to be licensed in discourse-initial position, is not general enough. The methodological problem is that, his protests to the contrary, Merchant’s view multiplies hidden structure without necessity.
.
Introduction
It appears that speakers can produce ordinary words and lexically headed phrases and thereby perform speech acts. What’s more, it appears that they can do so when no appropriate linguistic antecedent is in place. For instance, to introduce some attested examples, Benigno got into a taxi and said To Segovia. To the jail.1 With these discourse-initial words, he thereby ordered the driver to take him to Segovia. Or again, a theorist was discussing whether humans in general suffer from a recently noticed cognitive deficit. Dirk leaned over to a friend and whispered, Just him. Dirk here joked that it is just the theorist himself who suffers from the deficit in question. Dirk seemingly got this across with a mere phrase, however, not with a full sentence; and Dirk was not replying to an interrogative, nor to any other linguistic expression that one would usually think of as licensing Just him. In yet another example, a father was worried that his daughter was going to spill her chocolate milk. The glass was very full, and she was quite young, and prone to accidents. He said, or appeared to say, Both hands. The father thereby instructed his daughter to use both hands, using not a sentence but a mere phrase. As with the second example, there was talk going on prior to this utterance, so in one sense there was prior discourse; but there was not an appropriate linguistic trigger for Both hands. So, in that sense, we seemingly have what I will call a “linguistically discourse-initial” usage of a mere phrase in a speech act.
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.2 (122-187)
Robert J. Stainton
Some theorists have rejected these appearances. In particular, some have contended that even linguistically discourse-initial examples are really cases of syntactic ellipsis of a familiar sort. This is precisely the view of Jason Merchant (2004, and Chap. 3, this volume). It is Merchant’s account of linguistically discourse-initial cases of (what appear to be) nonsentential speech acts that will be my focus here. The game plan is as follows. I will quickly rehearse Merchant’s clever and insightful proposal, and will also explain its very significant advantages compared to previous attempts to reject the appearances by appeal to syntactic ellipsis. I explain both his general account (for cases in which there is a linguistic antecedent of some kind), and his more specific suggestion for the kinds of case that have been my own focus. My aim here is threefold. First, I want to make clear why (to my mind anyway) Merchant’s is the most sophisticated attempt so far to deal syntactically with the phenomenon at issue. Fairness demands as much. Second, though Merchant’s own discussion is admirably clear, it is directed at professional linguists. Given the multidisciplinary audience of this volume, I hope to present his proposal in a way that philosophers and others can readily understand. Third, I fear that some of the seeming advantages vis-à-vis cases that do have linguistic antecedents (e.g., the answer in a question-answer pair) don’t ultimately hold up, once all the facts about discourse-initial cases are on the table. That is, some familiar objections to ellipsis accounts that seem to have been definitively overcome by Merchant really are not. So, it’s important to rehearse those familiar objections, and Merchant’s treatment of them. Having explained the various facets of Merchant’s view, I provide empirical and methodological considerations against it. I conclude that the appearances noted earlier cannot be set aside by appeal to syntactic ellipsis. More than that, not only is there not ellipsis going on, there aren’t genuinely “fragments” at play either (hence the title of the chapter). Instead, precisely as the appearances suggest, people are using plain old words and phrases, rather than “fragments” of anything, to perform speech acts. Before moving forward, I should issue some caveats. To begin with, I am not defending the idea that syntactic ellipsis never happens between sentences. I do not do so because both Morgan (1973, 1989) and Merchant (2004, and Chap. 3, this volume) provide examples that seem, at present, to call for a syntactic account. My burden is merely to argue that syntactic ellipsis is not the whole story; I’m happy for it to be part of the story. I should also stress that the present chapter is almost entirely negative. I present and criticize a view, but I don’t here propose an alternative account. Nor do I here consider the important consequences of rejecting, as a general account, a syntactic ellipsis story – for linguistic theory, for cognitive science more broadly, or even for philosophy of language/mind. Those interested in my positive view and its implications should consult my Words and Thoughts (2006). (I also discuss syntactic ellipsis accounts far more thoroughly there.)
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.3 (187-233)
Neither fragments nor ellipsis
. Merchant’s proposal and its advantages Merchant’s innovative account can be divided into two facets. There is a part that is needed to handle “fragments” of all kinds, whether in linguistically discourse-initial position or not. This part essentially extends and revises Merchant’s (2001) theory of sluicing from within-sentence cases to between-sentence cases. The second facet is specifically introduced to account for “fragments,” as he calls them, that occur without an antecedent. I will explain these two facets in turn, moving then to the advantages of Merchant’s story. A warning, however, before I proceed: what follows is a brief sketch of Merchant’s empirically and theoretically rich view. Given the purpose of this section, details are omitted. Maybe more importantly, trees are introduced that include elements that are no longer the “cutting edge” in theoretical syntax. At the same time, certain quite new posits are sometimes included, resulting in trees that are hybrids of various frameworks. This is inelegant, but the aim is to render the story more accessible to those not specializing in syntax.
. The [E] feature and movement Merchant’s account introduces a morphosyntactic feature that “serves as the locus of all the relevant properties that distinguish the elliptical structure from its nonelliptical counterpart” (2004: 670). He calls it the [E] feature. Let me briefly review its phonology, syntax, and semantics. Consider (1): (1) Jose lives in Canada, and we know why On Merchant’s story, what appears “below the surface” in such a construction is: (2) Jose lives in Canada, and we know why [E][IP Jose lives in Canada] The [E] feature tells the phonological component not to parse its complement. (This rule is easy enough to state. How to implement it in a parsing and production system is another matter. But I put that aside.) In the present example, the presence of the [E] feature as the sister to the embedded IP tells the phonological component not to pronounce this embedded sentence at all. So, what we hear is not the whole sentence We know why Jose lives in Canada, but only what’s not within the scope of [E], namely, we know why. Turning to syntax, there is variation across languages with respect to where [E] can appear. For instance, in English, the feature [E] can take a VP as its complement, as in Jose lives in Canada but we don’t [E][VP live in Canada]; in German, [E] cannot so appear. (There are issues in syntactic theory about what gives rise to such variation. Happily, such details needn’t detain us.) What will prove especially important about its syntax for our purposes is that [E] appears on a node in a tree, and operates on its sister node. Equally crucial is that the licensing condition for [E] is not that the precise syntactic material that follows it should appear in prior discourse. Instead, [E] is licensed when the right sort of “similar content” is present in prior discourse. This is
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.4 (233-268)
Robert J. Stainton
what allows for the well-formedness of sentences like John likes his own dog and Mary too: the omitted syntactic material, viz. likes her own dog, is not in fact present in the prior clause, but a similar content is (more on this later). Taking (1) as our example once again, the syntactic tree for the second conjunct, we know why, would thus be something like (3): (3)
IP NP
I’
we INFL
VP V
CP
know C[E] why
IP NP
I’
Jose INFL
VP V
PP
live
in Canada
We hear only we know why because [E] applies to its sister, the IP node to its right. And this is licit because the content JOSE LIVES IN CANADA is present. (Identical contents are a subcase of “similar” ones.) That [E] always operates on its sister node leaves one with an immediate problem: how is ellipsis of a nonconstituent to be accounted for? Consider, for example, what material would have to be omitted from a complete source sentence to account for the following attested case. Meera is putting jam on her toast. As she scoops out the jam, she says, Chunks of strawberries. It would seem that the source sentence would be something like This jam contains chunks of strawberries. But This jam contains, the part elided, cannot be the sister of [E], because This jam contains is not even a constituent. Nor is this problem merely a feature of the example This jam contains chunks of strawberries. A similar problem arises fairly generally: typically we need to elide a nonconstituent to arrive at what was actually pronounced. Thus take London as an answer to Where is Jim going to? The omitted material would seemingly be Jim is going to, patently not a constituent. The heart of the difficulty, of course, is that when one retains a constituent from a complete sentence, as in ellipsis, what’s taken away is not, as a general rule, itself a constituent: prune a tree to leave a single node, and what’s re-
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.5 (268-326)
Neither fragments nor ellipsis
moved won’t, even typically, be a constituent. But then how can Merchant’s [E] apply in the general case? One might take refuge in the idea that there is [E]-style ellipsis, for when a constituent is elided, and then other kinds of ellipsis. For instance, there are familiar apparent exceptions to constituent-only deletion, like John loves hot salty beans and Mary corn, where the omitted element seems to be the nonconstituent loves hot salty. But the familiar exceptions are limited in scope. In contrast, allowing rules to operate on the collection of nodes that would be required to derive Chunks of strawberries from This jam contains chunks of strawberries, and related cases, would leave us with no constraints at all on what things can be the targets of syntactic processes. In short, the “only constituents” generalization might admit of a few exceptions; but it’s worth saving what we can. Being well aware of this problem, Merchant proposes an ingenious solution: when we seem to have nonconstituent deletion, what is really going on is movement of the pronounced material to the front of the sentence, with [E] operating on the node out of which that material was moved. Taking the London example again, the idea is that the real source sentence is: (4) London1 [E][IP Jim is going to t1 ] Notice how nicely this works. [E] is licensed because the immediately prior question Where is Jim going to? affords the right sort of content. [E] applies to a constituent, its sister. And the only part predicted to be pronounced is London. Let’s take a second example of movement-then-deletion, and fill in some extra details. Merchant notes that (5) and (6) constitute a perfectly fine discourse: (5) Who did she see? (6) John Even such a non-discourse-initial example seems problematic, however, because it seems like the abbreviated answer (6) derives from the full sentence (7), with the omitted material being She saw: (7) She saw John The problem is that She saw is patently not a syntactic constituent of (7). Merchant’s suggestion, applied to this example, is that the source sentence for the answer John is actually (8), with his [E] feature taking the entire embedded sentence as its complement, and John having moved into focus position:
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.6 (326-368)
Robert J. Stainton
FP
(8) NP
F’ IP
John1 F[E] NP
I’
she INFL past
VP V
NP
see
t1
(A word about notation: “FP” is FocusP. FP is a structure that sits above an ordinary sentence, and that creates a spot into which a focused item can move. For the details, see Rizzi 1997.) Given that John has moved out of the structure to which the [E] feature applies, we hear that word and nothing else. Given that the thing omitted is an IP, however, we do not have deletion applying to a nonconstituent. I have rehearsed, albeit briefly, the phonology and syntax of [E]. I’ve also introduced Merchant’s suggestion that movement comes into play when we have (seeming) cases of nonconstituent ellipsis. I turn now to the semantics of [E]. Here, things are a bit complicated. In VP ellipsis, [E] stands for a function from a propositional function to a propositional function. In sluicing, [E] stands for a function from a proposition to a proposition. (That is, at the level of extension, [E] stands for something of Montagovian type , <e, t>> in VP ellipsis, and something of type in sluicing.) Taking sluicing as our example, this function works as follows. If the input proposition is appropriately related to the content of the linguistic antecedent, the function in question outputs the input proposition. Continuing with example (1), for instance, if the proposition input to the denotation of [E] is JOSE LIVES IN CANADA, then the function in question outputs this very proposition. That’s because this proposition is appropriately related to the content of the antecedent sentence, Jose lives in Canada. In particular, the two contents are identical; and, as noted earlier, that’s a subcase of “appropriately related.” Having been output, this content is then passed up the tree, ultimately serving as the argument for WE KNOW WHY. Crucially, the function denoted by [E] does not always output the very proposition input. It does so only when the input is appropriately related to the antecedent sentence. For example, if in (1) the argument to the function denoted by [E] had been FISH SWIM, then the function would fail to provide an output. That input would make the whole derivation of the semantic content crash. Thus does the “appropriately related” constraint on [E]’s de-
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.7 (368-424)
Neither fragments nor ellipsis
notation explain why sentence (1) cannot mean JOSE LIVES IN CANADA AND WE KNOW WHY FISH SWIM.
. The limited ellipsis strategy I said that Merchant’s story has two facets. Having just explained the first, viz. the application of [E], sometimes preceded by movement, I turn to the second facet. As noted, the mechanism about to be introduced is specifically designed to allow for the appearance of “fragments” in linguistically discourse-initial position. To see why this second facet is required, recall that the function that [E] denotes will only output a proposition if the argument is appropriately related to the proposition expressed by the linguistic antecedent – which seemingly requires that there be one. To handle this kind of case, Merchant offers another astute suggestion. He says, in effect, that there are two exceptions to his generalization about what the content of the input proposition must be appropriately related to. First, when [E] appears as sister to the verb phrase [VP do it], [E] can output a propositional function as long as the linguistic or nonlinguistic context resolves what it refers to. Second, when [E] appears as sister to a sentential constituent that consists of (a) a demonstrative or expletive subject and (b) the copula, then [E] can output a proposition if a referent for the subject is salient and the “existence predicate” is manifest. (Let’s not pause to consider what the latter would amount to.) Merchant writes: In short, I’m proposing a kind of “limited ellipsis” analysis, one in which a demonstrative (such as this/that or a pronoun in a demonstrative use) or expletive subject and the copula are elided – given the appropriate discourse context, which will be almost any context where the speaker can make a deictic gesture, and where the existence predicate can be taken for granted. (2004: 725)
It may help to illustrate both kinds of case. So, suppose that two people look into a room, and see a horrible mess on the floor. Alison says, Lauren. Alison here gets across that Lauren did it, even though there is no linguistic antecedent. Merchant’s story, applied to this case, is that Alison really produced (9): (9) [IP Lauren [[E][VP did it]]] The [E] feature is licensed here, despite the lack of a linguistic antecedent, precisely because an awful mess is a very salient thing indeed – enough so that the context determines that it refers to the mess. That’s how the verb phrase [VP do it] works. Here are two examples of full (embedded) sentences being omitted. Suppose Juhani holds up a letter and says From Spain, thereby asserting of the letter that it is from Spain. Merchant will claim that what Juhani actually produced is not the bare phrase [PP from Spain], but rather the sentence in (10) – a sentence that has been subject not just to ellipsis but to movement (viz., focus fronting) as well:
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.8 (424-463)
Robert J. Stainton FP
(10)
F’
PP1
IP
From Spain F[E] NP
I’
this INFL
VP
3SG.PRES V
PP
be
t1
To give another example, if Juhani points at a woman and says Sam’s mother, thereby asserting of the woman that she is Sam’s mother, Merchant’s view would have it that Juhani produced the sentence (11): FP
(11)
F’
NP1 Sam’s mother
IP
F[E] NP
I’
that INFL 3SG.PRES
VP V
PP
be
t1
I should address, on Merchant’s behalf, a natural objection to this means of dealing with the lack of a linguistic antecedent. Syntactic ellipsis is something that happens within the language faculty. It isn’t a matter of the agent as a whole “guesstimating” what sentence the speaker had in mind. Rather, as one used to say, it’s a matter of specifically linguistic competence. But, goes the objection, Merchant seems precisely to be letting nonlinguistic context, real-world facts, feed syntactic ellipsis. There is, however, a ready reply to this imagined worry. On Merchant’s (2004) view there are only two exceptional cases. So, when the language faculty encounters syntactic ellipsis without an appropriate linguistic antecedent, there are only two possibilities for what
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.9 (463-519)
Neither fragments nor ellipsis
must be filled in: [VP do it] or [IP this/that [I’ is t]]. Picking between these may indeed require appeal to the extralinguistic environment; but disambiguation, which is essentially what would be going on here, has never previously been thought to threaten the autonomy of the language faculty. What’s more, it’s part of the proposal that these unpronounced structures will only be employed when appropriate referents for their indexicals are salient. And, here again, this kind of assignment of reference to indexicals is something that everyone must grant to be linguistically tractable, even in linguistically discourse-initial situation – because we find it in sentential speech. (See King & Stanley 2005 and Stanley 2000 for related points.) In short, we have one exception in a VP ellipsis construction, and one in a sluicingtype clausal deletion: the argument to [E]’s content in these two exceptional circumstances does not need to be appropriately related to a propositional content expressed by a prior linguistic item.
. Advantages Some advantages of Merchant’s proposal are already clear. Unlike just about every previous attempt to treat between-sentence “fragments” as syntactic ellipsis, Merchant can handle cases that seem to involve nonconstituent deletion. That’s the role of movement. What’s more, as he notes, appeal to movement explains some otherwise quite puzzling “anticonnectivity” effects. These are cases in which the word/phrase used in isolation behaves differently than it does in the seeming source sentence. Here is a first example. Notice that certain items that have both negative and “free-choice” readings cannot occur unembedded with their negative reading in linguistically discourse-initial position – even when they can so occur in a supposed source sentence. Thus if the issue of what Max will choose not to read is in the air, I cannot assert, with anything, that MAX WILL CHOOSE TO READ NOTHING. I cannot do this even if the hearer could use the context to infer that this was meant: regardless, anything cannot be used in its negative sense here. Even in such a discussion, anything could only be used to mean that Max will just randomly pick, among the options, which book will go unread. The question is: why is the former interpretation not available? There’s a simple answer on the story according to which subsentence use is genuine, and the appearances reflect the reality: anything is occurring in isolation, unembedded; hence it appears without a negative polarity item as a licensor; so, it can only be read as “free choice,” that is, as “whatever.” In short, because there is no licensor if subsentential speech is genuine, anything simply can’t be read negatively in the situation described. Apparently, however, there is no answer forthcoming on the ellipsis account, since the sentence Max will choose not to read anything is fine when read negatively (indeed, when read in that way, the sentence sounds better); and this could serve as a source sentence. Why, given this, can’t one simply elide all the words but the last, thereby yielding the negative existential reading? This is the (first) puzzle of anticonnectivity.
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.10 (519-593)
Robert J. Stainton
Merchant’s move-then-delete account affords an elegant solution. To get anything on its own, it must first be fronted. Otherwise deletion would have to apply to the nonconstituent Max will choose not to read __. Thus the source sentence is not Max will choose not to read anything but rather (12). However, in sentence (12), with anything fronted, we only find the free-choice reading. The negative existential reading is not available. (12) Anything, Max will choose not to read So, what at first looks like a problem for ellipsis accounts, turns out to be evidence in favor of Merchant’s story: such anticonnectivity is what we should find. Here is a second example. The case assignment one finds on subsentences is often different from what appears on the corresponding item within sentences. Thus in reply to the question, Who likes Elvis best?, the sentential answer is I like Elvis best. But if the question hasn’t been explicitly asked, though it’s clear that it’s at issue, the “fragment” answer is Me! not I! Or again, Korean has both case-marked noun forms and caseless forms, each apparently occurring unembedded.2 Morgan (1989) notes that when embedded in a sentence, the caseless NPs are ill formed – hence they presumably cannot appear in the supposedly sentential “source” from which a fragment is derived. Yet those forms can be used on their own, to perform speech acts. These examples initially suggest that (syntactic) ellipsis is not going on. Otherwise, these items would carry the case that they do in the source sentence. Instead, the words and phrases are appearing unembedded. Thus, again, anticonnectivity effects pose a problem. As before, however, Merchant’s view elegantly explains why case markings differ in certain fragments. It’s because fronted elements exhibit different case, and on his view the things “left over” have been fronted. The foregoing advantages derive from the movement aspect of Merchant’s view. Consider now what the licensing conditions on [E] account for. One of the most important objections facing syntactic ellipsis accounts is that bare phrases can occur as speech acts without a linguistic antecedent. Yet, from the earliest treatments of ellipsis in generative grammar, theorists have insisted on “the recoverability of deletion.” That is, whatever the speaker omits via ellipsis must be recoverable by the hearer. This, in turn, seemed to require that the right kind of linguistic material be present – that, it seemed, was precisely what permitted the recovery of the elided structure. Some even urged that identical linguistic material must be present if ellipsis is to be licit. For, short of that, it would not be possible to recover precisely the material deleted. (See Sag 1976 and Barton 1990 for discussion.) All of this suggests very strongly indeed that, whatever is going on when people start a discourse with a word or phrase, it cannot be syntactic ellipsis. Merchant improves the situation for the syntactic ellipsis theorist on two fronts. First, he suggests, on independent grounds, that there need not be the exact same syntactic material in prior discourse. Instead, similar content can do the trick. Second, as explained, he introduces two kinds of ellipsis constructions that don’t need even that kind of licensor: [VP do it] and [IP this/that [I’ is t]].
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.11 (593-633)
Neither fragments nor ellipsis
Consider now two additional advantages. First, as with syntactic ellipsis stories generally, because the “fragments” derive from full sentences, the view easily handles certain familiar “connectivity effects”: that the case markings on subsentences frequently mirror those on sentence constituents; that we find binding, and binding violations, of kinds that also appear in sentences; and so on. These parallels are easily explained if appearance seemingly in isolation is really just a subcase of appearance in a sentence. (See Morgan 1973, 1989 and Ludlow 2005 for related points.) Second, Merchant notes a series of otherwise very curious syntactic island effects that are automatically explained if fronting-then-deletion is indeed taking place. Finally, Merchant contends that his account is more conservative and simpler, on the grounds that it doesn’t necessitate “a revision of the systems of form-meaning mappings” (2004: 663), nor does it require enriching “the pragmatic interpretive component” (2004: 717). Since on his view hearers actually recover complete sentences when they (appear to) encounter mere words and phrases, the usual semantic rules that assign propositions to sentences apply here as well. Thus no extra pragmatic processing is required. And, insofar as the ellipsis that goes on is independently attested, there are no additional changes to the grammar either. I end this section with what may be the most important, yet seldom noticed, advantage of Merchant’s account. Any syntactic ellipsis account of the appearances must meet what I call the “Not-Just-Recasting” constraint. This is the requirement that what the theorist means by “syntactic ellipsis” is incompatible with a genuine word or phrase being used to perform a speech act. That is, the constraint is that the appeal to “ellipsis” not be just a soothing deployment of this word, while essentially restating the appearances with which I began. Merchant’s story, unlike some others, really does meet this vital constraint: if he is right about linguistically discourse-initial cases, then nonsentential accounts are plain wrong.
. Empirical objections Those who wish to resist the appearances with which I began, by appeal to syntactic ellipsis, have faced an uphill battle. Merchant makes important strides forward. The familiar worries about syntactic ellipsis without a linguistic antecedent can be met, says Merchant, if we simultaneously: (a) lessen the restrictions on what counts as an acceptable antecedent, to something about content rather than about form (a lessening that he argues is independently motivated) and (b) posit a “limited ellipsis” account for linguistically discourse-initial cases. He argues further that some data that are troublesome on familiar syntactic ellipsis accounts – licensing of negative polarity items, deviant case assignment, apparent omission of nonconstituents, and so on – can be dealt with if we recognize that deletion sometimes occurs only after movement to focus position has taken place. Despite these tactical successes, I think the battle to treat all (speech act realizing) “fragments” as elliptical is ultimately lost. (More specifically, I remain especially unconvinced by his treatment of linguistically
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.12 (633-703)
Robert J. Stainton
discourse-initial cases – though I remain open to the idea that he is right about direct answers to questions, for instance.) In fact, I think there are empirical problems with each of the key features of Merchant’s account: the [E] feature, movement when one seemingly has nonconstituent deletion, and the “limited ellipsis” gambit. (I will take them in turn.) In addition, the alleged advantage of simplicity is illusory.
. An empirical problem about [E] The central interest of Merchant’s account for my purposes is his attempt to extend his analysis to utterances in linguistically discourse-initial position. That’s how he attempts to explain away the appearances with which I began. Nevertheless, my discussion will extend beyond those difficulties with his proposal that arise specifically for the discourse-initial subcases. In particular, the next few objections apply to Merchant’s overarching account of ellipsis, wherever it is applied. Beginning with [E], the fundamental problem that I want to raise has to do with its distribution. On Merchant’s view, as on ellipsis accounts generally, the same sound can correspond here to a word or lexically headed phrase, there to a sentence. There is the phrase [NP Several men from Segovia], but there is also the same-sounding sentence [IP [NP Several men from Segovia][E][I’ did it]]. Of course this rampant ambiguity of sound patterns might itself strike one as an unhappy feature of the view. But let’s put that aside. What I want to focus on is a prediction that follows from the postulation of elliptical sentences that sound just like words/phrases: they ought to embed where sentences, elliptical or otherwise, can. That’s because, on this proposal, the posited items just are (elliptical) sentences. In this regard, notice the ill-formedness of (13) and (14) as compared to the perfectly grammatical (15) and (16): (13) *If there is graffiti on the wall, then several men from Segovia (14) *If several men from Segovia, then the job will be poorly done (15) If there is graffiti on the wall, then several men from Segovia did it (16) If several men from Segovia did it, then the job will be poorly done Merchant and other ellipsis theorists can explain why (13) and (14) are bad on one reading, of course, namely the one where the embedded element is an NP. But they cannot explain why it isn’t good on another, viz. the (supposed) reading on which an elliptical sentence appears. Another example may help make the point. On Merchant’s view, there is an elliptical sentence of the form [IP [John][E][VP did it]]. It is pronounced just like the ordinary name John, but it is not syntactically a name at all. Given this, here once again is the problem. The counterpart of this sentence without [E], namely [IP [John][VP did it]], embeds just fine in the context (17). Witness (18). (17) I couldn’t eat the worm, but [IP __] (18) I couldn’t eat the worm, but John did it
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.13 (703-770)
Neither fragments nor ellipsis
In contrast, when we substitute the alleged sentence [IP [John][E][VP did it]] in (17), the result is terrible: (19) *I couldn’t eat the worm, but John The obvious explanation for this is that there really is no “elliptical sentence” [IP [John][E][VP did it]] with feature [E] in it. To be clear, the point is not that apparent words and phrases can never embed in what look like sentential contexts. To the contrary, there are some cases where just this sort of thing seems to happen. Thus if a garden expert is asked how to eradicate grubs, she could say: if they cover less than 30% of the lawn, a mild pesticide; if they cover more than that, DDT. The point, rather, is that what appear to be words and phrases frequently cannot embed even when, on Merchant’s syntactic ellipsis account, there is a perfectly appropriate same-sounding sentence that should embed. Another note, to clarify: it matters that I have picked cases of [VP do it] omission to pose the objection, since in the move-then-sluice cases, Merchant’s view will predict the right result at least in English. English sentences that have undergone movement to focus position do not embed easily, so we don’t expect the (alleged) elided versions to do so either. For example, If from France that is, then it’s dangerous is very awkward at best; this alone would account for the unacceptability of If from France, then it’s dangerous in English. But when movement doesn’t happen before deletion, this explanation is not available.3 Speaking of movement, I turn now to another objection.
. An empirical problem about movement It is crucial for Merchant’s view that movement to the left periphery frequently occur in “fragments” – whether in an answer to an explicit question, or in linguistically discourse-initial position. At least in many cases, what moves is what we hear, for Merchant. The problem is that in numerous cases the content asserted can’t be accounted for by fronting-then-deletion, because what we hear could not have been so moved. Consider first an example from English. A teacher has been reading a story to her pupils about how the Pope has a fondness for beer mixed with tomato juice. To test her students’ comprehension skills, she asks: The Pope likes beer and what? The correct answer could be given as Tomato juice. Using just this bare phrase is a perfectly fine way to assert the proposition THE POPE LIKES BEER AND TOMATO JUICE. But consider what the sentential source would have to be for this answer, on Merchant’s account: (20) [FP [NP Tomato juice]1 [F’ [E] [IP [NP the Pope][I’ 3SG.PRES [VP like [NP beer and t1 ]]]]]] The corresponding nonelliptical version, Tomato juice the Pope likes beer and, is patently ungrammatical. So surely (20), which merely adds [E], is ungrammatical too. But then why is it possible to use this phrase to assert this proposition? Or again, suppose according to the story being read that the Pope sleeps on a rock-hard bed. The
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.14 (770-821)
Robert J. Stainton
question The Pope sleeps on a hard what in the story? can receive the answer Bed. A proposition is conveyed here without difficulty. But Bed, the Pope sleeps on a hard in the story, from which Merchant would derive the fragmentary answer, is nearly word salad. The examples can easily be multiplied. One might complain about such examples that they aren’t genuinely answers to interrogatives. What’s used, instead, are declaratives with focal stress – echo questions. That’s true enough, but it’s entirely beside the point. The issue, after all, is not which background conditions allow one to assert THE POPE LIKES BEER AND TOMATO JUICE with Tomato juice – interrogatives, echo questions, or something else again – but the mere fact that one can state this proposition at all, with Tomato juice. To see this, note that it would be just as problematic for Merchant if, as seems perfectly possible, nonlinguistic context made clear that the teacher wanted to know what the Pope likes with beer, and the mere phrase was apparently used in reply. Merchant could not account for that either. A similar kind of point arises in languages that don’t allow fronting of certain kinds of expressions, no matter where they are to be extracted from. For example, Malagasy permits lexical and phrasal answers to questions: In reply to the Malagasy version of Who does Rabe respect?, the answer can be the bare name Rasoa. More than that, such answers carry the case marking that one expects. In the example just given, for instance, the name Rasoa must carry accusative case. In contrast, the question Who is respected by Rabe? can also receive the one-word answer Rasoa; but in that circumstance, the name must be marked nominative. These two points are illustrated by (21) and (22): (21) Lexical/Phrasal Answers in Malagasy, Accusative Required: Q: Manaja an’iza Rabe? respect acc-who Rabe ‘Who does Rabe respect?’ A: An-dRasoa acc-Rasoa ‘Rasoa’ A’: *Rasoa Rasoa(-nom) ‘Rasoa’ (22) Lexical/Phrasal Answers in Malagasy, Nominative Required: Q: Iza no hajain-dRabe? who foc respect(-passive)-Rabe ‘Who is respected by Rabe?’ A: Rasoa Rasoa(-nom) ‘Rasoa’
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.15 (821-892)
Neither fragments nor ellipsis
A’: *An-dRasoa acc-Rasoa ‘Rasoa’ So, in Malagasy words and phrases can occur unembedded, with the speaker thereby performing a speech act. And when they so occur, they even carry the requisite case markings. Merchant, of course, will want to claim that these are cases of ellipsis. More specifically, to avoid nonconstituent deletion, he will want to claim that there is ellipsis after movement. Crucially, however, as Keenan (1976) first observed, Malagasy does not allow nonsubjects to front. While (23a) is fine, (23b) is ungrammatical: (23) a.
Subject fronting: Rabe no manaja an-dRasoa Rabe foc respect acc-Rasoa b. Nonsubject fronting: *An-dRasoa no manaja Rabe acc-Rasoa foc respect Rabe ‘Rasoa, Rabe respects’
(Note that example (23b) involves focus fronting, the kind of particular interest here. This is merely a special case of the broader generalization that Malagasy does not allow fronting of any nonsubject.) Applying this point to our earlier example, even though the bare name An-dRasoa is well formed, and can be used in this context to say that Rabe respects Rasoa, Merchant’s view cannot capture this fact – because (21A) cannot be derived, as he would wish, from (23b). The sentential source that Merchant’s view requires simply is not generated in Malagasy.4 Nor is Malagasy a unique language in this regard. Merchant (personal communication) has suggested that both the English and the Malagasy cases might be the result of repair effects. The latter is an interesting phenomenon in which the “elliptical version” of a sentence strikes us as more grammatical than the “full” version. For instance, contrast the ill-formed (24) with the more grammatical elliptical version of it, in (25): (24) *They want to hire someone who speaks a Balkan language, but I don’t remember which they want to hire someone who speaks. (25) They want to hire someone who speaks a Balkan language, but I don’t remember which. (The examples are from Merchant 2001: 4–5.) My response to this is that repair effects may well appear in highly constrained circumstances; however, one cannot appeal to them whenever an ellipsis account seems to make the wrong predictions – at the risk of making “repair effects” a get-out-of-counterexample-free card. What is needed to really mount a rebuttal to these sorts of cases, then, is a positive reason for thinking that these fall under the constrained set of such “escape hatches” – not just the mere existence of such a class. So far, no such reasons have been provided.
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.16 (892-951)
Robert J. Stainton
What should we conclude from the foregoing? That syntactic ellipsis never happens between sentences? I myself consider that too strong a conclusion, since there remain many cases where the nonsentential analysis seemingly predicts the wrong result: binding violations, island effects, and so on. (I thus disagree with Casielles, Chap. 5, this volume.) Neither can we merely conclude, however, that there are a few extra wrinkles to be accounted for. That’s because, in each case where movement-plus-[E] cannot be the mechanism at work, many (if not all) of the earlier problems facing an ellipsis treatment are reinstated: nonconstituent deletion, anticonnectivity with respect to case, and negative polarity.
. Two empirical problems about limited ellipsis The next empirical objection is specific to subsentences in linguistically discourseinitial position, rather than to Merchant’s move-then-delete approach as a whole. It has to do with generality. Merchant provides a special exception for linguistically discourse-initial cases in which the elements elided are [VP do it] or [IP this/that [I’ is t]]: [E] can function without a linguistic antecedent when its complement is either of those, as long as it’s clear from context what it and this/that refer to, and as long as be is manifest. But there are a whole host of linguistically discourse-initial examples that do not involve the omission of [VP do it] or [IP [NP this/that] [I’ is t]]. Recall some of the attested examples with which I began. To Segovia, said to the cab driver as his passenger enters the car, did not mean TO SEGOVIA DO IT. Assuming this order even makes sense, that gets the content wrong. Nor, of course, did To Segovia mean THIS IS TO SEGOVIA.5 That proposition, unlike the former one, seems a coherent enough content – one might point at a bus, and assert with This is to Segovia that the bus’ destination is Segovia – but, patently, that was not the content of the passenger’s speech act. What the passenger conveyed was something like TAKE ME TO SEGOVIA. Hence neither of Merchant’s exceptions applies to this example. Here is another. Two black coffees, if used discourse initially (in the relevant sense) to order something, cannot be assimilated to either of [VP do it] or [IP [NP this/that] [I’ is t]]. Similar points hold for Just him and Both hands: Merchant’s two exceptions do not account for these either. Finally, if a driver yells out Jerk! to someone who has cut him off, he does not mean, pace Merchant, either Do it jerk or Jerk this is. A possible response would be to increase the inventory of exceptions: it’s not just [VP do it] and [IP [NP this/that] [I’ is t]] that can be elided without a linguistic antecedent. For Both hands, we could say that that the structure employed was something like (26). For Two black coffees, the rough structure would be (27). And Jerk would be something like (28): (26) [FP [NP Both hands]1 [F’ [E][IP pro [I’ 2SG.PRES [VP use t1 ]]]]] (27) [FP [NP Two black coffees]1 [F’ [E][IP pro [I’ 2SG.PRES [VP [VP give2 me][VP [V t2 ] [NP t1 ]]]]]]]
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.17 (951-1026)
Neither fragments nor ellipsis
(28) [FP [NP Jerk]1 [F’ [E][IP you [I’ 2SG.PRES [VP be t1 ]]]]] Spelling this out informally, the crude idea is that speakers produce, in essence, the sentences Use both hands, Give me two black coffees, and You are a jerk – but without pronouncing certain words. They can do this discourse initially (i.e., in the sense of “without an appropriate linguistic antecedent”) because ellipsis of use, give me, and you are are also licensed exceptionally by nonlinguistic context: like do it and this is, these contents are “thin enough” to be recoverable from a nonlinguistic source. Less crudely, to ensure that what is elided is a constituent, Merchant will insist that what’s strictly speaking employed are the fronted versions, Both hands [E] use, Two black coffees [E] give me, and Jerk [E] you are. We thus have three further exceptions to the general rule about how [E]’s semantics works. But this doesn’t really solve the problem. To begin with there are some minor worries. First, the nonelliptical versions of these fronted sentences are grammatically peculiar: Both hands use, Two black coffees give me, and Jerk you are are Yoda-speak, not idiomatic English, whereas Both hands, Two black coffees, and Jerk! are perfectly fine. Second, specifically with respect to (28), there’s the issue of why we get Jerk! rather than A jerk!, since it’s the latter that must appear in object position of the corresponding sentence. (You are jerk is ill formed, as is Jerk, you are.) The larger problem is how ad hoc this quickly becomes. For each new kind of case, we need to introduce another exception. To drive the point home, consider some further attested examples. (I leave it to the reader to find the requisite “sources”): (29) [Bob is looking a bit lost, trying to find the office that issues keys. He encounters Walt, who clearly works in the building:] Walt: Can I help you? Bob: Keys Walt: Up this way. Just around the corner to your left. (30) [Beto notices that Debbie is missing. He crosses the room to her husband Gord:] Beto: Have we lost Debbie? Gord: Washroom (31) [(From The Hudsucker Proxy.) Barnes gets off an elevator, looking for Mr. Musburger’s office. He addresses Al, who is working in the hallway:] Barnes: Mr. Musburger’s office? Al: [Points] Al: Not that way. Through the door Notice too that the more exceptions we add, the more work the language faculty has to do on the basis of extralinguistic context. To see the point, recall the worry about Merchant’s “limited ellipsis” strategy that I set aside earlier: syntactic ellipsis is a language-internal process, went the objection, hence the language faculty can’t use nonlinguistic context to sort out what’s missing in linguistically discourse-initial cases. I responded on Merchant’s behalf that all the language faculty has to do is choose be-
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.18 (1026-1092)
Robert J. Stainton
tween [VP do it] versus [IP [NP this/that] [I’ is t]]. But the initial objection returns with real force once the list of exceptions expands. Things start to look progressively less like a grammatical derivation, and progressively more like the agent as a whole assessing, on an all-things-considered basis, whether she should fill in do it, this is, use, give me, you are, or something else again. (Nor, pace Merchant (2004), will appeal to “scripts” help here. For it’s surely the person, and not her subpersonal syntactic competence, that is able to decide which scripts are actually in play.) Another strategy for dealing with Both hands and the other examples, which would avoid the ad hoc charge, would be to introduce a general rule for when [E] can function without a linguistic antecedent. The rule might go like this. Whenever the context, linguistic or nonlinguistic, supplies enough content, [E] can function in a nonstandard way. In particular, if the proposition input into the function denoted by [E] is appropriately related to one of the contextually salient ones (rather than being related to the proposition encoded by the linguistic antecedent, as in the nonexceptional cases) then [E] outputs the very proposition input. Such a general rule is not ad hoc, at least not in the sense of positing new rules for each new case. But Merchant now faces a dilemma. If it’s the whole agent that is applying this rule, rather than narrow syntax/semantics, then we have arrived at a convoluted anti-ellipsis view. That is, if [E] gets reconceived not as a rule within the language faculty, but rather as a pragmatic constraint on what we interpreters may take speakers to be asserting – viz., what they are asserting must be among the propositions that are salient – then the “Not-Just-Recasting” constraint isn’t met after all. The appearances are accepted, though they are redescribed using obfuscating language. That horn of the dilemma is clearly unacceptable. Suppose, then, that it is the language faculty that does the job of checking whether the content input is among the salient propositions. Yet it cannot have access to all the propositions that are salient in the context: only the agent as a whole has access to that. Nor would the language faculty even have access to a significant subset of them, because that’s just not something that narrow syntax/semantics works on. Suppose then that the language faculty deploys just one, or maybe two or three, propositions that are salient – just as happens in the ordinary case, except that here the proposition comes from nonlinguistic context. But if that’s right, the operation of reconstructing a sentence in order to interpret the speaker becomes effectively otiose. In order for the entry condition for applying [E] to be met, something else must have pretty much preselected the asserted proposition from among all those that are salient. But then there’s (almost) no point in constructing an elliptical sentence and then interpreting it, to arrive back at a proposition that has antecedently been identified. In summary, either Merchant’s view, though baroquely reworded in terms of [E], collapses into an anti-ellipsis account, or it’s an ellipsis account that includes what are essentially otiose formal rules. Having just stressed the wide variety of contents that can be asserted subsententially, and without a linguistic antecedent, let me introduce another problem facing Merchant. As I’ve noted in the past (Stainton 1998), bare words and phrases do not license elliptical constructions in subsequent discourse as easily as sentences do. For instance, if The man from Paris is said when a knock is heard at the door, this cannot
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.19 (1092-1153)
Neither fragments nor ellipsis
be followed by the elliptical sentence And Betty is too. (It can be followed by other fragmentary replies, e.g., Betty too. But those don’t test in the right way for sentencehood.) In contrast, the complete sentence The man from Paris is at the door can be followed by this VP ellipsis construction. I inferred, on the basis of this contrast in licensing properties, that what appeared to be a bare phrase really was one – otherwise it should license subsequent VP ellipsis. Now, Merchant (2004: 727–729) resists this and related examples by suggesting, in light of his claim that only [VP do it] and [IP [NP this/that] [I’ is t]] can be used without a linguistic antecedent, that what The man from Paris must be elliptical for is [IP [The man from Paris]1 [IP that [I’ is t1 ]]]. If that were right, my data would be antecedently accounted for within an ellipsis account, since this latter sentence cannot license And Betty is too. This strategy for resisting my point about licensing fails, however, if Merchant is wrong that only [VP do it] and [IP this/that [I’ is t]] can be used without a linguistic antecedent – and I’ve just argued that he is wrong. To see the point, consider the example in more detail. Suppose, with Merchant, that an elliptical sentence is produced. Suppose, with me and contra Merchant, that what one can assert with nonsentences in linguistically discourse-initial position goes well beyond what [VP do it] and [IP [NP this/that] [I’ is t]] can account for. I’ll try to show that on these suppositions, ellipsis theorists still don’t have an explanation for why And Betty is too cannot appear after The man from Paris. Let’s take this in two steps. First, suppose that The man from Paris can be used to assert a richly descriptive proposition, for example, THE MAN FROM PARIS IS AT THE DOOR. To allow precisely this content to be asserted, the elliptical sentence used, employing [E], would need to be (32): (32) [IP [The man from Paris] [E][I’ is at the door]] However, if that sentence had been used, then the elliptical sentence And Betty is [[E] [at the door]] too should be licensed on Merchant’s view. That’s because the content AT THE DOOR is appropriately related to a propositional function expressed by a part of the linguistic antecedent, viz. the elliptical sentence (32). Thus [E] is predicted to output just this propositional function. So, on this first case, the mystery of why And Betty is too isn’t licensed remains. Second step. In response, one might insist – implausibly in my view – that The man from Paris cannot be used to assert such descriptively rich propositions. Goes the idea, in the spirit of Merchant’s “limited ellipsis” strategy, the descriptive material in (32) should be replaced by a deictic. Even supposing this, however, the problem remains. If it’s not just [VP do it] and [IP this/that [I’ is t]] that can appear without a linguistic trigger, then the thing used could be (33): (33) [IP The man from Paris [E][I’ is there]] But taking (33) to be the sentence used equally leaves it a mystery why And Betty is too is not licensed. After all, if there really were such a nondescriptive sentence in prior discourse, even it should license And Betty is too. Once again, this is because [E] is predicted to output the propositional function IS THERE, on the grounds that this content is encoded by part of the supposed linguistic antecedent.
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.20 (1153-1211)
Robert J. Stainton
The dialectic, in summary, is this: I raised an objection to ellipsis treatments on the basis of licensing facts; Merchant tried to explain away those facts by insisting that there wouldn’t be the right kind of sentential licensor on the correct ellipsis view, because only [VP do it] and [IP this/that [I’ is t]] can be used without a linguistic antecedent; I’ve responded that once this limitation is given up (which it needs to be, to make the approach more generally applicable), the original objection stands. Indeed, it stands even if one insists that the omitted material consist only in “thin” material such as light verbs and deictics. Here is the second objection to the “limited ellipsis” strategy. Merchant addresses the problem of subsentences appearing in linguistically discourse-initial situation by noting that VP ellipsis and sluicing can so occur as well. Such sentences can be and are used with only a “pragmatic controller.” This is just what Hankamer and Sag (1976) and Sag and Hankamer (1977) pointed out, and a host of other authors have followed suit (e.g., Schachter 1977, 1978; Stanley 2000). Thus, Merchant concludes, it’s not true that genuine syntactic ellipsis requires a linguistic antecedent; so, in turn, one shouldn’t argue from subsentences occurring on their own to their not being elliptical. Merchant is right about the failings of that specific argument. It is too crude. But there is a more subtle argument that he fails to address. Merchant seems, in the end, to grant that using He won’t or I don’t know why in linguistically discourse-initial situations is rather awkward. This contrasts with using these as replies to direct questions (hence not in discourse-initial position); and it contrasts with using nonelliptical expressions, for example, full sentences, in discourse-initial position. Neither of those usages is the least bit awkward. It’s true, Merchant seems to allow, that there is a “felt difference” between VP ellipsis and sluicing in linguistically discourse-initial position on the one hand, and discourse initial full sentences or elliptical sentences with linguistic antecedents on the other. And whether or not Merchant himself concedes the point, it’s surely true that there is a difference between the former usage and the latter two kinds of case. Taking this as background, compare how apparent words and phrases pattern: putting aside prescriptive sensibilities, subsentences can occur without a linguistic antecedent as freely as nonelliptical sentences can, and without awkwardness. The question that arises, then, is this: why is there this notable difference between the comparative awkwardness of VP ellipsis and sluicing in linguistically discourse-initial position, and the comparative naturalness of apparent words and phrases in the same position? Put in terms of examples, Merchant leaves us wondering why there should be any difference at all between a linguistically discourse-initial use of From Spain about a salient letter or Sam’s mother about a salient woman, versus saying I don’t know why out of the blue, about someone climbing a hill. For, on his view, both are straightforward cases of ellipsis. This is the more subtle argument from facts about usage in discourse-initial position that Merchant fails to address.
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.21 (1211-1264)
Neither fragments nor ellipsis
. A methodological objection: “Fragments” and simplicity So far, I’ve highlighted five empirical problems with the move-then-delete view: the inability of alleged “elliptical sentences” to embed in certain sentential contexts; nonsentence uses even when focus fronting isn’t possible; the insufficient generality of allowing only [VP do it] and [IP [NP this/that] [I’ is t]] to be licensed directly by the larger nonlinguistic context; the reappearance of an older objection, about licensing, in light of this; and a difference, not captured by Merchant’s account, between the awkwardness of linguistically discourse-initial sluicing and VP ellipsis, versus the naturalness of linguistically discourse-initial subsentence use. I end my discussion with issues about simplicity. Merchant suggests that extra machinery is required if we have subsentences being interpreted “directly” (as accepting the appearances requires), rather than via the reconstruction of an elliptical sentence. He thus claims that his ellipsis-based view is simpler and more conservative. This is just not the case. We know that subsentences can be used and understood nonpropositionally: they show up in grocery lists, book titles, and dictionary entries; on street signs, maps, currency, and so on. So, the competence required for assigning a subpropositional semantic content to such items is independently attested. (Merchant grants this, but suggests putting these aside, as another topic (2004: 731–732). Say I, that’s not appropriate: such uses are directly relevant to the present debate precisely because of parsimony issues.) We also know that hearers have the pragmatic ability to go from a semantic content that patently isn’t meant by the speaker, to the proposition that the speaker did mean. This happens in conversational implicatures, indirect speech acts, irony, and the like. And many would argue that it also occurs with various kinds of pragmatic determinants of what is said: precisifying vague words, sense expansion, and so on. (See Carston 2002 for an overview.) So, all of the necessary syntactic, semantic, and pragmatic competences required by those who take the appearances at face value are independently attested. In particular, given word and phrase meanings, the “pragmatic interpretive component” is antecedently rich enough to do what’s required to arrive at a proposition asserted. Thus it’s just not true that extra devices need to be introduced on a pragmatics-oriented approach, specifically to handle subsentence use in linguistically discourse-initial position. It can seem otherwise if one conceives the phenomenon in the wrong terms. This takes me to the title of my chapter. I’ve long maintained that what occurs in the kind of examples at play is not ellipsis – at least not in any sense that would help avoid the appearances. (See Stainton 2005, 2006 for extended discussion.) I would now add that we shouldn’t even think of the cases at issue as fragments. There are two reasons for this. To begin with, selecting the term fragment threatens to beg the question. My trusty Canadian Oxford Dictionary defines fragment as: “1. a part broken off; a detached piece. 2. an isolated or incomplete part. 3. the remains of an otherwise lost or destroyed whole... 4. a scrap; a left over piece.” Now, for those who accept the appearances, what are used are words and lexically headed phrases. But words and lexical phrases aren’t detached from a whole, incomplete, remains or leftovers. For instance, the word jerk is
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.22 (1264-1315)
Robert J. Stainton
patently not a “remnant,” a “leftover piece”; the phrase To Segovia isn’t either. The only difference between words and phrases on the one hand, and sentences on the other, is that the latter are headed by things like Tense and Agreement. So, since words/phrases are complete, but “fragments” are not, to think of the things used as “fragments” is ipso facto not to think of them as words/phrases. Thus is the question implicitly begged, by the choice of a label. The second problem with the term fragment is that it suggests that there is one thing, the “fragment,” that is both propositional and (apparently) subsentential. It then seems that those of us who take subsentential speech to be genuine need to introduce something new into the grammar, viz., formatives that are genuine subsentences but that nevertheless express propositions. As Merchant points out, this would “require a revision of the systems of form-meaning mappings” (2004: 663). Looked at that way, it does indeed seem more theoretically conservative to maintain, of the one thing, the “fragment,” that it isn’t actually subsentential after all. The problem with this line of reasoning is precisely that it misconceives the alternative way of seeing things. Rather than one thing with two seemingly conflicting properties, those who take subsentence use to be genuine maintain that there are two things. On the one hand, there are the subsentences, which are nothing more than our familiar words and phrases, with the standing meaning and syntax thereof. On the other hand, there are fully propositional speech acts that we perform with those subsentences. The former aren’t sentential, but they aren’t propositional either – so no new mapping is required. The latter are propositional, yes, but they become so in the usual way: by the interaction of the language faculty with lots of other things. No new grammatical mapping is required – any more than being able to conversationally implicate something about a man’s honesty with Grice’s (34) requires a special form-meaning mapping for this sentence type: (34) He likes his colleagues, and he hasn’t been to prison yet In summary, neither the items deployed nor the actions performed merits the label fragment. And neither of them demands new machinery. More than that, it’s really Merchant’s view that introduces new devices: sentences that contain [E], some of which have undergone movement, and that nevertheless do not need linguistic antecedents.
. Concluding remarks I’ve just argued that Merchant’s view is not superior as far as simplicity goes. It’s also deficient empirically, both with respect to cases with a linguistic antecedent (Tomato juice in response to The Pope likes beer and what?) and with respect to linguistically discourse-initial cases (including the three with which I began). I thus conclude that, though a brave and immensely admirable attempt to overcome the uphill battle, Merchant’s proposal to reject the appearances of nonsentential speech acts in
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.23 (1315-1379)
Neither fragments nor ellipsis
linguistically discourse-initial position, by appeal to movement and syntactic ellipsis, ultimately fails. Ellipsis alone still cannot do the trick. This conclusion settles the issue of whether the appearances with which I began reflect the reality. They do. Genuinely nonsentential speech acts do occur. It does not settle, however, the larger question of whether syntactic ellipsis ever happens between sentences. (As noted, I myself suspect syntactic ellipsis of that kind does exist.) Still less does it answer what should now become a guiding question in this domain, viz., which account best applies to which subvarieties? This is an issue of great interest. It’s also terrifically challenging, both empirically and conceptually. To my mind, it can only be addressed on a case-by-case basis, deploying the evidence and the arguments that continue to come in.
Notes * My first thanks go to Ellen Barton, whose work has so informed my own. I am also grateful to Dan Blair, Catherine Wearing, and especially Ileana Paul for discussion and for comments on a previous draft. Finally, thanks to both anonymous reviewers and to the editors for suggestions that substantially improved the chapter. . This example was actually uttered in Spanish: A Segovia. A la cárcel. But that complication can be set aside, at least for now. . Interestingly, the caseless forms require no linguistic context, though they are also permitted to occur within a discourse context; in contrast, the case-marked forms require overt linguistic context, for example, a question asked. To give an example from Morgan (1989), nae cha! (‘my car-no.case’) can be used by a person on returning to a parking lot, and finding her car stolen. But both nae cha-ka (‘my car-nom’) and nae cha-rul (‘my car-acc’) are ill formed in that discourse-initial circumstance. . There are languages that are more forgiving about embedding fronted sentences, but that still don’t allow embedding of what seem to be bare PPs in sentential positions. Spanish is an example. Thus Si de María viene, no lo abras (‘If it’s from María, don’t open it’) is quite good, but Si de María, no lo abras (‘If from María, don’t open it’) is quite bad. These provide additional problems for Merchant’s [E]. . I am very grateful to Ileana Paul both for the specific Malagasy examples, and for extended discussion of subsentence use in languages with fronting restrictions. By the way, an anonymous referee raised the question of whether these “fragments” might be derived in another way, namely, from Malagasy pseudo-clefts. The answer is (a) that the construction presented here just is the closest Malagasy has to pseudo-clefts and (b) that appeal to clefts would seemingly reintroduce deletion of nonconstituents in any case. . It’s also worth noting that the [VP do it] subcase cannot be applied to languages that lack VP ellipsis. For instance, the attested example To Segovia was actually uttered in Spanish. Yet Spanish does not have VP ellipsis – so what’s going on in this example couldn’t be ellipsis of [VP do it].
JB[v.20020404] Prn:25/07/2006; 12:31
F: LA9304.tex / p.24 (1379-1461)
Robert J. Stainton
References Barton, E. (1990). Nonsentential Constituents. Amsterdam: John Benjamins. Carston, R. (2002). Thoughts and Utterances. Oxford: Blackwell. Hankamer, J. & Sag, I. (1976). Deep and surface anaphora. Linguistic Inquiry, 7, 391–428. Keenan, E. (1976). Remarkable subjects in Malagasy. In C. Li (Ed.), Subject and Topic (pp. 247– 301). New York, NY: Academic Press. King, J. & Stanley, J. (2005). Semantics, pragmatics and the role of semantic content. In Z. Szabo (Ed.), Semantics versus Pragmatics (pp. 111–164). Oxford: OUP. Ludlow, P. (2005). A note on alleged cases of nonsentential assertion. In R. Elugardo & R. Stainton (Eds.), Ellipsis and Nonsentential Speech (pp. 95–108). Dordrecht: Springer. Merchant, J. (2001). The Syntax of Silence. Oxford: OUP. Merchant, J. (2004). Fragments and ellipsis. Linguistics and Philosophy, 27, 661–738. Morgan, J. (1973). Sentence fragments and the notion ‘sentence.’ In B. J. Kachru, R. B. Lees, Y. Malkiel, A. Pietrangeli, & S. Saporta (Eds.), Issues in Linguistics (pp. 719–751). Urbana, IL: University of Illinois Press. Morgan, J. (1989). Sentence fragments revisited. Chicago Linguistics Society: Papers From the Parasession on Language in Context, 25, 228–241. Rizzi, L. (1997). The fine structure of the left periphery. In L. Haegeman (Ed.), Elements of Grammar (pp. 281–337). Dordrecht: Kluwer. Sag, I. (1976). Deletion and Logical Form. Ph.D. dissertation, Massachusetts Institute of Technology. Sag, I. & Hankamer, J. (1977). Syntactically vs. pragmatically controlled anaphora. In R. Fasold & R. Shuy (Eds.), Studies in Language Variation (pp. 121–135). Washington, DC: Georgetown University Press. Schachter, P. (1977). Does she or doesn’t she? Linguistic Inquiry, 8, 763–767. Schachter, P. (1978). English propredicates. Linguistic Analysis, 4, 187–224. Stainton, R. (1998). Quantifier phrases, meaningfulness ‘in isolation,’ and ellipsis. Linguistics and Philosophy, 21, 311–340. Stainton, R. (2005). In defense of non-sentential assertion. In Z. Szabo (Ed.), Semantics Versus Pragmatics (pp. 383–457). Oxford: OUP. Stainton, R. (2006). Words and Thoughts. Oxford: OUP. Stanley, J. (2000). Context and Logical Form. Linguistics and Philosophy, 23, 391–434.
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.1 (47-123)
chapter
Big questions, small answers* Eugenia Casielles This paper considers elliptical and nonelliptical accounts of short answers to questions and suggests pursuing a nonelliptical approach. Despite some connectivity effects that hint at the presence of extra-sentential structure, I show that short answers also exhibit robust anticonnectivity effects such as case-mismatched NPs or tenseless VPs, which suggest a nonsentential, phrasal analysis. I argue (contra Merchant 2004) that a nonelliptical view does not run into syntax-semantics mapping problems and that it can also account for the focal nature of short answers without costly operations such as movement and deletion. Furthermore, this approach can explain why children can produce short answers early, which would be unexpected if short answers involved complicated operations akin to sluicing or clefting.
.
Introduction
Short or small answers to questions in exchanges such as that in (1) differ in a number of ways from other nonsentential data. (1) A: Which movie did you see? B: Casablanca. As opposed to telegrams in (2), recipes in (3), diary entries in (4), or newspaper headlines in (5), short answers to questions such as that in (1) do not seem to have a special register flavor. (See Paesani, Chap. 6, this volume, for the features of nonsentential data in special registers.) (2) Car problem. (3) . . .place in bowl with mushrooms. . .
(Barton & Progovac 2005) (Haegeman 1988: 236)
(4) Felt a joy yesterday. Soon clouded. (Sylvia Plath, 28.I.I959, p. 293, quoted in Haegeman 2000: 130) (5) Sharon to End Some Settlements
(New York Times, June 4, 2003)
However, as Morgan (1989) points out, speakers tend to consider cases such as (6) ungrammatical unless preceded by an appropriate question, as in (7):
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.2 (123-193)
Eugenia Casielles
(6) I used to think lobster. (7) A: What is the most delicious seafood dish? B: I used to think lobster.
(Morgan 1989: 239)
Taking this kind of example into account, we could ask if question-answer pairs (QAPs) such as those in (1) and (7) might also be considered a special register, namely, conversational English. Yet, QAPs seem to have different features from those present in what is referred to as informal spoken English, exemplified in (8): (8) a. b. c. d.
Told you so. Don’t know what to say. Serves you right. Looks like rain.
(Haegeman 2000: 132)
As (8) shows, informal spoken English is partly characterized by the use of null subjects in an otherwise non-null-subject language and is, in this sense, similar to diary entries, such as (4). Small answers to questions, however, do not omit only recoverable subjects, they omit all noninformative elements (including verbs). In the case of (1), we omit all recoverable information which in this case includes the subject and the verb (I saw) and we state only the most informative or focal element: Casablanca. QAPs also differ from other instances of nonsentential utterances such as those in (9): (9) a. One hotdog, please. b. Fire! c. From Spain. (uttered while holding a letter) Similar to QAPs, these cases are not considered to be a special register. However, while the nonsentential utterances in (9) do not need a linguistic antecedent, in QAPs, the question preceding the small answer is often the crucial linguistic antecedent that provides all the information that can then be left out in the answer. Importantly, since the missing elements can be recovered from the context, a nonsentential answer such as that in (11) can be immediately compared with its full counterpart in (12): (10) What do you see in the window? (11) A large yellow bird. (12) I see a large yellow bird in the window.
(Morgan 1989: 228)
The fact that in QAPs we can easily recover the missing elements might have contributed to an analysis of small answers in terms of ellipsis. That is, an elliptical account of nonsentential utterances such as those in (9) faces the problem of having to guess or stipulate what is missing. However, in the case of QAPs, the preceding question contains all the elements that seem to be missing in the small answer and therefore an elliptical account seems prima facie most adequate. This view would place small an-
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.3 (193-249)
Big questions, small answers
swers to questions next to other well-known cases involving ellipsis, such as VP ellipsis in (13), stripping or bare argument ellipsis in (14), sluicing in (15) or gapping in (16). (13) Laura took a nap, and Lena did too. (14) a. Some people smoke but not many. b. Laura left Texas, and Lena as well. c. Laura drank the milk last night, or perhaps the juice. (Heim & Kratzer 1998: 249) (15) He bought something but I don’t know what. (16) Laura likes to take a nap and Lena to go for a swim. From this perspective, and based on recent analyses of sluicing, Merchant (2004, and Chap. 3, this volume) has recently analyzed small answers to questions as a type of ellipsis preceded by movement of the small answer to a left-peripheral, focal position. As we will see in more detail later, this proposal has many advantages and elegantly accounts for a crucial feature of small answers to questions: their focal nature. However, this chapter will suggest that a nonelliptical account of small answers is still possible. The rest of this chapter is organized as follows. Section 2 considers some of the classical evidence adduced for an elliptical account of small answers to questions, which has to do with connectivity effects. In Section 3 nonelliptical analyses are examined. These proposals show that there are also anticonnectivity effects such as small answers with case-mismatched NPs or tenseless VPs. Section 4 examines Merchant’s recent approach based on sluicing, which avoids the problem of nonconstituent deletion that previous elliptical approaches faced and points out some problems with this analysis. Section 5 explores how a nonelliptical view can account for the focal nature of short answers without costly operations such as movement and deletion and in Section 6 some conclusions and remaining problems are considered.
. The elliptical view As Barton (Chap. 1, this volume) points out, Morgan (1973) provides early evidence for an elliptical account of small answers, mostly based on matching or connectivity effects (see also Hankamer 1979). His proposal is further developed in Morgan (1989), where he considers several cases that point to dependency relations between the nonsentential answer and the corresponding sentential equivalent. One case is complementizer choice. He suggests: “A fragment is well-formed just in case it satisfies complementizer requirements in the relevant position in the linguistic context” (230). This is shown by the contrast between (17) and (18): (17) A: What does John want? B: To come over after dinner./*Come over after dinner. (18) A: What did John help you do?
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.4 (249-329)
Eugenia Casielles
B: *To wash my car./Wash my car.
(Morgan 1989: 231)
As these examples show, the short answer has to match the complementizer requirements of the main verb in the preceding question. It is worth pointing out that these facts on complementizer choice seem to hold cross-linguistically. In (19) through (24) we have some examples from French, German, Serbian, and Spanish. (19) A: Que faut-il qu’elle fasse? what is-necessary-cl that-she do-3sg.subj ‘What does she need to do?’ B: Qu’elle travaille plus./ *Travailler plus that-she work-3sg more/ *work-inf more ‘That she work more./*To work more.’ (Paesani, personal communication) (20) A: Was möchte John? what want-3sg John ‘What does John want?’ B: *Vorbeizukommen nach dem Essen./Vorbeikommen nach dem Essen. *to come-inf after the meal/come after the meal ‘*To come over after dinner./Come over after dinner.’ (Work, personal communication) (21) A: Šta treba da radiš? what need that do-2sg ‘What do you need to do?’ B: Da kreˇcim zid./*Kreˇciti zid. that paint-1sg wall/*paint-inf wall (22) A: Šta ´ceš raditi? what will-2sg do-inf ‘What will you do/be doing?’ B: Kreˇciti zid./*Da kreˇcim zid. paint-inf wall/*that paint-1sg wall (Progovac, personal communication) (23) A: ¿Qué quiere Juan? what want-3sg Juan ‘What does Juan want?’ B: Venir después de cenar./*A venir después de cenar. come-inf after dinner/*part come-inf after dinner ‘To come over after dinner.’ (24) A: ¿A qué te ayudó Juan? part what cl helped Juan ‘What did Juan help you to do?’ B: A lavar el coche./*Lavar el coche. part wash-inf the car/*wash-inf the car
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.5 (329-392)
Big questions, small answers
Morgan also offers binding facts, as in (25), where John cannot corefer to he in either answer, and facts about adjectives that can only occur in predicative position, as in (26): (25) A: Where is he staying? B: In John’s apartment. B: He is staying in John’s apartment. (26) A: Is he an unhappy boy? B: No, *content/contented. B: No, he’s a *content/contented boy. These examples show that at least some short answers are well formed only if they would be well formed in the corresponding larger expression. Finally, according to Morgan (1989), small answers also show some island effects, where the fragment cannot be in a position that is in an island in the corresponding full sentence. Thus, the short answers in (27) through (29) are not equivalent to the corresponding full responses: (27) A: Did John and Bill leave this morning? B: No, Harry. B: No, John and Harry left this morning. (28) A: Was the man who shot Lincoln a Marine? B: No, Kennedy. B: No, the man who shot Kennedy was a Marine. (29) A: Did John’s seeing Martha upset the President? B: No, Thelma. B: No, John’s seeing Thelma upset the President. It should be pointed out, however, that part of the problem with the examples in (27) through (29) is the fact that these yes/no questions do not unambiguously mark the intended contrast. These improve if the contrasted element is focused in the question as in Did John’s seeing MARTHA upset the President? No, THELMA. If so, rather than island effects, these examples seem to show a focus structure mismatch, which can be partly rescued by prosodically marking narrow focus on the element being questioned. Culicover and Jackendoff (2005) have also noticed this focus-related effect. They point out that, given the appropriate focus intonation, they find the examples in (30) and (31), which Merchant (2001) finds ungrammatical, reasonably acceptable: (30) A: Does Abby speak the same Balkan language that BEN speaks? B: No, Charlie. (31) A: Did Abby vote for a GREEN PARTY candidate? B: No, Reform Party. They also offer the following examples, where the acceptable short answer would violate island constraints, if extracted:
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.6 (392-452)
Eugenia Casielles
(32) A: John met a guy who speaks a very unusual language. B: Which language?/Yes, Albanian. (*Which language did John meet a guy who speaks t?) (*Albanian, John met a guy who speaks t.) These examples are adapted from parallel sluicing examples in Chung, Ladusaw, and McCloskey (1995). We come back to these issues in Section 4, where we consider Merchant’s sluicing-based analysis of short answers and their focal interpretation.
. The nonelliptical view Despite the matching effects we have observed in the previous section, it has been pointed out (Barton 1998) that not all small answers show such effects.1 In fact, in some cases the small answer cannot be considered to be a subset of the corresponding complete answer. First, some questions such as that in (33) can be answered with a bare infinitive, as in (34b). (33) What did John do? (34) a. He played baseball. b. Play baseball. This should not be possible if small answers are elliptical since the nonelliptical source of (34b) illustrated in (35), is not possible: (35) *John play baseball. As one reviewer notes, a sentential counterpart of (34b) would be the pseudocleft in (36): (36) What John did was play baseball. As pointed out by den Dikken (2006), many works on pseudoclefts have stressed the links between these constructions and QAPs (see Akmajian 1979; Higgins 1979; Seuren 1985; den Dikken, Meinunger, & Wilder 2000, among others). One obvious link is focus. As mentioned earlier, a short answer encodes the unique focal element. A pseudocleft is one of the mechanisms that languages use to mark narrow focus on a particular phrase. Thus, in both (34b) and (36) the unique focus is Play baseball. As we will see shortly, (pseudo)clefts and small answers share other interesting properties. However, they are not interpreted in the same way. Thus, as an answer to the question in (33), a cleft or a pseudocleft would have a different interpretation than the answers in (34).2 (37) a. What did John do? b. It is play baseball that he did./What he did is play baseball.3
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.7 (452-520)
Big questions, small answers
Infinitival short answers such as that in (34b) do not occur only in English, although there seems to be some cross-linguistic variation in this regard. Languages like German and Spanish behave like English, as in (38) and (39) respectively, while others like French in (40) or Serbian in (41) do not allow for the infinitive. (38) A: Was macht John? B: Spielt Baseball./Baseball spielen. (39) A: ¿Qué hace Juan? B: Juega al béisbol./Jugar al béisbol. (40) A: Que fait Jean? B: Il joue au baseball./*Jouer au baseball. (41) A: Šta Jovan radi? B: Igra bezbol./*Igrati bezbol. Interestingly, there doesn’t seem to be a perfect correlation between allowing the infinitival answer in (34b) and the pseudocleft in (36). As might be expected, the German and Spanish counterparts of (36) are grammatical in these two languages. However, while neither French nor Serbian allows an infinitival answer, infinitival pseudoclefts are impossible only in Serbian. The French counterpart of (36) is possible. (42) Ce que John fait c’est jouer au baseball./Jouer au baseball, c’est ce que John fait.4 ‘What John does is play baseball./Play baseball is what John does.’ Although we come back to the links between short answers and pseudoclefts later, the cross-linguistic data on bare infinitives examined here do not point to a perfect correlation between the two. More evidence for a nonelliptical account of small answers comes from negative responses of the sort illustrated in (43): (43) A: Who did you talk to? B: Not to John./Not John.5 Again, the short answer in (43) is not just an elliptical version of the complete answer or of the pseudocleft counterpart: (44) *I talked not (to) John./*What I did was talk not (to) John.6 Furthermore, short answers can also show some interesting case mismatches. Short answers usually show case connectivity, where the short answer appears in the same case as it would have appeared in the corresponding sentential answer. This is shown in (45) for Korean. (See Merchant 2004 for similar facts in Japanese and Culicover & Jackendoff 2005 for similar examples in German.) (45) Nu-ka ku chaek-ul sa-ass-ni? Yongsu-ka./*Yongsu-rul7 ‘Who bought the book?’ nom/*acc However, in English the question in (46) can be answered as in (47) or as in (48):
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.8 (520-595)
Eugenia Casielles
(46) Who can eat another piece of cake? (47) I can (eat another piece of cake). (48) Me. Notice that the answer in (48) cannot be the elliptical version of a complete answer, as shown in (49): (49) *Me can eat another piece of cake.8 As a reviewer points out, me is also the form that would occur in a dislocated structure, as in (50a) and in a cleft, as in (50b): (50) a. Me, I can’t eat anymore. b. It is me who can eat another piece of cake. I consider dislocated structures such as that in (50a) in detail in Section 4. In regard to (50b), although I do not think it would be plausible that a cleft could be the source or base for the short answer Me in (48) (see n. 6), it is no coincidence that this form appears in all these cases. What all three positions have in common is that they are all nonargument positions, positions where structural case is not assigned. It is interesting to note that these nonnominative pronouns also occur in child language, as in (51), and so-called Mad Magazine utterances in (52). (See Potts & Roeper, Chap. 7, this volume, for the additional meaning involved in these structures.) (51) a. Me do it. b. Him gone.
(Radford 1990)
(52) Me stop smoking? Never! Progovac (Chap. 2, this volume) suggests that these pronouns bear default case and notes that what is crucial about examples such as (48) and (52) is the correlation between absence of tense and absence of nominative case.9 In examples involving lexical rather than pronominal subjects, this correlation can be seen in the concomitant absence of tense and determiners, as shown in the following examples: (53) a. Battery dead. (*Battery is dead./*The battery dead.) b. Problem solved. (*Problem is solved./*The problem solved.) Progovac (Chap. 2, this volume) offers a nonelliptical analysis of nonsentential utterances such as these, which elegantly accounts for this correlation between absence of tense and absence of nominative case/determiner. From this perspective, a small answer such as that in (48) is analyzed as a base-generated phrase, and not as an elliptical fragment, while examples such as those in (52) and (53) are analyzed as root small clauses. Thus, the contrast between (47) and (48) could in fact be seen as a contrast between an elliptical answer that contains a tensed auxiliary and nominative case, and a nonelliptical phrasal answer that shows no tense and default case.10
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.9 (595-630)
Big questions, small answers
It should be mentioned that even if we put aside these cases, which are particularly difficult for an elliptical account, there is still a problem with deriving more straightforward cases such as (10), repeated here as (54): (54) A: What do you see in the window? B: A large yellow bird. As has been pointed out in the literature, deriving a short answer such as that in (54) from the corresponding sentential (I saw a large yellow bird) faces the following problem: what seems to be elided is not a constituent, as shown in (55): (55)
TP T’
SPEC I
VP
T Spec
V’ V see
DP a large yellow bird
Recently, however, Merchant (2004, and Chap. 3, this volume) has defended an elliptical analysis that does not run into this problem. We now turn to this view.
. The sluicing view Merchant (2004, and Chap. 3, this volume) offers a very elegant elliptical analysis of small answers that avoids some of the problems of previous elliptical views (i.e., Morgan 1973). Thus, he does not analyze a short answer such as that in (56) as involving deletion of the subject and verb She saw. (56) A: Who did she see? B: John. Instead, and based on recent analyses of sluicing, he suggests that the short answer John is A-bar moved to a left-peripheral (focal) position and that it is the remaining TP constituent (in angle brackets) that is elided, as shown in (57):
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.10 (630-717)
Eugenia Casielles
(57)
CP Spec John
C’ C
she saw
This nicely accounts for the fact that small answers are focal elements. That is, when we reply to a question, we may leave out everything that is considered to be noninformative in that context, she saw in the case of (56). In fact, question-answer pairs have been the traditional test used in the literature to identify the focus of a sentence. From this perspective, a sentence such as Peter saw John could have a different focus structure depending on the previous question, as shown in (58) through (60), where the focal element appears in capital letters. (58) Who did Peter see?
Peter saw JOHN.
(59) Who saw John?
PETER saw John.
(60) Did Peter talk to John?
Peter SAW John.
It should be pointed out that although this type of long answer abounds in the literature on focus, it is sometimes mentioned that that is not how people would answer a question. They would just answer JOHN, in the case of (58). This is probably why short answers to questions have sometimes been considered to be the main type of answer and not a derivative of the longer version. The noninformative part in cases of narrow focus such as these is usually referred to as the background/ground/open proposition and according to Merchant is not just left unexpressed, but is first projected and then deleted after the focal element has moved out to a left-peripheral position.11 Although this proposal solves some difficulties of previous elliptical analyses, and nicely accounts for the focal nature of small answers, it faces some empirical and theoretical problems. First, it should be pointed out that a simpler analysis of small answers such as that in Barton and Progovac (2005) and Progovac (Chap. 2, this volume), if tenable, is a priori preferable under economy considerations (see later discussion). Second, this movement-plus-ellipsis analysis would mean that when we are leaving out everything that seems unnecessary for effective communication, when we are being maximally economical, we are in fact doing much more complicated operations than when we produce the longer version, since the short answer would involve everything required in the longer version plus A-bar movement and then deletion.12 While this seems certainly counterintuitive for adult short answers, proposing that small answers are not simpler but rather more complex than long answers becomes highly improbable if we take into account that children produce such short answers quite effortlessly from early on. That is, as opposed to clear cases of ellipsis such as
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.11 (717-764)
Big questions, small answers
those exemplified in (13) through (16), which, as expected, do not occur in early child language, children have no problem producing a short answer to a question at early stages when they are known not to have acquired either structures involving A-bar movement, such as wh-questions, or cleft and pseudocleft constructions. (See Guasti 2002 and the references therein.) If long answers were less complex than short ones, we would expect children to produce long answers before short ones, which is not the case. Thus, unless there is some evidence that adult small answers differ in a significant way from children’s small answers, a unified account that considers these short answers as simple phrases, possibly generated by the type of Minimal Default Grammar (MDG) proposed by Roeper (1999), seems certainly more adequate.13 This argument also applies to patients with agrammatism, who, if capable of answering a question at all, would produce a short rather than a long answer. (See Avrutin 2004; Kolk Chap. 9, this volume; and Siple Chap. 10, this volume, for details.) Third, short answers with mistmached case, such as that in (48), are still a problem for this type of movement-plus-ellipsis analysis. Even if moved, me has the wrong case. Merchant (Chap. 3, this volume) suggests that this might not be a problem since “it is this form of the pronoun that similarly can bear focus on the left periphery of the clause as well.” He gives the examples in (61) and (62): (61) A: Who watered the plants? B: Me. (62) Me, I watered the plants. However, the problem with this argumentation is that a structure like (62) does not involve focus or movement. If intended as an answer to the question in (61), where both ME and I are prosodically prominent and semantically focal, the only possible analysis would be one in which we have two different utterances, as in (63): (63) A: Who watered the plants? B: ME. I watered the plants. That is, (62) cannot be a case of a single structure exemplifying Focus-Movement, as Merchant seems to imply. Focus-Movement involves movement of a focal element without any resumptive pronoun and it is represented by examples such as those in (64). (64) a. FIDO they named it. b. THIRTY POUNDS it weighs. c. Did you say he was Japanese? KOREAN I said he was. As Prince (1981) points out, Focus-Movement in English is restricted to the preposing of nonreferential NPs such as those in (64).14 Although the focal preposed element is not usually a pronoun, given the referential restrictions of this type of preposing in most dialects of English, some speakers might accept a sentence such as (65) where the preposed focal element is a pronoun.
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.12 (764-831)
Eugenia Casielles
(65) ME they wanted to nominate.15 Note that, in this case, the focus-moved element does not co-occur with a pronoun inside the sentence. Thus, the example in (62) with a comma is only possible as an instance of left dislocation, as in (66), similar to that in (67) where a sentence-initial topic co-occurs with a resumptive pronoun: (66) (As for) Me, I watered the plants. (67) Your brother, he is a nice guy. It is worth pointing out that contrary to what the discussion in Merchant (Chap. 3, this volume) seems to imply, the left dislocation in (66), whose focal element would be the predicate, would not be an appropriate answer to the question in (63), as shown in (68): (68) A: Who watered the plants? B: #Me, I WATERED THE PLANTS. The structures in (66) and (67) do not involve a focal-initial element. One possible context for (66) would be an answer to a question such as What have you guys been doing since I left?, where the focus would be watered the plants. The initial element introduces a discourse-new topic before including it in the predication.16 We have seen that the structure in (62), with a comma, does not involve a sentenceinitial focal element. Furthermore, as Merchant (2004) acknowledges, it is not clear that these structures involve movement either. (See Anagnostopoulou, van Riemsdijk, & Zwarts 1997; Grohmann 2003, for recent base-generation analyses of these structures.) From this perspective, the fact that in English we find the tonic (default case) pronoun me both in isolation (small answers such as (61)) and in dislocated topic positions (as in (62)) might be related to the weak nature of the nominative form I in English, which seems unable to stand on its own. However, this can hardly be used as an argument in support of a Focus-Movement-plus-ellipsis analysis of small answers, particularly when the allegedly similar dislocated structures do not involve a focusrelated movement, and possibly no movement whatsoever. Once we carefully look at the intonation and interpretation of the answer in (63) we have to conclude that, as an answer to Who watered the plants?, it can only involve two different phrases: the short answer with default case and the longer answer. Thus, a movement-plus-ellipsis analysis fails to account for short answers such as those in (48) and (61). Fourth, even if we put aside these instances of case mismatch that seem particularly difficult for an elliptical analysis, it is still not clear how this movement-plusellipsis analysis would account for the type of bare infinitive answer we saw in (34b), repeated here as (69).17 (69) What did John do? Play baseball.
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.13 (831-904)
Big questions, small answers
Furthermore, moving and deleting does not seem to help to account for the following cases, either. (70) A: Is he an unhappy boy? B: No, contented. (*Contented he is a boy.) (71) A: How do you like your coffee? B: Sweet./Not sweet. (*Sweet/not sweet I like my coffee.) (72) A: Is Mary a good teacher? B: No, bad. (*Bad Mary is a teacher.) Finally, and with regard to more straightforward short answers such as the one we saw in (56), a movement-plus-ellipsis analysis still needs to explain why Focus-Movement could freely apply in all instances of small answers in English (as long as it is followed by ellipsis) while it is in fact a highly restricted (and not very productive) type of movement in this language. That is, (73) is not acceptable for most speakers of English. (73) *JOHN she saw. This contrasts with sluicing, where the nonelliptical counterpart is also fine. That is, we can have both (74) and (75): (74) He bought something but I don’t know what he bought. (75) He bought something but I don’t know what. According to Merchant’s analysis of small answers, we would be getting to (76) through the unacceptable (77): (76) JOHN. (77) *JOHN I saw. As mentioned earlier, most dialects of English restrict Focus-Movement to nonreferential phrases such as those in (64), and do not allow for referential NPs to be focus moved. Other languages, however, are not so restricted, as the following examples from Spanish and Catalan show. (78) a.
El CUCHILLO puse en el cajón de arriba. ‘The KNIFE I put in the top drawer.’ b. El GANIVET, vaig ficar al calaix de dalt. ‘The KNIFE I put in the top drawer.’
(Spanish) (Catalan) (Vallduví 1990)
These contrast with the English counterpart in (79): (79) ??The KNIFE I put in the drawer. From this perspective it is improbable that a highly restricted type of A-bar movement in English is involved in all instances of short answers to questions. Furthermore, if this analysis is not only for small answers in English, but for small answers in any language, a complicating factor would need to be taken into account: the fact that
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.14 (904-965)
Eugenia Casielles
in many languages Focus-Movement behaves like wh-movement triggering subjectpredicate inversion, as in the following examples from Spanish: (80) EN PRIMAVERA visitó Juan Leningrado. in spring visited Juan Leningrad ‘In the SPRING John visited Leningrad.’ (81) *EN PRIMAVERA Juan visitó Leningrado. This shows that the nature of Focus-Movement in English might be different from this type of A-bar movement in other languages, which means that we would have to be more specific as to what type of Focus-Movement might be involved in small answers to questions. Taking into account the lack of sufficient evidence of complex mechanisms such as focus-related movements and ellipsis, and the inability of the movement-plus-ellipsis analysis to explain recalcitrant data involving case mismatches such as that in (61), and other types of small answers, such as the ones in (69) through (72), I will assume that short answers are not more complex than longer answers, but simpler. From this perspective I explore their syntax and semantics under a nonelliptical view in the next section.
. Toward a phrasal analysis of small answers As pointed out by Progovac (Chap. 2, this volume), considering nonsentential utterances as phrases (rather than sentences) follows structure-building procedures under Minimalism where there is not a priori construction, but rather phrases that are merged as needed in a bottom-up fashion. Furthermore, from a syntactic point of view, a phrasal analysis of small answers is not only licit but in fact more appropriate given economical principles of derivation and/or representation such as the Shortest Derivation Requirement or Boškovi´c’s (1997a) Minimal Structure Principle in (82): (82) The Minimal Structure Principle Provided that lexical requirements of relevant elements are satisfied, if two representations have the same lexical structure and serve the same function, then the representation that has fewer projections is to be chosen as the syntactic representation serving that function. With particular regard to QAPs, Radford (1997: 86), for instance, chooses the small answer in (83) to exemplify how phrases are formed and assigns it a VP structure, as in (84): (83) A: What’s the Government planning to do? B: Privatize hospitals. (84) [VP [V privatize] [N hospitals]]
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.15 (965-1044)
Big questions, small answers
As we have seen, a phrasal analysis of small answers can also account for case mismatches such as that in (85), left unaccounted for in elliptical approaches. (85) A: Who watered the plants? B: Me. Answers such as that in (85) are not an idiosyncrasy of the English language. This type of answer is hard to see in those languages whose default case is nominative rather than accusative (e.g., German, Spanish, Serbian). However, as noted already in Morgan (1989), other languages have similar case alternations. Thus, in Korean answers to questions can either match the case of the question as in (86) or show no case whatsoever, as in (87). (86) Nu-ka ku chaek-ul sa-ass-ni? Yongsu-ka./*Yongsu-rul.18 ‘Who bought the book?’ nom/*acc (87) Nu-ka ku chaek-ul sa-ass-ni? Yongsu.19 no.case ‘Who bought the book?’ From a syntactic point of view, a phrasal analysis of short answers like those in (85) and (87) is certainly viable. However, advocates of elliptical approaches to small answers have claimed that a phrasal approach poses a problem from the point of view of the syntax-semantics mapping. That is, a phrasal approach would need to explain how a nonsentential object has a fully propositional content while a sentential view does not run into this problem. According to Merchant (Chap. 3, this volume), for instance, a sentential perspective does not require a revision of usual form-meaning mappings, does not need a new method of generating and interpreting these nonsentential data, and gets the propositional content from the usual mechanisms. However, there are some problems with this argumentation. First, mismatches between form and interpretation occur in other contexts than small answers to questions. They seem to occur in other adult registers such as titles, labels, headlines, and telegrams, and also in language produced by children, speakers of pidgin languages, or patients with agrammatism. Thus, rather than trying to eliminate them we need to investigate them further for independent reasons. Second, there are mechanisms already in place to account for mismatches between form and interpretation. These are particularly well known in the semantics literature. In the late 1970s and 1980s, for instance, important work tried to reconcile the evidence for a systematic correspondence between syntactic categories and semantic types with the evidence of actual flexibility in natural languages (see Grimshaw 1977, 1979; Heim 1979). Focusing on the interpretation of verbs, Partee and Rooth (1983) propose to give up Montague’s (1973) uniform category-to-type mapping, the requirement that each syntactic category have a single semantic type associated with it. In Montague’s system, a given syntactic category is assigned the “highest” type needed. Instead, Partee and Rooth allow for each category to correspond to a family of types, rather than just a single type, and propose type-lifting rules. Crucially, expressions are
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.16 (1044-1087)
Eugenia Casielles
assigned the “simplest” type adequate to capture their meaning and additional higher types are among the options available, if needed. Furthermore, they propose that there is a general processing strategy of trying lowest types first. Thus, “sentences which are intuitively simpler now have simpler translations involving lower types than sentences which are intuitively more complex” (368). Partee (1986) extends this idea of type-shifting principles to Noun Phrase interpretation. Instead of treating NPs uniformly as generalized quantifiers, as Montague did, she proposes three basic NP types: e (“referential”), <e, t> (“predicative”) and ,t> (“quantificational”). This proposal succeeds in reconciling flexibility in category-type mappings with a systematic correspondence between categories and types, elegantly resolving the apparent conflict among the different interpretations that had been assigned to NPs in the previous literature. From this perspective a phrase such as every man is always interpreted as a generalized quantifier while a phrase such as John has the basic e interpretation and is only lifted to a generalized quantifier interpretation if needed, for instance in the conjunction John and every woman. Finally, and based on Partee (1986), Groenendijk and Stokhof (1989) proposed a flexible approach to the relationship between syntactic categories and semantic types with particular regard to the semantics of questions. Thus, a flexible approach to the correlation between syntactic categories and semantic types is certainly not new, and has in fact been long suggested for independent reasons. From this perspective, considering a nonsentential utterance such as Problem solved a VP, as proposed by Progovac (Chap. 2, this volume), makes sense not only because of the concomitant absence of tense and case (no is and no the), but also because it means assigning an intuitively simpler utterance the simplest, lowest structure compatible with the semantic expression of an argument-function relation. It is worth noting that to assign a phrase with propositional content a VP structure (instead of a CP structure) would not go against usual syntax-semantics mappings, particularly if CP and IP are just extended projections of VP, as proposed by Grimshaw (1991). Third, with regard to the interpretation of small answers in QAPs such as John in (88), we might not even need to say that the phrasal answer is interpreted as a proposition. (88) A: Who did you see? B: John. As pointed out by Barbara Partee (personal communication), in addition to Kartunnen’s (1977) classical interpretation of questions as sets of propositions, other views (e.g., Hausser 1983) see the answer as the argument, and the question as a function. From this point of view we get a proposition from the QAP, not from the answer alone. (See also Hamblin 1973; Higginbotham 1996; Ginzburg 1996.) Thus, we have seen that a phrasal analysis of small answers does not necessarily run into syntax-semantics interface problems. Finally, can this approach explain the focal nature of these phrases? As mentioned before, Merchant’s movement-plus-ellipsis analysis elegantly accounts for the focal nature of short answers. However, we have seen that a simpler, phrasal approach might
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.17 (1087-1175)
Big questions, small answers
be theoretically and empirically preferable, so what we need is a way of getting the focal interpretation without the movement or the ellipsis. Although there is much research in semantics about the interpretation of focus in QAPs, most of it deals only with long answers (e.g. Erteschik-Shir 1997; Krifka 1991; Partee 1991; Rooth 1985). In general, we do not need to move a phrase to a preverbal focus position in order to get a focal interpretation. In fact, the preverbal focus position is usually associated with a contrastive focus interpretation not present in small answers to questions such as that in (88). The default focus position in English and many other languages is inside the VP. That is, unless we emphatically stress another element as in the sentence in (89) or use some other syntactic mechanisms such as a cleft (90) or a right dislocation (91), an in situ object such as that in (92) is interpreted as focal by default. (89) SHE saw John. (90) It was Mary who saw John. (91) Anne saw him today, John. (92) She saw John. Thus, John in (92) would be interpreted as focal due to its rightmost position inside the VP. From this perspective, it would make sense if an isolated phrase gets the same type of default focal interpretation be it as an answer to a previous question, as in (93), or without any linguistic antecedent, as in (94): (93) A: Who did she see? B: John. (94) John. (uttered while holding a letter) It is worth pointing out that early research on language acquisition noticed that children are aware of new versus old information from early on (Baker & Greenfield 1988) and that they choose the most informative element when they are producing one-word utterances (Bates 1976; Greenfield & Smith 1976; Greenfield, Reilly, Leaper, & Baker 1985). Interestingly, Leonard and Schwartz (1978) discovered that children who were capable of multiword utterances used single-word utterances as a focalizing device, to emphasize a new aspect of a situation. The children they studied used single words more often than longer utterances to express new information.20 From this perspective, it is plausible to assume that one-word utterances or short answers to questions are interpreted as focal by default. If so, we do not need to place them in any particular syntactic position for them to have a focal interpretation. Notice that it would be possible to account for the focal nature of these short answers and pursue a sentential analysis without any movement or ellipsis. One possibility would be to analyze the small answer in (93) as an isolated focal phrase, which gets interpreted as the answer to the preceding question when the nonfocal part (she saw in (93)) is copied (possibly at LF) forming a Focus-Background structure similar
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.18 (1175-1210)
Eugenia Casielles
to Merchant’s but without A-bar movement or ellipsis. Similar LF-copying mechanisms have been proposed for the analysis of pseudoclefts (e.g., Boškovi´c 1997b). In fact, regardless of how small answers are analyzed (either as isolated focal phrases or as Focus-Background structures where the Background has been deleted), a nonmovement (simple merger/adjunction) analysis of Focus-Background structures would in fact make more sense from the point of view of language acquisition, since FocusBackground structures have been found in children’s early utterances (see Bates 1976; Bowerman 1973; Sinclair de Zwart 1973, for early studies). As Bates notices for Italian, sometimes the child blurts out the new information first and then the rest in decreasing order of interest. I have also found these structures in child Spanish data and in child-directed speech in Spanish (Casielles, Andruski, Kim, Nathan, & Work 2006). MacWhinney (1984) also points out other studies that report that children begin their sentences with new information (Leonard & Schwartz 1978; O’Shea 1907, among others) and suggests that if that is the case, the words that follow the first word must be viewed as afterthoughts. Needless to say, more research is needed to evaluate the importance of these Focus-Background structures both in the input that the child receives and in children’s early productions. If further investigation confirms the early presence of these structures and shows that they follow one-word focal phrases, their syntactic analysis in terms of A-bar movement will have to be revised, since they might involve a much simpler merger mechanism.
. Conclusions and further issues This chapter has argued that it is possible to pursue a phrasal, nonelliptical analysis of simple short answers such as those in (95) through (97), repeated here for convenience: (95) A: Who ate the cake? B: Me. (96) A: Who did you see? B: John. (97) A: What did John do? B: Play baseball. As pointed out earlier, this type of short answer occurs early in the language acquisition process and can be accounted for without appealing to any hidden sentential structure. What about some of the cases mentioned in Section 2, which showed connectivity effects? The following two cases involving complementizer choice and anaphors certainly point to the presence of extra-sentential structure: (98) A: What does John want? B: To come over after dinner./*Come over after dinner. (99) A: Who does John love?
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.19 (1210-1281)
Big questions, small answers
B: Himself./*Him. It is possible that short answers that include dependencies such as that between a pronoun and its antecedent or that between a verb and its clausal complement are indeed different and more complex than those in (95) through (97).21 Nevertheless, even for these cases, connectivity effects are not always perfect. Culicover and Jackendoff (2005) point out that choice of complementizer and tense in short answers is not as restricted as in full sentences. They provide the following examples: (100) A: What did you want? B: To leave./That Robin leave./That Robin should leave. (*I want that Robin leave./*I want that Robin should leave.) In regard to binding relations, anticonnectivity effects have also been pointed out. Thus, Ginzburg and Sag (2000) show that both short answers and sluices can disobey binding constraints. They offer the examples in (101) and (102): (101) A: Who will punish Billi if he fails. B: He himself./Himself. (#He himselfi ./*Himselfi will punish Billi if he fails.) (102) A: Who appeared to be the cause of John and Mary’s problem? B: Each other. (*Each other appeared to be the cause of John and Mary’s problems.) Culicover and Jackendoff (2005) also point out that connectivity effects of this type also show up in clefts and pseudoclefts, which according to some analyses do not involve movement, and that as shown in Higgins (1979), the same relation would hold if the first conjunct is a relative clause from which extraction would not be possible, as in (103). (103) The only one that Johni has become enamored of is himselfi . In addition, they show that the short answer in (104) cannot be a reduction of a full sentence: (104) A: Ozzie loves Harriet. B: But not himself. (*But not Ozzie loves himself.) Ginzburg and Sag (2000) note that both short answers and matrix sluices such as that in (105) are tied to a conversational context which involves multiple agents and different information states. (105) A: I can find someone to do the job. B: Who? They point out that, as (106) shows, we cannot always recover the syntactic antecedent. (106) a.
A: Did anyone see Mary?
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.20 (1281-1351)
Eugenia Casielles
b. B: Yes. c. A: Who? That is, since B’s answer in (106b) cannot be the syntactic antecedent of (106c), the only possibility would be (106a) and it is not clear how we could extract from (106a) the syntactic reconstructed element for (106c). Putting aside these difficulties, an elliptical account is not the only option even for the cases that do show the expected connectivity effects. With special regard to dependencies of the type in (98), Laury and Ono (2005) point out that although speakers are aware of structures produced by other speakers and frequently model their own utterances so that they syntactically fit previous utterances, this doesn’t necessarily point to a sentential representation, but could easily be attributed to memory and what has been called socially shared cognition.22 In the framework of Head-Driven Phrase Structure Grammar (HPSG), Ginzburg and Sag (2000) offer an account based on an independently motivated theory of dialogue context, which does not rely on syntactic reconstruction. Culicover and Jackendoff (2005) also offer a solution that accounts for connectivity effects. They suggest what they refer to as “indirect licensing.” This account is nonelliptical in that it does not assume any extra-syntactic structure, but differs from other nonelliptical accounts in that the syntactic structure of the antecedent is relevant not only to the interpretation of the nonsentential utterance, but also to its syntactic well-formedness, because when the antecedent is processed these syntactic features get activated. From this perspective, syntactic features can be assigned without licensers in the local context if the licensers are in a previous sentence, in a part of the lexicon activated by the previous sentence, in a part of the lexicon activated by the nonlinguistic context, or in the rules of grammar that establish a syntactic connection appropriate to the nonsentential’s semantic role in the antecedent (Culicover & Jackendoff 2005: 263). Nevertheless, Merchant (2004, and Chap. 3, this volume) offers cross-linguistic data showing robust connectivity effects, which certainly hint at the presence of extrasentential structure at least for some cases. A particularly interesting set of data, which has not been considered in this chapter, is the case of prepositionless answers to whphrases governed by a preposition in preposition-stranding languages such as English, as in (107): (107) A: Who was he talking with? B: Mary. As Merchant points out, this kind of prepositionless answer is not possible in nonpreposition-stranding languages. In (108) we have an example from Spanish: (108) A: ¿Con quién hablaba?/*¿Quién hablaba con? B: Con María./*María. Merchant interprets this as evidence of a movement analysis of the short answer in (107). These contrasts are important and deserve further investigation of several issues,
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.21 (1351-1411)
Big questions, small answers
including the different nature of prepositions in different languages.23 The contrast between (107) and (108) certainly shows that in non-preposition-stranding languages an answer to a wh-phrase governed by a preposition has to match the wh-phrase and also has to be a phrase headed by a preposition, but I am not convinced that it is clear evidence of a movement analysis of the answer. Merchant (2004) also points out that in preposition-stranding languages the preposition is optional in sluices of the type in (109): (109) He was talking to somebody, but I don’t know (with) who. Interestingly, this contrast with non-preposition-stranding languages does not seem so clear-cut. The Spanish counterpart of (110) seems quite acceptable: (110) Hablaba con alguien, pero no sé quién. Thus, there is definitely a considerable amount of complex data that remains to be carefully studied. However, a constructive dialogue between elliptical and nonelliptical views is expected to give fruitful results and advance our understanding of the nature of small and not so small answers to questions. Certainly, I remain open to the possibility that a nonelliptical analysis of short, simple answers that this chapter has defended can be extended to account also for more complex answers involving syntactic dependencies. From this perspective, I depart from Merchant’s suggestion to try to find a way to prohibit a nonsentential analysis in contexts where an elliptical analysis is possible, and instead I would suggest going the other way around and pursue an elliptical analysis only for those contexts where a nonelliptical analysis is shown to be inadequate.
Notes * Parts of this chapter were presented at the 2003 Workshop on Nonsententials at Wayne State University. I would like to thank the WSU Humanities Center for funding our working group on nonsententials since 2002 and the workshop participants and audience for their comments. Special thanks go to the coeditors of this volume, Ellen Barton, Kate Paesani, and Ljiljana Progovac, and to the other members of our working group, Walter Edwards, Pat Siple, and Nicola Work. I would also like to thank the following friends and colleagues for their judgements: Enoch Aboh (Gbe); Michel DeGraff, Caroline Jumel, and Kate Paesani (French); Ljiljana Progovac (Serbian); Cristina Schmitt (Brazilian Protuguese); and Nicola Work (German). I am especially grateful to Michel DeGraff and Barbara Partee, who gave me detailed and very helpful comments and suggestions. Usual disclaimers apply. . For other nonelliptical approaches see Barton (1990), Barton and Progovac (2005), Carston (2002), Culicover and Jackendoff (2005), Ginzburg and Sag (2000), Jackendoff (2002), Stainton (1998), and Yanofsky (1978), among others. . A detailed study of clefts and pseudoclefts is beyond the scope of this chapter, but for relevant recent works see Boškovi´c (1997b), Collins (1991), den Dikken et al. (2000), and den Dikken (2006).
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.22 (1411-1483)
Eugenia Casielles . A reviewer points out that in some languages like Portuguese and Gbe, clefts can be answers to questions. This is also true of French. Thus, as Lambrecht (2001) points out in French a question such as How do you know? can be answered with the cleft in (i): (i)
C’est HUMA qui me l’a DIT. it-is Huma who cl cl-has told ‘It is Huma who (has) told me.’
Even in this type of language, however, clefts have the usual uniqueness interpretation that is absent in a simple small answer, and that makes a cleft inappropriate in a context where a uniqueness interpretation is not expected, as in the following example adapted from Lambrecht (2001): (ii) (Mother, looking around dinner table) A: Qui veut encore de la viande? ‘Who wants some more meat?’ B: Moi/#C’est moi. ‘Me.’/#‘It’s me.’ In a similar vein, according to my Brazilian Portuguese informant, the following sentence can only be interpreted exhaustively and contrasting with another phrase: (iii) (Group of friends in a restaurant) A: Quem pediu mais uma garrafa de vinho? who asked more a bottle of wine ‘Who asked for another bottle of wine?’ B: Foi o Pedro que pediu, nao fui eu. it-was det Peter that asked not was I ‘It was Peter that asked, not me.’ Gbe differs from French and Portuguese in the way focus is marked. According to Aboh (2004), in this language a focal element appears in sentence-initial position immediately followed by the focus particle wÜ7, as in (iv): (iv) SÜ7ná wÜ7 xìá wémà Sena foc read book ‘SENA read a book.’
(Aboh 2004: 237)
In regard to cleft constructions, Aboh notes that according to Ameka (1992), although both the focus-marked construction in (iv) and clefts are related in terms of their information structure, they should be distinguished due to their different grammatical, semantic, and pragmatic features. . As a reviewer rightly points out, the counterpart of the English cleft is ungrammatical in French: (i) It is play baseball that John does. (ii) *C’est jouer au baseball que Jean fait. However, this is also the case in Spanish, a language that allows for infinitival short answers: (iii) *Es jugar al béisbol que hace John.
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.23 (1483-1546)
Big questions, small answers
In these and other languages the closest to an it-cleft is a structure that does not have the complementizer that (que in both French and Spanish) but the counterpart of what (ce que in French and lo que in Spanish): (iv) a. b.
C’est jouer au baseball ce que Jean fait. Es jugar al béisbol lo que hace Juan. ‘*It is play baseball what John does.’
Lambrecht (2001) argues that this structure in French is a right-dislocated free relative clause that resembles an it-cleft and is also possible in English and German: (v) a. b. c.
C’est du champagne, ce qu’il a acheté. It’s champagne, what he bought. Das ist Champagner, was er gekauft hat.
. As one reviewer points out, an elliptical view might argue that what is missing in this type of answer is the presupposition of the question, namely, I talked to somebody. From this perspective, a small answer such as Not to John might be seen as the elliptical version of I talked to somebody, but not (to) John. . A reviewer points out that both the short answers John and Not to John could be derived from the clefts in (i) and (ii): (i) (It is) to John (that I talked). (ii) (It is) not to John (that I talked). As far as I know, there is not an elliptical analysis of small answers in the literature that explicitly proposes clefts or pseudoclefts as the base or source of small answers. It would certainly seem counterintuitive to view such complex constructions as the base or source of a short answer. Notice that, although prima facie deleting the parenthetical material in (i) and (ii) might look like nonconstituent deletion (of the subject + copula (It is) on the one hand, and the free relative clause (that I talked) on the other hand), this problem disappears if clefts are analyzed along the lines first proposed by Chomsky (1977), where this construction involves movement of a whoperator to the specifier of CP and base generation of the clefted element in a position adjoined to CP. The case in (ii) might still need to move Not to John, but in general such a hypothetical elliptical analysis would be similar to Merchant’s solution, and face the same problems, which are considered in Section 4. Notice that elliptical views such as these would predict, contrary to fact, that children would have to acquire clefts such as that in (ii), before they are able to produce short answers such as that in (iii): (ii) It was Mommy who gave me that. (iii) A: Who gave you that? B: Mommy. . As we will see later (example (87)), the same question can be answered in Korean with no case whatsoever. . Merchant (2004, and Chap. 3, this volume) offers a solution to this problem. His proposal is considered in the next section. . See Schütze and Wexler (1996) for an analysis of nonnominative pronouns in children’s utterances such as that in (i) as involving default accusative case caused by the absence of Agr(eement).
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.24 (1546-1619)
Eugenia Casielles
(i)
Him fall down.
(Schütze & Wexler 1996: 670)
. In languages like Spanish or German where default case is not the accusative, but the nominative, all those examples would have a form indistinguishable from a nominative case form. . Not all sentences have this type of information structure involving narrow focus. There are other information structures such as Topic-Comment, which usually involve wide focus. The sentence He cooked lasagna, for instance, can have narrow focus on LASAGNA, as in (i) or wide focus as in (ii). See Casielles (2004) for a detailed survey of dichotomies such as Topic-Comment and Focus-Background and the different primitives that they involve. (i)
A: B: (ii) A: B:
What did Mark cook? He cooked [LASAGNA]. What did Mark do? He [cooked LASAGNA].
. As a reviewer points out, a simpler and more economical syntax means a more elaborated pragmatics module. This is in fact a desirable result, since an elaborated pragmatics module is needed for independent reasons in order to interpret all kinds of indirect answers, for instance, and more generally to understand how natural conversation works, as pointed out by Grice (1975). (See Epilogue, this volume, for further discussion.) . This assumption that adult small answers might be as simple as children’s small answers is independent of and compatible with different accounts of children’s telegraphic utterances such as Radford’s (1990) small clause analysis, or more recent analyses that assume truncation/omission or underspecification of INFL (see Radford 2000; Rizzi 1994, 2000; Schütze & Wexler 1996; Wexler 1998, for some proposals). However, if, as pursued in different chapters of this volume, adults still have access to a type of nonsentential grammar, which can generate simple phrases and root small clauses, we don’t have to explain how children’s nonsentential grammar disappears: it does not. In this sense, if correct, this view supports continuity, not so much because children’s grammar is viewed as essentially the same as the adult’s sentential grammar but because adults still have access to a nonsentential grammar. Thanks to Ljiljana Progovac (personal communication) for bringing this to my attention. . As Prince (1981) points out, Focus-Movement (FM) contrasts with so-called YiddishMovement (YM), which does allow for referential preposed focal phrases, and from yet another preposing mechanism, Topicalization, which does not prepose a focal element: in (i) through (iii) we have an example of each: (i) MACADEMIA NUTS I think they’re called. (ii) MACADEMIA NUTS I should buy yet. (iii) Macademia nuts I can’t AFFORD.
(FM) (YM) (Topicalization) (Prince 1981)
Although Topicalization and Focus-Movement differ in the focal versus nonfocal nature of the preposed element and in the type of accent this preposed element gets (A-accent in FocusMovement vs. B-accent in Topicalization), both involve the preposing of an often contrastive element with no resumptive pronoun. (See Casielles 1998; Prince 1981; Ward 1988, for more details.) . Topicalization can also prepose a pronoun, usually when this element is a contrastive topic. As (i) shows, no resumptive pronoun would appear either:
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.25 (1619-1692)
Big questions, small answers
(i)
Me they haven’t CALLED.
. In fact, as opposed to Merchant (Chap. 3, this volume), Merchant (2004) does not relate (62) to focus, but to this type of left dislocation, which involves a nonfocal left- (or right-) dislocated DP with a resumptive pronoun. . Thanks to Ljiljana Progovac (personal communication) for pointing this out to me. . These case-matching answers might be similar to short answers in English that show nominative case and tense such as I did. See Ginzburg and Sag (2000), however, for an account for this type of matching effect without ellipsis. . Morgan (1989: 237) points out that while there is an option between no case and the corresponding case in short answers to questions, in the type of nonsentential utterances that do not require a preceding linguistic context such as the ones we saw in (9), the only possibility is absence of case as in (i): (i)
Phyo han-cang (no case) ticket one ‘One ticket!’ (to order a ticket)
. This might be similar to one-word adult focal utterances, such as the following: (i) (ii) (iii) (iv) (v) (vi)
Softer. Sugar? Beautiful. Here. Half. Fire!
Thanks to Barbara Partee (personal communication) for some of these examples. . It is worth pointing out that both (98) and (99) involve two phenomena that are not found in early child language: complex structures involving complementizers and proper interpretation of pronouns. As is well known, children show what has sometimes been referred to as a “Delay of Principle B Effect” and perform poorly on the interpretation of pronouns even up to the age of 6;6 (Grimshaw & Rosen 1990; Chien & Wexler 1987). Thus, they allow pronouns to be bound inside their governing category as in (i). (See Baauw 2000; Chien & Wexler 1990; Grodzinsky & Reinhart 1993, for details.) (i)
Mama Beari is washing heri .
Interestingly, Avrutin (2001) reports that this is also the case for agrammatic aphasics. . They refer to the following works: Schegloff (1991), Fox (1994), and Ono and Thompson (1995). . See Epilogue (this volume) for related issues.
References Aboh, E. O. (2004). The Morphosyntax of Complement-Head Sequences. Clause Structure and Word Order Patterns in Kwa. Oxford: OUP. Akmajian, A. (1979). Aspects of the Grammar of Focus in English. New York, NY: Garland.
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.26 (1692-1806)
Eugenia Casielles
Ameka, F. (1992). Focus constructions in Ewe and Akan: A comparative perspective. MIT Working Papers in Linguistics, 17, 1–25. Anagnostopoulou, E., van Riemsdijk, H., & Zwarts, F. (Eds.). (1997). Materials on LeftDislocation. Amsterdam: John Benjamins. Avrutin, S. (2001). Linguistics and agrammatism. GLOT International, 5(3), 87–97. Avrutin, S. (2004). Optionality in child and aphasic speech. Lingue e Linguaggio, 1, 67–97. Baauw, S. (2000). Grammatical Features and the Acquisition of Reference. A Comparative Study of Dutch and Spanish. Utrecht: LOT. Baker, N. D. & Greenfield, P. M. (1988). The development of new and old information in young children’s early language. Language Sciences, 10(1), 3–34. Barton, E. (1990). Nonsentential Constituents: A Theory of Grammatical Structure and Pragmatic Interpretation. Amsterdam: John Benjamins. Barton, E. (1998). The grammar of telegraphic structures: Sentential and nonsentential derivations. Journal of English Linguistics, 26, 37–67. Barton, E. & Progovac, L. (2005). Nonsententials in minimalism. In R. Elugardo & R. J. Stainton (Eds.), Ellipsis and Nonsentential Speech (pp. 71–93). Dordrecht: Kluwer. Bates, E. (1976). Language and Context: The Acquisition of Pragmatics. New York, NY: Academic Press. Boškovi´c, Ž. (1997a). The Syntax of Nonfinite Complementation. Cambridge, MA: The MIT Press. Boškovi´c, Ž. (1997b). Pseudoclefts. Studia Linguistica, 51, 235–277. Bowerman, M. (1973). Early Syntactic Development: A Cross-Linguistic Study with Special Reference to Finnish. Cambridge: CUP. Carston, R. (2002). Linguistic meaning, communicated meaning and cognitive pragmatics. Mind and Language, 17, 127–148. Casielles, E. (1998). FOCUS PREPOSING (it is called). University of Massachusetts Occasional Papers, 21, 51–64. Casielles, E. (2004). The Syntax-Information Structure Interface: Evidence from Spanish and English. London: Routledge. Casielles, E., Andruski, J., Kim, S., Nathan, G., & Work, R. (2006). Syntactic and discourse features of subjects in child Spanish: Evidence from Spanish/English bilingual acquisition. In D. Bamman, T. Magnitskaia, & C. Zaller (Eds.), Proceedings of the 30th Annual Boston University Conference on Language Development (pp. 72–83). Somerville, MA: Cascadilla. Chien, Y.-C. & Wexler, K. (1987). Children’s acquisition of the locality condition of pronouns and reflexives. Papers and Reports on Child Language Development, 26, 30–39. Chien, Y.-C. & Wexler, K. (1990). Children’s knowledge of locality conditions in binding as evidence for the modularity of syntax and pragmatics. Language Acquisition, 1, 225–295. Chomsky, N. (1977). On Wh-movement. In P. W. Culicover, T. Wasow, & A. Akmajian (Eds.), Formal Syntax (pp. 71–132). New York, NY: Academic Press. Chung, S., Ladusaw, W., & McCloskey, J. (1995). Sluicing and logical form. Natural Language Semantics, 3, 239–282. Collins, P. C. (1991). Cleft and Pseudo-Cleft Constructions in English. London: Routledge. Culicover, P. W. & Jackendoff, R. (2005). Simpler Syntax. Oxford: OUP. den Dikken, M. (2006). Specificational copular sentences and pseudoclefts. A case study. In M. Everaert & H. van Riemsdijk (Eds.), The Blackwell Companion to Syntax, Volume IV (pp. 292–409). Oxford: Blackwell. den Dikken, M., Meinunger, A., & Wilder, C. (2000). Pseudoclefts and ellipsis. Studia Linguistica, 54, 41–89.
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.27 (1806-1923)
Big questions, small answers
Erteschik-Shir, N. (1997). The Dynamics of Focus Structure. Cambridge: CUP. Fox, B. A. (1994). Contextualization, indexicality, and the distributed nature of grammar. Language Sciences, 16, 1–38. Ginzburg, J. (1996). Interrogatives: Questions, facts and dialogue. In S. Lappin (Ed.), The Handbook of Contemporary Semantic Theory (pp. 385–422). Oxford: Blackwell. Ginzburg, J. & Sag, I. (2000). Interrogative Investigations: The Form, Meaning, and Use of English Interrogatives. Stanford, CA: CSLI. Greenfield, P. M., Reilly, J., Leaper, C., & Baker, N. (1985). The structural and functional status of single-word utterances and their relationship to early multi-word speech. In M. Barrett (Ed.), Children’s Single-Word Speech (pp. 233–267). London: Wiley. Greenfield, P. M. & Smith, J. H. (1976). The Structure of Communication in Early Language Development. New York, NY: Academic Press. Grice, P. (1975). Logic and conversation. In P. Cole & J. L. Morgan (Eds.), Syntax and Semantics, Vol. 3, Speech Acts (pp. 41–58). New York, NY: Academic Press. Grimshaw, J. (1977). English Wh-constructions and the Theory of Grammar. Ph.D. dissertation, University of Massachusetts, Amherst. Grimshaw, J. (1979). Complement selection and the lexicon. Linguistic Inquiry, 10, 279–326. Grimshaw, J. (1991). Extended projection. Ms., Brandeis University. Grimshaw, J. & Rosen, S. T. (1990). Knowledge and obedience: The developmental status of the binding theory. Linguistic Inquiry, 21, 187–222. Grodzinsky, J. & Reinhart, T. (1993). The innateness of binding and coreference. Linguistic Inquiry, 24, 69–102. Groenendijk, J. & Stokhof, M. (1989). Type-shifting rules and the semantics of interrogatives. In G. Chierchia et al. (Eds.), Properties, Types and Meaning, Volume II: Semantic Issues (pp. 21–68). Dordrecht: Kluwer. Grohmann, K. (2003). Prolific Domains: On the Anti-Locality of Movement Dependencies. Amsterdam: John Benjamins. Guasti, M. T. (2002). Language Acquisition: The Growth of Grammar. Cambridge, MA: The MIT Press. Haegeman, L. (1988). Register variation in English: Some theoretical observations. Journal of English Linguistics, 20, 230–248. Haegeman, L. (2000). Adult null subjects in non pro-drop languages. In M.-A. Friedemann & L. Rizzi (Eds.), The Acquisition of Syntax (pp. 129–169). London: Longman. Hamblin, C. L. (1973). Questions in Montague Grammar. Foundations of Language, 10, 41–53. Reprinted in (1976) B. H. Partee (Ed.), Montague Grammar (pp. 247–259). New York, NY: Academic Press. Hankamer, J. (1979). Deletion in Coordinate Structures. New York, NY: Garland. Hausser, R. R. (1983). The syntax and semantics of English mood. In F. Kiefer (Ed.), Questions and Answers (pp. 97–158). Dordrecht: Reidel. Heim, I. (1979). Concealed questions. In R. Bauerle et al. (Eds.), Semantics from Different Points of View (pp. 51–60). Berlin: Springer. Heim, I. & Kratzer, A. (1998). Semantics in Generative Grammar. Oxford: Blackwell. Higginbotham, J. (1996). The Semantics of questions. In S. Lappin (Ed.), The Handbook of Contemporary Semantic Theory (pp. 361–383). Oxford: Blackwell. Higgins, F. R. (1979). The Pseudo-Cleft Construction in English. New York, NY: Garland. Jackendoff, R. (2002). Foundations of Language: Brain, Meaning, Grammar, Evolution. Oxford: OUP. Kartunnen, L. (1977). Syntax and semantics of questions. Linguistics and Philosophy, 1, 3–44.
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.28 (1923-2033)
Eugenia Casielles
Krifka, M. (1991). A compositional semantics for multiple focus constructions. Semantics and Linguistic Theory, 1, 127–158. Lambrecht, K. (2001). A framework for the analysis of cleft constructions. Linguistics, 39, 463– 516. Laury, R. & Ono, T. (2005). Data is data and model is model: You don’t discard the data that doesn’t fit your model! Language, 81, 218–225. Leonard, L. B. & Schwartz, R. G. (1978). Focus characteristics of single-word utterances after syntax. Journal of Child Language, 5, 151–158. MacWhinney, B. (1984). Grammatical devices for sharing point. In R. Schiefelbusch & J. Pickar (Eds.), The Acquisition of Communicative Competence (pp. 323–374). Baltimore, MA: University Park Press. Merchant, J. (2001). The Syntax of Silence. Oxford: OUP. Merchant, J. (2004). Fragments and ellipsis. Linguistics and Philosophy, 27, 661–738. Montague, R. (1973). The proper treatment of quantification in ordinary English. In K. J. J. Hintikka et al. (Eds.), Approaches to Natural Language (pp. 221–242). Dordrecht: Reidel. Reprinted in Montague, R. 1974, 247–270. Morgan, J. (1973). Sentence fragments and the notion ‘sentence.’ In B. Kachru et al. (Eds.), Issues in Linguistics (pp. 719–751). Urbana, IL: University of Illinois Press. Morgan, J. (1989). Sentence fragments revisited. In B. Music, R. Graczyk, & C. Wiltshire (Eds.), Chicago Linguistic Society 25. Parasession on Language in Context (pp. 228–241). Chicago, IL: Chicago Linguistic Society. Ono, T. & Thompson, S. A. (1995). What can conversation tell us about syntax? In P. W. Davis (Ed.), Alternative Linguistics: Descriptive and Theoretical Models (pp. 213–271). Amsterdam: John Benjamins. O’Shea, M. V. (1907). Linguistic Development and Education. New York, NY: Macmillan. Partee, B. H. (1986). Noun phrase interpretation and the type-shifting principles. In J. Groenendijk et al. (Eds.), Studies in Discourse Representation Theory and the Theory of Generalized Quantifiers (pp. 115–143). Dordrecht: Foris. Partee, B. H. (1991). Topic, focus and quantification. Semantics and Linguistic Theory, 1, 159– 187. Partee, B. H. & Rooth, M. (1983). Generalized conjunction and type ambiguity. In R. Baüerle et al. (Eds.), Meaning, Use and the Interpretation of Language (pp. 362–383). Berlin: de Gruyter. Prince, E. F. (1981). Topicalization, focus-movement and Yiddish-movement: A pragmatic differentiation. Berkeley Linguistics Society, 7, 249–264. Radford, A. (1990). Syntactic Theory and the Acquisition of English Syntax. Oxford: Blackwell. Radford, A. (1997). Syntactic Theory and the Structure of English: A Minimalist Approach. Cambridge: CUP. Radford, A. (2000). Children in search of perfection: Towards a minimalist model of acquisition. Essex Research Reports in Linguistics, 34, 57–74. Rizzi, L. (1994). Early null subjects and root null subjects. In T. Hoekstra & B. Schwartz (Eds.), Language Acquisition Studies in Generative Grammar (pp. 151–176). Amsterdam: John Benjamins. Rizzi, L. (2000). Remarks on early null subjects. In M.-A. Friedemann & L. Rizzi (Eds.), The Acquisition of Syntax (pp. 269–292). London: Longman. Roeper, T. (1999). Universal bilingualism. Bilingualism: Language and Cognition, 2(3), 169–186. Rooth, M. (1985). Association With Focus. Ph.D. dissertation, University of Massachusetts, Amherst.
JB[v.20020404] Prn:25/07/2006; 12:33
F: LA9305.tex / p.29 (2033-2075)
Big questions, small answers
Schegloff, E. A. (1991). Conversation analysis and socially shared cognition. In L. B. Resnick, J. M. Levine, & S. D. Teasley (Eds.), Perspectives on Socially Shared Cognition (pp. 150–171). Washington, DC: American Psychological Association. Schütze, C. & Wexler, K. (1996). Subject case licensing and English root infinitives. In A. Stringfellow, D. Cahana-Amitay, E. Hughes, & A. Zukowski (Eds.), Proceedings of the 20th Annual Boston University Conference on Language Development (pp. 670–681). Somerville, MA: Cascadilla. Seuren, P. A. M. (1985). Discourse Semantics. Oxford: Blackwell. Sinclair de Zwart, H. (1973). Language acquisition and cognitive development. In T. Moore (Ed.), Cognitive Development and the Acquisition of Language (pp. 9–25). New York, NY: Academic Press. Stainton, R. J. (1998). Quantifier phrases, meaningfulness ‘in isolation’ and ellipsis. Linguistics and Philosophy, 21, 311–340. Vallduví, E. (1990). The Informational Component. Ph.D. dissertation, University of Pennsylvania, Philadelphia. Ward, G. (1988). The Semantics and Pragmatics of Preposing. New York, NY: Garland. Wexler, K. (1998). Very early parameter setting and the unique checking constraint: A new explanation f the optional infinitive stage. Lingua, 106, 23–79. Yanofsky, N. (1978). NP utterances. In D. Farkas, W. Jacobsen, & K. Todrys (Eds.), Chicago Linguistic Society 14 (pp. 491–502). Chicago, IL: Chicago Linguistic Society.
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.1 (49-117)
chapter
Extending the nonsentential analysis The case of special registers* Kate Paesani This chapter highlights nonsentential properties in special registers and extends the syntactic analysis developed in Progovac (Chap. 2, this volume) to account for them. Certain special register data exhibit the co-occurrence of tenselessness and caselessness typical of nonsententials in conversational adult speech. These data are analyzed as base-generated phrasal projections and root small clauses that lack the syntactic Tense node. Data from English headlines and cooking recipes provide the bulk of the evidence for this analysis; however, cross-linguistic data from other languages provide additional empirical support. This chapter also examines some unexpected properties of tense, aspect, and negation in certain special registers and argues that the expression of time in headlines, in particular, provides further support for the nonsentential analysis.
.
Introduction
This chapter analyzes special register data that exhibit clear nonsentential properties: missing or incomplete tense on lexical verbs, the absence of copular/auxiliary verbs, bare nouns in subject position, and the co-occurrence of tenselessness and caselessness. English serves as the basis for this analysis; however, data from other languages, including Arabic, Dutch, French, German, Spanish, and Serbian, are supplied to support certain claims. Some of these data, in particular those characteristic of headlines and recipes, will be argued to follow from the syntactic analysis proposed in Progovac (Chap. 2, this volume). Progovac notes three important tendencies present in a nonsentential grammar: (a) the systematic absence of Tense, (b) the systematic absence of nominative case, and (c) the systematic correlation between the absence of Tense and the absence of nominative case. Due to these facts, Progovac argues that nonsentential utterances lack the syntactic Tense node (TP), which necessitates the absence of structural Case assignment on subjects. According to this correlation, nonsentential subjects do not receive structural Case marking, and thus need not be DPs. The absence of structural Case can be signaled by default case (accusative case in English), or by the (possible) absence of determiners. Nonsententials are thus analyzed as basegenerated phrases (e.g., VP, PP, AP) that do not project a TP. If these phrases have a
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.2 (117-197)
Kate Paesani
subject in their specifier position, they are analyzed as root small clauses that also do not project a TP. The primary goal of the chapter is to explore how much data in which special registers may be accounted for by the nonsentential analysis. Nonsentential properties present in special registers are identified, with a particular emphasis on the co-occurrence of tenselessness and caselessness, as illustrated in the English headline data in (1) through (3): (1) Huskies dominant inside and out
(ESPN.com, April 6, 2004)
(2) Proof in the pudding (3) Young inmate disciplined after game
(Metro Times, March 26, 2004) (Detroit Free Press, March 4, 2005)
As pointed out by Progovac and other authors in this volume, the co-occurrence of the absence of Tense and the absence of nominative case on subjects is typical of nonsentential utterances, and provides evidence for a base-generated analysis of these utterances as phrases and small clauses.1 Other nonsentential properties found in conversational contexts among typical adult speakers (see Barton, Chap. 1, this volume; Progovac, Chap. 2, this volume) are present in isolation in certain special registers. These include the absence of copular and auxiliary verbs (4), the absence of tense on lexical verbs (5), the absence of determiners in subject position (6), and the absence of subjects (7). (4) Ireland waking up to Woods’ kicking prowess
(Engel 2000)
(5) Cook and drain tortellini as directed on package (6) Leader takes a bus to summit (7) Am ill
(Engel 2000) (Barton 1998)
An additional characteristic of certain special registers is the absence of objects (8).2 (8) Mix together and refrigerate overnight for a few hours. Finally, some special registers provide evidence for the default time interpretation characteristic of nonsententials. In some cases, the default time frame is present, as in the cooking recipes in (5) and (8), in other cases, the default time frame is past, as in the headlines in (1) through (3). All of these properties will be explored in detail throughout the chapter. It should be assumed from the outset that none of the special registers considered in this chapter is required to use nonsententials; just as in adult conversation, special registers may alternate between full sentences and nonsententials. Therefore, the nonsentential grammar is considered an optional alternative in special registers; the claims made herein are thus specific to individual occurrences of properties of this grammar within the special registers considered. The chapter is organized as follows: Section 2 provides a brief overview of the nonsentential properties of special registers and the treatment of these properties in
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.3 (197-259)
The case of special registers
previous literature; Section 3 argues that some special registers are more strongly characterized by nonsentential properties than others; Section 4 considers clear cases of nonsententials in headlines and recipes, and develops a syntactic analysis to account for them; Section 5 extends the nonsentential analysis to root infinitives and the expression of time in headlines; finally, Section 6 presents general conclusions.
. An overview of special registers Registers are language varieties used to communicate for specific purposes. The special register contexts considered in this chapter are headlines, cooking recipes (a subset of the instructional register), telegrams, diaries, note taking, labels, and titles. Register, or field of discourse, is defined by Quirk, Greenbaum, Leech, and Svartvick (1985) as follows: The field of discourse is the type of activity engaged in through language. A speaker of English has a repertoire of varieties according to field and switches to the appropriate one as occasion requires. The number of varieties that speakers command depends upon their profession, training and interests. (23)
Ferguson (1982: 57–58) identifies five assumptions with respect to register variation: (9) a. b. c. d. e.
Register variation is universal; Registers exist; Register systems differ cross-linguistically and change diachronically; A given register is variable in the degree of its distinctiveness; Competence in register variation is acquired as part of language development.
Special registers differ from the nonsententials typical of conversational adult speech in two ways. First, special registers are used to communicate for specific purposes. Generally, the switch from one register to another requires turning to a set of lexical items and grammatical correlates specific to the field or register in question. For example, in English cooking recipes, one typically uses a set of lexical items (e.g., to stir, to mix, to chop) in the imperative form (e.g., stir mixture, mix all ingredients, chop onions).3 Second, special registers are characteristic of written rather than oral discourse. As such, they are subject to certain restrictions (e.g., length, style). The discussion that follows is organized based on the typical nonsentential properties found across special registers: the absence of tense, determiners, and subjects. A fourth characteristic, the absence of objects, is discussed at length in much of the previous literature on special registers, and is thus overviewed here.4 These properties appear to be productive in some special registers, yet are subject to certain syntactic and semantic restrictions.5 The data presentation is coupled with a discussion of how these phenomena are treated in previous research. This research tends to provide elliptical analyses of nonsentential properties, relying on full-sentence sources and using
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.4 (259-338)
Kate Paesani
constructs like exceptional empty categories to account for the absence of subjects and objects, for instance. The theoretical and empirical shortcomings of these elliptical analyses are highlighted throughout the discussion, and are revisited in Sections 3 and 4, which argue in favor of a nonsentential analysis of certain forms typical of some special registers.
. Absence of Tense Like nonsentential examples from conversational adult speech, special registers often exhibit utterances that lack a copula, an auxiliary, or inflection on the lexical verb, suggesting that at least some registers do not access Tense in all cases. This property is present in several special register contexts, including diaries, note taking, telegrams, and headlines. Examples (10) and (11) illustrate the absence of a copula in the English diary and note-taking registers, respectively.6 (10) Audience good (11) Word order pretty loose
(Williams 1993: 200–201) (Janda 1985: 445)
The absence of copulas is particularly prevalent in telegraphese and headlinese and seems to be available across languages.7 The telegraphese data below illustrate the absence of a copula in English (12) and German (13).8 (12) Car broken down (13) Auto defect car defective
(Barton 1998: 50, 56)
Additionally, (10) through (13) illustrate co-occurrence of a bare noun in subject position and the absence of a copula in the same utterance. The following English (14) and French (15) data illustrate headlines that lack a copula. (14) Proof in the pudding (15) EDF prête à solder sa coûteuse aventure en Italie ‘EDF ready to sell off its costly venture in Italy’
(Le Figaro, May 3, 2005)
Headlines may also lack an auxiliary in passive (16) and progressive (17) constructions, as the following English examples illustrate. (16) Failed vasectomies claim rejected (17) Embassy siege: 79 facing charges Once again, the absence of Tense in (14) through (16) co-occurs with a bare noun in subject position. (The absence of the determiner in subject position will be correlated to the absence of Tense through a generalization regarding structural Case in Section 4.)
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.5 (338-387)
The case of special registers
In the constructions in (10) through (17), a predication relation exists between the subject and the following adjective, preposition, or verb phrase. Vinet (1993) argues that in headlines this predication is not realized syntactically by Tense (TP), but rather by Aspect (AspP). In her analysis, the subject of the predication is realized in Spec,AspP, and the adjective or preposition moves to the head of AspP to create a predication relationship. Vinet argues that evidence for AspP in headlines comes from constructions in which the predicated Aspect head (adjective or preposition) is modified by an adverb that aspectually marks the predication relation between the subject and the head (e.g., L’actrice Sophia Loren encore belle ‘The actress Sophia Loren still beautiful’). The projection of AspP in Vinet’s analysis, however, is not independently motivated. Progovac (Chap. 2, this volume) points out that although AspP may be a relevant projection for some English nonsententials, without some sort of independent motivation, AspP is not a defining property of their syntax. Finally, cooking recipes, which are typified by imperative (18) and infinitive (19) constructions across languages, exhibit impoverished tense or the absence of tense inflection on lexical verbs.9 The following examples illustrate cooking recipe data from French and Spanish, respectively. (18) Otez le bout terreux des champignons. remove-imp the end earthy of-the mushrooms ‘Remove the stem from the mushrooms.’ (19) Lavar las patatas con su piel. wash-inf the potatoes with their skin ‘Wash the potatoes with their skin (on)’ In each of these examples, tense is not overtly realized on the verb. These data thus suggest that cooking recipes lack or may not have full access to Tense and that the syntactic realization of inflection may be different than in full sentential correlates. Indeed, the imperative is often considered to be morphologically meager cross-linguistically (cf. Platzack & Rosengren 1998; Sadock & Zwicky 1985; Zhang 1990). Zhang, for instance, found that none of the 46 languages he surveyed displays tense features in the imperative paradigm. Nevertheless, Zhang argues that imperatives are TPs, in spite of the lack of overt tense features in the paradigm. Akmajian (1984), on the other hand, argues that imperatives lack an AUX (or TP) node, and Progovac (Chap. 2, this volume) suggests that at least English imperatives may not merge beyond VP. Platzack and Rosengren argue that a lack of Tense in Germanic imperatives falls out from a syntactic analysis in which imperatives lack FinP in the C-domain, and thus TP and the feature [finite]. A consequence of this analysis is that imperatives also lack a prototypical subject. (See Potsdam (1995, 1998) for an alternative analysis in which imperatives are argued to have finite features.)
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.6 (387-451)
Kate Paesani
. Absence of Determiners Nouns with missing determiners are common in special registers, and occur in both subject and object position.10 English data from the diary (10), note-taking (11), telegraphese (12), and headline (1)–(3) registers presented earlier illustrate bare nouns in subject position.11 Bare nouns are an available property of French headlines (20) as well, although full-DP subjects are also prevalent (21). (20) Risque d’avalanche maximal ce weekend ‘Risk of avalanche maximal this weekend’
(Engel 2000)
(21) Un employé légèrement contaminé ‘An employee slightly contaminated’
(Engel 2000)
In (21), the determiner may be present to highlight that one of several employees was contaminated; whereas in (20), there is a general avalanche risk that need not be specified. Note also that in (20), the bare noun in subject position co-occurs with the absence of Tense. Bare nouns occur in other French special registers, usually when the noun occurs on its own or as part of a list.12 The data in (22) and (23) illustrate the absence of a determiner in labels and titles, respectively. (22) a.
Moutarde forte au vin blanc ‘Strong mustard with white wine’ b. eau, graines de moutarde, vin blanc, sel, épices ‘water, mustard seeds, white wine, salt, spices’
(23) a. Législation électorale fédérale b. La législation électorale fédérale13 ‘(The) federal electoral legislation’
(Curat 1999: 200)
Bouchard (2002), following Curat (1999), claims that to satisfy conditions on semantic and pragmatic reference, and to be able to refer on their own without a determiner, French bare nouns must fall into one of the six categories in (24). (24) a. b. c. d. e. f.
Contact of the word with the referent Deictic NPs Proper names A single determiner and coordinated Ns Determinerless coordinated Ns De + pronominal adjective
(Bouchard 2002: 290–297)
Bouchard argues that “the referential potential of these [determinerless] NPs arises not from atomization, but from some restricted semantic/pragmatic conditions” (2002: 289). In the case of labels and titles, the bare noun is in direct contact with its referent:
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.7 (451-519)
The case of special registers
When the N is in direct contact with the referent or its representation, there is no need to have a Det with Number to induce atomization in order to pick a particular referent from the class of N: the direct contact already identifies a referent. . . [W]hen the N is an actual label on an object, its graphic representation is stuck onto the referent or its container. (Bouchard 2002: 290)
Indeed, pragmatic context, specifically the physical presence of the object, appears to obviate the need for the determiner in (22) and (23) in some way.
. Absence of Subjects A characteristic of some special registers is an inflected verb lacking an overt subject. This property is typical of the diary register, as illustrated in the following English (25), Dutch (26), French (27), and German (28) examples.14 (25) Telephoned Louie (26) Hoorde over de TV dat Franco dood zou zijn. . . ‘Heard on the TV that Franco should be dead’ (27) Revient à l’affaire Alb. . . Me demande si devant la menace. . . ‘Returns to the affair Alb. . . Asks me if in front of the threat. . . ’ (28) Beschloss, ihn zu kaufen, äderte diesen Beschluss ‘Decided to buy it, changed this decision’
(Haegeman 1990: 164–188)
The empty subject in diary contexts tends to be interpreted as a first- or third-person singular or plural pronoun.15 Haegeman and Ihsane (1999) postulate that the lack of second-person interpretation for the empty subject is due to the fact that the writer rarely addresses someone directly in diary writing. Empty subjects also appear in the note-taking register (29) and in telegraphese (30).16 (29) Want to capture this generalization (30) Am ill
(Janda 1985: 445) (Barton 1998)
According to Janda (1985), the empty subject in note taking usually corresponds to a subject pronoun (rather than a full NP or DP). In his corpus, although a number of subject pronouns were overt, omission of the subject was significantly more common. Barton (1998) argues that the empty subject in telegraphese examples like (30) is the result of an ellipsis rule deleting first-person subjects from full-sentence sources. She claims that the presence of the inflected verb is evidence that these phrases are sentential and are thus associated with a Tense node (or S node in Barton’s analysis) in the syntax. Her principle of Functional Deletion, in which all and only those functional categories that are recoverable from context are deleted, accounts for the missing subject in (30).
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.8 (519-583)
Kate Paesani
Subject omission is found in embedded clauses in diaries (31), note taking (32), and telegrams (33), although the absence of subjects in this syntactic context appears to be restricted (see Haegeman & Ihsane 2001).17 (31) Wonder if ever make it so?
(Haegeman & Ihsane 1999: 128)
(32) What did was take “sha” and. . .
(Janda 1985: 445)
(33) Regret cannot attend
(Barton 1998: 54)
Haegeman (1997, 2000) argues that in the diary register, null subjects in main clauses are empty categories that occur in the highest NP position (Spec,IP) of the root clause containing the inflected verb.18 Haegeman argues that English null subjects in the diary register are not preceded by any position that may host an antecedent. Such null subjects can only survive in a context where no linguistic antecedent is available; when they occur in the highest position possible in a root clause, this condition on antecedence is met. According to this analysis, then, reference of null subjects in the diary register is determined by the discourse context. A problem for this root clause analysis is that it predicts that empty subjects are not allowed in embedded clauses, since the matrix clause may provide the syntactic possibility for an antecedent. Yet null subjects in embedded clauses in English diary contexts are attested, as illustrated in (31). To account for such data, Haegeman and Ihsane (1999, 2001) argue that there are two dialects of the diary register: the majority dialect, in which subject omission is only allowed in matrix clauses, and for which a strictly root clause analysis is sufficient; and the minority dialect (mainly manifested in fictional writing), in which subject omission is allowed in matrix and embedded contexts, and for which a root clause analysis is insufficient.19 To account for the minority dialect, Haegeman and Ihsane develop an ellipsis analysis, in which the diary register is equivalent to a full-sentence grammar, but for which economy considerations lead to ellipsis of pronouns that can be recovered from the discourse context.
. Absence of Objects Although nonsententials in conversational adult speech are not typically characterized by the absence of objects, this characteristic is common in some special registers, in particular cooking recipes.20 The following examples illustrate object omission in English (34), French (35), German (36), Serbian (37), and Spanish (38) cooking recipes. (34) Rinse Ø with cold water. [Ø = tortellini] (35) Salez Ø, poivrez Ø, laissez suer Ø sur feu doux 10 minutes. [Ø = onions] ‘Salt, pepper, and let sweat on a low flame for 10 minutes.’ (36) Äpfel in feine Würfel schneiden und Ø mit dem Zitronensaft verrühren. [Ø = apples] ‘Cube the apples and mix with the lemon juice.’
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.9 (583-651)
The case of special registers
(37) Péci Ø u tepsiji dobro namazanoj maš´cu i posutoj brašnom. [Ø = cake] ‘Bake in pan well greased with lard and sprinkled with flour.’ (38) Retirar Ø y dejar enfriar Ø . [Ø = potatoes] ‘Remove and allow to cool.’ In each of these examples, the antecedent is a nonovert topic; context (linguistic or nonlinguistic) determines the referent of the null object (cf. Haegeman 1987; Zwicky 1990). This context is usually the list of ingredients at the start of the recipe, and the first mention of the particular ingredient in the instructions. No repetition of the ingredient (topic) is required after it is first introduced into the discourse context (cf. Sadock 1974). Culy (1996: 102) notes that objects are optionally absent in recipes: “The most important conclusion to be drawn concerning the form of the object [in recipes] is that zero anaphors and pronouns perform the same functions. The choice of zero versus pronoun in contemporary cookbooks is largely one of style.” French exhibits exactly this kind of variation between overt versus null objects. Indeed, in some contexts, French favors no object drop, using instead the object pronoun.21 (39) Trancher les deux aubergines. . . Les poser sur la plaque du four. . . les imbiber d’huile d’olive et les faire dorer quelques minutes sous le gril en les retournant délicatement. Les réserver. ‘Slice the two eggplants. . . Place them on a baking sheet. . . Sprinkle them with olive oil and brown them for a few minutes under the broiler, turning them delicately. Set them aside.’ Massam and Roberge (1989) and Haegeman (1987) claim that the absence of objects in the recipe register corresponds to an empty category in the syntax.22 Massam and Roberge argue that missing objects in recipes must be represented syntactically as some kind of empty category, either a variable (wh-trace) or an NP trace, bound by a null discourse topic. They point out that this object receives a specific rather than an arbitrary interpretation, and that this interpretation is not dependent on a linguistic antecedent, but rather on the discourse context. Similarly, Haegeman (1987) argues that the implied objects in cooking recipes must be represented structurally; they can control subjects of infinitival clauses (40) and can be modified by adjuncts (41). (40) Allow to set. (41) Serve covered by the vegetables.
(Haegeman 1987: 237)
Because these objects are syntactically present, yet covert, Haegeman argues that they are empty categories with the properties of a wh-trace. The antecedent for the wh-trace is a nonovert, sentence-peripheral topic position. Haegeman supports her analysis with research by Huang (1984) and Raposo (1986), in which Chinese, Japanese, Korean, and Portuguese are argued to have topic-linked object drop as well. Haegeman further argues, following Huang’s idea that topic-prominent languages are closely
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.10 (651-715)
Kate Paesani
linked to discourse orientation, that the recipe register “exhibits characteristics of a topic prominent and discourse oriented language” (1987: 245).
. Summary As the preceding data illustrate, certain special registers possess the properties typical of nonsententials in conversational adult speech: absence of tense, absence of determiners in subject position, and absence of subjects. These properties appear to be available to all of the special registers surveyed here to varying degrees. In addition, some special registers, in particular cooking recipes, exhibit the absence of objects. Interestingly, some special registers exhibit the co-occurrence of tenselessness and caselessness that is characteristic of nonsententials in conversational adult speech. Some of these co-occurrence phenomena provide clear evidence for the availability of a nonsentential grammar in certain special registers, and fall out from the syntactic analysis developed in Progovac (Chap. 2, this volume). These phenomena are outlined next, and discussed in more detail in Section 4, where a syntactic analysis of the data is developed. Data from the headline (42), telegram (43), and note-taking (44) registers provide strong support for the correlation between tenselessness and caselessness noted by Progovac (Chap. 2, this volume). (42) Proof in the pudding (43) Battery dead (44) Word order pretty loose In each of these examples, utterances that lack a verb (tenselessness) may co-occur with a bare noun in subject position (caselessness). In some cases, this co-occurrence is available cross-linguistically (e.g., in (13), (15), and (20)).23 Other clearly nonsentential data are instances in which a nonfinite verb co-occurs with a bare noun in subject position or with no subject at all. In general, these data include progressives (45) and passives (or participials) (46). (45) a.
Alaskan volcano showing signs of erupting (Detroit Free Press, December 16, 2005) b. Pondering a wall and trying not to hit one (New York Times, December 19, 2005)
(46) Young inmate disciplined after game These data provide further support for the co-occurrence of tenselessness and caselessness – given that neither of the verbs in (45) or (46) is inflected for tense – and thus for the nonsentential analysis. The co-occurrence of tenselessness and caselessness may also be linked to the absence of objects in recipes (47).
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.11 (715-775)
The case of special registers
(47) Rinse with cold water. In such cases, the unavailability of structural Case assignment may not only explain bare nouns in subject position or the absence of subjects, it may also explain their co-occurrence with the absence of objects. Other tendencies in the data are less straightforward and may provide less clear support for the nonsentential analysis developed in Progovac (Chap. 2, this volume). These data typically involve tensed verbs. In headlines, for instance, there are examples that seem to be finite and fully inflected, but that nonetheless involve a default time interpretation. (48) Deserted boy, two, dials 999
(Engel 2000)
In (48), the tensed verb is present by form, but is interpreted as past, given the obvious pragmatic context of headlines reporting on past events. The importance of the default expression of time in such examples suggests that a nonsentential grammar may still play a role, although not in a straightforward fashion.24 The syntactic properties presented throughout this section reveal that there is a range of data available across special registers. Some of these data are very clearly nonsentential, providing strong evidence for the Tense-Case correlation noted by Progovac (Chap. 2, this volume); other data are less clearly nonsentential, exhibiting overt Tense and Case in certain contexts. To explain this variation, it appears that special registers have access to a continuum of possibilities with respect to sentential and nonsentential grammars, and that some special registers more readily resort to one grammar or the other, while others are more difficult to classify. I argue for this possibility in detail in Section 3 before moving on to a syntactic analysis of the data in Section 4.
. The (non)sentential nature of special registers As the discussion thus far has shown, a variety of syntactic characteristics typify certain special registers. Certainly, the degree to which a register exhibits these properties varies. Ferguson (1982; see (9)) claims that a given register is variable in the degree of its distinctiveness. In other words, variation is evident within a single register. This type of variation is noted by Haegeman and Ihsane (2001) for the diary register, for example. They claim that there exists a majority diary dialect and a minority diary dialect, and that this classification accounts for variation in the type and frequency of features across diary contexts. Within a special register, then, the property of being nonsentential holds of individual utterances; some special registers seem to resort to these nonsentential utterances more easily than others. Ferguson’s claims about variation may be extended beyond a single register: not only does there exist register-internal variation, but cross-register variation exists as well. Registers, then, may vary from one another in their degree of distinctiveness, where distinctiveness can be measured in terms of degree of similarity to a full-sentence grammar. Following this
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.12 (775-846)
Kate Paesani
argumentation, it is possible to claim that those registers that more easily resort to a sentential grammar are less distinctive, whereas those registers that more easily resort a nonsentential grammar are more distinctive. Syntactically, the diary register appears to resort to a sentential grammar more easily than the headline and recipe registers, for instance. Although diaries exhibit some of the properties that characterize a nonsentential grammar such as null subjects and the absence of tense inflection, rarely do these properties co-occur. Indeed, the diary register is typified by utterances having a null subject and a tensed verb. Similarly, the note-taking register, although characterized by more nonsentential properties than the diary register, appears to be less distinctive than headlines and recipes. The note-taking register exhibits most of the nonsentential properties outlined earlier, yet, like the diary register, does not productively exhibit the co-occurrence of tenselessness and caselessness typical of nonsententials in casual adult speech. The headline and recipe registers, on the other hand, appear to be more distinctive because they resort to a nonsentential grammar more easily than do the diary and note-taking registers. Headlines and recipes are typified by the types of base-generated phrases and root small clauses that form the basis of the nonsentential analysis proposed by Progovac (Chap. 2, this volume). In particular, both registers provide strong support for the correlation between tenselessness and caselessness characteristic of adult nonsententials. One constant across the syntactic characteristics presented in Section 2 is the clear link between discourse/pragmatic context and missing elements in the syntax. Identification of subjects, objects, determiners, and tense is most commonly with a nonlinguistic antecedent. Indeed, the research surveyed here highlights the important role of discourse context in the interpretation and licensing of each of these features. Yet even at the level of pragmatics and discourse, a distinction is made between those special registers that tend to exhibit more sentential properties and those that tend to exhibit more nonsentential properties: it appears that the more closely a register is linked with its discourse/pragmatic context, the more it accesses the nonsentential grammar. Indeed, the headline and recipe registers are highly discourse dependent. This idea is consistent with the arguments developed in Barton (1998) for telegraphese, and Progovac (Chap. 2, this volume) for casual adult speech. According to Barton, “a nonsentential analysis, particularly one that points toward a robust pragmatic component for interpretation, offers a satisfactory explanation for the grammatical structure and pragmatic interpretation of . . . telegraphese” (1998: 57–58). And Progovac claims that nonsentential utterances are highly context dependent, in particular for establishing temporal reference and making assertions, and thus may establish reference through nonlinguistic means. The idea of pragmatic robustness and of nonlinguistic reference may be extended to offer further support for the fact that headlines and recipes more readily access a nonsentential grammar. I therefore propose, based on the syntactic and pragmatic evidence, that there are two types of special registers: those that have a preference for sentential properties and whose typical examples are derived from a full-sentence grammar, and those that have a preference for nonsentential properties and whose typical examples are
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.13 (846-907)
The case of special registers
derived from a nonsentential grammar. The registers considered in this chapter form a sort of sentential-to-nonsentential continuum, with diaries and note taking falling more clearly on the sentential end, and recipes and headlines falling more clearly on the nonsentential end. Of the registers considered here, the diary register most easily resorts to a sentential grammar, given that it drops topical and expletive subjects (in main and embedded clauses), yet usually maintains tensed verbs. The note-taking register falls more toward the middle of the continuum; this register exhibits a variety of nonsentential properties, yet is also characterized by embedded clauses and a lack of productive co-occurrence phenomena. Headlines and recipes, on the other hand, not only resort to each of the nonsentential properties illustrated in Section 2, they frequently exhibit the co-occurrence of tenselessness and caselessness characteristic of the nonsentential grammar.25 Moreover, these registers rarely if ever allow embedding (see the Epilogue for a discussion of (the lack of) embedding in nonsentential speech). The idea of positing two grammars to account for special registers is not novel. Barton (1998) proposed, and ultimately rejected (cf. Barton, Chap. 1, this volume; Barton & Progovac 2005), nonsentential and deletion analyses for a variety of telegraphese data (e.g., Battery dead versus Am ill), and Casielles (Chap. 5, this volume) argues that a nonsentential analysis accounts for some question-answer pair phenomena (e.g., Who did it?, Him), while an ellipsis analysis accounts for others (e.g., Who did it?, He did). Indeed, it seems intuitive to argue that two different grammars may account for variation within and across registers. Merchant (2004: 732), for instance, proposes that speakers generally command several “domain-specific grammars,” one or more of which may account for variation in special registers. Speakers thus access a sentential grammar to produce and comprehend some utterances in certain registers, and a nonsentential grammar to produce and comprehend others. This reasoning is consistent with Roeper’s (1999) concept of Universal Bilingualism. Roeper argues that adult speakers are bilingual; they have access to a full-sentence grammar as well as to one or more grammars that are incompatible with the full-sentence grammar in one way or another. Indeed, speakers possess “a set of mini-grammars for different domains” (1999: 169), and resort to these grammars depending on the discourse context.26 The focus of the remainder of this chapter will be on the data summarized in Section 2.5: the co-occurrence of tenselessness and caselessness, and the expression of default time. I develop a syntactic analysis of these data, with a particular focus on headlines and recipes.
. The syntax of nonsententials in special registers As illustrated in Sections 2 and 3, some of the data typical of certain special registers provide evidence for single phrase projections and root small clauses of the type proposed in Progovac (Chap. 2, this volume). These data are analyzed as structures smaller than a full sentence, that is, structures that lack a Tense Phrase (TP). Due the lack of
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.14 (907-937)
Kate Paesani
Tense, nonsentential subjects do not receive structural Case marking, and thus need not be DPs. The absence of structural Case can be signaled by default case (accusative case in English), or by the (possible) absence of determiners. Nonsententials in special registers are thus argued to be base-generated phrases that lack a TP or, if these phrases include a subject in their specifier position, as root small clauses. This analysis explains the following empirical facts about nonsententials in special registers, facts that mirror casual adult speech: the absence of tense on lexical verbs, the absence of copular and auxiliary verbs, the possible absence of determiners on subjects, and the co-occurrence of tenselessness and caselessness. Furthermore, the analysis may potentially be extended to account for the co-occurrence of missing objects with the absence of Tense and structural Case assignment. A sentential analysis cannot predict these data because it cannot explain the absence of articles on subjects, and thus the lack of structural Case assignment, nor can it explain the correlation between tenselessness and caselessness. Data from headlines and recipes, special registers that more readily resort to a nonsentential grammar, provide the bulk of the evidence for the analysis developed here. Examples of root small clauses in headlines are presented as evidence for a nonsentential analysis, followed by a potential extension of the nonsentential analysis to account for the absence of objects in cooking recipes.
. A root small clause analysis English headlines are characterized by root small clauses that show the co-occurrence of bare nouns in subject position and missing copular or auxiliary verbs. The data in (49) and (50) provide clear evidence of the Tense-Case correlation noted by Progovac (Chap. 2, this volume); in both of these examples, the headline is marginal or ungrammatical with the addition of a tensed verb (49b), (50b) or a subject with a determiner (49c), (50c). (49) a. Huskies dominant inside and out27 b. ??Huskies were dominant inside and out c. *The Huskies dominant inside and out (50) a. Proof in the pudding b. ??Proof is in the pudding c. *The proof in the pudding Following the nonsentential analysis, these data are analyzed as root small clauses that do not merge to TP: (49a) is an example of a small clause AP and (50a) is an example of a small clause PP.28 The absence of a determiner on the subject is directly related to the absence of Tense. TP assigns structural Case; when TP is absent, so is structural Case assignment to the subject. The tree structures in (51) and (52) illustrate the syntax of the nonsententials in (49a) and (50a).
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.15 (937-992)
The case of special registers
AP
(51) Huskies
A’
A
PP
dominant
inside and out
PP
(52) Proof
P’ P
NP
in
the pudding
As argued in Progovac, TP and DP are interrelated in syntactic theory. A DP is visible as a subject argument when it is assigned abstract (nominative) Case; TP then checks the abstract Case of the subject DP. As a result of this interrelationship, utterances that lack Tense cannot check structural Case. The absence of structural Case may be manifested through the use of default case (accusative case in English), or through the (possible) omission of determiners on subjects. The root small clauses in (51) and (52) lack finite tense, and thus may lack determiners on their subjects. The French headline examples in (53) and (54) provide cross-linguistic evidence for the root small clause analysis. In (53a) and (54a), the copular/auxiliary verb is absent. Because Tense is absent, structural Case assignment is not possible. These examples therefore have a bare noun in subject position. Vinet (1993) also notes that the presence of a bare noun in subject position in French headlines is directly linked to the fact that these nouns coincide with the absence of Tense. She further points out that in tensed constructions, determiners are required; the tensed examples in (53b) and (54b) are thus infelicitous with a bare noun in subject position because structural Case is not realized. (53) a.
Risque d’avalanche maximal ce weekend ‘Avalanche risk maximal this weekend’ b. *Risque d’avalanche est maximal ce weekend Irak: enfants en danger ‘Iraq: children in danger’ b. *Enfants sont en danger en Irak ‘Children are in danger in Iraq’
(Engel 2000)
(54) a.
(Vinet 1993: 93)
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.16 (992-1038)
Kate Paesani
The French headlines in (53a) and (54a) are thus analyzed as root small clauses that do not merge to TP.29 These adjectival and prepositional small clauses mirror the English headlines in (49) and (50), and are represented structurally in (55) and (56). AP
(55)
A’
Risque d’avalanche
(56)
A
NP
maximal
ce weekend
PP Enfants
P’ A
NP
en
danger
Negation facts provide further evidence that nonsententials in French headlines do not merge to TP. Adjectival root small clauses and phrases allow the negative adverb pas (57a), (58a) but not the negative clitic ne (57b), (58b). (57) a.
Le Premier Ministre Bourassa pas très en forme ‘Prime Minister Bourassa not in good shape’ b. *Le Premier Ministre Bourassa ne pas très en forme Féal, pas vassal ‘Fearful, not servile’ b. *Féal, ne pas vassal
(Vinet 1993: 88)
(58) a.
(Le Monde, January 2, 2006)
Following Pollock (1989), Vinet (1993) argues that examples like (57b) and (58b) are ungrammatical because ne must move to Tense where it cliticizes to the verb. (57a) and (58a), on the other hand, are grammatical because pas is not required to move to Tense; it may adjoin to the adjective phrase over which it has scope. The ungrammaticality of ne in (57b) and (58b), then, is evidence that NegP and thus TP are not projected in the structure, and that the negated APs in (57a) and (58b) are base-generated nonsententials. This kind of evidence provides clear support for a nonelliptical analysis over an elliptical analysis. Adjectival root small clauses are semantically restricted in French: adjectival predicates indicating permanent or individual traits are not permitted (cf. Carlson 1977; Vinet 1993). Vinet points out the following examples, where the permissible data in (59) are in clear contrast with the ungrammatical data in (60).
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.17 (1038-1084)
The case of special registers
(59) a.
Marie Curie victorieuse ‘Marie Curie victorious’ b. La chanteuse Céline Dion malade ‘The singer Celine Dion sick’
(60) a. *Marie Curie intelligente ‘Marie Curie intelligent’ b. *La chanteuse Céline Dion québécoise ‘The singer Celine Dion Québécoise’
(Vinet 1993: 86)
Vinet’s observation also holds true for adjectival root small clauses for which there is evidence of the correlation between tenselessness and caselessness (i.e., a bare noun in subject position and a missing auxiliary/copular verb). The headline data in (61) and (62) mirror Vinet’s data in (59) and (60). (61) a.
Participation française incertaine ‘French participation uncertain’ b. Risque d’avalanche maximal ce weekend ‘Avalanche risk maximal this weekend’
(Le Monde, January 2, 2006)
(62) a. *Participation française intelligente ‘French participation intelligent’ b. *Risque d’avalanche permanent ce weekend ‘Avalanche risk permanent this weekend’ This contrast appears to be due to a distinction between stage-level (59), (61) and individual-level (60), (62) predicates.30 Stage-level predicates are eventive in nature, focusing more on the “here and now” typical of nonsententials, whereas individuallevel predicates are stative, focusing on states that held at a point prior to the speech time, and continue to hold at a time beyond the speech time.31 The root small clause analysis of nonsententials in headlines also explains the cooccurrence of bare nouns in subject position with nonfinite verbs. Data like those in (3) and (45a), for example, are simply analyzed as small clause VPs. The following tree structures illustrate. VP
(63)
V’
Young inmate V
PP
disciplined
after game
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.18 (1084-1142)
Kate Paesani
(64)
VP Alaskan volcano
V’
V
NP
showing
signs of erupting
Interestingly, there exists a subset of data, particularly prevalent in the headline register, in which a tenseless phrase co-occurs with a DP subject. (65) A Down Payment on New Orleans
(New York Times, December 19, 2005)
This type of construction seems to be more prevalent in French than in English, as evidenced in (21), (57), and (59). At first sight, these data are problematic for Progovac’s (Chap. 2, this volume) analysis, which argues for the unavailability of structural Case assignment due to the lack of a TP node in the syntax. Yet, Progovac, following Longobardi (1994), argues further that in addition to their formal function as indicators of Case, determiners also have a referential function (see also Section 2.2). It may be that in headline examples like (21), (57), (59), and (65), the determiner is used solely for referential purposes, rather than for formal, Case-assigning purposes. This marked use of the determiner might help explain why examples like (65) are rare in English: given that determiners are not required in positions where structural Case is not assigned, determiners should be used much less often in subjects of nonsententials than in subjects of sententials. If this argument is on the right track, then nonsententials in which a tenseless phrase co-occurs with a determiner + noun in subject position do not necessarily violate the Tense-Case correlation crucial to the analysis developed here.
. The Tense-Case correlation and the absence of objects in cooking recipes As was noted in Section 2, missing objects are a common characteristic of the recipe register. The absence of objects in this context may be linked to the co-occurrence of tenselessness and caselessness characteristic of the nonsentential grammar. Recipes are typified by nontensed verbs, as illustrated in the English (5), French (18), and Spanish (19) data. These data, characterized by imperative and infinitive constructions, suggest that because tense is not overtly realized on the verb, and because subjects are not realized, the Tense node may be absent from the syntax. The research summarized in Section 2.1 supports this suggestion for imperative constructions: Akmajian (1984), Platzack and Rosengren (1998), and Progovac (Chap. 2, this volume) all argue that imperatives lack TP. Interestingly, there appears to be a link between the co-occurrence of tenselessness and caselessness and the absence of objects in cooking recipes. Indeed, the data in (34)
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.19 (1142-1225)
The case of special registers
through (38), repeated here, show that object omission is attested in cooking recipes cross-linguistically. (66) Rinse with cold water. (67) Salez, poivrez, laissez suer sur feu doux 10 minutes. ‘Salt, pepper, and let sweat on a low flame for 10 minutes,’ (68) Äpfel in feine Würfel schneiden und mit dem Zitronensaft verrühren. ‘Cube the apples and mix with the lemon juice.’ (69) Péci u tepsiji dobro namazanoj maš´cu i posutoj brašnom. ‘Bake in pan well greased with lard and sprinkled with flour.’ (70) Retirar y dejar enfriar. ‘Remove and allow to cool.’ In each of these examples, a missing object co-occurs with a missing subject and a tenseless verb (imperative in (66) and (67) and infinitive in (68)–(70)). In English, objects tend to be focal in nature; however, in recipes, objects appear to be topical.32 Objects are introduced in the list of ingredients, and then are mentioned in the instructions the first time they play a role in the recipe; they therefore need not be mentioned again in the discourse and can be dropped. Moreover, not only do recipes presume that the object is known from the list of ingredients, they also presume that the ingredients are physically present as one carries out the instructions. This discourse context is thus not unlike that of labels, which also presume the physical presence of the object.33 Haegeman (1987) also argues that object drop in recipes is topic linked; the antecedent of a missing object is a sentence-peripheral topic position. Massam (1992) points to the link between missing subjects and missing objects in cooking recipes, and claims that the missing object is linked to an empty discourse topic in the subject position (Spec,IP) for imperatives. That recipes are topic linked is intuitive given the important discourse dependence of this special register. Huang (1984) has argued that object drop in languages such as Chinese, Japanese, and Korean may in part be explained by the discourse orientation of these languages. Chinese, Japanese, and Korean are topic prominent, whereas languages like English and French are subject prominent. Huang argues that there is a link between topic prominence and discourse orientation and subject prominence and sentence orientation. Given that English recipes are characterized by topic prominence, as the data suggest, then this special register may be discourse oriented (Haegeman 1987). If this suggestion is on the right track, then it may also be the case that nonsentential grammars are discourse oriented in general. Given the preponderance of imperative and infinitive constructions in the recipe register, and the fact that the pragmatic frame for recipes is the “here and now,” the lack of a Tense node may create the syntactic context for missing objects, as it does for the absence of structural Case assignment on subjects. Massam and Roberge (1989) note a link between the absence of subjects and objects in cooking recipes, stating that “it is difficult to tell whether the imperative aspect or the no-subject aspect of the sentences
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.20 (1225-1277)
Kate Paesani
is relevant to the [missing object]” (135). They point out the following contrasts with respect to the absence of subjects and objects in recipes: (71) a.
First, you take 2 eggs, ½ cup of flour and three tablespoons sugar. You beat them well and you cook the mixture for 5 minutes. Then you serve it while it is still warm. b. First, Ø take 2 eggs, ½ cup of flour and three tablespoons sugar. Ø Beat them well and Ø cook the mixture for 5 minutes. Then Ø serve it while Ø still warm. c. First, Ø take 2 eggs, ½ cup of flour and three tablespoons sugar. Ø Beat Ø well and Ø cook Ø for 5 minutes. Then Ø serve Ø while Ø still warm. d. *First, you take 2 eggs, ½ cup of flour and three tablespoons sugar. You beat Ø well and you cook Ø for 5 minutes. Then you serve Ø while it is still warm.
The recipe in (71a) provides the base example, where both overt subjects and overt objects are present. (71b) shows that null subjects may co-occur with overt objects, and (71c) shows that null subjects may co-occur with missing objects. In (71d), however, overt subjects co-occur with missing objects, a configuration that is ungrammatical. There is thus a one-way correlation between the absence of objects and the absence of subjects: if the object is omitted, then the subject must also be omitted, but not vice versa. An explanation of the data in (71) may be linked to the Tense-Case correlation, and, specifically, to the absence of structural Case assignment.34 If structural Case is not assigned to the object position, then neither the object pronoun nor the determiner is necessary. This one-way correlation may follow from the nonsentential analysis if one assumes that once the verb and its object are merged without structural Case features, then the derivation must proceed as a nonsentential rather than a sentential one. The subject, when merged, therefore lacks structural Case features as well. This one-way relationship is similar to the generalization made by Stowell (1991) regarding the omission of determiners in subject and object position in headlines (see n. 11). The pattern of determiner omission in (72) is similar to that for subject and object omission in (71). (72) a. b. c. d.
An old man finds a rare gold coin Old man finds a rare gold coin Old man finds rare gold coin *An old man finds rare gold coin
The headline data in (72) show a one-way correlation between the omission of determiners on subjects and objects: if the determiner is omitted on the object, then it is also omitted on the subject, but not vice versa. Given that the verb in (72) is inflected for present tense, however, these data do not follow straightforwardly from the nonsentential analysis and I leave their specific analysis open for future research.35 Other research supports the claim that cooking recipes, in particular those characterized by imperative constructions, lack Tense, and that tenselessness is linked to both
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.21 (1277-1328)
The case of special registers
the absence of subjects and the absence of objects. Visser (1996), for instance, points out the syntactic link between nonfiniteness and the licensing of object drop in imperatives in Dutch. To capture the fact that imperatives do not show tense differentiation, Visser claims that these constructions lack a TP. She further assumes that imperative subjects in Spec,vP (if overt) receive default Case since they cannot receive nominative Case due to the absence of Tense. Finally, Visser argues that empty objects get reference from a discourse operator that is adjoined to CP. Pirvulescu and Roberge (1998) claim that in imperative constructions, the complement (object) position has a special status that is directly linked to the absence of Tense. Assuming that nonsententials in cooking recipes, like those in casual adult speech, are associated with the here and now, it is possible to argue that recipes receive temporal interpretation from the discourse context, rather than from a syntactic Tense node. In addition, Pirvulescu and Roberge argue that the imperative “provides an impoverished predicative context for the lexical verb, which in turn triggers a ‘shift’ toward the complement position” (1998: 223). This “shift” explains the special status of the complement, and the option of omitting the object in specific contexts. Pirvulescu and Roberge propose a v–VP configuration for imperatives, in which an implicit subject is the subject of the light verb.36 The facts observed in the recipe register – the co-occurrence of tenselessness, caselessness, and missing objects – seem to be linked to the absence of syntactic Tense characteristic of the nonsentential grammar. In addition, the co-occurrences typical of cooking recipes may be related to the fact that both the subject (i.e., the person reading the recipe) and the objects (i.e., the list of ingredients) are known. As a result, what minimally needs to be expressed is the action performed by the subject on the objects. If it is indeed the case that nonsententials in recipes lack syntactic Tense, then they, like the headline data analyzed in Section 4.1, are base-generated phrasal projections of the type proposed in Progovac (Chap. 2, this volume). These phrases may only project as high as VP (or to vP, as in Pirvulescu & Roberge’s (1998) analysis), or to a functional category beyond VP but below TP. I leave the specific details of the syntactic analysis open for further research. What the discussion here seeks to illustrate is the significance of the empirical evidence to co-occurrence of nontensed verbs, missing subjects, and missing objects in recipes as support for the nonsentential analysis.
. Expressing time in special registers It is generally assumed that the typical default time frame for nonsententials is present or “here and now.” This default time frame certainly applies to the recipe register, as pointed out in the previous section. This register is typified by imperative and infinitive constructions across languages (see (18) and (19)). Because tense inflection is absent, and because these constructions lack TP, temporal reference is achieved nonlinguistically through context. Moreover, because the default time frame for nonsententials is generally present, then cooking recipes are interpreted as having present-time ref-
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.22 (1328-1395)
Kate Paesani
erence. Indeed, Progovac (Chap. 2, this volume) argues that Tense is not required to anchor a predication in time. Instead, time can be established independently through pragmatic/discourse context. Yet the default “here and now” time frame may be overridden by adverbials, for instance, or by some other time frame that typifies a particular discourse context. Such is the case with headlines, which appear to have a default time frame of past given the obvious pragmatic situation of reporting about past events.37 As with other nonsententials for which the default time is “here and now,” the default time frame for headlines is established independently of a Tense node through extra-linguistic context. The absence of syntactic Tense explains why root small clause headlines are characterized by missing copular or auxiliary verbs (see Section 4.1), as illustrated in (1) through (3), and repeated here. (73) Huskies dominant inside and out (74) Proof in the pudding (75) Young inmate disciplined after game In these examples, Tense is absent, and past time is established through context. It is in fact the context of reporting about past events, typical of newspapers, as well as the link between the information relayed in the headline and the information reported in the newspaper article itself, that creates the pragmatic frame of past time. Interestingly, even when overt tense is expressed in headlines, the pragmatic frame is still past time. For instance, use of the present tense in English headlines usually corresponds to a historical present interpretation.38 (76) Deserted boy, two, dials 999 (77) Blair tells pro-labour MPs to sell electoral reform
(Engel 2000)
That the headlines in (76) and (77) have a past interpretation is reinforced by the content and time frame of the article itself. Indeed, Engel (2000) cites a portion of the article text for each of the headlines in (76) and (77), which further establishes their temporal location in the past. (78) A frightened two-year-old dialed 999 after he was left home alone. (79) At a meeting . . . last week Mr. Blair told the backbenchers to go out and evangelise. . . In headlines like (76) and (77) Tense is not required for temporal reference, which is established contextually. It is unclear, however, whether these headlines are required to merge as far as TP for syntactic reasons. The purpose of this discussion is to highlight the saliency of the default time interpretation. I thus leave the precise syntactic analysis of these data open as an area for future research. In some English headlines, the simple past is used to express tense overtly, as illustrated in (80) and (81).
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.23 (1395-1462)
The case of special registers
(80) They used our babies and never told us (81) Woman: Jackson plotted smear
(Engel 2000) (Detroit Free Press, March 4, 2005)
According to Engel (2000), the simple past is used to report direct (80) and indirect (81) discourse. Moreover, the use of the simple past may underscore the importance of the information conveyed by the quote. The difference between the use of the simple past and the absence of copular/auxiliary verbs may in fact be an aspectual one. Simple past is a clear expression of perfective aspect, which focuses on the culmination point (or end point) of an event. Headlines that lack an inflected verb, on the other hand, seem to express perfect aspect. The perfect differs from the perfective in that it is associated with a consequent state (or result) that begins when the event culminates and holds for a time period after the event’s culmination.39 When the copula or auxiliary is absent from headlines, this consequent state holds until (or beyond) the time the newspaper article is published, and thus links the event with the present. In (75), for instance, the culmination point of the inmate being disciplined results in certain consequences or changes in behavior; these consequences or changes are the consequent state that holds up to the present time. In contrast, when headlines occur with the simple past, as in (80) and (81), there is a focus on the culmination point of the event, rather than on a consequent state.40 Although a larger database of headlinese would lend further support to this argument, the perfect-perfective distinction could provide insight into the syntactic analysis of headlinese. When headlines lack an inflected verb, the temporal frame is past, and the aspectual frame is perfect. It may be that in addition to having a default time frame, headlines also have a default aspectual frame. This default aspectual frame is overridden when headlines occur with the simple past; whereas the default time frame remains past, the aspectual frame, determined linguistically by the verb form, shifts to perfective. The default past time frame of headlines may be overridden by linguistic mechanisms. To report future events, for instance, English headlines use to-infinitives, as the following data illustrate. (82) Hazardous waste site to open near I-94 (Detroit Free Press, December 27, 2005) (83) Israeli leader to leave hospital on Tuesday (New York Times, December 19, 2005) The use of to-infinitives in examples like (82) and (83) is not simply an indication of what might happen in the future, but rather the indication of a fixed plan. The implication is that the decision to carry out the future action has been made, and all that remains is for that action to be carried out at some point after the time the headline appears in print.41 This interpretation of to-infinitives requires that headlines refer to the future from a past pragmatic frame; the decision to carry out the action was made in a past (default) time frame, and the execution of the action will happen in a future time frame. Avrutin (1997) makes similar claims about to-infinitives in English
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.24 (1462-1529)
Kate Paesani
headlines, arguing that the infinitival particle to locates the event in time. The event time is future (nonpast and nonpresent), and will occur after the current moment in time (i.e., the time of publication). That to locates the event in time explains why it cannot be omitted in the headline register. (84) *Hazardous waste site open near I-94 (85) *Israeli leader leave hospital on Tuesday Moreover, Avrutin argues, “there seems to be (at least intuitively) a presupposition of the existence of another event (usually described in the article under the headline) that has already taken place and that is somehow relevant for the event described by the headline” (1997: 76). This type of temporal interpretation appears to be restricted to headlines; indeed, in casual adult speech, a to-infinitive would not be used to refer to a fixed plan to be carried out in the future. The equivalent of to-infinitives to express future time is also evident in French headlines. According to Engel (2000), French infinitives express future when they are preceded by the preposition à, as illustrated in (86). (86) Des arbres à abattre dans le Gers some trees to fell-inf in the Gers ‘Trees to fell in Gers’ Chinese employs a similar mechanism to express future in headlines. The construction jiang + verb, illustrated in the Chinese headlines in (87) and (88), is roughly equivalent to a to-infinitive. (87) CAAC jiang jinxing chunji hangban tiaozheng CAAC will carry-out spring flight adjustment ‘CAAC will carry out adjustments on spring flights’ (88) Huabei jiang dafeng jiangwen Northern-China will big-wind drop-temperature ‘Northern China will have big winds and temperature will drop’ This construction does not occur outside of the headline context. In fact, when jiang is omitted, the headline reverts back to the default past time frame. Jiang is therefore a specific form used to express future time and override the default past time frame.42 The linguistic mechanism used to override the default past time frame of headlines must be expressed in the syntax. In English, it is the infinitival particle to that linguistically overrides this time frame and provides a future interpretation. The fact that similar types of constructions are used cross-linguistically to express future time in headlines suggests that some sort of node beyond VP is activated for these nonsententials. Yet, because future time is expressed through the particle to rather than through verbal inflection, Tense may be absent from to-infinitives. The absence of structural Case, as suggested by the presence of bare noun subjects in (82) and (83), provides further evidence for this claim. These data suggest that the co-occurrence of tense-
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.25 (1529-1597)
The case of special registers
lessness and caselessness characteristic of a nonsentential grammar is also present in headlines that resort to the to-infinitive construction to express future time. Avrutin (1997) highlights constraints on subjects in Russian root infinitives that are mirrored in English headlines. Russian root infinitives, or “Princess sentences,” are illustrated in (89) and (90). (89) Carevna xoxotat’ princess to-laugh ‘The Princess started to laugh’ (90) Zriteli applodirovat’ spectators to-applaud ‘The spectators started to applaud’
(Avrutin 1997: 66)
These constructions are similar to to-infinitives in headlines in that they are uninflected for Tense and they lack structural Case assignment on the subject.43 In addition, Russian root infinitives may only occur with stage-level predicates (89), (90); individual-level predicates are impossible in these constructions (91). (91) *Carevna znat’ anglijskij princess to-know English ‘The Princess started to know English’ This contrast appears to hold for English to-infinitives as well. (92) a. President to visit Russia b. *President to be fat (93) a. President to speak Russian b. *President to know Russian
(Adapted from Avrutin 1997)
The data in (92a) and (93a) illustrate that to-infinitives may occur with stage-level predicates, whereas the data in (92b) and (93b) illustrate that they cannot occur with individual-level predicates. Avrutin argues that this contrast is related to the absence of overt Tense inflection. Indeed, headlines like (92b) and (93b) improve when they are associated with an inflected verb. (94) President will be fat (95) President will know Russian This contrast may be accounted for by the idea, developed in Section 4.1, that stagelevel predicates are eventive in nature, and thus more closely linked to the pragmatic/discourse context than individual-level predicates. Given the robust role played by context across nonsententials, this link seems plausible. Moreover, a strong link to the discourse/pragmatic context seems to correlate in some way with the absence of a syntactic Tense node. To account for the role of to in establishing temporal reference, and for the cooccurrence of tenselessness and caselessness, I adopt Hyde’s (2000) analysis of to-
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.26 (1597-1684)
Kate Paesani
infinitives in complement positions. In such constructions, Hyde analyzes to as a preposition that takes a bare infinitive (VP) complement.44 The following structure illustrates: (96) [PP to [VP PRO [V’ verb [ XP ]]]] This structure may account for to-infinitives that occur without a lexical subject, as in (97), or with a lexical subject, as in (98). (97) Agnes tried [to win the race]. (98) Agnes expected [Bill to win the race]. The to-infinitive clause in (98) is identical to the root infinitive headline data in (82) and (83). I therefore argue, following Progovac (Chap. 2, this volume), that infinitive structures of the type in (98) may occur in isolation as root small clauses. This claim is supported by the co-occurrence of tenselessness and caselessness evidenced in toinfinitives and typical of nonsententials in general. This analysis accounts nicely for negation in to-infinitives.45 Consider the following data: (99) a. President will not visit Russia. b. *President not will visit Russia. (100) a. *President to not visit Russia. b. President not to visit Russia. These data show that while negation in English cannot precede clear Tense elements, such as the modal verb in (99b), it can precede the infinitive particle to. These data support the argument that to, and thus the infinitive, are not in Tense. Negation in these constructions may possibly occur in an adjoined position inside PP (see Progovac, Chap. 2, this volume); TP is thus not required to account for these negated nonsententials.
. Conclusion This chapter has presented data from a variety of special registers across languages as support for the nonsentential analysis developed in Progovac (Chap. 2, this volume), and has provided an alternative to previous research on special registers. Indeed, the claims made in this chapter regarding the distinctiveness of certain special registers, negation facts, the default time frame in headlines, and the extension of the analysis to objects in recipes, resolves many of the problems present in the previous literature. A number of properties typical of certain special registers – the absence of tense on lexical verbs, the absence of copular/auxiliary verbs, the absence of determiners on subjects, and the role of default time – support the analysis developed herein and the basic claim that nonsententials are base-generated phrases and root small clauses that
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.27 (1684-1717)
The case of special registers
lack a syntactic Tense node. English headlines, in particular, provide clear examples of root small clauses, and negation data from French and English lend strong empirical support to the analysis. These data show that the ungrammaticality of the negative clitic ne, the placement of the negative adverb not, and thus the absence of NegP follow directly from the absence of syntactic Tense. The co-occurrence of tenselessness and caselessness in certain special register contexts also follows from the analysis developed here. Because Tense assigns structural Case, and because special registers may lack structural Case on subjects (as evidenced by the possible absence of determiners), then these constructions lack TP. Finally, temporal reference in special registers supports the analysis developed here. Whereas some special registers have the default time frame of “here and now” typical of nonsententials in casual adult speech, other special registers have a different default time frame, such as the past frame of headlines. This default time frame, whether past or present, is determined by the discourse/pragmatic context, and is not dependent on syntactic Tense for interpretation. The claims made in this chapter regarding negation and the default time frame, as well as the extension of the analysis to objects in recipes, resolve many of the problems present in previous literature on special registers. This analysis of nonsententials is consistent with Chomsky’s (1995) economy of representation: to be economical, the syntax must project a minimal amount of structure. The base-generated phrases and root small clauses typical of a nonsentential analysis do just this. An elliptical analysis cannot account for the nonsentential properties typical of certain special registers in this unified and economical way. Given that the nonsentential grammar typical of some special registers is economical and less explicit, it may be an example of Roeper’s (1999) Minimal Default Grammar (MDG), where “the set of MDG structures reflects principles of economy. That is, they project fewer nodes than elaborated particular grammars” (173). The forms that exist in the MDG are often incompatible with the sentence grammar (or elaborated particular grammar). The choice of MDG is directly linked to context; if the features used are contextually clear and comprehensible to the hearer (or reader), then the speaker (or writer) may choose to use the MDG. Adults are able to produce and comprehend the forms of the MDG because they are bilingual; that is, they can manipulate both nonsentential and full-sentence grammars. If this conclusion is correct, then strong parallels may be drawn between the nonsentential grammar used in certain special register contexts and other potential MDG varieties, including child language (see Potts & Roeper, Chap. 7, this volume), adult learner varieties (see Work, Chap. 8, this volume) and agrammatic aphasic speech (see Kolk, Chap. 9, this volume; Siple, Chap. 10, this volume).
Notes * My deepest gratitude is extended to my coeditors, Ljiljana Progovac, Eugenia Casielles, and Ellen Barton, who helped shape this chapter and commented on its many drafts. I am also grate-
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.28 (1717-1779)
Kate Paesani
ful to the other members of the Humanities Center Nonsentential Working Group at Wayne State University: Walter Edwards, Haiyong Liu, Pat Siple, and Nicola Work. Much of the data presented in this chapter were collected in collaboration with these group members. Earlier versions of this chapter were presented at the two workshops sponsored by this working group (Paesani 2003, 2004), and I am grateful to the participants for their input. Finally, my thanks go to the two reviewers of this chapter, whose detailed comments contributed significantly to its improvement. . Throughout this chapter, I use the term nonsentential to refer to nonsentential utterances and to reflect a commitment to a nonsentential analysis via base generation. See also the Introduction for further discussion of terminology. . Missing objects are not noted by Progovac (Chap. 2, this volume) as a salient feature of nonsententials. They are included in this list of features because missing objects figure prominently in previous research on special registers. The data in (4) through (8) come from the following special registers: (4) is an example of the diary register; (5) and (8) are examples of the recipe register; (6) is an example of the headline register; and (7) is an example of the telegram register. . Infinitival forms are also characteristic of cooking recipes, particularly in languages such as French, German, Spanish, and Serbian. See Sections 2.1 and 4.2 for further discussion. . Each of these features has also been noted in child language. See, for example, Radford (1990). . Again, it is important to emphasize that in this chapter, tendencies to use certain nonsentential features in certain special registers are highlighted; generalizations regarding the productivity of nonsentential features in special registers are not possible without significant statistical data. . In a study of be-deletion in Virginia Woolf ’s diary, Ihsane (1998) found that be was absent from 184 out of 363 examples. . Tesak et al. (1995) found that the copula (understood as be, have, and become) was absent from telegrams in six languages as follows: in German, the copula was absent in 283 (92.2%) tokens; in Dutch, 234 (55.1%) tokens; in Swedish, 343 (76.1%) tokens; in Italian, 372 (96.9%) tokens; in Finnish, 202 (78.0%) tokens; and in Hungarian, 43 (32.1%) tokens. They further found the following results with respect to main verbs: in German, the main verb was absent in 41 (3.2%) tokens; in Dutch, 17 (2.4%) tokens; in Swedish, 12 (2.0%) tokens; in Italian, 12 (4.3%) tokens; in Finnish, 31 (4.8%) tokens; and in Hungarian, 22 (2.4%) tokens. Barton (1998) found that auxiliaries were deleted in 27% of 380 English telegrams surveyed. In a survey of four English-language and four French-language newspapers, Engel (2000) found that, on average, the finite verb (copular, auxiliary, or main verb) was absent from 68.6% and 44.1% of headlines, respectively. . Absence of the copula is also evidenced in Swedish telegrams: (i)
Bilen sonder car-def broken-down
As pointed out by a reviewer, however, this example is problematic due to the -en suffix on the noun, which is often taken to be a D head. Constructions such as these are usually analyzed as evidence for N-to-D movement, thus implying that TP is not absent from these examples, contrary to the nonsentential analysis argued for in this chapter. . The following verb constructions are permitted in cooking recipes in the five languages surveyed: English and French allow imperatives and infinitives; German allows infinitives; Serbian
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.29 (1779-1856)
The case of special registers
allows imperatives and infinitives; and Spanish allows infinitives. Thanks to Ellen Barton for English recipe data, Eugenia Casielles for Spanish recipe data, Ljiljana Progovac for Serbian recipe data, and Nicola Work for German recipe data. . In Tesak et al.’s (1995) telegraphese database, 97.3% of German determiners were absent; 93.8% of Dutch determiners were absent; 35.7% of Swedish determiners were absent; 100% of Italian determiners were absent; and 14.5% of Hungarian determiners were absent. Tesak et al. do not specify the position (subject vs. object vs. other) of these determinerless nouns, however. In other special registers, such as headlines (i) and recipes (ii), bare nouns occur frequently in object position: (i) Villanueva honors aunt with title (ii) Add blueberries. Because bare nouns in object position do not provide strong support for the nonsentential analysis developed in Progovac (Chap. 2, this volume), they are not discussed in this chapter. . Stowell (1991) points out that bare nouns in subject and object position may co-occur in English headlines. (i)
Old man finds rare gold coin
This use of bare nouns in subject and object positions is constrained. A bare noun in subject position may co-occur with a noun plus determiner in object position (ii); however, a bare noun in object position may not co-occur with a noun plus determiner in subject position (iii). (ii) Old man finds a rare coin (iii) *An old man finds rare coin This one-way correlation is revisited in Section 4.2. . French bare nouns are not restricted to special registers, and may occur in full sentences. Curat (1999: 177–179) points out the use of determinerless nouns in numerous contexts, including archaisms (e.g., Ø Chat échaudé craint l’eau froide ‘(A) burned cat fears cold water’), verbal expressions (e.g., Elle perd Ø patience ‘She’s losing patience’), attributive subjects (e.g., Redevenue Ø oiseau, perchée dans un arbre, elle attend ‘Transformed into a bird again, perched in a tree, she waits’), and coordinated nouns (e.g., . . .je cherchais Ø retraite et Ø paix ‘I was looking for rest and relaxation’), among others. . According to Curat (1999), when a title occurs without a determiner, as in (23a), it names the content of the work; when a title occurs with a determiner, as in (23b), it introduces its subject matter. . In a study of subject omission in Virginia Woolf ’s diary, Ihsane (1998) found that 11.83% of all declarative utterances occurred without a subject. Haegeman and Ihsane (1999) surveyed six traditional diaries and three fictional diaries. They found that subject omission in declarative utterances ranged from 2.07% to 59.27% across the authors surveyed. . In the majority of French diary data with missing subjects, the subject is third-person singular. This may be due to the fact that in typical adult conversational contexts, the third-person pronoun may be omitted in certain French utterances. Commonly, this pronoun is impersonal il ‘it’ (whereas omission of the first-person subject pronoun is generally not permitted). (i)
Faut le faire. ‘(It) must be done.’
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.30 (1856-1920)
Kate Paesani
Moreover, a missing subject often co-occurs with an overt pronominal clitic, as illustrated in the following examples: (ii) Puis se colle à moi then self-3sg clings to me ‘Then clings to me’ (iii) De nouveau me tend sa bouche of new to-me-1sg presents her mouth ‘Again offers her mouth to me’
(Haegeman 1997: 249)
. Tesak et al. (1995) found that in German, Dutch and Swedish, the first person singular subject pronoun was absent in 100%, 89.2%, and 92.4% of the telegraphese data surveyed. Barton (1998) found that in English telegrams, the first-person singular subject pronoun was deleted in only 21% of examples. Tesak et al. further found that subject pronouns, other than the firstperson singular, were absent as follows: 45.5% of German examples; 23.0% of Dutch examples; 24.7% of Swedish examples; 67.9% of Finnish examples; and 3.8% of Hungarian examples. . Of the six traditional diaries and three fictional diaries surveyed in Haegeman and Ihsane’s (1999) study, one traditional diary and one fictional diary showed subject omission in embedded clauses. Of the 55 pages of the traditional diary surveyed, 3.66% of embedded clauses lacked an overt subject; of the 70 pages of the fictional diary surveyed, 11.45% of embedded clauses lacked a subject. Moreover, Becquet (2000) notes that when the subject in the embedded clause is absent, the subject of the main clause tends to be absent as well. Indeed, Becquet finds no attested examples of an overt subject in the matrix clause followed by an absent subject in the embedded clause (see also Haegeman 2002; Haegeman & Ihsane 1999, 2001). . In an earlier analysis, Haegeman (1990) argues that null subjects in the diary register are whtraces. For interpretation, the referent of the empty category is a nonovert TOPIC that serves as the antecedent for the wh-trace, and is topicalized to a pre-S A-bar position. . Haegeman and Ihsane (1999) further observe that in the minority dialect, omission of a reflexive pronoun often co-occurs with omission of the pronominal subject. This dependency provides empirical motivation for an analysis of omitted embedded subjects based on spec-head agreement. . Missing objects are also allowed in the English note-taking register: (i)
a. b.
Room acoustics make different Stop using for a while
. French allows imperative and infinitive forms in cooking recipes, as illustrated in (35) and (39), respectively. The choice to use one or the other appears to depend on the style of the author or the work in which the recipe is published (cf. Engel 1997). . See also Culy (1996) for a discussion of analyses of object drop and the utility of an empty category analysis. . A problem raised by both reviewers is the occurrence of proper names, quantifiers, or demonstratives in subject position when the utterance lacks a verb. (i) Duncan in France (ii) Nobody home (iii) This all derivational morphology
(Haegeman 1990: 193) (Metro Times, January 7, 2004) (Janda 1985: 445)
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.31 (1920-1988)
The case of special registers
These examples are potentially problematic because it is unclear whether the subject should be analyzed as a bare noun, or as a DP that requires structural Case assignment. According to Progovac (Chap. 2, this volume), nothing in the proposed analysis of nonsententials excludes the occurrence of proper nouns in subject position. An analysis of these data is beyond the scope of this chapter; the reader is referred to Progovac for further discussion. . It is interesting to note that in (48), the tensed verb co-occurs with a bare noun in subject position. This type of example is problematic for the nonsentential analysis, which assumes the systematic, one-way correlation between tenselessness and caselessness. Equally problematic are examples in which a tensed verb co-occurs with an absent subject, as in (25) through (30). The co-occurrence of tensed verbs and bare noun subjects does not follow from the basic correlation between tenselessness and caselessnes and is a topic of future research. I will thus limit my discussion of examples like (48) to the role of default time only. The reader is referred to Section 5 of this chapter and to Progovac (Chap. 2, this volume) for further discussion of default time, and to analyses by Haegeman (1997, 2000, 2002), Haegeman and Ihsane (1999, 2001), and Rizzi (1994), among others, for a discussion of null subjects with tensed verbs. . Telegraphese also falls on the nonsentential end of the continuum. I refer the reader to Barton (1990, 1991, 1998) and Barton and Progovac (2005) for nonsentential analyses of this register. . One reviewer suggests that the notion of two grammars to account for the continuum of sentential-nonsentential utterances may be similar to truncation analyses for null subjects developed by Rizzi (1994) and Haegeman (1997). In these analyses, the syntactic structure is truncated (i.e., projection ends) such that null subjects are the leftmost constituent in the clause. Structures where both a subject and a verb are absent might be accounted for by truncating at a lower point in the structure. To explain the continuum, the reviewer suggests that there may be variation in the amount of truncation that is applied, with typically nonsentential utterances/registers using more truncation. . As noted by Progovac (Chap. 2, this volume), in English a bare noun must be a single count noun in order to clearly indicate the lack of nominative case assignment. In the case of (49), Huskies refers to a sports team, not to a group of husky dogs, and therefore requires a determiner in the full-sentence correlate (e.g., The Huskies were dominant. . . ). . Similar nonsentential examples are attested in the telegram (i), note-taking (ii), and diary registers (iii). (i) Battery dead (ii) Word order pretty loose (iii) Audience good These data are also analyzed as root small clause APs that do not merge to TP. . Note that in some small clause APs in French, agreement may be realized on the adjective. (i)
Glycine prête à ceuillir ‘Wisteria ready to be picked’
In (i) the adjective prête is marked with feminine agreement morphology. The presence of agreement suggests that the functional projection AgrP may be present in the syntax. If this is the case, TP is still absent from such nonsententials, as evidenced by the co-occurrence of a bare noun in subject position and the absence of Tense. The reader is referred to Progovac (Chap. 2, this volume), where this point is addressed.
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.32 (1988-2052)
Kate Paesani . According to Vinet (1993), adjectival root small clauses are also available in Moroccan Arabic (in normal adult speech), but they are not subject to the same semantic restrictions on predicates as in French; both stage-level (i) and individual-level (ii) predicates are permissible in copulaless phrases. (i)
Ahmed mrïd ‘Ahmed sick’ (ii) Ahmed mtuwwar ‘Ahmed intelligent’ . Progovac (Chap. 2, this volume) makes a similar argument for stage-level and individuallevel predicates in Hebrew. For a discussion of the stage-level/individual-level distinction in Russian and English see Avrutin (1997) and in Haitian Creole see DeGraff (2001). . See Casielles (Chap. 5, this volume) for a discussion of focus in nonsentential questionanswer pairs. See also Section 2.4. . Thanks to Eugenia Casielles (personal communication) for pointing out the link between missing objects and discourse topics, and to Ellen Barton (personal communication) for pointing out the link between recipes and labels. . Thanks to Ljiljana Progovac (personal communication) for pointing out this explanation. . Indeed, as one reviewer points out, a direct correlation between the Case of the object, the Case of the subject, and tenselessness is not straightforwardly captured by syntactic theory. In current theory, it is usually argued that accusative case assignment on objects is linked to a small v head, not to Tense. This argument is supported by the fact that nonfinite verbs in languages like English and French take accusative objects even though nonfinite verbs presumably lack Tense. The ultimate explanation of this co-occurrence is thus a topic for future research. . According to one reviewer, negation facts suggest that, at least for French, a higher structural position may be needed in the syntax because the negative particle ne is available in imperatives. (i)
Ne prenez pas celle-là! neg take-3pl neg this-there ‘Don’t take that one!’
Zanuttini (1996) argues that in languages in which the imperative verb follows a negative particle like ne, the structure has a TP dominated by the negation (adjoined or phrasal). Platzack and Rosengren (1998), on the other hand, argue that in Germanic languages, negation dominates VP (either as an adjoined phrase or a NegP taking VP as its complement). I will continue to assume that, in general, imperatives do not project a TP, and leave the specific analysis of negated imperatives open for future research. . Past may also be the default time frame for other special registers, including telegraphese and diaries. . Engel (2000) discusses the use of the historical present (and other tenses) in English and French headlines. Her survey of headlines in four British newspapers reveals that headlines that lack a copula/auxiliary and headlines in the present tense account for nearly 90% of her data. The results for French newspapers were similar, with French headlines having a strong preference for missing copula/auxiliary (with an overall average of 69%) over the present tense (overall average of 24%). These data support the claim that past is the default time frame for the headline register. (See also Monville-Burston & Waugh (1991) for a discussion of the historical present in French journalistic discourse.)
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.33 (2052-2133)
The case of special registers . See Paesani (2001) for an overview of the semantics of perfect and perfective aspect. . This analysis may also be relevant for present-tense headlines. In (76), for instance, the use of the present tense seems to underscore the present consequence of the boy’s behavior, specifically that it was an amazing feat for a two-year-old boy to have had the wherewithal to dial 999, rather than the culmination point of the event. . English headlines also employ modal will to express future actions, as illustrated in (i) and (ii): (i) Governors contend Medicaid cuts will strain budgets (ii) Family will pay for justice in hit-and-run
(Detroit Free Press, March 4, 2005)
The difference between the use of will and the use of to-infinitives to express future time may be due to immediacy or determinacy. Whereas to-infinitives express a fixed plan, will implies that the event might take place, as in (i), or that it should take place, as in (ii). It should be noted that in (ii) a bare noun occurs in subject position, despite the presence of Tense (in the form of modal will). This fact does not follow from the basic Tense-Case correlation and is a topic of future research (see also n. 24). . Future time may also be expressed with time adverbials in Chinese, as illustrated in (i), although time adverbials used for future reference are not restricted to this context. (i)
GMD mingri fangwen dalu GMD tomorrow visit Mainland ‘The National Party will visit Mainland tomorrow’
Thanks to Haiyong Liu (personal communication) for the Chinese data and insights about the expression of future time in headlines. . Radford (1990) points to the correlation between bare nouns in subject position and infinitives in child language. See also Roeper and Pérez-Leroux (1997). . See Los (2005) for a historical perspective on the prepositional status of infinitival to. . My thanks go to one of the reviewers for pointing out these data.
References Akmajian, A. (1984). Sentence types and the form-function fit. Natural Language and Linguistic Theory, 2, 1–23. Avrutin, S. (1997). Events as units of discourse representation in root infinitives. MIT Working Papers in Linguistics, 12, 65–91. Barton, E. (1990). Nonsentential Constituents: A Theory of Grammatical Structure and Pragmatic Interpretation. Amsterdam: John Benjamins. Barton, E. (1991). Nonsentential constituents and theories of phrase structure. In K. Leffel & D. Bouchard (Eds.), Views on Phrase Structure (pp. 193–214). Dordrecht: Kluwer. Barton, E. (1998). The grammar of telegraphic structures: Sentential and nonsentential derivation. Journal of English Linguistics, 26, 37–67. Barton, E. & Progovac, L. (2005). Nonsententials in Minimalism. In R. Elugardo & R. Stainton (Eds.), Ellipsis and Non-Sentential Speech (pp. 71–93). Dordrecht: Kluwer.
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.34 (2133-2247)
Kate Paesani
Becquet, K. (2000). L’omission du sujet grammatical dans les journaux intimes. DEA-paper, Université Charles de Gaulle, Lille III. Bouchard, D. (2002). Adjectives, Number and Interfaces: Why Languages Vary. Amsterdam: Elsevier. Carlson, G. N. (1977). Reference to Kinds in English. Ph.D. dissertation, University of Massachusetts, Amherst. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: The MIT Press. Culy, C. (1996). Null objects in English recipes. Language Variation and Change, 8, 91–124. Curat, H. (1999). Les déterminants dans la référence nominale et les conditions de leur absence. Geneva: Droz. DeGraff, M. (2001). On the origin of Creoles: A Cartesian critique of Neo-Darwinian linguistics. Linguistic Typology, 5, 213–310. Engel, D. (1997). The recipe for success: Syntactic features of la chronique gastronomique. Journal of French Language Studies, 7(2), 195–207. Engel, D. (2000). Syntaxe à la une: La structure des titres de journaux français et britanniques. The Web Journal of French Media Studies, 3(1): http://wjfms.ncl.ac.uk. Ferguson, C. A. (1982). Simplified registers and linguistic theory. In L. K. Obler & L. Menn (Eds.), Exceptional Language and Linguistics (pp. 49–66). New York, NY: Academic Press. Haegeman, L. (1987). Register variation in English: Some theoretical observations. Journal of English Linguistics, 20, 230–248. Haegeman, L. (1990). Understood subjects in English diaries: On the relevance of theoretical syntax for the study of register variation. Multilingua, 9(2), 157–199. Haegeman, L. (1997). Register variation, truncation, and subject omission in English and French. English Language and Linguistics, 1, 233–270. Haegeman, L. (2000). Adult null subjects in non pro-drop languages. In M.-A. Friedmann & L. Rizzi (Eds.), The Acquisition of Syntax (pp. 129–169). London: Addison, Wesley, and Longman. Haegeman, L. (2002). Nonovert subject pronouns in written English. In S. Scholz, M. Klages, E. Hantson, & U. Römer (Eds.), Language, Context and Cognition. Papers in honour of Wolf Dietrich Bald’s 60th Birthday (pp. 135–149). Munich: Langenscheidt/Longman. Haegeman, L. & Ihsane, T. (1999). Subject ellipsis in embedded clauses in English. English Language and Linguistics, 3, 117–145. Haegeman, L. & Ihsane, T. (2001). Adult null subjects in the non-pro-drop languages: Two diary dialects. Language Acquisition, 9, 329–346. Huang, J. (1984). On the distribution and reference of empty pronouns. Linguistic Inquiry, 1, 531–574. Hyde, B. (2000). The structures of the to-infinitive. Lingua, 110, 27–58. Ihsane, T. (1998). The Syntax of Diaries: Grammar and Register Variation. MA thesis, University of Geneva. Janda, R. D. (1985). Note-taking English as a simplified register. Discourse Processes, 8, 437–454. Longobardi, G. (1994). Reference and proper names: A theory of N-Movement in syntax and Logical Form. Linguistic Inquiry, 25, 609–665. Los, B. (2005). The Rise of the To-Infinitive. Oxford: OUP. Massam, D. (1992). Null objects and non-thematic subjects. Journal of Linguistics, 28, 115–137. Massam, D. & Roberge, Y. (1989). Recipe context null objects in English. Linguistic Inquiry, 20, 134–139. Merchant, J. (2004). Fragments and ellipsis. Linguistics and Philosophy, 27, 661–738.
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.35 (2247-2354)
The case of special registers
Monville-Burston, M. & Waugh, L. R. (1991). Multivalency: The French historical present in journalistic discourse. In S. Fleischman & L. R. Waugh (Eds.), Discourse-Pragmatics and the Verb (pp. 86–119). London: Routledge. Paesani, K. (2001). The Semantics and Syntax of the Passé Surcomposé in Modern French. Ph.D. dissertation, Indiana University. Paesani, K. (2003). Nonsententials in Recipes. Paper presented at the First Wayne State Workshop on Small Utterances, Detroit. Paesani, K. (2004). Small Utterances, Nonsententials, and Small Clauses: Extensions. Paper presented at the Second Wayne State Workshop on Small Utterances, Detroit. Pirvulescu, M. & Roberge, Y. (1998). Objects and the structure of imperatives. In J.-M. Authier, B. E. Bullock, & L. A. Reed (Eds.), Formal Perspectives on Romance Linguistics: Selected Papers from the 28th Linguistic Symposium on Romance Languages (LSRL XXVIII) (pp. 211–226). Amsterdam: John Benjamins. Platzack, C. & Rosengren, I. (1998). On the subject of imperatives: A minimalist account of the imperative clause. Journal of Comparative Germanic Linguistics, 1, 177–224. Pollock, J.-Y. (1989). Verb movement, Universal Grammar and the structure of IP. Linguistic Inquiry, 20, 364–424. Potsdam, E. (1995). Phrase structure and the English imperative. In L. Gabriele & R. Westmoreland (Eds.), Proceedings of the Sixth Annual Meeting of the Formal Linguistic Society of Midamerica (pp. 143–154). Bloomington: Indiana University Linguistics Club. Potsdam, E. (1998). Syntactic Issues in the English Imperative. New York, NY: Garland. Quirk, R., Greenbaum, S., Leech, G., & Svartvick, J. (1985). A Comprehensive Grammar of the English Language. London: Longman. Radford, A. (1990). Syntactic Theory and the Acquisition of English Syntax: The Nature of Early Child Grammars of English. Cambridge: CUP. Raposo, E. (1986). On the null object in European Portuguese. In O. Jaeggli & C. Silva-Corvalán (Eds.), Studies in Romance Linguistics (pp. 373–390). Dordrecht: Foris. Rizzi, L. (1994). Early null subjects and root null subjects. In T. Hoekstra & B. Schwartz (Eds.), Language Acquisition Studies in Generative Grammar (pp. 151–177). Amsterdam: John Benjamins. Roeper, T. (1999). Universal bilingualism. Bilingualism: Language and Cognition, 2(3), 169–186. Roeper, T. & Pérez-Leroux, A. (1997). The interpretation of bare nouns in semantics and syntax: Inherent possessors, pied-piping, and root infinitives. MIT Working Papers in Linguistics, 12, 114–133. Sadock, J. (1974). Read at your own risk: Syntactic and semantic horrors you can find in your medicine chest. In M. W. LaGaly, R. A. Fox, & A. Bruck (Eds.), Papers from the Tenth Regional Meeting of the Chicago Linguistic Society (pp. 599–607). Chicago, IL: Chicago Linguistic Society. Sadock, J. & Zwicky, A. (1985). Speech act distinctions in syntax. In T. Shopen (Ed.), Language Typology and Syntactic Description (pp. 155–196). Cambridge: CUP. Stowell, T. (1991). Empty Heads in Abbreviated English. Paper presented at GLOW 1991. Tesak, J., Ahlsén, E., Györi, G., Koivuselkä-Sallinen, P., Niemi, J., & Tonelli, L. (1995). Patterns of ellipsis in telegraphese: A study of six languages. Folia Linguistica, 29, 297–316. Vinet, M.-T. (1993). L’aspect et la copule vide dans la grammaire des titres. Langue Française, 100, 83–100. Visser, J. (1996). Object drop in Dutch imperatives. In C. Cermers & M. den Dikken (Eds.), Linguistics in the Netherlands (pp. 257–268). Amsterdam: John Benjamins. Williams, K. (1993). The Kenneth Williams Diaries. Glasgow: HarperCollins.
JB[v.20020404] Prn:26/07/2006; 12:24
F: LA9306.tex / p.36 (2354-2369)
Kate Paesani
Zanuttini, R. (1996). On the relevance of tense for sentential negation. In A. Belletti & L. Rizzi (Eds.), Parameters and Functional Heads: Essays in Comparative Syntax (pp. 181–207). Oxford: OUP. Zhang, S. (1990). The Status of Imperatives in Theories of Grammar. Ph.D. dissertation, University of Arizona. Zwicky, A. (1990). Empty NPs in English and government in unexpected places. In J. A. Edmonson, C. Feagin, & P. Mühlhäusler (Eds.), Development and Diversity: Language Variation Across Time and Space. A Festschrift for Charles-James N. Bailey (pp. 255–263). Dallas, TX: Summer Institute of Linguistics.
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.1 (47-118)
chapter
The narrowing acquisition path From expressive small clauses to declaratives* Christopher Potts and Tom Roeper We analyze expressive small clauses like you fool (and their counterparts in other languages) as contributors of expressive content. Independently known restrictions on expressive content in turn allow us to derive their limited distribution. The theory has ramifications for child language. It correctly predicts which root-level small clauses will survive into adult grammar and which will be blocked by the acquisition of higher functional projections. It also opens the way to an analysis of children’s one- and two-word utterances as denoting expressive, rather than straightforwardly propositional, content.
.
Expressive small clauses
If I forget to buy a crucial item on my grocery list, I might, frustrated with myself, mutter one of the sentences in (1) to express mild self-disapprobation. (1) a. Oh, you fool! b. You idiot! c. You nincompoop/dumbass/screwball! If my coauthor overheard my muttering, he would recognize that I was in a heightened emotional state, but he would do me a disservice if he reported to others that I thought I was a fool (idiot, nincompoop). I did not, after all, use any of the fully sentential forms in (2). (2) a. Oh, you are a fool! b. You are an idiot! c. You are a nincompoop/dumbass/screwball! With the sentences in (1), I express a momentary attitude linked to a situation. With the sentences in (2), if I intend myself to be the addressee, I characterize myself in more general terms. We call the clauses in (1) expressive small clauses (ESCs). They represent a productive construction type: the predicate can be just about anything that can take on
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.2 (118-179)
Christopher Potts and Tom Roeper
emotive force and, though the nominal is you in English, we will shortly see that there is considerable cross-linguistic variation and flexibility on this point. In choosing the term small clause, we intend mainly to recognize that these clauses are necessarily verbless. An ellipsis analysis is a potential competitor, but we assume, with Merchant (Chap. 3, this volume) that this option is not viable. The semantic contrast between (1) and (2) argues against assimilating the short forms to the long ones via phonological reduction. However, there are important differences between the clauses in (1) and those found in, for example, the complement to consider. For instance, if I want to tell you about my self-perceptions, I can use (3a) for this purpose. But the unembedded versions, (3b) through (3d), are ungrammatical in adult English. (3) a. b. c. d.
I consider myself at peace. *Myself at peace. *Me at peace. *I at peace.
The paradigm reverses, in a sense, for ESCs, which are grammatical only as root-level clauses. The heart of our theoretical proposal is that the examples in (1) involve a restricted subclass of lexical items that we call expressives (Kaplan 1999; Kratzer 1999; Potts 2003, 2005). The class includes predicates like fool (on some of their uses), the descriptive content of epithets, attributive-only adjectives like damn, and (outside of English) formal and familiar pronouns and honorifics, among many others. We develop a theory of semantic composition that not only allows, but in fact ensures, that ESCs appear only in unembedded positions and without functional material. Our ESCs further subdivide into self-disapprobation and incredulity types. The first allow no determiners or copular verbs, and the second allow only determiners to intervene between the nominal and the predicate. Our study interacts in novel ways with the theory of language acquisition. Children in the early stages of acquisition are competent with only a restricted set of syntactic structures but are nonetheless able to express a vast range of meanings. This expressivity demands that the surface forms of child grammars be significantly more polysemous than those of adult grammars. The acquisition task involves narrowing the possible meanings for structures. Researchers have identified a variety of syntactic principles that contribute to this narrowing. We argue that the acquisition of semantic distinctions also plays a significant role here. In particular, we seek to support the following generalizations about small clause acquisition: (4) Syntax The acquisition of the IP projection helps define the distribution of root-level small clauses.
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.3 (179-239)
The narrowing acquisition path
(5) Semantics The recognition of a realm of expressive semantic content helps define the distribution of root-level small clauses. Recognizing expressive content can bring us one step further – we can begin to explain why ESCs are so limited in their distribution.
. Small clauses in adult English Small clauses in adult English are well documented and characterized in the literature. In this section, we briefly review their core features, with the aim of highlighting similarities and contrasts with children’s small clauses and indicating why ESCs are special. As noted above, adult English small clauses are generally embedded. We illustrate in (6). (6) a. I consider her a genius. b. *Her/She a genius. ESCs are different: they are unembeddable. Progovac (Chap. 2, this volume) identifies a range of other root-level small clauses. Though we do not here explore the complex interrelations between her data and expressives, it is worth noting one unifying feature: root-level small clauses seem invariably to have a semantics not reproducible with fully sentential forms. The phrase Ali in Nepal, set below a photograph, is subtly different from Ali {is/was/depicted/photographed} in Nepal, and different in ways that transcend tense marking and the like. This observation feeds into our overarching hypothesis that root-level small clauses survive into adult English only if their content is not expressible using a fully sentential form. Small clauses are invariably predicational constructions (Heycock & Kroch 1999): (7) a. I consider Clark to be Superman. b. *I consider Clark Superman. Syntactically, this means that small clauses have the form [DP PredP], where PredP is usually AP but can sometimes be PP (see Huddleston & Pullum 2002). Semantically, the two elements are always put together by functional application; the default composition probably involves applying the Pred’s meaning to the DP’s meaning. However, since the DP can be quantificational, another possible composition scheme has the DP’s meaning applying to the Pred’s meaning. In contrast, the syntax is more rigidly directional: the predicate is always in second position, as we see in (8). (8) a. b. c. d.
I consider Ali intelligent. *I consider intelligent Ali. I consider Ali a genius. *I consider a genius Ali.
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.4 (239-303)
Christopher Potts and Tom Roeper
For additional discussion of English small clauses, see the above cited work and also Stowell (1981) and Svenonius (1994).
. Small clauses in child English In child English, small clauses are freer in their syntax and in the relationship between the nominal and predicational elements. We illustrate in (9) and (10) (Bloom 1973): (9) a.
Baby highchair. ‘Baby is in the highchair.’ b. Mommy sock. ‘Mommy’s sock.’ ‘Mommy wants a sock.’ c. Mommy eggnog. ‘Mommy had her eggnog.’ d. Sweater chair. ‘Sweater is on the chair.’
[agent–location] [possession] [agent–object] [agent–object] [object–location]
(10) a. Me big. b. Me happy. In (9), we see that a small clause can express much more than a simple predication relation between the nominal and the predicate. In both (9) and (10), we see that these clauses can appear unembedded. There is, though, sufficient overlap in the properties of child and adult small clauses to justify connections between them. For instance, children’s small clauses always involve a pair of constituents and no sentential embedding: (11) *Said ate. ‘I said I ate it.’ What’s more, children’s small clauses are invariably predicate-second (Bloom 1990): (12) a. It big. b. *Big it. The predicate-second requirement on small clauses strikes us as important to their overall characterization. It’s for this reason that we have excluded clauses like (13) from our present domain of inquiry: (13) a. b. c. d. e.
Silly me. *Me silly. Silly you. Silly Chris. ?Goofy/Smelly/Grumpy me.
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.5 (303-363)
The narrowing acquisition path
It is tempting to include these in our analysis. But they differ significantly from the selfdisapprobation clauses that are our focus. In (13e), we see that the predicate is essentially fixed as silly. Perhaps more importantly, they have the order predicate–nominal, which is basically unattested in the realm of English small clauses.
. A note on expressive content Our semantic analysis of ESCs is based in the theory of expressive content, and we call upon general restrictions on such meanings to explain what is special about ESCs in particular. We therefore pause now to establish some background on expressive content in general. Potts (2003, 2005) provides a semantically multidimensional analysis of expressive content. Strands of this work trace to Cruse (1986), Kaplan (1999), and Kratzer (1999), and the ideas have recently been applied more widely by McCready (2004), Kratzer (2004), Potts and Kawahara (2004), and others. The present section is an overview of expressive content. The criteria are taken from Potts’s work, but we tailor them specifically to ESCs. We remain at a descriptive level in this section; in Section 8, we provide a precise theoretical formulation that aims to capture these properties. The most prominent feature of expressive content is its strong tie to the utterance situation. For instance, as Potts and Kawahara (2004) show, the content of Japanese honorifics and antihonorifics is always speaker oriented, even when it is expressed inside a belief context, where one might expect it to be oriented toward the subject of the belief predication. Example (14) illustrates. (14) John-wa [Mary-ga nesugoshi-chimat-ta] -koto- oshitteiru. John-top Mary-nom oversleep-anti.hon-past -fact know a. ‘John knows that Mary overslept.’ b. ‘It sucks that Mary overslept.’ The antihonorific expression chimat- expresses the speaker’s disapprobation of the propositional content of the clause containing it. Though it here appears inside the knowledge ascription to John, its content is attributed to the speaker. John might be pleased, or indifferent, that Mary overslept. It is the speaker who looks askance on this fact. The facts are the same for formal and familiar pronouns, which indicate something about the speaker’s relationship to his addressee. Similarly, expressive modifiers like damn are always speaker contributions. One can felicitously utter the sentence Bush thinks the damn Republicans deserve public support because damn trickles up to root, attaining the same standing as the overall assertion (Potts 2003). We cannot really test to see whether ESCs have this property in full, because they are syntactically unembeddable, as we see in (15). (15) a. *I consider you fool/nincompoop/screwball.
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.6 (363-425)
Christopher Potts and Tom Roeper
b. *I consider myself/me fool. c. *Ed saw you fool (so I was embarrassed). Similarly, the predicates involved in ESCs cannot be modified (16a–c), except by other expressive modifiers (16d). (16) a. b. c. d.
*You nonfool. *You unfool. *You complete idiot. You fucking idiot.
We are at present unsure of how the composition works for (16d). But, in Section 8, we derive the limitations evident in (16a) through (16c) from some basic facts about the semantic composition of these clauses. For now, we need only take note of the descriptive generalization: ESCs can’t mingle with the regular descriptive material around them. From this, it follows that they are semantically unembeddable. So ESCs are semantic isolates. What content do they isolate? This turns out to be extremely hard to pin down. In our opening, we described a situation in which one would use an ESC, and we showed that the fully sentential form would express something much different in that context. But what do ESCs mean? If they express propositions, then which ones? If they impose definedness conditions on the context, in the manner of presuppositions, then what are those conditions? At best, these questions are hard to answer. But they might be fundamentally misguided. For expressives, we have strong intuitions about where and how they are properly used, but we hem and haw when pressed to say what they mean. This is the descriptive ineffability property, and it is nowhere stronger than with the expressions of mild disapprobation that are our focus here. Expressive content has at least one other important property: it is always intuitively independent of the descriptive content around it. I can assent to the content of That bastard Kresge is famous without thereby endorsing the content of bastard. I assent to the proposition that Kresge is famous; whatever bastard contributes, it remains with the speaker. However, ESCs don’t permit this kind of test. In virtue of the fact that they are syntactically and semantically isolated, their content never arrives wrapped, so to speak, in descriptive content. They are purely emotive. This sets them somewhat apart from epithets, honorifics, and the like, both descriptively and at a technical level (Section 8). But ESCs have enough in common with these other classes of expression to justify a treatment that unifies them under the broader expressives heading.
. When speaking expressively, we’re all children Sections 2 and 3 showed that adult and child small clauses have different distributions. Part of our descriptive claim, though, is that a limited class of root-level small clauses survives into adulthood. We illustrated for English in (1), repeated here:
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.7 (425-492)
The narrowing acquisition path
(17) a. Oh, you fool! b. You idiot! c. You nincompoop/dumbass/screwball! The next few subsections show that ESCs are cross-linguistically common. Though they manifest themselves in a variety of ways, their important shared property is this: none of them contains a verb. Nothing mediates between the composition of the expressive predicate and its argument. Our account, developed in Section 8, handles this nicely. But we first want to explore the factual terrain more thoroughly. Our taxonomy of ESCs divides into two distinct classes. We have the selfdisapprobation clauses illustrated in (17). In addition, there are incredulity small clauses, as in (18) and (19). (18) Tom: Chris:
You’re really argumentative. Me argumentative? I am not!
(19) What, me worry? Our analysis attributes a quite different structure to these two clause types, and their semantics differs as well. We return to the incredulity type in Section 9. The examples in (18) and (19) display a number of puzzling morphosyntactic properties, some in common and some that distinguish them from each other. First, the self-disapprobation types can involve second-person pronouns even though they are easily used in a self-directed manner. Second, and similarly, they can involve proper names that pick out the speaker, though this is normally a highly marked choice in discourse. Third, the incredulity cases involve a pre-predicative accusative form (just as in ESCs in general) and first person (me) rather than second person (you), although the referent is the same. This is just the start of the morphological variation. As we will see, few of these properties hold constant across languages. But the ones that do are revealing. In particular, it is significant that no functional material of any kind – no verbs, no determiners, and so on – can appear in these self-disapprobation clauses: (20) a. *Chris, you are idiot. b. *Chris, you an idiot. c. Chris, you are an idiot.
[not expressive]
In addition, ESCs don’t ever embed, as discussed in Section 4.
. Self-directed disapprobation cross-linguistically We begin with an extremely safe universal: (21) Every human language provides the means for calling oneself a fool.
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.8 (492-553)
Christopher Potts and Tom Roeper
The next few subsections seek to document some of the variation in form that ESCs can have.
. Auf Deutsch German differs from English in allowing the nominal in the ESC to be a first-person nominative form: (22) a. Ich idiot! b. Ich arm-er Mensch! I poor-masc.nom man It is also possible to use a second-person form. I might say (23) to myself if I fell asleep going 90 mph on the highway and touched the left railing before I woke up again.1 (23) Du idiot, das war gefährlich! you idiot that was dangerous As far as we know, the distribution of these clauses is otherwise the same: they do not really embed, and they arrive without any functional material.
. Op Afrikaans In Afrikaans,2 we find more variation than in English and German. The first noteworthy property of Afrikaans ESCs is that the nominal form is in the accusative case. A nominative pronoun (the norm for subjects) is ungrammatical unless the copular verb and the article appear, turning the clause into a regular predication, as in (24c). (24) a.
Jou idioot! you-acc idiot b. *Jy idioot! you-nom idiot c. Jy is ’n idioot! you-nom are an idiot
So far, this is in accord with the generalizations that we seek to capture theoretically below. We should mention, though, that expressive meanings seem also to arise via different structures in Afrikaans. For instance, a reduced copular verb can produce an expressive meaning of some kind (25a), as can a vocative (23b) and a combination of these two elements (25c). (25) a. Jy’s ’n idioot! b. Andries, jy is ’n idioot! c. Andries, jy’s ’n idioot!
[vocative expressive] [reduced copula and vocative]
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.9 (553-622)
The narrowing acquisition path
We do not here attempt to account for these facts in our semantic analysis. We assume for now that expressive meanings can arise in a variety of different ways. The vocative examples in (25b) and (25c) suggest an alternative to our analysis, one that a few people have suggested to us: we might seek to assimilate ESCs to vocatives. However, it seems clear that vocatives and ESCs are different constructions. First, vocatives can easily occur without any disapprobation: (26) You, waiter! Could you bring me the bill? This is quite different from, and carries a different intonation from, example (27). (27) You (mere) waiter/jerk! What’s more, the vocative in this form does not allow self-reference in English at least. In addition, one can combine the vocative with this form of ESC, as in Chris, you idiot! So we think that those examples that informants give us from other languages that involve the vocative might be inherently different, although we are not able to pursue the matter further at the moment.
. Po-Russki Russian ESCs and regular small clauses differ when it comes to predicate agreement. For an expressive meaning, speakers simply use the predicate inflected for gender: (28) a.
Idiot idiot-male.spkr b. Idiotka! idiot-fem.spkr
If one uses a pronominal in addition to an inflected predicate, the result sounds somewhat like a medical diagnosis (Anna Verbuk, personal communication): (29) Ja idiotka! I idiot-fem.spkr Scrambling, though, can bring back the expressive meaning that the pronoun removes in the canonical order: (30) Idiotka ja! idiot-fem.spkr I
. Nihongo-De Japanese ESCs are of particular interest both from the point of view of the present study and for our understanding of honorification more generally. Both (31a) and
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.10 (622-682)
Christopher Potts and Tom Roeper
(31b) have expressive meanings, but (31b) is noteworthy for involving an antihonorific that is self-directed. (31) a.
Ore me b. Ore me
tte topic tte topic
baka da na. idiot copula part o-baka-san. hom-idiot-hon
Potts and Kawahara (2004) provide a semantics for honorifics and antihonorifics in Japanese. Like the present analysis, theirs is based in the notion of expressive content as importantly distinct from regular content.
. General thoughts on variation and interfaces The variation attested above invites some commentary on the nature of linguistic interfaces, in particular, on what general properties govern the interface and how those general properties might assist children in acquiring language. It is clear from the above that different languages make different choices when it comes to realizing expressive predications. But, as noted earlier, they display a kind of common core: none of them contains a verbal element (excepting the reduced copular clause in the apparently expressive Afrikaans example in (25)). We suggest that this absence arises from the following principle: (32) No verb meaning is expressive. We will see below that this has a straightforward formal counterpart in our restrictive theory of semantic types. We can in addition make some sense of the attested variation in the form of the nominal in these predicates. The generalization seems to be that a language will pick its least marked surface realization for the expressive nominal. In German, the default case is clearly nominative, hence the form ich (‘I’) in clauses like Ich idiot! and Du (‘you-nom’) in Du idiot! The default case in English is accusative – it shows up just about everywhere but the canonical subject position – but we can really only speculate that it is the case of sentences like You idiot, since the second person does not inflect for case. It seems telling, though, that the potentially related expressive Silly me! clearly involves the accusative. The generalization seems to extend to Russian. Though Russian’s silent presenttense copular verb seems to prevent it from becoming a true pro-drop language, many of its pronominal subjects end up being phonologically silent, an observation that is in line with its rich inflectional morphology. It is therefore not surprising that Russian ESCs are generally just inflected predicates.
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.11 (682-740)
The narrowing acquisition path
. The acquisition of expressive small clauses Small clauses are ideal for pursuing our distinct but related aims of tracking the path of acquisition and identifying what is special about expressives. This is so because it is common to adopt (33) in some version or another. (33) Acquisition Stage 1 All two-word forms are small clauses
(Lebeaux 1988; Radford 1990).
That is, in the earliest stages of language acquisition, all clauses that children utter have the structure of small clauses. This means, of course, that the construction is pressed into duty for everything that children wish to communicate. It is thus no surprise that we find them in matrix position and also that they can be used to assert that something other than a predication relation holds between the nominal and the predicate. Stage 2 can be characterized as follows: (34) Acquisition Stage 2 As learners acquire more functional projections, they begin also to move toward a one-to-one syntax–semantics connection This recalls a prominent claim of learnability theory: the acquisition process involves expansion but not retreat. That is, the acquisition path involves adding functional projections and features, monotonically, building from a small clause base. The nature of this growth is influenced by language-particular information inferable from the speech community. We propose that the syntax-to-semantics mapping also undergoes a systematic development. In early stages, the mapping is far from functional – each structure is associated with a large set of meanings. As children acquire more structures, they work to reduce the one-to-many nature of the interface, by assigning meanings to the new structures and disassociating those meanings from the old structures. For instance, when children acquire the verbal projections necessary for transitive clauses like Mommy drank milk, the small clause Mommy milk loses its ability to express the proposition that Mommy drank milk. And so forth. The general trend is toward a oneto-one mapping from structures into meanings. The more structures one has at one’s disposal, the closer one can come to this ideal. Thus, increased syntactic sophistication correlates with decreased semantic ambiguity. What is ultimately required is a stepby-step account of how each misprojected small clause meaning is moved to another structure. It follows from (34) that the range of syntactic and semantic options for the small clause enters into decline as children acquire additional functional projections. For example, the clause Mommy eggnog loses its ability to express the proposition that Mommy had her eggnog, because the extended verbal projections of the VP and the IP have assumed the role of expressing transitive clauses of this sort. But, for reasons discussed in Section 8, ESCs are left out of this competition. Higher structure is incompatible with their semantics, which essentially forces a di-
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.12 (740-796)
Christopher Potts and Tom Roeper
rect application of the expressive predicate to the small clause nominal. We expect this to be true of all root-level small clauses that survive into adult English, for instance, those discussed by Progovac (Chap. 2, this volume). For instance, the small clause John in New York is possible precisely because its semantics differs from that of any fully sentential form. Our analysis invites us to reconsider the nature of the semantic content of many utterances of child English. For instance, does the child’s exclamation of He big! denote a proposition? The theory of expressive content suggests that it might not, that it might instead denote purely expressive content, in other words, that it is an ESC. As Kaplan (1999) shows, this does not mean that it couldn’t have propositional implications. It might implicate the proposition that he is big. But its denotation is potentially something else entirely. As the child acquires the ability to manipulate higher functional structure, the more direct and unambiguous alternative He is big crowds out the expressive variant. But the basic facility for ESCs survives in you fool and the like – sentences with meanings that are importantly distinct from their fully sentential counterparts, as we have shown. In a sense, Wexler (1998) anticipates this alternative and seeks to block it. Wexler argues that root infinitives must have a higher TP phrase. The primary motivation for this claim is the assumption that TP is the sole locus of propositional denotations. However, we reject this assumption. Propositional denotations arise not only from TP nodes, but also via conversational implicatures, via semantically multidimensional constructions (Potts 2005), and via presupposition triggers. So the apparently assertive force of children’s subsentential utterances does not, by itself, argue against the idea that the content of those utterances is purely expressive. We think that this avenue remains open and worth exploring.
. Analysis of self-disapprobation small clauses An analysis of self-disapprobation clauses must account for at least the following related properties: (35) a. Self-disapprobation clauses are unembeddable. b. Self-disapprobation clauses lack functional material. The examples in (36) illustrate each of these properties in turn: (36) a. *I regard you/yourself/myself/me idiot! b. *You are idiot. c. *You an idiot. The vehicle of our analysis is the theory of semantic types. Here is a broad overview (described more fully below): we define a special expressive type, E, that is never an input type. Semantically, this means that nothing takes an expressive-typed thing as an
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.13 (796-864)
The narrowing acquisition path
argument. This ensures that such content fails to embed, and it also explains why functional material (determiners, copular verbs, etc.) cannot come between the expressive and its argument. Our analysis could be paired with a more clearly syntactic view of expressive content, perhaps connecting in important respects with the syntax of point of view (Speas 2004; Hollebrandse 2000; DeVilliers 2003; Hollebrandse & Roeper 1998) and their role in acquisition. To do this, one would define a special expressive feature that could select for certain items but that could never be selected for, even by abstract functional heads. Indeed, this approach is so close to our own type-theoretic formulation that readers are free to interpret it as such.
. A meeting at the interface: Type theory The set of semantic types for our theory is defined in (37). (37) a. b. c. d. e.
e and t are regular types. E is an expressive type. If a and b are regular types, then is a regular type. If a is a regular type, then is an expressive type. Nothing else is a type.
Clauses (a) and (c) (along with the extremal clause (e)) define the usual sort of type theory in semantics. The new clauses are (b) and (d). Clause (b) specifies a new expressive type, symbolized E. It can enter into just the limited class of functional types defined by clause (d). In short, E can be an output type – it can appear in types like , where a is a regular type. But we have no types like > or <E,>. In these, the expressive type is an input type – just what we exclude with (37). We also have no types in which E is embedded inside an output type. For example, is not in this type space. In thinking about these types, one should also consider their corresponding semantic domains, which we specify in (38). (38) a. b. c. d.
The domain for type e is De , a set of entities. The domain for type E is DE , the set of expressive meanings. The domain for type t is Dt , a set of propositions. The domain for type is D , the set of all functions from Da into Db .
For the most part, this definition is standard. The only really noteworthy clause is (b), which essentially leaves the nature of expressive content unanalyzed. The exact nature of this content is difficult to specify. We do not venture a specific hypothesis here. A variety of answers have appeared in the literature. In Potts (2005), it is taken to be regular semantic content; expressive and regular content are distinguished, in that theory, only by their semantic types, a largely formal division. Potts (2003) seeks to connect expressives with speech acts, and Potts and Kawahara (2004) argue that it should be modeled
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.14 (864-913)
Christopher Potts and Tom Roeper
using the real numbers in the interval [–1,1]. We refer also to McCready (2004), an analysis that takes this content to be importantly dynamic. For the purposes of this chapter, though, we require only the type-theoretic division suggested by (37). Our inability to say more about the domain DE is a direct result of our current lack of understanding of the descriptive ineffability property discussed in Section 4.
. Expressive composition Expressive-typed expressions create an impasse of sorts when we look at semantic composition. Suppose, for instance, that we let composition proceed by function application alone. That is, suppose (39) is our only composition principle. (39) Functional application (order independent) A(B) : b B:a
A :
This would allow us to have semantic parsetrees like the following: (40) You fool! fool (you) : E you : e
fool : <e,E>
The expression on the root node in this tree is of type E. Thus, nothing can take it as an argument. This is a direct consequence of definition (37), which ensures that we cannot have expressions of type <E,a> for any type a. To put it another, more specific, way: because consider is a function from the type of regular properties (or propositions), it cannot appear as the sister to You fool! The result would be semantically uninterpretable for the simple reason that it would induce a type mismatch between sisters. We can similarly block versions of these clauses that contain, say, a copular verb: (41) *You are fool! undefined you : e
undefined
be : <e,<e,t>>
fool : <e,E>
The same logic block structures with determiners. We need only assume that they are not expressively typed, that is, that they have the usual semantics, delimited by the type ,e> or the type ,,t>>.
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.15 (913-967)
The narrowing acquisition path
We should note that this theory of expressive composition differs from that of Potts (2005) and related work. There, expressive content (and, more generally, conventional-implicature content), is governed by the following rule: (42) Expressive composition (Potts 2005) B : a, A(B) : E B:a
A :
In this rule, expressives involve functional application in one dimension of meaning. But the semantic value that an expressive determines for its mother node is multidimensional – it denotes both the expressive-type meaning and the regular-type meaning given by the expressive’s argument. If we used this rule for ESCs, we would have trees like the following: (43)
you : e, fool (you) : E you : e
fool : <e,E>
In a sense, the claim would be that You fool! denotes both an expressive meaning and picks out the addressee. This is an incorrect analysis, though: ESCs do not behave in any sense like nominals. This is easily brought out by minimal pairs like the following (for which we are indebted to Caroline Heycock, personal communication): (44) a. You fools should read more carefully. b. *You fool should read more carefully. The plural form permits a reading on which it is a kind of integrated appositive like we linguists or you cyclists. Such appositives require plural pronouns, and hence you fool is forced into its ESC reading, where it of course cannot act as a grammatical subject. We capture this semantically if ESCs have one-dimensional expressive meanings (a possibility that Potts (2005) does not discuss at all). We fail to capture it if we import Potts’s theory of expressives directly to the present work, by relying on (42) to put together the parts of ESCs. We thus propose that (42) does not characterize ESCs. This means, though, that we must be somewhat careful about how we assign semantic types. It would be a mistake to allow the theory to be nondeterministic, that is, it would be a mistake to allow that daughters of the sort in (45) could have either (45a) or (45b) as the value for their mother. (45) A : B : a a. A(B) : E b. A(B) : E , B : a
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.16 (967-1027)
Christopher Potts and Tom Roeper
Thus, we must allow that the predicates in ESCs have a slightly different type than those that appear in phrases like the damn linguists, which can appear in embedded positions but which also have an element of expressivity to them (Potts 2005). To be fully precise, we assume that the domain of expressives divides into two subdomains: the kind that determine multidimensional meanings and the kind that determine one-dimensional meanings (like those that appear in ESCs).
. Analysis of incredulity small clauses We have so far left incredulity clauses out of the discussion. But they bear a striking resemblance to self-disapprobation clauses. In particular, they too are unembeddable and become ungrammatical as soon as we introduce any functional structure into their minimal design. (46) a. (What,) Me worry? b. *I wonder/doubt me worry. c. *Me are/is/be worry? Our semantic explanation for these facts is, though, different from the one we gave in Section 8 for self-disapprobation clauses. In this case, we are guided by the typical way of expressing these meanings in German. We illustrate in (47). (47) a.
Ich und Angst haben? I and fear have ‘Me afraid?’ b. Ich und ein Professor (sein). I and a professor be ‘Me a professor?’
The und (‘and’) that appears in these cases is evidently not the usual Boolean coordinator or anything like it. For one thing, it conjoins two things that are of different syntactic category and semantic type. For instance, ich is presumably an entity-level expression. It appears here to be coordinated with Angst haben, a predicate. In addition we find articles are not only possible but required: (48) a. Me an idiot? b. *Me idiot? Moreover, the notion that a more abstract pairing is involved comes through because appositives are possible inside the incredulity small clauses where they are on the outside of self-disapprobation clauses: (49) a. Me, Tom, an idiot? b. *You, Tom, an idiot! c. Tom, you idiot!
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.17 (1027-1095)
The narrowing acquisition path
Thus, we propose that this coordinator is merely a pair-formation operator in the semantics. The denotation of Ich und Angst haben is actually more like (50). (50) < the speaker, the property of being afraid > The idea is that the speaker, in using this object instead of the propositional denotation obtainable by applying the second member to the first, means to say that there is something inappropriate about that act of functional application. The speaker presents these two objects independently as a way of signaling the infelicity of combining them. This provides an immediate account of why these clauses do not embed: there are presumably no lexical items whose semantics allows for this kind of object as one of its arguments. Like expressive-typed items, objects like (50) thus in effect put an end to the semantic composition – they have to be on the root, because nothing can have them as a sister. Incredulity clauses are not known in early child language for reasons that are expected under our account. Conjunction involves a higher order projection and is not found among early two-word meanings. Children, surprisingly, do not normally say things like (51). (51) a. *Mommy daddy (‘Mommy and Daddy’) b. *Meat rice (‘Meat and rice’) c. *Knife fork (‘Knife and fork’) These are natural meanings, but they would be captured with and, exactly the form we are attributing to incredulity clauses. So although this and is different from the incredulity and, it is part of a family of meanings not initially employed by children.
. Summary and conclusions It is common to assume that small clauses provide a foundation for children to acquire (build) the more complex structures of their language. They are therefore an excellent starting point if one wishes to track the path of acquisition. It is especially fruitful to link this starting point with the development of children’s facility with semantic interpretation. In this chapter, we assumed two kinds of progression: children seek to acquire additional structures, and they strive for something like a bijective relationship between forms and meanings. Thus, as they acquire new structures, the one-to-many mapping that they were once forced to by their impoverished syntax gets closer and closer to being functional. This has distributional consequences. In particular, the higher structure limits small clauses in two ways: they lose their main clause status (because IP can serve that purpose), and they become invariably predicational (because verbal projections handle the other available relations). In a sense, this chapter began with the following idea: suppose there were a type of clause (or content) that couldn’t project any higher than the small clause, for prin-
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.18 (1095-1157)
Christopher Potts and Tom Roeper
cipled reasons. On the above view of acquisition and interpretation, we would expect it to remain childlike, in the sense that its meaning would never get taken over by more complex syntax. ESCs seem to instantiate just this clause type. Because expressive content is inherently unable to act as a semantic argument, a small clause with an expressive type for its predicate acts as a barrier to further composition. ESCs are thus unembeddable and lack the verbal and functional material that we expect to find with main clauses. The semantic relationship between function and argument in ESCs seems inflexible (we claim that it is predication), but these clauses are otherwise a kind of throwback to early child grammar, before higher structures and the pressures of a deterministic syntax-to-semantics mapping forced small clauses into a merely supporting role.
Notes * Our thanks to Alena Anishchanka, Doreen Bryant, Manfred Bierwisch, Andries Coetzee, Shai Cohen, Ilaria Frana, Ewald Lang, Hubert Haider, Caroline Heycock, Shigeto Kawahara, Angelika Kratzer, and Anna Verbuk, as well as the crowd at the Barbara Partee Retirement Party (UMass Amherst, September 18, 2004), where we presented an early version of this work and gathered many useful examples. Thanks also to the members of the Workshop at Wayne State University on Nonsentential Syntax, in particular, Ellen Barton, Eugenia Casielles, Ljiljana Progovac, Robert Stainton, and the audience at the 2005 meeting of the German Linguistic Society. . Our thanks to Hubert Haider for this example, and also to Jan Anderssen, Manfred Bierwisch, Doreen Bryant, Angelika Kratzer, Ewald Lang, and Florian Schwarz for helping us identify the relevant class of German examples. . Our thanks to Andries Coetzee for these examples and observations.
References Bloom, L. (1973). One Word at a Time: The Use of Single Word Utterances Before Syntax. The Hague: Mouton. Bloom, P. (1990). Syntactic distinctions in child language. Journal of Child Language, 17, 343– 356. Cruse, D. A. (1986). Lexical Semantics. Cambridge: CUP. DeVilliers, J. (2003). Getting complements on your mental state (verbs). In J. van Kampen & S. Baauw (Eds.), Proceedings of GALA 2003 (pp. 13–26). Utrecht: LOT. Heycock, C. & Kroch, A. (1999). Pseudocleft connectivity: Implications for the LF interface level. Linguistic Inquiry, 30, 365–397. Hollebrandse, B. (2000). The Acquisition of the Sequence of Tense. Ph.D. dissertation, University of Massachusetts, Amherst. Hollebrandse, B. & Roeper, T. (1998). Point of View Operators, Features, and a Theory of Barriers. Paper presented at University of Pennsylvania Roundtable in Linguistics. Huddleston, R. & Pullum, G. K. (2002). The Cambridge Grammar of the English Language. Cambridge: CUP.
JB[v.20020404] Prn:20/07/2006; 10:28
F: LA9307.tex / p.19 (1157-1216)
The narrowing acquisition path
Kaplan, D. (1999). What is Meaning? Explorations in the Theory of Meaning as Use. Ms., University of California, Los Angeles. Kratzer, A. (1999). Beyond Ouch and Oops: How Descriptive and Expressive Meaning Interact. A Comment on David Kaplan’s Paper. Paper presented at the Cornell Conference on Theories of Context Dependency. (http://semanticsarchive.net/Archive/WEwNGUyO/) Kratzer, A. (2004). Interpreting Focus: Presupposed or Expressing Meanings? Theoretical Linguistics, 30, 123–136. Lebeaux, D. (1988). Language Acquisition and the Form of Grammar. Ph.D. dissertation, University of Massachusetts, Amherst. McCready, E. (2004). Two Japanese adverbials and expressive content. In K. Watanabe & R. B. Young (Eds.), Proceedings of the 14th Conference on Semantics and Linguistic Theory (pp. 163–178). Ithaca, NY: CLC. Potts, C. (2003). The Performative Nature of Expressive Content. Paper presented at the University of Connecticut (November 17) and the University of Rochester (November 18). Potts, C. (2005). The Logic of Conventional Implicatures. Oxford: OUP. Potts, C. & Kawahara, S. (2004). The performative content of Japanese honorifics. In K. Watanabe & R. B. Young (Eds.), Proceedings of the 14th Conference on Semantics and Linguistic Theory (pp. 303–322). Ithaca, NY: CLC. Radford, A. (1990). Syntactic Theory and the Acquisition of English Syntax. Oxford: Blackwell. Speas, M. (2004). Evidentiality, logophoricity and the syntactic representation of pragmatic features. Lingua, 114, 255–276. Stowell, T. (1981). Origins of Phrase Structure. Ph.D. dissertation, Massachusetts Institute of Technology. Svenonius, P. (1994). Dependent Nexus: Subordinate Predication Structures in English. Ph.D. dissertation, University of California, Santa Cruz. Wexler, K. (1998). Very early parameter-setting and the unique checking constraint. Lingua, 106, 23–79.
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.1 (47-116)
chapter
Nonsententials and second language acquisition* Nicola Work This chapter applies the nonsentential analysis (Progovac, Chap. 2, this volume) to early L2 learner data. Adult nonsententials and similar structures found in learner varieties share several syntactic features such as the absence of Tense, the absence of Case, the co-occurrence of tenselessness and caselessness, and the use of phrasal instead of sentential negation. Therefore, I argue that certain L2 structures are indeed nonsententials and that L2 learners employ this nonsentential grammar early in the acquisition process before potentially developing a full grammar. Teacher talk, on the other hand, seems to be an artificially created language by teachers to address L2 learners that is characterized by deletion of certain elements, and can therefore not be analyzed by the nonsentential approach.
.
Introduction
This chapter seeks to extend the nonsentential analysis developed by Progovac (Chap. 2, this volume) to account for early second language (L2) learner data. The main goal of the chapter is to identify similarities between nonsentential utterances typical of adult speech and those present in the L2 Basic Variety (BV) as well as in L2 postbasic learner languages. Following Progovac (Chap. 2, this volume), nonsententials are defined as utterances smaller than a full sentence, as tenseless phrases or small clauses. Native adult nonsententials are characterized as base-generated NPs, PPs, VPs, APs, and DPs and are built bottom up based on the Minimalist principle of Merge (see, e.g., Chomsky 1995). English nonsententials in normal adult speech are characterized by the following features: i.
Absence of Tense (either the complete absence of a copula or auxiliary verb, or the omission of Tense on the lexical verb) ii. Absence of structural (nominative) Case (instead use of Default accusative Case in English) iii. Co-occurrence of the absence of Tense and the absence of Case
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.2 (116-192)
Nicola Work
iv. Use of phrasal negation (not) instead of sentential negation (don’t, didn’t, isn’t) The following examples of adult nonsententials demonstrate these features. (1) Problem solved (2) Not urgent In (1) the absence of Tense is illustrated by the absence of a copula verb and the absence of Case is illustrated by the absence of a determiner on the subject. (2), on the other hand, shows that negation in nonsententials is expressed through phrasal negation (not) as opposed to sentential negation (didn’t, doesn’t, isn’t). Certain nonsententials found in the BV or later learner languages have similar features. (3) Girl go
(Klein & Perdue 1992)
(4) No too hard (3) demonstrates the same co-occurrence of tenselessness and caselessness demonstrated in (1). Furthermore, (4) illustrates that a form of phrasal negation (no) is found in L2 learner varieties instead of its sentential counterparts (isn’t, doesn’t). In order to establish whether the nonsentential analysis can indeed be extended to early L2 data, Section 2 reviews relevant L2 acquisition theories. Section 3 then introduces Klein and Perdue’s (1997) Basic Variety and presents relevant L2 nonsentential data. Section 4 discusses similarities and differences between adult and L2 nonsententials, explores whether the nonsentential analysis can account for these data, and points out any difficulties this account poses for L2 acquisition studies. Section 5 analyzes L2 acquisition theories compatible with this view. The final section presents teacher talk data and demonstrates the structural differences between these utterances and nonsententials typical of adult speech and L2 acquisition data.
. Background The study of L2 acquisition was rooted in the behavioristic tradition from the 1940s to the 1960s and was associated with Contrastive Analysis (CA), which focused on the learner’s native language as well as the target language (TL) to be acquired. According to this theory, learners are strongly influenced by their native language (L1). Therefore, whenever L1 and L2 were structurally the same or very similar, learning would be facilitated (positive transfer), whereas as soon as L1 and L2 differed, learning would be impeded (negative transfer). Some researchers (e.g., Lado 1957) believed that learners would only produce errors when the L1 and the TL differed and hence these errors were the result of interference or transfer from the L1. CA not only impacted the field of L2 acquisition, but also influenced classroom teaching practices at that time. In the 1970s, when studies found that not all learner errors originated in the L1, the focus of L2 ac-
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.3 (192-244)
Nonsententials and second language acquisition
quisition shifted and Error Analysis (EA) replaced CA. Whereas CA studied a learner’s L1 and the L2 to be acquired, EA investigated learner language (i.e., language produced by the learner) for the first time. Even though errors had been studied in the past, it was not until the 1970s that EA was accepted as a methodology for L2 acquisition. In his seminal paper, Corder (1967) viewed learners’ errors as crucial to L2 learning and no longer as bad or wrong habits to be avoided at all cost, as was the case in CA. According to Corder, errors (a) show the teacher how much the learner has learned, (b) provide evidence of how languages are learned, and (c) help learners discover the TL rules. L2 research not only studied the errors that learners made, but also the overall L2 system produced by learners. Corder, for instance, calls learner languages “transitional competence” and implies that these systems are constantly changing in the process of L2 acquisition: It will be useful therefore hereafter to refer to errors of performance as mistakes, reserving the term error to the systematic errors of the learner from which we are able to reconstruct his knowledge of the language to date, i.e. his transitional competence. (167)
Selinker (1972) coined the term interlanguage to refer to the language produced by learners that is independent of both the learner’s L1 and the L2 to be acquired. This term has several interrelated meanings such as (a) the learner system resulting from systematic rules at any given point during the acquisition process as well as (b) a series of interlocking systems that demonstrate learners’ progression. Hence, interlanguage is considered to be a system that is both rule governed and dynamic. Many researchers in the field of L2 acquisition found that learner languages or learner varieties show many similarities cross-linguistically. Syntactically, these early varieties often include utterances that lack auxiliaries and sometimes even verbs, and are deficient of inflectional morphology (such as case marking, agreement, and tense) and articles. However, these learner languages are not at all chaotic or without organization; simply different principles for sentence organization and structure seem to be at work. As generative grammar developed and with the introduction of Chomsky’s (1981a) Principles and Parameters approach to Universal Grammar (UG), L2 researchers started to investigate how UG works together with transfer from the L1. This approach is primarily a theory of L1 acquisition and attempts to explain how children can learn their L1 in a fairly short period of time and with only a limited amount of input. According to Chomsky (1981a), every child comes equipped with an innate “domain-specific language faculty,” the language acquisition device (LAD), which includes UG (Johnson 2004: 30). According to Chomsky (1981b), “UG consists of a highly structured and restrictive system of principles with certain open parameters, to be fixed by experience” (38). The fact that all languages have nouns and verbs, for instance, constitutes a universal principle, and so does Subjacency, the principle that restricts movement possibilities. Parameters, on the other hand, are “principles with a limited number of built-in options (settings or values), which allow for crosslinguistic
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.4 (244-306)
Nicola Work
variation” (White 2003: 9). One example is the pro-drop parameter, which specifies if, in a given language, a subject is always realized (as in English or French) or is optional (as in Spanish or Italian). Even though this specific parameter has only two settings, there are others that allow multiple options. This knowledge of the properties of the L1 grammar is implicit or subconscious. Regarding the aspects of grammar acquired through UG, Johnson (2004) points out: UG does not assist the child in the acquisition of the entire grammar. It is responsible only for guiding the process of acquiring core grammar, the unmarked features of the child’s native language grammar.. . . The marked rules, the periphery of the grammar, need to be learned. (36)
Since the mid-1990s L2 acquisition research in the generative tradition has started to focus on the L2 initial state. According to White (2003), the term initial state is variously used to describe the “kind of unconscious linguistic knowledge that the L2 learner starts out with in advance of the L2 input, and/or to refer to characteristics of the earliest grammar” (58). To study and explain what happens during the L2 initial state, generative L2 researchers investigate transfer in conjunction with theorizing about the extent of accessibility to UG. Transfer and access to UG are important concepts in L2 acquisition research because – depending on the theory adopted – one or both of them constitute, to different extents, the initial state. Sauter (2002) has recently offered a useful six-way division of the different theories, which takes into account the extent of L1 transfer (No Transfer, Partial Transfer, Full Transfer) as well as the relative accessibility to UG, that is, the degree to which L2 acquisition is constrained by UG (No Access, Full Access).1 The body of research on the relationship between transfer and access is vast and presents a variety of conclusions. The reader is referred to Sauter for a detailed summary. This research on L2 acquisition in general and the nature of interlanguage in particular provides the foundation for the detailed discussion of the syntax of learner language. One of the goals of this chapter then is to compare L2 utterances that are very similar to adult nonsententials and to determine whether the nonsentential analysis in Progovac (Chap. 2, this volume) can be extended to account for them. The next section lays the groundwork for this comparison and analysis by presenting relevant L2 nonsentential data.
. Early L2 acquisition data: The Basic Variety Learner language or interlanguage represents the learner’s knowledge of the L2 to be acquired at any point during the acquisition process. In L2 research the characteristics of learner language are studied in order to gain information on how an L2 is acquired. According to Ellis (1994), four areas of learner varieties have received attention in L2 research: (a) errors, (b) acquisition orders and developmental sequences, (c) variability, and more recently (d) pragmatic features relating to how language is used in
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.5 (306-359)
Nonsententials and second language acquisition
context for communication. Apart from investigating learners’ errors (see discussion above), researchers have found evidence of a sequence of developmental stages in the acquisition of certain syntactic features when analyzing learner varieties. Morpheme studies (as discussed in Krashen 1977) suggested the possible existence of a natural order when acquiring grammatical features. Furthermore, researchers discovered that learners go through various stages during the acquisition process, that is, their learner language and the rules of this system are constantly changing. However, when studying a learner’s interlanguage, Bley-Vroman (1983) reminds researchers that the continuous comparison of the learner language with the TL might obscure the internal logic of this system – namely its rules and constraints, which often are based on neither the L1 nor the L2.2 When acquiring an L2, learners go through several different stages at individual speeds and fossilize at any given stage. Dietrich et al. (1995), for example, demonstrated that early on learners, regardless of what L2 is to be acquired, use what is called a pre-Basic Variety, which has four properties: (a) it is lexical; (b) it lacks functional inflection; (c) it lacks complex constructions; and (d) it is heavily dependent on context. Klein and Perdue (1997), as part of a longitudinal, cross-linguistic study by the European Science Foundation, analyzed the learner language of 40 adult learners of different L2s (English, French, German, Swedish, Dutch) in an immersion setting with different L1 backgrounds (Punjabi, Italian, Turkish, Moroccan Arabic, Spanish, Finnish). The authors found that all learners in the study used a type of learner language beyond the pre-Basic Variety identified by Dietrich et al. (1995), which they call the Basic Variety (BV). Klein and Perdue and Dietrich et al. sum up the characteristics of the BV as follows: utterances in the BV generally consist of uninflected verbs, their arguments, and optionally some adverbials; by and large there is no case marking, no morphological inflection, and no subordination. Lexical items usually occur in a single invariant base form, which can differ from language to language (Dietrich et al. 1995). As illustrated below, L2 learners can use verbs, for instance, in their bare stem (example (3) repeated in (5)) or V-ing form (6), or as an infinitival form (7). Subjects may also appear as bare nouns in (5), (6), and (8); as nonnominative (tonic) pronouns (7); and in the case of languages with overt case morphology, as caseless nouns (8).3 (5) Girl go
(Klein & Perdue 1992)
(6) Police coming (7) Toi faire ça you-tonic do-inf that ‘You do that’ (8) Baustell vielleicht construction-no.case maybe ‘Maybe construction’
(French; Prévost & White 2000a)
(German; Klein & Perdue 1992)
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.6 (359-437)
Nicola Work
Generally, the main source for lexical items in the BV is the TL, although there might be some borrowings from the L1. The lexicon of the BV also consists of a minimal system of content and function words. As the data below illustrate, the L2 BV contains a minimal pronoun system, that is, mainly first- and second-person singular pronouns (10), (13b); some determiners (12), (13b); a few quantifiers; some adjectives and adverbs (9), (11); generally one word for negation (11); a few prepositions (presumably often denoting space and time, but in this data only the preposition with (13)); no complementizers; and usually no expletive elements (e.g., English there).4 (9) Very good student (10) Ich schreiben I write-inf ‘I write’ (11) But no possible (12) Die meisje honger the-fem girl hunger ‘The girl is hungry’
(Lee 2000)
(German; Prévost & White 2000a) (Klein & Perdue 1992)
(Dutch; Klein & Perdue 1992)
(13) a.
Mit Lehrer oder mit Mutter spielen? With teacher or with mother play-inf ‘Play with the teacher or with the mother?’ (German; Prévost & White 2000a) b. Moi jouer avec le train me-tonic play-inf with the-masc train ‘I play with the train’ (French; Prévost & White 2000a) c. Girl play with toy (Ionin & Wexler 2002)
To put these lexical items together, speakers of the BV apply three constraints for utterance organization: (a) phrasal constraints, (b) semantic constraints, and (c) pragmatic constraints (Klein & Perdue 1997). Phrasal constraints govern the form and relative order of the constituents, semantic constraints are associated with the case role properties of arguments, and pragmatic constraints affect the organization of information in connected text, such as topic-focus structure (Klein & Perdue 1997: 313). The phrasal constraints observed in the BV allow three basic phrasal patterns with some subvariants: PH1a NP1 -V PH1b NP1 -V-NP2 PH1c NP1 -VNP2 -NP2 Adj PH2 NP1 -Cop- NP2 PP
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.7 (437-501)
Nonsententials and second language acquisition
PH3
V
-NP2
Cop
(Klein & Perdue 1997: 314)
Example (5) is a case of PH1a, example (7) of PH1b, and (18) of PH3. According to Klein and Perdue (1997), utterances in the BV contain verbs, but there are no finite verbs in whatever function, as was attested in (5) through (7), (10), and (13). Since the BV lacks the grammatical means to indicate tense and aspect, speakers compensate by using temporal adverbials, as in (14a) and (14b). Otherwise the utterances are assumed to refer to the here and now (see, e.g., (5) and (6)). (14) a. She make already a cake b. Charlie get up first
(Ionin & Wexler 2002) (Klein & Perdue 1992)
A final characteristic of the BV is weak features. According to the Minimalist Program (Chomsky 1995), lexical items carry features that need to be checked by raising to certain functional categories. Strong features on functional categories can trigger movement whereas weak ones cannot. The Opacity or Verb Raising Parameter illustrates the difference between strong and weak features cross-linguistically. In languages such as French, Spanish, and German, which are considered to have rich verbal morphology, verbs raise overtly, whereas in English, for instance, they do not. Hence, a raised verb in French leaves behind the adverb and the negation pas, whereas the adverb and negation remain to the left of the English verb.5 Regarding feature strength in early learner language, Klein and Perdue (1997) claim that all features in the BV are initially weak and that therefore no inflectional morphology and no complex structures are found since they would require movement. The BV then does not project functional projections and also lacks the raising of elements to check features in higher categories. For this reason, L2 acquisition beyond the BV involves a process of selecting the appropriate features to be made strong – namely, those that happen to be strong in the TL. Nevertheless, many L2 learners never progress beyond the BV and plateau at this stage of acquisition. For Klein and Perdue (1997), the BV represents a highly efficient system of communication, a real language, which proves to be highly structured. However, the notion of the cross-linguistic existence of the BV and some of its properties has been criticized by many researchers (e.g., Bierwisch 1997; Comrie 1997; DeGraff 2001; Jordens 1997a; Meisel 1997; Schwartz 1997). DeGraff, for instance, claims that a language that lacks recursion – the possibility to generate complex structures via hierarchical embeddings (via the operation Merge) – and movement cannot be called a highly efficient, natural, human language. Bierwisch and Meisel, on the other hand, critique the notion of the BV being an I-language, that is, a cognitive system internalized within the brain (Chomsky 1986). They point out different properties of I-languages and other natural languages such as, for example, developmental differences between L1 and L2 acquisition. Comrie, furthermore, explains that natural languages have both a social (communicative) function as well as a cognitive function. The BV, however, only has a communicative function. Comrie also raises the question whether the phrasal con-
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.8 (501-555)
Nicola Work
straint PH1b (NP-V-NP) is typical of all L2 settings and suggests that it may reflect the salient word order of the target languages of the Klein and Perdue (1992) project. Finally, Schwartz addresses the fact that not all interlanguages are the same and that variation among these learner languages occurs due to L1 influence.6 In addition, it should be noted that many learners progress beyond the BV to a post-BV stage (Dietrich et al. 1995) where irregular morphology precedes regular morphology, grammatical tense marking precedes grammatical aspect marking, and various morphological forms coexist at the same time. As mentioned before, learners will proceed through these various stages of language acquisition at different speeds and will fossilize at any one of them. Instances of nonsententials can be found in the BV as well as in post-BV stages of learner language. Hence, an issue for Klein and Perdue’s (1997) BV analysis is the problem of initial learners who produce more complex structures early on and seem to have passed beyond the BV stage. It should be pointed out, however, that even though nonsententials occur in the BV and other early stages of learner language, they need not be the only type of utterances present. Furthermore, it is important to realize that the BV or any other stage of learner language is not a simplified, incomplete, or faulty version of the TL. Corder (1981: 149) points out that a learner cannot simplify the TL since it has not been acquired yet. The occurrence of nonsententials in the BV or any other early stage of learner language can be triggered by several factors. First, learners lack most verbal and other bound morphology, as well as function words, such as auxiliaries, in early stages of L2 acquisition. Hence, in order to communicate, L2 learners will create a learner variety without these categories. This process bears resemblance to the development of pidgins, which represent a form of communication among people with mutually unintelligible languages and can also hardly be interpreted as simplification of the dominant language (see Winford, Chap. 11, this volume). As soon as L2 learners progress to the next stage of acquisition, more and more nativelike principles and syntactic functions will be applied. Second, according to Chomsky’s (1995) Minimalist Program, the economy principle requires that syntactic representations should contain as few constituents as possible, and syntactic derivations as few grammatical operations as possible. Herschensohn (2000), for instance, applies this principle to L2 acquisition: From the perspective of language learning and language use, it would be reasonable to assume that direct and succinct linguistic computations would be preferable to lengthy and redundant operations. (67)
Finally, certain pragmatic conditions might license the use of nonsententials in L2 acquisition, as has been attested for adult nonsententials (see Epilogue, this volume; Haegeman 1995; Lasser 1997). According to Lasser, for example, adult nonsententials with nonfinite verbs are used “for asserting, seeking information and requesting action on the part of the addressee” (33).
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.9 (555-616)
Nonsententials and second language acquisition
(15) Aber erst Nachrichten gucken but first news watch-inf ‘But first I/we will watch the news’
(Lasser 1997)
Here the adult nonsentential represents a declarative sentence used for assertion. This discussion clearly illustrates that learner languages at the BV level and beyond seem to exhibit properties that are syntactically very similar to nonsententials in casual adult speech. The following section will compare the properties of these two varieties in detail.
. Learner language and native adult nonsententials The discussion that follows will show that learner language at or beyond the BV stage is typified by some of these same features of adult nonsententials, namely the absence of Tense, the absence of Case, the co-occurrence of both, and the use of phrasal instead of sentential negation. There are, however, certain structural differences between the two. When comparing native adult nonsententials to similar structures found in the early interlanguage of L2 learners, generally similar characteristics can be attested. Early L2 structures by and large lack verbal inflectional morphology (18) through (20), nominal inflectional morphology (23), and determiners (9) (repeated in (16)) and (24). In addition, L2 learners generally use a type of phrasal negation rather than a sentential one (11) (repeated in (22)). The following are examples from early L2 learner data: NP: English: (16) Very good student
(Lee 2000)
VP: English: (17) Benny washing
(Richards & Lockhart 1996)
French: (18) Jouer de hockey play-inf of hockey ‘Play hockey’
(Prévost & White 2000a)
German: (19) Aber du sprechen italienisch but you-sg speak-inf Italian ‘But you speak Italian’
(Parodi 2000)
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.10 (616-687)
Nicola Work
(20) Nein, sprechen mit no, speak-inf with ‘Don’t speak with’
(Parodi 2000)
AP: English: (21) Mary so funny
(Ionin & Wexler 2002)
(22) But no possible German: (23) Schön diese Haus beautiful that-pl.fem house ‘Beautiful that house’
(Klein & Perdue 1992)
Dutch: (24) Meisje weg girl gone ‘The girl is gone’
(Klein & Perdue 1992)
These utterances produced by L2 learners can be analyzed as base-generated NPs, PPs, VPs, and APs, which constitutes – according to Progovac (Chap. 2, this volume) – their most economical derivation. An explanation of the absence of Tense, the absence of determiners on NPs, and the choice of phrasal negation in the BV all fall out from this basic analysis. One of the similarities between adult English NP nonsententials and those of L2 learners is that they lack Case. (25) Very good student (26) a. Nice lady! b. Old grudge Both the early L2 data in (25) and the native adult nonsententials in (26) occur without a determiner, exhibiting only an adjective (phrase) modifying the following hereby caseless noun.7 Caselessness is further illustrated by the German L2 learner data in (27): (27) Baustell vielleicht construction-no.case maybe ‘Maybe construction’
(Klein & Perdue 1992)
In this example, caselessness is evidenced by the absence of a determiner and by the absence of the ending -e. A second similarity between L2 (28) and adult (29) nonsententials is tenselessness. In each of these examples, the lexical verb lacks tense morphology. (28) Girl go
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.11 (687-759)
Nonsententials and second language acquisition
(29) Heart throb? Interestingly, each of these examples also has a bare noun in subject position, thus reflecting the co-occurrence of tenselessness and caselessness (see more detailed discussion below). Tenselessness in adult and L2 nonsententials is also illustrated by the absence of copula/auxiliary verbs in participial verb constructions, as illustrated in (30) and (31), respectively. (30) Problem solved (31) Car gone
(Klein & Perdue 1992)
L2 nonsententials in French and German typically use verbs in their infinitive form, as illustrated in (32) and (33), respectively. (32) a.
Toi aller à Guy you-tonic go-inf to Guy ‘You go to Guy’s place’ b. J’aller au cinéma I go-inf to the-masc movie-theater ‘I go/I am going to the movie theater’
(33) Ich schreiben I write-inf ‘I write/I am writing’
(Grondin & White 1996)
(Herschensohn 2004)
(Prévost & White 2000a)
Note, however, that in (32b) and (33), the subject occurs in its nominative form, rather than in a nonnominative form, as in (32a) (cf. Schütze 1997). Third, the data below (see also (5), (30), and (31)) show that like adult nonsententials, L2 nonsententials are also characterized by the co-occurrence of tenselessness and caselessness. In the learner data in (34) and the adult data in (35), the utterance lacks a copula/auxiliary verb and a determiner on the subject. (34) Meisje weg girl gone ‘The girl is gone’
(Klein & Perdue 1992)
(35) Battery dead This co-occurrence of tenselessness and caselessness provides further support for the claim that certain L2 utterances are nonsententials. Fourth and last, adult nonsententials and those found in L2 learner data both use a type of phrasal instead of sentential negation. The data in (2) and (4) above, repeated here in (36a) and (37), illustrate this feature. (36) a. Not urgent b. It isn’t urgent (37) No too hard
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.12 (759-791)
Nicola Work
Both the adult nonsentential in (36a) and the L2 nonsentential in (37) exhibit the use of phrasal negation, either in the form of not or no. Due to the aforementioned similarities between adult nonsententials and those produced by L2 learners, they may both be represented as base-generated small clauses (see Progovac, Chap. 2, this volume, for the details of this analysis). L2 learner nonsententials are represented in (38) and (39) and native adult nonsententials in (40) and (41). This basic analysis is also pursued by Radford (1990), for instance, for L1 acquisition. (38)
VP DP
V’
toi
V aller
(39)
AP Meisje
A’ A weg VP
(40)
V’
DP V
DP
stop
smoking
Him
AP
(41) Battery
A’ A dead
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.13 (791-852)
Nonsententials and second language acquisition
The L2 learner utterances presented thus far are clearly nonsentential, and fall out from the analysis developed in Progovac (Chap. 2, this volume). There are, however, characteristics of L2 learner data that distinguish it from nonsententials in adult speech. Specifically, the BV sometimes exhibits structural Case on subjects, variation in the availability of tenseless utterances, and the use of a different negative particle when using phrasal negation. First, according to Progovac (Chap. 2, this volume), subject pronouns will appear in the default case (accusative case in English) in native adult nonsententials due to the absence of the Tense node and hence the lack of a syntactic mechanism to assign structural nominative case to the subject. (42) a.
Him stop smoking? *He stop smoking? b. Me first *I first
The examples in (42) illustrate that adult native speakers of English will have to resort to the default case for English, namely the accusative, when using nonsententials. It is interesting to note, however, that nominative subject pronouns are frequently used in L2 nonsententials.8 (43) She no old (44) J’aller au cinéma I go-inf to-the-masc movie-theater ‘I go/I am going to the movies’ The L2 learner of English uses the subject pronoun in the nominative case (43) instead of the default accusative case. This sentence would, however, be ungrammatical in adult speech as illustrated by example (42). The French L2 learner also resorts to the nominative clitic pronoun je in (44). Haznedar and Schwartz’ (1997) study of Erdem, an L1 Turkish speaker learning English, shows that he employed subject pronouns almost entirely in the nominative case as well (see also other studies, e.g., Haznedar 2001, 2003). There are several possible explanations for this phenomenon. First, it may be that there exists a natural order for the acquisition of Case when learning an L2 and that the nominative case – whether it is the default case in the TL or not – will be acquired initially since in many languages it is the least marked form of the noun/determiner/pronoun. In other words, early in the acquisition process L2 learners may not have access to the default case of the L2 (only to that of their L1) since the default case can vary from language to language. Furthermore, as will be detailed below, default case may be absent from the L2 input learners receive, given that native speakers do not consistently employ nonsententials. In addition, nominative subjects are statistically much more frequent in the input L2 learners receive than pronouns in the default case (for a more detailed discussion on default case, see Epilogue, this volume).
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.14 (852-912)
Nicola Work
Data from French, furthermore, show that learners use default strong tonic pronouns and nominative clitic subject pronouns interchangeably, a feature that is absent from adult nonsententials of the type illustrated in (42). (45) Toi faire ça you-sg.tonic do-inf that ‘You do that’ (46) J’aller au cinéma I go-inf to-the-masc movie-theater ‘I go/I am going to the movies’ This interchangeable use of certain features is not uncommon in L2 learner language. In the L2 literature, for instance, researchers distinguish between true root infinitives (explained by Rizzi’s 1993/1994 Truncation Hypothesis) as in (45), and raised verbs with missing or default inflection as in (46).9 Herschensohn (2001) explains that French root infinitives are characterized by the use of a null or tonic pronoun, an infinitival verb, and preverbal negation. If the L2 phrase contains a nominative subject clitic, an inflected verb (i.e., with default inflection), and postverbal negation, it is, according to Herschensohn, an instance of a raised verb. Many researchers have found in their L2 data that learners use both root infinitives and raised infinitives simultaneously (Eubank & Beck 1998; Herschensohn 2001; Pierce 1992; Prévost 1997; Prévost & White 1999; Rizzi 1993/1994; Wexler & Weinberg 1994).10 In native adult speech, however, strong pronouns in subject position are ungrammatical if used without a subject clitic, as illustrated in (47).11 (47) a.
Moi, je vais au cinéma I-tonic I go to-the-masc movie-theater ‘Me, I am going to the movies’ b. Je vais au cinéma I go to-the-masc movie-theater ‘I go/I am going to the movies’ c. *Moi vais au cinéma I-tonic go to-the-masc movie-theater ‘I go/I am going to the movies’
In learner language, however, strong pronouns without resumptive clitics as well as both clitic and tonic pronouns with nonfinite verbs are attested (see (45) and (46)). A second difference between adult and L2 nonsententials in English is the distribution and availability of tenseless verb constructions. English native adult nonsententials often contain verbs in their participle forms; these forms are both present (48) and past (49), regular (49a) and irregular (49b). (48) How about staying? (49) a. Mailed package b. Left town
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.15 (912-982)
Nonsententials and second language acquisition
The L2 learner of English, on the other hand, seems to have a much smaller repertoire of possible verb forms. (50) Charlie get up first (51) Benny washing
(Klein & Perdue 1992) (Richards & Lockhart 1996)
Most of the nonsententials in early L2 English data contain either nonfinite verbs (50), or verbs in the present participle (51). Few, if any, verbs in the past participle – regular or irregular – are attested in L2 nonsententials. Krashen’s (1977) “natural order” of morphemes for English L2 acquisition might shed light on the reason why. According to Krashen, L2 learners go through different stages when acquiring inflectional morphemes. (52)
-ing plural copula ↓ auxiliaries articles ↓ irregular past ↓ regular past
Therefore, it seems that early L2 learners (when using the BV or post-BV) might not yet have mastered the inflectional morphemes for the past participle forms, and may have a smaller repertoire of possible verb forms in their nonsentential grammar than adult native speakers do. As pointed out in, for example, Bardovi-Harlig (2000), the acquisition of perfect tenses lags behind the acquisition of past and progressive. Native adult and L2 nonsententials also differ in their use of a negative particle. As was shown by Progovac (Chap. 2, this volume) and previously in this section, English native adult nonsententials employ – just like early L2 learners do – a type of phrasal negation (such as not) instead of sentential negation (such as didn’t). (53) a. Not sweet b. Not urgent Both adult APs in (53) include the negative particle not as part of the phrasal negation before the adjective. However, even though both adults and L2 learners use a type of phrasal negation, they differ in the choice of the negative particle. Instead of the particle not, which is found in the adult English nonsentential data above, L2 learners of English utilize the particle no, as illustrated in (54).
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.16 (982-1049)
Nicola Work
(54) a. She no old b. But no possible c. No too hard Interestingly the L2 learner example with the negative particle no instead of not would be ungrammatical in normal adult speech, as illustrated in (55). (55) a. Not sweet b. *No sweet The reason for this discrepancy may again be the order of acquisition of certain grammatical items and structures. Ellis (1994) demonstrates the different steps involved in the acquisition of negation for L2 English speakers: (a) learners acquire external negation first and mainly use the negative particle no; (b) internal negation is employed, that is, the negative particle moves inside the utterance; (c) the negative particle is attached to the modal verbs; and (d) the TL rule regarding negation is acquired and not is regularly used as the negative particle. Therefore, English native adult and L2 nonsententials differ regarding negation because the L2 learner only has access to the negative particle no at this early stage. The same holds true for negation in L2 German. (56) a.
Nein helfen no help-inf ‘Don’t help’ b. Nein spielen Katze no play-inf cat ‘Don’t play with the cat’
The L2 German data above demonstrate that at this stage of acquisition learners use negation with nein (56) instead of using the negative particle nicht. In adult nonsentential speech nicht must be used for phrasal or sentential negation (57a), whereas nein is ungrammatical (57b). (57) a.
Nicht helfen not help-inf ‘Don’t help’ b. *Nein helfen no help-inf ‘Don’t help’
As was the case regarding the acquisition of English, this discrepancy may again stem from the order of acquisition. In German L2 acquisition learners employ sentenceexternal negation (nein) initially, before using sentence-internal negation with nicht in postverbal and preverbal positions interchangeably in the next stage. In the following stage learners eliminate preverbal negation before acquiring the TL rules regarding negation placement with direct objects and free adverbials in the final stage (Ellis
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.17 (1049-1105)
Nonsententials and second language acquisition
1994). Hence, the nonsententials of native adult speakers and nonnative speakers share similar syntactic structures but differ in their use of the negative particle.12 In conclusion, early learner language shows many similarities with native adult nonsententials, including the use of nonfinite verb forms, the absence of structural Case, and the possible absence of determiners on subjects. However, there are also some differences between early L2 language and adult nonsententials. First, nominative case is frequently used in L2 learner language, even though the default case in English is accusative. My suggestion was that this is due to the fact that the learner may not yet know that the default case in English is accusative, given that nominative is the default case in a variety of other languages. Also, the fact that external negation is used in the learner language (i.e., no in English and nein in German) can likewise be ascribed to the fact that the learner has not yet acquired internal negation (i.e., not in English, and nicht in German). Similarly, the relatively late acquisition of the past participle forms may be a factor in the paucity of past participle nonsententials in learner language. The next section addresses the significance of these findings.
. Nonsentential grammar and theories of L2 acquisition As established in the previous section, early L2 learner language shows many similarities with native adult nonsententials, including the use of nonfinite verb forms and the possible absence of determiners. However, there are also some differences, including the use of the nominative case in learner language; the use of external, rather than internal, negation; and the paucity of root small clauses with past participles. Some of these differences were argued to follow from independent principles having to do with the general stages of L2 acquisition. All in all, given that L2 learner data share many similarities with adult nonsentential speech, and given that most of the differences between the two varieties can be ascribed to independent factors, it seems that Progovac’s (Chap. 2, this volume) general analysis of nonsententials can be extended to account for nonsententials present in early learner varieties. According to Progovac, nonsententials found in adult speech are systematic and rule governed, and represent products of a nonsentential grammar that coexists with a full grammar and is used only under certain circumstances. These nonsententials are arguably base-generated phrases, small clauses, or infinitives, rather than full sentential structures that have undergone deletion/ellipsis. There are two types of data discussed in this chapter that speak against an ellipsis analysis: the first involves the use of root infinitives in French and German, as illustrated in (32a) and (33), repeated below: (58) Toi aller à Guy you-tonic go-inf to Guy ‘You go to Guy’s place’
(Grondin & White 1996)
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.18 (1105-1176)
Nicola Work
(59) Ich schreiben I write-inf
(Prévost & White 2000a)
It is not clear what kind of full sentence such infinitive clauses would be derived from. The second argument has to do with the use of caseless subjects in German, which can also not be derived from a full sentential source. The example from (8) (and (27)) is repeated here: (60) Baustell vielleicht construction-no.case maybe ‘Maybe construction’
(Klein & Perdue 1992)
In a full German sentence, one would have to use a form such as die Baustelle, in which both the determiner die and the ending -e on the noun would be obligatory, and in which both forms would seem to be responsible for identifying Case. This kind of nonsentential/nonelliptical approach to L2 is compatible with, for example, Vainikka and Young-Scholten’s (1994, 1996a, 1996b) Minimal Trees Hypothesis (see also Radford 1990 for a similar analysis of L1 acquisition). According to Vainikka and Young-Scholten (1994, 1996a, 1996b), the initial grammar of L2 learners lacks functional categories, that is, neither functional categories from the L1 nor from UG will be present. Hence, the initial grammars only contain lexical categories and their projections (e.g., NP, VP, PP, AP) and possibly one underspecified functional projection (FP). Det, Infl, and Comp and their associated projections then emerge gradually as more and more L2 input becomes available. That means that learners progressively add functional categories and their projections to their interlanguage grammars in a bottom-up fashion so that IP exists before CP, for example. While there is L1 transfer of certain lexical items, no transfer from the L1 is predicted in the domain of functional categories. Furthermore, the absence of structural Case on subjects (as evidenced by the possible absence of determiners, for instance) correlates with the absence of Tense and agreement on their predicates. This also goes hand in hand with Vainikka and Young-Scholten’s (1994, 1996a, 1996b) prediction that during the first stage of L2 acquisition little subject-verb agreement occurs and learners resort to the use of infinitives, bare stems, or default suffixes instead of finite verbs. Our account of nonsententials is also compatible with Rizzi’s (1993/1994) and Haegeman’s (1995) truncation approach to acquisition. The Truncation Hypothesis originally constituted a theory of L1 acquisition but was applied to L2 acquisition by Prévost and White (2000a, 2000b).13 According to this hypothesis, every projection above the truncation site is stripped away, which means that projections cannot be removed from the middle of a syntactic tree. Even though all functional categories are assumed to be present in the child’s L1 grammar, it does not mean that all functional categories are projected. Hence, the child’s structure may be truncated at any point below CP, so that root VPs, root TPs, root NegPs, and root AgrPs are possible in addition to CPs. According to Rizzi and Haegeman, truncation occurs in L1 child grammars and children are assumed to lack the Root Principle dictating that the root
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.19 (1176-1223)
Nonsententials and second language acquisition
clause must be a CP. Prévost and White (2000a) make several predictions about L2 acquisition regarding the Truncation Hypothesis. First, it is assumed that truncation may, but does not necessarily have to, occur in L2 acquisition, especially in adult L2 acquisition. Therefore, VP, IP, and CP roots could be expected in the same stages of learner language. If, for example, a root VP is projected, the verb should occur in the nonfinite form. Should the verbs, on the other hand, be in IP and CP roots, they will be finite. However, Prévost and White (2000a) add that adult learners were found to “include nonfinite verbs in CPs, null subjects in CPs, subject clitics in root infinitives and nonfinite verbs preceding negative markers” (224). Hence, these results for adult L2 learners might be more consistent with the missing inflection hypothesis. In addition, Progovac (Chap. 2, this volume) maintains that native adult nonsententials result from feature underspecifications, namely the fact that Tense and Case can be selected with unspecified/default features and therefore make a projection of TP unnecessary. The Valueless Features Approach by Eubank (1993/1994, 1996) also claims that the features in L2 learner interlanguage are valueless or inert before later being adjusted to the TL values by means of more input.14 Feature strength impacts word order: in languages like English, where I has weak V-features, finite verbs remain within the VP, whereas in languages like French, where I is strong, the verb raises to I to check its features. Since, in accordance with Eubank’s approach, feature values are neither strong nor weak, finite verbs can alternate between raising or not raising. Furthermore, inertness is a temporary phenomenon that only exists in early interlanguage grammars. As soon as inflectional morphology emerges, feature strength (strong or weak) will be acquired.
. Teacher talk When addressing L2 or foreign language (FL) learners, teachers and native speakers tend to modify their speech, creating a special kind of discourse called teacher talk (in a classroom setting) or foreigner talk (in an immersion setting) respectively. Several studies (cited in Ellis 1994) have investigated the modifications made when speaking to nonnative speakers learning an L2. Modifications such as slower rate of speech, higher number of pauses, and syntactic adjustments denote teacher/foreigner talk. Generally, teachers will use shorter utterances and less subordination, less marked forms such as the past tense, and more declaratives and statements rather than questions. Chaudron (1988: 55) affirms that “the adjustments in teacher speech to nonnativespeaking learners serve the temporary purpose of maintaining communication – clarifying information and eliciting learners’ responses.” This discourse variety used by teachers might not sound natural outside the classroom due to its modifications and context specific to a classroom setting. And, even though ungrammatical teacher talk is rare, it does occur from time to time. In comparing teacher talk to early learner language it appears that different mechanisms must be at work to create and analyze these types of utterances. Even though
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.20 (1223-1298)
Nicola Work
the utterances produced by teachers might at first glance resemble native adult nonsententials and L2 learner language, and even though they follow a similar syntactic structure, their analysis may, however, differ. NP: (61) No examples VP: (62) Necessary you speak
(Hawkins 2001)
(63) You have book
(Hawkins 2001)
(64) Say again
(Chaudron 1988)
AP: (65) He not here
(Chaudron 1988)
(66) The writing not important
(Chaudron 1988)
(67) Necessary for school
(Chaudron 1988)
Whereas I argued in the previous section that certain data in early L2 learner language are nonsententials, it seems that English teacher talk data are more sentential in nature. If this is the case, then the analysis would be based on the generation of fullfledged sentences, followed by the intentional deletion/omission of certain elements. The utterances in (68) and (69), instances of teacher talk, support this possible analysis. (68) Necessary you speak [It is] necessary [that] you speak
(Hawkins 2001)
(69) You have book You have [a/the] book
(Hawkins 2001)
In (6) the teacher uses a subordinate clause, a complex structure that is not attested in the early interlanguage of L2 learners, but omits the subject pronoun and the copula in the first clause as well as the complementizer that in the second clause. Following a bottom-up analysis, sentences such as (68) and (69) are not expected, given that there are omissions from the middle of a syntactic structure. These types of utterances therefore may be better analyzed with a sentential approach and subsequent deletion of certain categories and projections. When comparing teacher talk data in more detail to nonsententials typical of casual adult speech, some further differences are attested. The data below illustrate the fundamental differences between nonsententials in casual adult speech (70) and the teacher talk utterances (61) through (69). (70) a. b. c. d.
Him in jail? Me dancing Heart not healthy? Car dead
(Progovac, Chap. 2, this volume)
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.21 (1298-1349)
Nonsententials and second language acquisition
Whereas a native speaker will resort to the default case for English (the accusative case) in (70a) and (70b), the teacher talk data show the use of the subject pronoun in the nominative case (65), which again points toward a sentential analysis of the teacher data. In summary, teacher talk, a type of discourse artificially created by foreignlanguage teachers, might be better analyzed following a sentential analysis with subsequent deletion of certain words/structures at the discretion of the teacher, given the obvious differences between this variety and adult and L2 nonsententials. More research is needed, however, to substantiate this basic proposal.
. Conclusion This chapter has explored a nonsentential analysis of early learner language, and has highlighted similarities between L2 nonsententials and nonsententials typical of casual adult speech. In the L2 data presented here, learners initially make use of the nonsentential grammar since the full grammar is not yet developed (see other L2 data collections/corpora: e.g., Klein & Perdue 1992; Prévost & White 1999, 2000a, 2000b; Vainikka & Young-Scholten 1994, 1996a, 1996b). Some L2 learners will improve beyond the nonsentential stage and hence develop a second system – the near-full grammar.15 Other learners will remain at the level of a nonsentential grammar in the L2 and may never develop beyond that stage. Even though I argue that nonsententials in L2 learner data are base-generated phrases and small clauses, this approach does not seem to apply to similar structures found in teacher talk. Instead, a sentential analysis with subsequent deletions seems to provide a better account of the data. And since teacher talk seems to be (a) more of an artificially created type of discourse only relevant when talking to L2 students in the classroom or nonnative speakers in an immersion setting, and (b) different from nonsententials found in adult data, it seems appropriate to suggest a sentential analysis. It is then up to the teacher to decide what element(s) should be left out in discourse depending on the level of the L2 learners as well as the complexity and difficulty of the discourse and the topic to be discussed.
Notes * I would like to thank several people who have contributed to this chapter in various ways. First of all, I would like to thank the participants of the Wayne State University Humanities Center Working Group for inviting me to be part of their group and for fruitful discussions about the topic: Ellen Barton, Eugenia Casielles, Walter Edwards, Kate Paesani, Ljiljana Progovac, and Pat Siple. My special thanks go to the coeditors of this volume, Ellen Barton, Eugenia Casielles, Kate Paesani, and Ljiljana Progovac, for their careful reading and detailed comments. Thanks to two anonymous reviewers who helped me organize my chapter as well as disentangle my ideas. Last
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.22 (1349-1421)
Nicola Work
but not least, I am grateful to my husband for having interest in my research and proofreading this chapter. . Sauter (2002) adapts White’s (2000) extension of the terms used by Schwartz and Sprouse (1996), but does not use Partial Access to refer to any of the different positions. See Sauter for details. . White (2003) addresses the fact that some researchers found what are called “wild grammars” in the interlanguage of their L2 learners, that is, properties that are part of neither the learner’s L1 nor the L2. . Nominative case seems to be the unmarked case in German, as opposed to English, which has accusative case as its unmarked case (see Progovac, Chap. 2, this volume). . As Progovac (Chap. 2, this volume) points out, determiners have two functions: one structural, to check case, and one referential/quantificational. L2 learners most likely use determiners to quantify or to refer. . See L2 data collections/corpora for more examples illustrating the aforementioned features of L2 learner data (e.g., Klein & Perdue 1992; Prévost & White 1999, 2000a, 2000b; Vainikka & Young-Scholten 1994, 1996a, 1996b). . For a more detailed discussion of theoretical problems with Klein and Perdue’s Basic Variety see DeGraff (2001) and Jordens (1997b). . Some BV nonsententials do have NPs with determiners (i), some do not (ii): (i)
a.
b.
Die meisje honger the-fem girl hunger ‘The girl is hungry’ La table tout bon the-fem table all good-masc ‘The table is all fine’
(ii) Meisje weg girl gone ‘The girl is gone’
(Dutch; Klein & Perdue 1992)
(French; Klein & Perdue 1992)
(Dutch; Klein & Perdue 1992)
These determiners may be used for purely referential (rather than syntactic) purposes. The reader is referred to Progovac (Chap. 2, this volume) for further discussion. . It is possible that learners assume that nominative is the default case in such examples (e.g., if it is the default case in the L1, or if they have not had any evidence to the contrary). . For a detailed analysis of root infinitives in several different languages see, for example, Roberts (1999). . For a more detailed discussion of the literature see De Cat (2004). . Third-person singular tonic pronoun lui can occur in subject position in Canadian French. In European French lui can only be in that position if it bears contrastive stress (De Cat 2004: 52– 53). (i)
Et lui s’appelle Fernand le pélican and him-tonic refl-calls-3sg Fernand the pelican ‘And that one’s called Fernand le pelican’
(French; De Cat 2004)
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.23 (1421-1519)
Nonsententials and second language acquisition . There may be other differences between adult nonsententials and the BV. For example, adult speakers frequently create nonsententials that can be analyzed as base-generated PPs (e.g., Martha Stewart in jail), whereas few instances of such nonsententials seem to be found in the early L2 data. More research is needed to substantiate this claim, however. . See Guasti (2002) for a detailed explanation of this hypothesis regarding L1 acquisition. . However, Eubank (1993/1994, 1996) argues that all the functional projections are present even in the BV. . I call it “near-full grammar” since learners have been attested to achieve near-native fluency and therefore their full grammar might still differ to some degree from a native speaker’s full grammar.
References Bardovi-Harlig, K. (2000). Tense and Aspect in Second Language Acquisition: Form, Meaning and Use. Oxford: Blackwell. Bierwisch, M. (1997). Universal Grammar and the Basic Variety. Second Language Research, 13, 348–366. Bley-Vroman, R. (1983). The comparative fallacy in interlanguage studies: The case of systematicity. Language Learning, 33, 1–17. Chaudron, C. (1988). Second Language Classrooms: Research on Teaching and Learning. Cambridge: Cambridge University Press. Chomsky, N. (1981a). Lectures on Government and Binding. Dordrecht: Cinnaminson. Chomsky, N. (1981b). Principles and Parameters in syntactic theory. In N. Hornstein & D. Lightfoot (Eds.), Explanation in Linguistics: The Logical Problem of Language Acquisition (pp. 32–75). London: Longman. Chomsky, N. (1986). Knowledge of Language: Its Nature, Origin and Use. New York: Praeger. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: MIT Press. Comrie, B. (1997). On the origin of the Basic Variety. Second Language Research, 13, 367–373. Corder, S. P. (1967). The significance of learners’ errors. International Review of Applied Linguistics, 5, 161–169. Corder, S. P. (1981). Formal simplicity and functional simplification. In R. Andersen (Ed.), New Dimensions in Second Language Research (pp. 146–152). Rowley, MA: Newbury House. De Cat, C. (2004). Apparent non-nominative subjects in L1 French. In P. Prévost & J. Paradis (Eds.), The Acquisition of French in Different Contexts: Focus on Functional Categories (pp. 51–88). Amsterdam: John Benjamins. DeGraff, M. (2001). On the origin of Creoles: A Cartesian critique of Neo-Darwinian linguistics. Linguistic Typology, 5, 213–310. Dietrich, R., Klein, W., Noyau, C., & Coenen, J. (1995). The Acquisition of Temporality in a Second Language. Amsterdam: John Benjamins. Ellis, R. (1994). The Study of Second Language Acquisition. Oxford: Oxford University Press. Eubank, L. (1993/1994). On the transfer of parametric values in L2 development. Language Acquisition, 3, 183–208. Eubank, L. (1996). Negation in early German-English interlanguage: More valueless features in the L2 initial state. Second Language Research, 12, 73–106.
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.24 (1519-1626)
Nicola Work
Eubank, L. & Beck, M.-L. (1998). OI-like effects in adult L2 acquisition. In A. Greenhill, M. Hughes, H. Littlefield, & H. Walsh (Eds.), Proceedings of the 22nd Annual Boston University Conference on Language Development (pp. 189–200). Somerville, MA: Cascadilla Press. Grondin, N. & White, L. (1996). Functional categories in child L2 acquisition of French. Language Acquisition, 5, 1–34. Guasti, M. T. (2002). Language Acquisition: The Growth of Grammar: Cambridge, MA: MIT Press. Haegeman, L. (1995). Root infinitives, tense, and truncated structures in Dutch. Language Acquisition, 4, 205–255. Hawkins, R. (2001). Second Language Syntax: A Generative Introduction. Oxford: Blackwell. Haznedar, B. (2001). The acquisition of the IP system in child L2 English. Studies in Second Language Acquisition, 23, 1–39. Haznedar, B. (2003). Missing surface inflection in adult and child L2 acquisition. In J. M. Liceras et al. (Eds.), Proceedings of the 6th Generative Approaches to Second Language Acquisition Conference (GASLA 2002) (pp. 140–149). Somerville, MA: Cascadilla Press. Haznedar, B. & Schwartz, B. D. (1997). Are there optional infinitives in child L2 acquisition? In E. Hughes, M. Hughes, & A. Greenhill (Eds.), Proceedings of the 21st Annual Boston University Conference on Language Development (pp. 257–268). Somerville, MA: Cascadilla Press. Herschensohn, J. (2000). The Second Time Around: Minimalism and L2 Acquisition. Amsterdam: John Benjamins. Herschensohn, J. (2001). Missing inflection in second language French: Accidental infinitives and other verbal deficits. Second Language Research, 17, 273–305. Herschensohn, J. (2004). What Can Acquisition Studies Contribute to Instruction of French Inflection? Paper presented at the Association of French Language Studies. Ionin, T. & Wexler, K. (2002). Why is ‘is’ easier than ‘-s’? Acquisition of tense/agreement morphology by child second language learners of English. Second Language Research, 18, 95–136. Johnson, M. (2004). A Philosophy of Second Language Acquisition. New Haven: Yale University Press. Jordens, P. (1997a). Introducing the Basic Variety. Second Language Research, 13, 298–300. Jordens, P. (Ed.). (1997b). Introducing the Basic Variety. Second Language Research, 13 [Special Issue]. Klein, W. F. & Perdue, C. (1992). Utterance Structure: Developing Grammars Again. Amsterdam/Philadelphia: John Benjamins. Klein, W. F. & Perdue, C. (1997). The Basic Variety: Or, couldn’t natural languages be much simpler? Second Language Research, 13, 301–347. Krashen, S. (1977). Some issues relating to the Monitor Model. In H. D. Brown, C. Yorio, & R. Crymes (Eds.), On TESOL ’77 (pp. 144–158). Washington, DC: TESOL. Lado, R. (1957). Linguistics Across Cultures. Ann Arbor: University of Michigan Press. Lasser, I. (1997). The interpretation of root infinitive constructions in adult and child German. In J. C. Schaeffer (Ed.), The Interpretation of Root Infinitives and Bare Nouns in Child Language (pp. 26–64). Cambridge, MA: MIT Press. Lee, J. F. (2000). Tasks and Communicating in Language Classrooms. Boston: McGraw-Hill. Meisel, J. M. (1997). The L2 Basic Variety as an I-language. Second Language Research, 13, 374– 385. Parodi, T. (2000). Finiteness and verb placement in second language acquisition. Second Language Research, 16, 355–381.
JB[v.20020404] Prn:9/08/2006; 15:06
F: LA9308.tex / p.25 (1626-1727)
Nonsententials and second language acquisition
Pierce, A. (1992). Language Acquisition and Syntactic Theory: A Comparative Analysis of French and German Child Grammars. Dordrecht: Kluwer. Prévost, P. (1997). Truncation and root infinitives in second language acquisition of French. In E. Hughes, M. Hughes, & A. Greenhill (Eds.), Proceedings of the 21st Annual Boston University Conference on Language Development (pp. 453–464). Somerville, MA: Cascadilla Press. Prévost, P. & White, L. (1999). Finiteness and variability in SLA: More evidence for missing surface inflection. In A. Greenhill, H. Littlefield, & C. Tano (Eds.), Proceedings of the 23rd Annual Boston University Conference on Language Development (pp. 575–586). Somerville, MA: Cascadilla Press. Prévost, P. & White, L. (2000a). Accounting for morphological variation in second language acquisition: Truncation or missing inflection? In M.-A. Friedemann & L. Rizzi (Eds.), The Acquisition of Syntax: Studies in Comparative Developmental Linguistics (pp. 202–235). Harlow, UK: Longman. Prévost, P. & White, L. (2000b). Missing surface inflection or impairment in second language acquisition? Evidence from tense and agreement. Second Language Research, 16, 103–133. Radford, A. (1990). Syntactic Theory and the Acquisition of English Syntax: The Nature of Early Child Grammars of English. Oxford: Blackwell. Richards, J. & Lockhart, C. (1996). Reflective Teaching in Second Language Classrooms. Cambridge: Cambridge University Press. Rizzi, L. (1993/1994). Some notes on linguistic theory and language development: The case of root infinitives. Language Acquisition, 3, 371–393. Roberts, I. (1999). Verb movement and markedness. In M. DeGraff (Ed.), Language Creation and Language Change: Creolization, Diachrony and Development (pp. 287–327). Cambridge, MA: MIT Press. Sauter, K. (2002). Transfer and Access to Universal Grammar in Adult Second Language Acquisition. Groningen: University Library. Schütze, C. T. (1997). INFL in Child and Adult Language: Agreement, Case and Licensing. Ph.D. dissertation, Massachusetts Institute of Technology. Schwartz, B. D. (1997). On the basis of the Basic Variety. Second Language Research, 13, 386–402. Schwartz, B. D. & Sprouse, R. A. (1996). L2 cognitive states and the Full Transfer/Full Access Model. Second Language Research, 12, 40–72. Selinker, L. (1972). Interlanguage. International Review of Applied Linguistics, 10, 209–231. Vainikka, A. & Young-Scholten, M. (1994). Direct access to X’-theory: Evidence from Korean and Turkish adults learning German. In T. Hoekstra & B. D. Schwartz (Eds.), Language Acquisition Studies in Generative Grammar (pp. 265–316). Amsterdam: John Benjamins. Vainikka, A. & Young-Scholten, M. (1996a). The early stages in adult L2 syntax: Additional evidence from Romance speakers. Second Language Research, 12, 140–176. Vainikka, A. & Young-Scholten, M. (1996b). Gradual development of L2 phrase structure. Second Language Research, 12, 7–39. Wexler, K. & Weinberg, A. (1994). Optional infinitives, head movement and the economy of derivations. In D. Lightfoot & N. Hornstein (Eds.), Verb Movement (pp. 305–362). Cambridge: Cambridge University Press. White, L. (2000). Second language acquisition: From initial to final state. In J. Archibald (Ed.), Second Language Acquisition and Linguistic Theory (pp. 130–155). Oxford: Blackwell. White, L. (2003). Second Language Acquisition and Universal Grammar. Cambridge: Cambridge University Press.
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.1 (47-128)
chapter
How language adapts to the brain An analysis of agrammatic aphasia* Herman Kolk In this chapter I argue that the major symptoms of agrammatic aphasia reflect functional reorganization. The underlying deficit is characterized as a reduced possibility to keep sentence elements in computational simultaneity. Language adapts to this brain condition in two ways: by an increase in the frequency of elliptical1 utterances and by an increase in the frequency of covert repair behaviors. From these adaptations the language output of these patients derives its telegraphic character and its slow rate. Linguistic, experimental, computational, and brain-imaging evidence for this theory is reviewed. Finally, I discuss functional reorganization in other disorders.
.
Overview
Agrammatism is a disorder that leads to difficulties with sentences, either in production or in comprehension. That these difficulties concern sentences rather than words as such is evident from the fact that word comprehension and word production can be relatively spared. Agrammatism occurs in several clinical populations, but has been studied most systematically in patients with Broca’s aphasia. Whereas the large majority of these studies are devoted to comprehension, the present chapter is about production. I will start with describing the inherent variability of the symptoms of agrammatic production (Section 2). Then a number of theoretical approaches will be discussed that attempt to explain this variability. It is concluded from this discussion that among these approaches, only the temporal-window hypothesis explains both between- and within-patient variability (Section 3). Various sorts of empirical evidence for this temporal-window hypothesis will then be described. They relate to syntactic priming, conceptual number, and tense (Section 4). I will then argue that the temporal-window hypothesis is insufficient to explain the specific symptoms of agrammatic production and that, for a complete account, processes of adaptation should be invoked. I propose two such processes: corrective and preventive adaptation (Section 5). In case of corrective adaptation, the incomplete sentence representation is repaired by means of a covert restart of the computational process. This process is held responsible for the
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.2 (128-183)
Herman Kolk
slow rate of speech. Empirical evidence for this hypothesis is presented (Section 5.1). Preventive adaptation means that there is a bias to select simple types of construction, either simple but complete sentences, or elliptical constructions. Recent computational and brain-imaging work is discussed that is consistent with these notions (Section 5.2). In support of the hypothesis that there is a bias to select elliptical forms, I present evidence that agrammatic speech obeys the rules of normal (non-aphasic) grammatical ellipsis (Section 5.3), and that the relative frequencies with which the different types of elliptical construction occur in normal conversations are highly similar to the relative frequencies of comparable incomplete constructions in adult aphasics and in normal young children (Section 5.4). In support of the hypothesis that the selection of elliptical forms results from adaptation, I review a number of studies that all indicate that the selection of these forms is task dependent (Section 5.5.). I end up with a general conclusion (Section 6). In an appendix, I present evidence for the role of adaptation in disorders other than agrammatism.
. Variability of symptoms Symptoms of agrammatic production have traditionally been assessed by means of analysis of spontaneous speech (Goodglass & Kaplan 1983).2 Three main types of symptoms of spontaneous speech have been established in this way: (a) reduced variety of grammatical form; the sentences that are produced have little subordination or phrasal elaboration; as these symptoms relate to sentence form, we will call them syntactic symptoms, where the term syntactic is used in a purely descriptive way; (b) omission of function words – articles, pronouns, auxiliaries, copulas, prepositions, and the like – and inflections; all these symptoms relate to grammatical morphology and will therefore be referred to as morphological; and (c) slow rate of speech or nonfluency, to be referred to as the rate symptom. Whereas the above symptoms have been established for English-speaking patients, similar symptoms occur in many other languages (Menn & Obler 1990). It is important to realize that these symptoms do not appear in an all-or-none fashion. Some patients show these symptoms just a bit more often than a normal speaker; other patients have them in almost all their utterances. In a study with 22 Dutch-speaking Broca’s aphasics, Hofstede (1992) looked at the frequency of syntactic, morphological, and rate symptoms. With respect to syntactic symptoms, a mean percent of embedded clauses of 6% was observed, as compared to 22% in a normal control group, but the frequency of this syntactic symptom varied from 0% up to 21%. The same variability was apparent in the omission rate of grammatical morphology: It varied from 98% to 10%, almost as low as the control group, who omitted 8%. Finally, variability was also present in the rate symptom, which went from 23 up to 90 words per minute, with the control group producing an average of 145 words per minute. Later work with a group of 37 English-speaking patients demonstrated similar variability (Rochon, Saffran, Berndt, & Schwartz 2000). As the latter authors
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.3 (183-236)
An analysis of agrammatic aphasia
indicate, there appears to be continuity on syntactic, morphological, and rate symptoms, both within the patient group and between the patient group and the normal controls. The implications of these observations are twofold. First, there is betweenpatient variation in the degree in which symptoms are present in individual patients. Second, because very few patients exhibit a particular symptom 100% of the time, there is within-patient variation: a symptom may be present on one occasion and fail to appear on other occasions. This probabilistic character of aphasic symptoms is not limited to sentence production. As is evident from Goodglass (1993), inconsistency is the hallmark of aphasic behavior: a word that is appropriately produced or understood at one time will go wrong next time, and vice versa (see also Butterworth, Howard, & McLoughlin 1984). At the same time, inconsistency is not a necessary outcome of brain dysfunction. It is not present in Alzheimer’s disease, for instance, where if a particular word is no longer understood, it will remain so on subsequent trials (e.g., Chertkow & Bub 1990). To me this means that the variability of aphasic symptoms needs to be accounted for, as much as the symptoms themselves. One way to approach variability is to assume multiple deficits. In principle, each patient could have a different deficit and should be studied in her own right. By comparing individual cases, we can hope to find dissociations and double dissociations, and such findings could help us to identify components, which are independently spared or impaired. This approach has been strongly advocated by Badecker and Caramazza (1985). Although for areas like reading and semantic processing, this procedure has led to a number of well-established dissociations, this is not the case for agrammatic production. The most influential proposal is that within the group of agrammatic speakers, there would be a dissociation between syntactic and morphological aspects of agrammatism. Miceli, Mazzuchi, Menn, and Goodglass (1983) describe an Italian-speaking patient as having “an almost pure morphological disorder” (p. 75), syntax being almost entirely spared. Berndt (1987) claims that the same dissociation is present in the Dutch-speaking agrammatic patient K, described by Kolk, van Grunsven, and Keyser (1985). Such dissociations would indicate the existence of two independent deficits, a morphological and a syntactic one. However, the speech of both patients was characterized by a high number of nonfinite clauses, in which either the verb was lacking or the verb was used in the form of an infinitive or a past participle. For such clauses, it is hard to argue that they are syntactically “normal.” At least for languages like English, Dutch, and German, a “normal” sentence must contain an inflected verb. In a further effort to find support for this double-deficit hypothesis, Rochon et al. (2000) conducted a large-scale factor analytic study involving 37 English-speaking patients with Broca’s aphasia. All patients differed significantly from the controls in speech rate. With respect to syntactic and morphological symptoms, different patterns were observed. First of all, there were patients (the patients in “cluster I”) who differed from the controls in syntactic symptoms only. So the syntactic symptoms can appear in the absence of morphological symptoms, suggesting a dissociation. Second, with respect to morphology, there also appeared to be a dissociation. One group (cluster IV) showed more morphological than syntactic symptoms,
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.4 (236-299)
Herman Kolk
whereas the reverse was true for another group (cluster III). These results were taken to confirm the hypothesis that morphological and syntactic aspects of sentence production can be selectively impaired in agrammatism. I dare say, however, that there are several reasons to doubt this conclusion. First, there was no control for differences in severity. It is possible that the contrast observed between patients with only syntactic symptoms and patients with both syntactic and morphological symptoms is due to a difference in severity in one single underlying deficit. In fact, the theory proposed by Friedmann and Grodzinsky (1997) – to be further discussed below – would account for the contrast in this way. Furthermore, the second dissociation did not come out entirely as one would expect. Instead of a dissociation between syntactic and morphological symptoms, there was a dissociation between syntactic symptoms and symptoms related to inflection omission on the one hand and symptoms related to function word omission on the other. It seems that the case for the existence of two independent grammatical deficits has not yet been made in a convincing way. There may be a good reason for this failure to establish a dissociation between syntactic and morphological agrammatism. As argued, for example, by Barton and Progovac (2005) and Progovac (Chap. 2, this volume), syntactic and morphological “symptoms” in nonsentential speech are but two facets of the same coin. The basic argument is that the lack of finiteness, that is, inflection on the verb, correlates with the lack of nominative case assignment to the subject position, which in turn licenses subject drop, determiner drop in the subject position, or the default case assignment in the subject position (for further details of this syntactic3 analysis, and for its theoretical implications, see Progovac, Chap. 2).
. Theories of agrammatism as a variable phenomenon In the absence of a multiple-deficit theory of variability, we have to look for theories that assume a single deficit, while taking variability into account. One such theory comes from Friedmann and Grodzinsky (1997; see also Friedmann 2001, 2002). Studying verb inflection in Hebrew- and Arabic-speaking patients, they observed that tense inflection is much more impaired than agreement inflection. To account for this result, they proposed the tree-pruning hypothesis. This hypothesis is phrased in the context of the generative grammar tradition (e.g., Chomsky 1981, 1995). According to this tradition, content words and grammatical morphemes are represented in various nodes in a syntactic tree. Content words are represented in a lexical structure, whereas grammatical morphemes are part of a functional structure. For the correct production of the various types of grammatical morphology, the presence of specific nodes in the functional structure is required. Correct production of a tense inflection requires the presence of a Tense node, correct production of the agreement inflection requires the presence of an Agreement node, and correct production of complementizers and whwords (who, what, where, which) requires the presence of a Complementizer node. According to the tree-pruning hypothesis, agrammatic production is the result of dam-
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.5 (299-353)
An analysis of agrammatic aphasia
age to the functional structure. This damage can be highly selective and affect only specific nodes. It is further assumed that nodes located below the damaged node remain available, whereas the damaged nodes as well as nodes higher up in the tree can no longer be reached. So, if only the Tense node is damaged, tense inflection will be affected but agreement inflection will be spared because the Agreement node is located lower than the Tense node. On the other hand, there will be problems with embedding and with wh-words, because both are dependent on the Complementizer node, which is located higher than the Tense node. The tree-pruning hypothesis thus accounts for between-patient variability by assuming that the nodes at the various levels of the syntactic tree can be independently damaged. The lower in the tree the impairment, the more severe the deficit is, because only relatively few nodes can be reached and relatively few types of grammatical morphology can be produced. However, withinpatient variability (inconsistency) is not accounted for. Although the impairment can be located at different nodes in the syntactic tree, the consequences of damage at a particular node are still predicted to be all-or-none. For instance, problems with embedding are thought to arise from damage to the Complementizer node or a lower node. In all cases, embedding should be impossible. As we saw above, however, in most patients we observe reduced likelihood rather than complete absence of embedding. Similarly, the use of infinitives (finiteness omission) is assumed to result from damage to both Agreement and Tense nodes. Data from Hofstede (1992) however, show that this parameter also varies stochastically over patients. A second way to account for agrammatism and its variability is to assume that short-term memory problems play a role. Shallice and Butterworth (1977) have demonstrated that a severely reduced digit span does not lead to agrammatic production. It can be argued, however, that specific buffers exist to keep syntactic information available while further processing is taking place. In agrammatism, the size of this syntactic buffer – the number of slots – could be subject to varying degrees of limitation. The consequence of such a limitation would be that sentences of a certain degree of structural complexity can no longer be adequately processed, because the amount of structural information exceeds the number of available slots. A hypothesis like this has been proposed by Caplan and Hildebrandt (1988) to account for complexity effects in agrammatic comprehension, but it could be applied to agrammatic production as well. The syntactic buffer hypothesis can account for between-subject variation, by assuming that different patients suffer from different amounts of limitation. As was the case with the tree-pruning hypothesis, however, it cannot accommodate the existence of within-patient variability without making additional assumptions: a structure – say, a passive – that contains too much information in relation to the number of available slots will never be adequately processed. A third approach to agrammatic production has to do with processing time. Kolk and van Grunsven (1985) proposed that computational simultaneity is the bottleneck in agrammatic production – and in agrammatic comprehension as well. Two states of affairs can threaten computational simultaneity of syntactic information. First, decay of syntactic information can be too fast. Second, there can be delay in the retrieval of
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.6 (353-402)
Herman Kolk
syntactic information. Delay and decay are of course two sides of the same coin: the faster the decay, the greater even a small effect of delay, and vice versa. One way to describe their combined effects is phrasing them in terms of a reduction in the size of the temporal window within which all sentence processing has to take place. With such a reduction, computational simultaneity between elements of a syntactic representation often cannot be obtained and the sentence cannot be produced. This theory was embodied in a computer model, called Synchron (Haarmann & Kolk 1991). Synchron was able to simulate agrammatic comprehension data, obtained in two different studies. Synchron’s critical assumptions were that syntactic trees were built up in real time and that agrammatics suffered from faster decay or slower retrieval of syntactic information. In Synchron, delay and decay were equally successful in simulating the aphasic data. It is worth noticing that a delay-decay contrast is also a cornerstone of a more recent simulation model of aphasic naming. In this spreading activation model, developed a number of years after the original publication by Kolk and van Grunsven (1985), decay and delay represent two alternative deficits of the word-production system (Dell, Schwartz, Martin, Saffran, & Gagnon 1997). In case of a decay deficit, units lose their activation abnormally fast. A delay deficit is implemented in their model by a reduction of connection strength, leading to an abnormally slow rise of activation. How does the temporal-window hypothesis deal with variability? Between-patient variability was successfully implemented in Synchron by making average decay or retrieval rates differ between simulated patients. As a consequence, even relatively complex sentences were performed upon correctly from time to time. Within-patient variability was a consequence of the fact that both decay rate and retrieval rate were not always the same, but varied around a mean from moment to moment. However, within-patient variability can also be observed when the behavior in one languageproduction task is compared to another such task. For this task-related variability, processes of adaptation are assumed to be responsible, and these will be discussed below.
. Evidence for the temporal-window hypothesis To test applicability of the temporal-window hypothesis to agrammatic production, Hartsuiker and Kolk (1998) used a syntactic priming task. The paradigm was developed by Bock and colleagues and demonstrated to be sensitive to the phrase structure configurations of sentences (Bock 1989; Bock & Loebell 1990). In this paradigm, the participants had to repeat a sentence (e.g., The church was struck by the lightning). After this, they were presented with a picture they had to describe in one sentence (e.g., a picture of a cat chasing a dog). Normal participants tend to do this by employing the sentence form just presented to them. They do so, despite the fact that they are unaware of the purpose of the experiment, since they are told they are engaged in a study meant to test their memory for pictures. In a study with 12 Dutch-speaking agrammatics, Hartsuiker and Kolk found normal and in one condition even better than normal priming. Of particular interest are the results obtained with the production of pas-
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.7 (402-477)
An analysis of agrammatic aphasia
sives. In the spontaneous speech of these patients, passives are extremely rare. In a picture description pretest, there was only a single occurrence of a passive construction in the whole group. After priming however, passives appeared to be deblocked: Now 7 out of 12 patients produced one or more passive constructions. These results fit the temporal-window hypothesis quite well. If the computation of the constituent structure is delayed, priming speeds up this computation, because the structural units have already reached a certain level of activation, due to the previous repetition of this structure, and it will generally take less time to bring these units to threshold. Saffran and Martin (1990) have reported similar results with English-speaking agrammatics. It is important to realize that neither the tree-pruning nor the syntactic buffer hypothesis can accommodate the effect of sentence priming on agrammatic production. Another prediction of the temporal-window hypothesis is that the greater the complexity of the utterance to be produced, the lower the chance of correct production. Kolk and van Grunsven (1985) conducted a first relevant study. They investigated constituent ordering in a sentence anagram task. Dutch-speaking agrammatic speakers were presented with sentences, with each constituent written on a separate card. Single-clause sentences were compared to double-clause sentences. Double-clause sentences were elicited presenting the participants with an initial clause written on a card, with the instruction to always put this card in the first position (e.g., I am glad because/my son/his exam/passed). Although only the last three constituents have to be ordered, the patient also has to process the introductory clause, to know how to order. Various constituent orders were elicited, by introducing elements that made a particular order obligatory. Subordinate clauses in Dutch have an SOV (subject-objectverb) order. So by using a conjunction like because, an SOV order was elicited. Clauses that start with a temporal adverb require a VSO (verb-subject-object) order in Dutch. Therefore, forcing the patient to start with a temporal adverb made a VSO order obligatory. The agrammatics made very few errors in ordering single-clause sentences, but many more in ordering complex ones, although the number of constituents to be ordered did not differ. The type of word order that was required had little effect. The latter result goes against the prediction of the tree-pruning hypothesis of Friedmann and Grodzinsky (1997). In Dutch, the linguistic base order is generally considered to be SOV. To produce the SVO order, a movement transformation would be required, called verb-second. For this operation to be possible, Friedmann and Grodzinsky assume that the Complementizer node must be available. Since in agrammatism the Tense node is damaged, the Complementizer node cannot be reached and verb-second should be impossible. Using a spoken version of the sentence anagram task with Dutch-speaking agrammatic patients, Kok, Kolk, and Havekort (2006) similarly found no evidence for the hypothesis that verb-second is impaired in agrammatism. Within the reduced temporal window of the agrammatic speaker, all operations have to be carried out that are necessary for planning a grammatical sentence. This means that not only syntactic but also conceptual- or message-level operations could reduce the chance of computational simultaneity. Hartsuiker, Kolk, and Huinck (1999) carried out two experiments with Dutch agrammatic speakers in which they studied
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.8 (477-508)
Herman Kolk
agreement inflection production. They drew on a paradigm, developed by Bock and Miller (1991), in which participants are presented with sentence fragments that have to be repeated and completed (e.g., the king of the colonies – was powerful). One manipulation concerned conceptual number of the head noun. This could be singular, as its grammatical number (e.g., the baby on the blankets), or it could be plural (e.g., the label on the bottles), unlike its grammatical number. In the latter example, although the head noun is grammatically singular, it is in fact referring to a multitude of labels, one on each bottle. Although Bock and Miller originally found no effect of conceptual number in English-speaking participants, later work has consistently found such effects for English (Eberhard 1999), Italian, Spanish, French, and importantly Dutch (Vigliocco, Butterworth, & Semenza 1995; Vigliocco, Butterworth, & Garrett 1996; Vigliocco, Hartsuiker, Jarema, & Kolk 1996). In particular, they observed that in sentences with a head noun that is grammatically singular but conceptually plural (e.g., the label on the bottles), more agreement errors were made than in sentences without such a mismatch (e.g., the baby on the blankets). This indicates that normal speakers take conceptual information into account when constructing subject-verb agreement. In two experiments with agrammatic speakers and normal controls, Hartsuiker et al. (1999) replicated this conceptual number effect for the normal but not for the agrammatic speakers. In fact, in the second experiment, which was better controlled, the agrammatics made fewer agreement errors in the mismatch condition than the normal controls! A subsequent comprehension test showed normal sensitivity in the agrammatics to the conceptual number variable. It was concluded that agrammatic speakers do not take into account conceptual information while constructing subjectverb agreement. They are unable to reach computational simultaneity of conceptual and syntactic information. The hypothesis that agrammatic speakers do not take into account conceptual information while constructing subject-verb agreement has direct relevance to their difficulties with tense. As described above, tense inflection appears more impaired than agreement inflection. Friedmann and Grodzinsky (1997) explain this dissociation by assuming that the Tense node is unavailable to these patients. The temporalwindow hypothesis suggests a different account. Recent work in psycholinguistics (e.g., Vigliocco & Hartsuiker 2002) indicates that agreement production takes place in two steps. First, number4 information is transferred from the conceptual level to the level of syntactic encoding. Second, this information is transferred from the subject NP to the verb. An important piece of evidence for this hypothesis is that the chance to make an agreement error is dependent on the syntactic distance number features have to travel to influence verb agreement (Franck, Vigliocco, & Nicol 2002). Although the production of tense has received relatively little attention from psycholinguists, one important difference with agreement – at least in languages like English, French, and Dutch – appears to be that whereas agreement is encoded on both the subject NP and the verb, tense is only encoded on the verb. For the temporal-window hypothesis this means that agreement inflection can be produced as long as the number feature encoded on the subject NP is still available; the conceptual number information may have
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.9 (508-583)
An analysis of agrammatic aphasia
been decayed. For tense production, on the other hand, conceptual information must remain available up to the point that verb inflection has been produced. Since, according to the temporal-window hypothesis, agrammatic speakers have difficulty keeping syntactic and conceptual information simultaneously available, their greater difficulty with tense as compared to agreement becomes understandable.
. Adaptation In the discussion of the temporal-window hypothesis, I have given only rudimentary treatment of specific symptoms. A number of questions therefore have not been answered yet. First, although a reduced temporal window will make double-clause sentences harder than single-clause ones, how does the syntactic symptom “reduced variety of grammatical form” actually come into existence? Second, how do we account for the omission of all the different types of grammatical morphology? Third, what can we say about the rate symptom? All these questions have been ignored for a specific reason. As we will see below, the specific symptoms of agrammatic production are so manifold, and the way they are codetermined by the task and the conversational setting is so subtle, that any attempt to explain them on the basis of a deficit alone is doomed to fail. This is true for the temporal-window hypothesis of the agrammatic deficit as much as for its alternatives. None of these deficit hypotheses even begins to explain the agrammatic symptom complex. In the early 1980s, Caramazza formulated a central principle for the analysis of behavioral data obtained from brain-damaged patients. According to this transparency assumption: The cognitive system of a brain-damaged patient is fundamentally the same except for a “local” modification. . . This assumption rejects the possibility that brain damage results in the de novo creation of cognitive operations. (Caramazza 1986: 52; see also Caramazza 1992)
This seems logical enough and undoubtedly many cognitive neuropsychologists are taking this assumption for granted (although see Farah 1994). Nevertheless, the transparency assumption may have its limitations. Brain damage may not lead to completely new operations, but something may change nevertheless. In particular, rarely used cognitive operations could become more frequent. It is known for instance that after limb amputation major adaptations of learned motor patterns occur (see Latash & Anson 1996 for a review). Apparently, the brain has ways to recruit motor patterns that are within its reach, but that are normally not or only infrequently used. And if the brain can do this after amputation, there is reason to assume that this same capacity can be used after damage to the brain itself, unless of course brain parts necessary for this adaptive change are also damaged. This example suggests that whether the transparency assumption works in a particular case depends on the availability of alternative routes to the same ultimate goal, something referred to as multiple-route
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.10 (583-632)
Herman Kolk
plasticity (Hagoort, Wassenaar, & Brown 2003). Behavior appears to vary in this respect. For instance, it seems likely that there is only one way to stretch your finger, but there are certainly a number of ways to move your finger to a target. In the motor behavior literature, this state of affairs is referred to as functional equivalence or degrees of freedom. Functional equivalence also exists at the level of articulation: Even single speech sounds can be articulated in more than one way. That the availability of alternative articulatory routes is employed for adaptive purposes is shown by the phenomenon of “pipe speech”: when speakers are given bite blocks between their teeth, they immediately produce acceptable vowels or consonants, despite the fact that the normal target positions of their articulators can no longer be reached (e.g., Fowler & Turvey 1980). With respect to language behavior, functional equivalence seems abundant. Take for instance word reading, a popular topic for neuropsychologists. It has long been recognized that there are two ways to read aloud a word: via the lexical or via the nonlexical route. Recent evidence suggests that the selection of one or the other route is under control of the reader (Hendriks & Kolk 1997; Kello & Plaut 2003). I will come back to word reading below. At the sentence level, many sentences can be understood largely on the basis of individual word meaning, without the use of syntactic information: aphasic patients are known to make use of this possibility (cf. Hagoort et al. 2003). In all these cases, there is an alternative to not changing: the impaired system reorganizes to exploit alternative routes to the same goal. This is the core of the present approach to agrammatic sentence production. Under this account, the specific symptoms of agrammatic production do not stem from the temporal-window deficit itself, but from the way language output is adapted to the reduced processing possibilities. Kolk and van Grunsven (1985) have proposed two main forms that such adaptation could take: corrective and preventive adaptation. Whereas corrective adaptation was thought to lead to the rate symptom, preventive adaptation would be responsible for both the syntactic and the morphological symptoms. I will treat these two forms of adaptation separately. After a discussion of the conceptual background of these two claims, I will present the empirical evidence that supports them.
. Corrective adaptation is responsible for nonfluency As indicated above, reduced speaking rate or nonfluency is present in almost all agrammatic speakers. This does not imply of course that the rate symptom is functionally related to the syntactic and morphological symptoms. It could also result from the articulation difficulties of these patients. Most investigators of the agrammatic syndrome have probably implicitly assumed this, since as a topic nonfluency is invariably absent in research on agrammatism. Some of the nonfluency undoubtedly derives from articulatory difficulties. However, clinical observation suggests that this cannot be the whole story. The severity of articulatory impairment varies widely between patients and it is not uncommon to observe patients with only a mild articulation impairment, who still are highly nonfluent. The adaptation account of nonfluency that we have
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.11 (632-692)
An analysis of agrammatic aphasia
proposed makes use of assumptions about speech monitoring in normal speakers. According to Levelt’s influential model (Levelt 1983, 1989), normal speakers discover their output errors in the same way as they discover errors in the speech of others: via their speech comprehension system. They can do so in two ways. First, they can monitor their overt speech: When they detect an error, they stop and repair the error. This outer loop monitoring would thus be responsible for the occurrence of overt repairs (e.g., the red. . . no. . . the green circle). Second, they can monitor their inner speech, which Levelt assumes to be identical to the “phonetic plan.” The phonetic plan is the end product of the formulation processes; it serves as input to the articulatory systems. This inner loop monitoring is thought to be responsible for the occurrence of covert repairs (e.g., the. . . uh. . . the green circle). Levelt considers various speech disfluencies such as (filled) pauses, prolongations, and repetition to be signs of covert repair. The inner loop is also held responsible for some overt repairs, in particular for those that are too fast to be detected by the outer loop (e.g., the r- no the blue circle). For further discussion of and recent computational evidence for this theory, the reader could consult Hartsuiker and Kolk (2001) and Hartsuiker, Kolk, and Martensen (2005). Now if an agrammatic patient attempts to construct a sentence and a planning error occurs due to her reduced temporal window, the patient can covertly repair this planning error by inner loop monitoring. Of course repair will not be as easy as in the normal speaker. Normal speakers make speech errors only infrequently, so a second attempt is highly likely to be successful. Agrammatic aphasics make many planning errors, given their deficit. However, by restarting the computational process a number of times, repair is still possible. Since there is still some residual activation from previous attempts, later computations will be faster and the chances for simultaneity to occur will increase. Covert repair works as a kind of self-priming. It will thereby lead to improved sentence production but also to a reduced speaking rate. The corrective adaptation hypothesis predicts that agrammatic speakers demonstrate relatively many covert repairs in their speech. Employing a picture description task, Schlenck, Huber, and Wilmes (1987) confirmed this prediction. In a study with German-speaking aphasics they observed more covert repairs (disfluencies) than overt repairs, whereas the controls demonstrated more overt than covert repairs. Oomen, Postma, and Kolk (2001) replicated this finding in a study with Dutch-speaking agrammatic patients. These patients produced more covert than overt repairs, whereas the normal controls showed the reverse pattern. Admittedly, it could be argued that disfluencies reflect other processes besides covert repair, such as word-finding difficulties. So, is there other evidence that agrammatic patients are heavily involved in inner loop monitoring? The study by Oomen et al. provides such evidence. Oomen et al. presented white noise (90 DB) to the participants, while they were performing a picture description task. The presence of noise severely reduces the possibilities for auditory feedback and outer loop monitoring. This should have consequences for the number of overt repairs. In normal speakers, who would employ both the inner and outer loops, the number of overt repairs should go down. This prediction was confirmed. In aphasic speakers on the other hand, who primarily rely on the inner loop, the noise
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.12 (692-738)
Herman Kolk
manipulation should have a smaller effect on the overt repair rate than in the normal controls. This prediction was also confirmed; in fact, the noise effect in the aphasics did not reach significance. Oomen et al. also tested the prediction of another account for the dominance of covert repairs, suggested by Schlenck et al., who argued that the outer loop would be damaged in agrammatic patients, leaving them with the inner loop only. This hypothesis predicts that the patients would be more impaired than normal controls in detecting errors in speech produced by others. One could argue that the latter task should in general be easier than the task of detecting errors in selfproduced speech, because the latter task is a dual task: the participant has to speak and to monitor simultaneously. Indeed, detecting other-produced errors was easier than detecting self-produced errors. This was true, however, for the controls as well as for the patients and there was no difference in the size of the advantage. There is no evidence therefore that the emphasis on the inner loop in the patients is due to damage to the outer loop. Instead, we believe they make limited use of outer loop monitoring for strategic reasons, to optimize the quality of their output. Hartsuiker, Kolk, and Martensen (2005) have simulated the Oomen et al. data and come to the conclusion that the contribution of the outer loop to the overt and covert repairs of the aphasic patients is zero. All monitoring is done by the inner loop.
. Preventive adaptation is responsible for syntactic and morphological symptoms Besides corrective adaptation, which would partially restore the consequences of computational overload, syntactic simplification is another strategic option: It would prevent computational overload. Simplification can be obtained by selecting simple sentence frames, consisting of simple clauses without embedding and phrasal elaboration. This type of simplification leads to what I described above as a syntactic symptom: reduced variety of grammatical form. The adaptation theory, however, holds that syntactic simplification is also responsible for the occurrence of morphological symptoms. This is a more radical type of simplification. It consists of planning clausal frames in which slots for specific types of grammatical morphology tend to be absent. In ordinary conversations, we sometimes leave out grammatical morphemes that we supply in other situations (e.g., nice dress, instead of a nice dress; New York instead of to New York). These are planned omissions: it is hard to see them as speech errors as they are rarely subject to self-repair. Such planned omissions are referred to as cases of ellipsis. They point to the existence of syntactic planning frames in which slots for grammatical morphology tend to be lacking. It should be pointed out that many elliptical utterances do contain some grammatical morphology. In particular, if a verb is used, it will carry the infinitival or participle inflection (e,g., drinking coffee). Similarly, adjectival inflection and plural inflection may occur (e.g., mooie handschoenen ‘beautiful gloves’). Still, what distinguishes elliptical utterances from their standard counterparts is that they contain less grammatical morphology.
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.13 (738-804)
An analysis of agrammatic aphasia
The claim that agrammatic speakers select “syntactic frames in which slots for morphology tend to be lacking” is entirely consistent with the analysis of nonsententials offered elsewhere in this book. As argued by Progovac (Chap. 2), in these nonsententials, the lack of finiteness,5 that is, finite inflection on the verb, correlates with the lack of nominative case assignment to the subject position, which in turn licenses determiner drop in the subject position or the default Case assignment. To put it another way, the choice of the ultimate syntactic projection in nonsententials (smaller than TP) dictates the lack of morphology otherwise associated with this projection, including tense morphology on the verb, and case assignment on the pronoun (visible in English with accusative subjects). This also has other syntactic consequences, such as the lack of verb-second in Dutch. Determiners and prepositions are important examples of grammatical elements for which there often are no slots in an elliptical sentence frame. Another, even more important, example is the subject-verb agreement marker. Omission of this marker is the defining characteristic of a particular kind of elliptical construction: the nonfinite clause. Nonfinite clauses come in two kinds. First, there are clauses in which a nonfinite verb is used, either an infinitive (e.g., in Dutch: dingen voor elkaar krijgen ‘getting things done’), or a past participle (e.g., leuk gedaan ‘nicely done’). Second, there are clauses from which a verb is omitted altogether (e.g., iedereen naar buiten ‘everybody out’). Nonfinite clauses are important not only because they occur so frequently in agrammatism, but also because they involve morphology and word order that is different from the finite clause, at least in Dutch and German. This will become clear below. Finally, if no agreement marker has to be produced, the demands for computational simultaneity are considerably reduced. A crucial assumption underlying the ellipsis hypothesis is that elliptical constructions require less grammatical processing than nonelliptical ones. This is not selfevident, since elliptical constructions may also require more processing in general. From the listener’s point of view, elliptical utterances are more complex, as they require more inferential work to compensate for the omitted elements. From the speaker’s point of view, the frequent use of elliptical constructions may also require more processing, since the incomplete speech must be as informative as possible. A recent PET study by Indefrey et al. (2001), however, indicates that elliptical clauses do indeed require less grammatical work. These authors presented German-speaking participants with pictures of simple colored objects (squares, circles, and ellipses) in different spatial configurations. The task of the participants was to describe these pictures, using one of three different sentence formats. In the full-sentence condition, they had to produce a grammatical sentence, containing all relevant information (e.g., Das rote Viereck stößt die blaue Ellipse weg ‘the red square pushes the blue ellipse away’). In the noun phrase condition, they were required to use a nonfinite clause (i.e., to put the verb in the infinitive and put it at the end of the sentence) and to leave out the determiner (e.g., Rotes Viereck, blaue Ellipse, wegstoßen ‘red square, blue ellipse, pushing away’). Note that the adjective inflection of the first NP (rotes) has changed. This has to do with the grammatical rule of German, that if the determiner is omitted or takes the
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.14 (804-847)
Herman Kolk
form of an indefinite article, the case marking inflection goes to the adjective. Below, we will see that German-speaking agrammatic patients follow that rule. In the word condition, participants were also required to produce an elliptical clause, but this time they should also omit the inflection of the adjective and put the adjective after the noun (e.g., Viereck rot, Ellipse blau, wegstoßen ‘square red, ellipse blue, pushing away’).6 In this way, the amount of grammatical encoding was varied from maximal in the full sentence, to minimal in the word condition, which served as a baseline. As indicated above, these elliptical clauses constitute grammatical forms of ellipsis. Although it is highly unusual to employ them exclusively, as the participants were required to do in this task, all participants were able to follow the instruction after some practice trials. The blood flow response varied between these conditions in the left operculum, a region just behind Broca’s area. Furthermore, the strength of the response varied as expected: maximal in the full-sentence condition, less strong in the noun phrase condition, and still less strong in the word condition. If indeed the strength of the response reflects the intensity of grammatical processing, these results support the assumption that, in order to produce an elliptical clause, less processing capacity is required. As indicated above, speaking elliptically is unusual for normal speakers. So, agrammatic speakers must be able to recruit infrequent ways of responding in order to adapt to their deficit. The ability to do so is implicitly denied in the transparency assumption. But what would be the justification for such a denial? Participants in behavioral experiments make unusual responses all the time. In the well-known Stroop task, for instance, they name the color of a word rather than just reading it. This is certainly a highly unusual response, but most participants are able to do it, although their response is slowed down in comparison to naming the color of a color patch. Current theories of Stroop task performance assume a special role for attention (Norman & Shallice 1986; Cohen, Dunbar, & McClelland 1990; Phaf, Van der Heijden, & Hudson 1990). Goals or tasks are sources of attentional activation and bias activation levels of potential responses. However, selection of action does not make explicit reference to these goals. In an impressive computer simulation and experimental study, Roelofs (2003) has recently demonstrated that a wide variety of empirical findings with respect to the Stroop task can be simulated by a model that couples an already existing word-production algorithm, WEAVER, with explicit goal reference. The new model outperforms competing models, which operate without such a component. These results indicate that explicit goal setting is a fundamental cognitive process. It could be as important for aphasic as it is for normal speakers. Just as it is the goal for the participant in a Stroop experiment to name the color of the word, it could be the goal of the agrammatic speaker to use elliptical style. In the discussion of his paper, Roelofs makes an historical reference to the work of the Würzburg school of thought, in particular to their notion of Aufgabe. Another word for Aufgabe was Einstellung, which has most often been translated as ‘mental set.’ The Würzburgers demonstrated that such a notion was necessary to account for their results with various kinds of word association tasks (cf. Boring 1950). It was this very same notion that led Isserlin (1922; abbreviated translation into English 1985) to propose that telegraphic speech in aphasics
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.15 (847-917)
An analysis of agrammatic aphasia
was ein Einstellungserscheinung – a phenomenon of mental set (401 in original text; in the English translation, reference is made to attitude, 339). It was this specific proposal that led to the formulation of adaptation theory (Kolk et al. 1985; see also Kolk 1995), which holds that agrammatic patients, in order to circumvent processing problems that arise with complete syntactic constructions, make it their goal to produce elliptical sentence forms.
. Agrammatic speech has all the grammatical features of normal ellipsis Although normal elliptical speech deviates from the standard sentence format, it is subject to a number of grammatical constraints. If agrammatic speakers indeed overuse normal elliptical constructions, their output should obey these constraints. The following constraints hold for Dutch and German: 1. Omissions of function words are allowed. Normal ellipsis allows for the omission of function words. So we could have a legal elliptical construction like (1). (1) Pen en papier pencil and paper If the omitted function word happens to be a determiner, and if at the same time an adjective is used, German ellipsis has a special rule. It requires the adjective to carry the case marker, normally carried by the determiner. This regularity was also expressed in the materials used by Indefrey et al. (2001). So one could have article omission plus strong adjective inflection, as in (2), but not article omission plus weak adjective inflection, as in (3). (2) Rotes Viereck red-neut.nom.sg square (3) *Rote Viereck red-neut.nom.sg square 2. Omissions of inflections are not allowed. Normal ellipsis does not allow for the omission of an inflection. An utterance like (4) would thus be ungrammatical. (4) *Twee pen two pencil 3. Substitutions of function words or inflections are not allowed. Substitutions in elliptical constructions are speech errors just as in complete sentences. So we don’t have constructions like (5) and (6) as these are ungrammatical. (5) *Ik denkt I thinks (6) *Het vrouw the-neut.sg woman
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.16 (917-1000)
Herman Kolk
4. Use of nonfinite verbs is allowed. Normal ellipsis allows the use of nonfinite verb forms, in particular, the infinitive or the past participle, without the use of an accompanying auxiliary. So, both (7) and (8) constitute cases of regular ellipsis: (7) Koffie drinken coffee drink-inf (8) Portemonnai verloren wallet lost-past.part 5. Omission of the main verb is allowed. Although in complete sentences, a main verb needs to be present, this is not the case for ellipsis, as in (9): (9) Bloemen voor mijn vrouw flowers for my wife 6. Omission of the grammatical subject is allowed. Grammatical subjects, present in complete sentences, can be omitted in elliptical constructions, as in (10). (10) Naar het ziekenhuis geweest to the hospital been-past.part 7. Sentence-final position of the verb is required. Whereas in finite sentences the verb is in second position in both Dutch and in German, in nonfinite sentences, the verb should be clause final. So (11) is grammatical, whereas (12) is not. (11) Auto gekocht car bought-past.part (12) *Gekocht auto bought-past.part car 8. If subject pronouns are used, they should have a phonologically strong, rather than weak (clitic, reduced) form. So an elliptical construction such as (13) is allowed, whereas (14) is not. (13) Ik fietsen I-strong bike-inf (14) *‘K fietsen I-weak bike-inf Research with both Dutch- and German-speaking patients has shown that these constraints of normal ellipsis are also present in agrammatic speech (see Kolk & Heeschen 1992; Kolk 2001). Linguistic behaviors such as function word omission that are allowed in normal ellipsis were generally present, with frequencies ranging from about 40% to about 75%. Behaviors that are not allowed had frequencies between 1% and 3%. Furthermore, for a number of these features (in particular the features 4, 5, 6, 7, and 8) it was demonstrated that they were highly uncharacteristic of the finite sentences of these patients.7,8
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.17 (1000-1329)
An analysis of agrammatic aphasia
. Distributional properties of types of nonfinite constructions are highly similar in agrammatic speech, child language, and normal ellipsis The data discussed in the previous section support the conclusion that agrammatic speakers overuse normal ellipsis. This does not mean that the actual employment of nonfinite constructions is the same in normals and aphasics. In particular, it could be that the aphasics use only a subset of the normal repertoire. Relevant data are reported in Kolk (2001). They were taken from the Hofstede corpus of nonfinite clauses produced by normal and aphasic speakers during informal conversations (Hofstede 1992). Data from three Dutch-speaking children were added, with age levels 2, 2.5, and 3. Children of these age levels are interesting to compare with agrammatics, because their speech appears telegramlike, just as the speech of Broca’s aphasics. Furthermore, there could be a similar limitation of processing resources in children and aphasics (Avrutin 1999). So, there is reason to expect similarity between child language and agrammatic language and, as a consequence, between child language and normal ellipsis. The following conclusions could be drawn. First of all, there is overuse of nonfinite clauses in both children and aphasics. Whereas control speakers produced 10% nonfinite clauses, aphasics produced about 60%. In children, the overuse of nonfinite clauses decreases with age: from 83 % in the 2-year-old, to 60% in the 2.5-year-old, to 40 % in the 3-year-old. Secondly, the same 10 types of nonfinite clauses occur in agrammatic speech, child language, and normal ellipsis and with similar relative frequencies. Table 9.1 gives an overview of the 10 different nonfinite constructions with examples. A major subdivision is between isolated predicates, consisting of syntactic phrases occurring on their own, and subject-predicate connections, in which the same phrasal types occur in combination with an NP functioning as “subject.” So with isolated predicates, the predication is about a presupposed topic, which the listener has to derive from the conversational context. With the category of subject-predicate connections, both the topic and the predicate are expressed, albeit not integrated in a syntactic frame. In Figure 9.1, I present relative frequencies of these 10 categories for the three types of speakers. It is important to realize that these are relative frequencies. They are derived by taking all nonfinite clauses produced by a particular group and the percent of these clauses that falls within a particular category, such as ‘NP + PP’. Therefore, relative frequencies of a particular category can be very similar for normal adults and children for instance, despite the fact that children employ nonfinite clauses much more frequently than normal adults. The similarity in the distributions of the tree populations can be expressed by means of rank-order correlations. In all comparisons, significant correlations were obtained. In addition, the various ways in which these nonfinite clauses were elaborated (e.g., by an additional adverb or Noun Phrase) also had very similar relative frequencies in the three populations.
Abbreviation
Nf.V
NP
PP
Adj
Adv
NP+ Nf.V
NP + NP
Category
A. Isolated predicates 1. Nonfinite verb
2. Noun Phrase
3. Prepositional Phrase
4. Adjective
5. Adverb
B. Subject + Predicate 6. Noun Phrase + nonfinite verb
7. Noun Phrase + Noun Phrase
Broerstraat ’n kerk Broerstreet a church ‘There is a church in the Broerstreet’
Hij lachen he laugh-inf ‘He was laughing’
Jammer genoeg too bad ‘It was too bad’
Bloedheet bloody-hot ‘It was bloody hot’
Op zo’n manier in such-a way ‘It was done in such a way’
Geen kunstmest no fertilizer ‘They use no fertilizer’
Veel sporten much sport-inf ‘We did a lot of sporting’
Normal Adult
Table 9.1 Categories of elliptical constructions in three populations: Examples in Dutch
In laarze in boots ‘I am wearing boots’ Lekker nice ‘It was nice’
Naar zee to sea ‘We went to sea’ Fanatiek fanatical ‘I am fanatical’
Ik tabletje I pill ‘I got a pill’
Dokter ook weten doctor know-inf too ‘The doctor knows it too’
Pappa cola daddy coke ‘Daddy gets a coke’
Die afgebroken that broken-past.part ‘That was broken’
Nog ’n keer one-more time ‘Do it one more time’
Lege schoene empty shoes ‘Those are empty shoes’
Goed weer good weather ‘It was good weather’
Niet zo erg not so bad ‘It was not so bad’
Broteram eten sandwich eat-inf ‘I am eating a sandwich’
Child
Bome stekken trees slip-inf ‘I was busy slipping trees’
Agrammatic
JB[v.20020404] Prn:20/07/2006; 13:33 F: LA9309.tex / p.18 (1329-1329)
Herman Kolk
NP + PP
NP + Adj.
NP + Adv.
8. Nounphrase + Prepositional Phrase
9. Noun Phrase + adjective
10. Noun Phrase + adverb
Vroeger de Zeigerbaan formerly the Zeigerstreet ‘Formerly, there was a street here, called the Zeigerstreet’
Laveloos die jongelui sodden those youngsters ‘Those youngsters were really sodden’
Wij Genua in we Genua into ‘We went downtown Genua’
Normal Adult
Winter buiten winter outside ‘Outside it was winter’
Hersenen niet goed brain no good ‘My brain is not good’
Koffie d’r in coffee it into ‘The coffee went into it’
Agrammatic
Ik ook me too ‘I want to do that too’
Ogen dicht eyes closed ‘I kept my eyes closed’
Auto bij je car with you ‘You have a car with you’
Child
Note. Each column gives the utterance in Dutch, followed by a literal translation into English, followed by a paraphrase of its likely meaning.
Abbreviation
Category
Table 9.1 (continued)
JB[v.20020404] Prn:20/07/2006; 13:33 F: LA9309.tex / p.19 (1329-1329)
An analysis of agrammatic aphasia
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.20 (1329-1357)
Herman Kolk relative frequencies of ellipsis in children, aphasics and normal adults
proportion of elliptical type
50 40 child
30
aphasic
20
normal
10
Ad v N f.V N P+ N P N P+ PP N P+ A N dj P+ Ad v P+
N
PP
P N
Ad j
N
f.V
0
elliptical type
Figure 9.1 Distribution of 10 categories of elliptical constructions in three populations; for the meaning of the labels, see Table 9.1
. Employment of ellipsis is task dependent In the present analysis, the use of ellipsis is not a direct effect of the underlying agrammatic deficit, but results from goal setting. The immediate goal is to produce elliptical speech in order to prevent premature disintegration of sentence representations. The ultimate goal is to improve communication. This analysis leads one to expect effects of task and communicative setting on the use of ellipsis. Below I will discuss the effects of (a) computational demands, (b) communicative demands, and (c) interactivity of the communicative situation. Since the ultimate goal of elliptical speech is to improve communication, agrammatic speakers will produce elliptical speech less often when communicative demands are high. In particular, when forced to express a particular type of information, they will show less omission of the relevant morphology. At the same time, the number of true errors (substitutions) is expected to increase, because computational overload is less well prevented. Hofstede and Kolk (1994) have manipulated communicative demand in a study with 16 Dutch-speaking and 3 German-speaking agrammatic patients. They did so in two ways. First they compared free narratives, where it is up to the patient how much detail she is willing to provide with a picture description task in which the patient is asked to describe the picture “as well as possible.” The pictures were constructed to elicit various types of grammatical morphology. For instance, on one of the pictures a boy took a large knife from a table on which, besides the big knife, a number of smaller knives were also lying: this picture was meant to elicit a definite article and an inflected adjective (e.g., The boy takes the big knife). On another picture, a book was lying on a bed: this picture was meant to elicit a locative prepo-
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.21 (1357-1413)
An analysis of agrammatic aphasia
sition (e.g., The book is on the bed). As predicted, the group as a whole omitted fewer prepositions, fewer determiners, and less finiteness, as compared to the interview. The decrease in the number of omissions of prepositions went together with an increase in the number of substitutions. In the determiners, however, no increase in substitution rate was observed, suggesting that the computational costs of providing a determiner are limited. A second task employed by Hofstede and Kolk (1994) that was meant to manipulate informational demands went as follows. The patient was presented with a set of pictures depicting spatial relationships (e.g., a red circle on top of a blue square or a yellow square in front of a green circle). The pictures were invisible to the experimenter, who was sitting in front of a board with wooden tokens, of various forms and colors. The patient was asked to describe the picture in such a way that the experimenter could reproduce it. In this task situation, there was a dramatic effect on the omission rate of prepositions, as compared to the interview. Omission rate went down (from 49% to 1%), whereas substitution rate went up (from 4% to 28%), in accordance with the predictions. Determiner and finiteness omission did not change at the group level, but at the individual level, both significant increases and decreases were observed. It seemed that some patients indeed attempted to produce a complete finite sentence, including a determiner and a preposition, whereas others did the reverse and increased the use of ellipsis. How do we account for these contradictory findings? In the above, I argued that although the immediate goal of the patient would be the production of an elliptical construction, the ultimate goal is to optimize communication. Now in the second task the patient faces a dilemma. To enable the experimenter to reconstruct the spatial array, a sentence must be produced in which both the adjectives and the prepositions are correctly embedded in a sentence. So, if the patient employs an elliptical construction to express the spatial relations on the picture, she will provide the experimenter with insufficient information. On the other hand, if she attempts to produce a full sentence, she may provide the experimenter with erroneous information. How did patients resolve this conflict? In particular, how did the ones who increased elliptical style still manage to get the meaning across and enable the experimenter to reconstruct the spatial array? De Roo, Kolk, and Hofstede (2003) recently reanalyzed some of the Hofstede and Kolk (1994) data. They studied only the performance in the second task and included only those Dutch-speaking patients who in their spontaneous speech had more finiteness omission than the control group. Their reanalysis suggests an answer. Patients compensated for the loss of information consequent on the use of ellipsis by increasing the number of utterances. Although each individual construction carried only part of the information, taken together the sets of elliptical utterances were sufficiently complete to enable reconstruction in about 89% of the cases.9 So one would get strings like witte blokken, rooie cirkels boven (‘white blocks, red circles up’). The elliptical constructions they employed were mostly without a verb, rather than with infinitives. Spatial relationships were expressed by means of prepositions with a nominal complement (e.g., onder het vierkant ‘under the square’), prepositions with a pronominal complement (e.g., eronder ‘there
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.22 (1413-1485)
Herman Kolk
under’), and intransitive prepositions (e.g., boven ‘up’). It should be noted that the use of intransitive prepositions is lexically restricted in Dutch. In particular, the preposition op (‘on’) can only be used transitively, whereas other prepositions like boven (‘above’) and bovenop (‘on top of ’) can be used both transitively and intransitively. The constructions made by the patients reflected this regularity. Color information was expressed by means of attributive constructions (e.g., het blauwe vierkant ‘the blue square’), verbless predicative constructions (e.g., vierkant blauw ‘square blue’), and isolated adjectives (e.g., blauw ‘blue’). The -e inflection of the adjective is required in attributive constructions, but is ungrammatical in predicative constructions. The constructions made by the patients reflected this regularity as well. To further test the hypothesis that agrammatic patients are overusing normal ellipsis, a number of control speakers were tested. They were asked to describe the pictures with the instruction to do so while using either three or two words per utterance. In a control condition, they described the pictures without a limitation on the number of words. The limitation condition, in particular the two-word condition, was effective in eliciting elliptical style: utterances without a verb increased in number and so did all the types of elliptical constructions that we observed in the patient group, even the ones in which the adjective was used predicatively (cirkel geel, vierkant rood ‘circle yellow, square yellow’). Finally, the number of utterances doubled, as compared to a condition without word limitation. In about 96% of the cases, the spatial arrays could be reconstructed on the basis of these sequences of elliptical utterances (vs. 89% of the aphasic utterances). It seems that at the group level there are sizable task effects. At the individual level, some patients show no effect, most patients show some effect, and some patients show spectacular effects. Patients with an almost all-or-none shift of speech style have been described in the literature. One such patient was studied by Bastiaanse (1995). This Dutch-speaking patient switched from complete constructions with some occasional errors to full telegraphic speech in the middle of the interview. Another Dutch-speaking patient was reported on by Hofstede and Kolk (1994). In this patient, the use of nonfinite verb forms dropped to 0% in a standard picture description task, as compared to 86% in free conversation. Function word omission dropped from 68% to 2%. This spectacular shift to completeness had a cost however. First, function word substitutions increased from 2% to 16%. Second, speaking rate dropped from 30 to 19 words per minute. Finally, the number of (content) word repetitions increased from 11% to 37%. The latter two changes indicate that the patient shifted from preventive to corrective adaptation. He no longer simplified his sentence frames, but he repaired his more complete constructions. Since this patient was able to make such huge shifts, we were interested to know whether he could refrain from ellipsis on request. He did. Function word omission dropped from 68% to 18%, use of nonfinite verb forms from 86% to 24%, and words per minute from 30 to 15. Why do patients shift from elliptical to more complete speech in an ordinary picture description task? The possibility proposed above is that the amount and type of information contained in the pictures elicits greater completeness. However, there
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.23 (1485-1524)
An analysis of agrammatic aphasia
is another difference between the free conversation and the picture description task. This difference has to do with the social setting. The free conversation is typically interactive in that both parties make their contribution and share information. The picture description task is very different. Only the patient contributes and there is little sharing of information. To test whether this social factor makes a contribution, we compared picture description performance of agrammatic patients in two situations. One was noninteractive and involved picture description in front of an experimenter who largely remained silent during the test. The other situation was interactive: the patient described the picture together with his or her partner, who contributed to the description by asking questions and making suggestions. One could expect that, in the latter condition, helping the patient this way would lead to greater completeness. The contrary happened. Those patients (4 out of 8) who demonstrated a clear decrease in finiteness omission in the noninteractive picture description task, as compared to free conversation, showed an increase of finiteness omission in the interactive situation (Heeschen & Kolk 1994). Social factors therefore could make a contribution to the way patients adapt to their deficit.
. Conclusion The most general claim of this chapter is that language adapts to an abnormal brain condition by increasing the frequency of specific behaviors from the normal repertoire. To substantiate such a claim one should: (a) demonstrate that a subset of the normal repertoire is indeed increased in frequency after brain damage, (b) make plausible that this increase is adaptive, given the nature of this abnormal brain condition, (c) propose a likely mechanism for this adaptation, and (d) demonstrate that adaptations are sensitive to task variation. For the case of agrammatic aphasia, I believe all these requirements have been met. The findings also support the approach taken in this book that nonsententials of this kind are part of the normal language system. Although they appear incomplete, they are still well formed and obey specific grammatical rules. When they are overused by young children, whose linguistic system is not yet completely automatized, or by aphasics, whose linguistic system is “de-automatized,” this grammaticality is retained.
Notes * I am grateful to Angelique Hendriks for her careful reading of several previous versions of the manuscript. . The term elliptical is purely descriptive, and refers to well-formed incompleteness. It does not presuppose an elliptical/deletion analysis, but is consistent with the nonsentential analysis as outlined by Progovac (Chap. 2, this volume).
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.24 (1524-1587)
Herman Kolk . Agrammatism in production is often contrasted to paragrammatism. Whereas agrammatism is slow, simple, and incomplete, paragrammatism is fast, complex, and full of errors. Paragrammatism is typically observed in patients with Wernicke’s aphasia. Kolk and Heeschen (1992) argue that agrammatism and paragrammatism stem from the same underlying grammatical deficit. The difference between the two groups would be that the agrammatic patients adapt to their deficit whereas the paragrammatic patients do not. See also Haarmann and Kolk (1992). . It also seems that, in current theoretical programs such as the Minimalist Program, inflectional morphology cannot be so easily dissociated from syntax. In Minimalism, signals of morphological inflection are represented in the syntax as heads of phrasal projections. The heads are involved in syntactic processing such as movement. In such a framework, there is little room for a dissociation between syntax and grammatical morphology. . The same reasoning holds for person features but so far person agreement has not been investigated experimentally. . I have so far described elliptical utterances as lacking finiteness. Alternatively, they may be described as lacking Tense. . I assume that in these examples, the adjective serves as a predicate. So in the example given, we would have three different utterances, corresponding to three different propositions: The square is red, The ellipse is blue, and The square pushes the ellipse away. . Despite the strong relationship between subject omission and omission of finiteness, agrammatic patients omit subjects from finite clauses as well. Some of them do it more often than normal speakers (de Roo 1999, 2003). De Roo (2003) argues that these are instances of normal omission phenomena found in German languages, such as Topic drop. They are present in some patients, in addition to finiteness omission. Their function in aphasia is to reduce processing load in a finite sentence. . With respect to the low rates of substitution errors with function words and inflections that we observed, it must be pointed out that higher substitution rates in agrammatic speech have also been reported (e.g., Menn & Obler 1990). These higher rates however have been found in other tasks than spontaneous speech and/or in other languages. As we will see below, other tasks do indeed lead to more substitutional errors, as task requirements may make the selection of elliptical forms less appropriate. Other languages may offer less opportunity to avoid constructional problems by means of ellipsis. For instance, if a language does not have a separate verb form for the infinitive, if verbs are used, they must be inflected: In that case, substitution errors will arise. This appears to be the case in Hebrew (Grodzinsky 1984). . It is to be noted that these elliptical utterances contained a number of substitutional errors, in particular with respect to prepositions and adjectives. So, in order to assess completeness, these substitutional errors were replaced by correct items.
References Avrutin, S. (1999). Development of the Syntax-Discourse Interface. Dordrecht: Kluwer. Badecker, B. & Caramazza, A. (1985). On considerations of method and theory governing the use of clinical categories in neurolinguistics and cognitive neuropsychology: The case against agrammatism. Cognition, 20, 97–125.
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.25 (1587-1694)
An analysis of agrammatic aphasia
Barton, E. & Progovac, L. (2005). Nonsententials in Minimalism. In R. Elugardo & R. J. Stainton (Eds.), Ellipsis and Non-Sentential Speech (pp. 71–93). Dordrecht: Kluwer. Bastiaanse, R. (1995). Broca’s aphasia: A syntactic and/or a morphological disorder? A case study. Brain and Language, 48, 1–31. Berndt, R. (1987). Symptom co-occurrence and dissociation in the interpretation of agrammatism. In M. Coltheart, G. Sartori, & R. Job (Eds.), The Cognitive Neuropsychology of Language (pp. 221–233). Hillsdale, NJ: Lawrence Erlbaum. Bock, J. K. (1989). Closed-class immanence in sentence production. Cognition, 31, 163–186. Bock, J. K. & Loebell, H. (1990). Framing sentences. Cognition, 35, 1–39. Bock, J. K. & Miller, C. A. (1991). Broken agreement. Cognitive Psychology, 23, 45–93. Boring, E. G. (1950). A History of Experimental Psychology (2nd ed.). Englewood Cliffs, NJ: Prentice Hall. Butterworth, B., Howard, D., & McLoughlin, P. (1984). The semantic deficit in aphasia: The relationship between semantic errors in auditory comprehension and picture naming. Neuropsychologia, 22, 409–426. Caplan, D. & Hildebrandt, H. (1988). Disorders of Syntactic Comprehension. Cambridge, MA: The MIT Press. Caramazza, A. (1986). On drawing inferences about the structure of normal cognitive systems from the analysis of patterns of impaired performance: The case for single case-studies. Brain and Cognition, 5, 41–66. Caramazza, A. (1992). Is cognitive neuropsychology possible? Journal of Cognitive Neuroscience, 6, 80–95. Castles, A. & Coltheart, M. (1993). Varieties of developmental dyslexia. Cognition, 47, 149–180. Chertkow, H. & Bub, D. (1990). Semantic loss in Alzheimer-type dementia. In M. F. Schwartz (Ed.), Modular Deficits in Alzheimer-Type Dementia (pp. 207–244). Cambridge, MA: The MIT Press. Chomsky, N. (1981). Lectures on Government and Binding. Dordrecht: Foris. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: The MIT Press. Cohen, J., Dunbar, K., & McClelland, J. (1990). On the control of automatic processes: A parallel distributed account of the Stroop effect. Psychological Review, 97, 332–361. Dell, G. S., Schwartz, M. F., Martin, N., Saffran, E. M., & Gagnon, D. A. (1997). Lexical access in aphasic and non-aphasic speakers. Psychological Review, 104, 801–837. de Roo, E. (1999). Agrammatic Grammar. Functional Categories in Agrammatic Speech. Ph.D. dissertation, University of Leiden. de Roo, E. (2003). Null subject pronouns in Broca’s speech production. Aphasiology, 17, 1057– 1072. de Roo, E., Kolk, H. H. J., & Hofstede, B. (2003). Structural properties of syntactically reduced speech: A comparison of normal speakers and Broca’s aphasics. Brain and Language, 86, 99–115. Eberhard, K. M. (1999). The accessibility of conceptual number to the processes of subject-verb agreement in English. Journal of Memory and Language, 41, 560–578. Farah, M. J. (1994). Neuropsychological inference with an interactive brain: A critique of the ‘locality’ assumption. Behavioral and Brain Sciences, 17, 77–78. Fowler, C. A. & Turvey, M. T. (1980). Immediate compensation in bite-block speech. Phonetica, 73, 306–326. Franck, F., Vigliocco, G., & Nicol, J. (2002). Subject-verb agreement errors in French and English: The role of syntactic hierarchy. Language and Cognitive Processes, 17, 371–404.
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.26 (1694-1793)
Herman Kolk
Friedmann, N. (2001). Agrammatism and the psychological reality of the syntactic tree. Journal of Psycholinguistic Research, 30, 71–90. Friedmann, N. (2002). Question production in agrammatism: The tree-pruning hypothesis. Brain and Language, 80, 160–187. Friedmann, N. & Grodzinsky, Y. (1997). Tense and agreement in agrammatic production: Pruning the syntactic tree. Brain and Language, 56, 397–425. Goodglass, H. (1993). Understanding Aphasia. New York, NY: Academic Press. Goodglass, H. & Kaplan, E. (1983). The Assessment of Aphasia and Related Disorders (2nd ed.). Philadelphia, PA: Lea & Febiger. Grodzinsky, Y. (1984). The syntactic characterization of agrammatism. Cognition, 16, 99–120. Haarmann, H. J. & Kolk, H. H. J. (1991). A computer model of the temporal course of agrammatic sentence understanding: The effects of variation in severity and sentence complexity. Cognitive Science, 15, 49–87. Haarmann, H. J. & Kolk, H. H. J. (1992). The production of grammatical morphology in Broca’s and Wernicke’s aphasics: Speed and Accuracy factors. Cortex, 28, 97–112. Hagoort, P., Wassenaar, M., & Brown, C. (2003). Real-time semantic compensation in patients with agrammatic comprehension: Electrophysiological evidence for multiple-route plasticity. Proceedings of the National Academy of Sciences, 100, 4340–4345. Hartsuiker, R. J. & Kolk, H. H. J. (1998). Syntactic facilitation in agrammatic sentence production. Brain and Language, 62, 221–254. Hartsuiker, R. J. & Kolk, H. H. J. (2001). Error monitoring in speech production: A computational test of the perceptual lop theory. Cognitive Psychology, 42, 113–157. Hartsuiker, R. J., Kolk, H. H. J., & Huinck, P. (1999). Agrammatic production of subject-verb agreement: The effect of conceptual number. Brain and Language, 69, 119–160. Hartsuiker, R. J., Kolk, H. H. J., & Lickley, R. J. (2005). Stuttering on function words and content words: A computational test of the Covert Repair hypothesis. In R. J. Hartsuiker, R. Bastiaanse, A. Postma, & F. N. K. Wijnen (Eds.), Phonological Encoding and Monitoring in Normal and Pathological Speech (pp. 261–280). Hove: Psychology Press. Hartsuiker, R. J., Kolk, H. H. J., & Martensen, H. (2005). The division of labour between internal and external speech monitoring. In R. J. Hartsuiker, R. Bastiaanse, A. Postma, & F. N. K. Wijnen (Eds.), Phonological Encoding and Monitoring in Normal and Pathological Speech (pp. 187–205). Hove: Psychology Press. Heeschen, C. & Kolk, H. H. J. (1994). Adaptation bei Agrammatikern: Interaktional motiviert? [Adaptation in agrammatics: Motivated interactionally?]. In I. M. Ohlendorf, T. A. Pollow, W. Widdig, & D. B. Linke (Eds.), Festschrift zum 85. Geburtstag von Anton Leischner (pp. 123–135). Freiburg: Hochschul. Hendriks, A. W. & Kolk, H. H. J. (1997). Strategic control in developmental dyslexia. Cognitive Neuropsychology, 14, 321–366. Hofstede, B. T. M. (1992). Agrammatic Speech in Broca’s Aphasia: Strategic Choice for the Elliptical Register. Ph.D. dissertation, Catholic University of Nijmegen [available as NICI Technical Report 92–07]. Hofstede, B. T. M. & Kolk, H. H. J. (1994). The effects of task variation on the production of grammatical morphology in Broca’s aphasia: A multiple case study. Brain and Language, 46, 278–328. Indefrey, P., Brown, C. M., Hellwig F., Amunts, K., Herzog, H., Seitz, R. J., & Hagoort, P. (2001). A neural correlate of syntactic encoding during speech production. Proceedings of the National Academy of Sciences of the United States of America, 98, 5933–5936.
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.27 (1793-1891)
An analysis of agrammatic aphasia
Isserlin, M. (1922). Über Agrammatismus [On agrammatism]. Zeitschrift für die gesammte Neurologie und Psychiatrie, 75, 332–416. [Abbreviated English translation by H. Droller, D. Howard, & R. Campbell (1985), Cognitive Neuropsychology, 1, 308–345.] Kello, C. T. & Plaut, D. C. (2003). Strategic control over rate of processing in word reading: A computational investigation. Journal of Memory and Language, 48, 207–232. Kok, P., Kolk, H. H. J., & Havekort, M. (2006). Agrammatic sentence production: Is verb second impaired in Dutch? Brain and Language, 96, 243–254. Kolk, H. H. J. (1991). Is stuttering a symptom of adaptation or of impairment? In H. F. M. Peters, W. Hulstijn, & W. Starkwheather (Eds.), Speech Motor Control and Stuttering (pp. 189–204). Amsterdam: Elsevier. Kolk, H. H. J. (1995). A time-based approach to agrammatic production. Brain and Language, 50, 282–303. Kolk, H. H. J. (2001). Does agrammatic speech constitute a regression to child language? A threeway comparison between agrammatic, child and normal ellipsis. Brain and Language, 77, 340–350. Kolk, H. H. J. & Heeschen, C. (1992). Agrammatism, paragrammatism and the management of language. Language and Cognitive Processes, 7, 82–129. Kolk, H. H. J. & Postma, A. (1997). Stuttering as a covert-repair phenomenon. In F. Curlee & G. M. Siegel (Eds.), Nature and Treatment of Stuttering: New Directions (2nd ed.) (pp. 182–203). Boston: Allyn & Bacon. Kolk, H. H. J. & van Grunsven, M. F. (1985). Agrammatism as a variable phenomenon. Cognitive Neuropsychology, 2, 347–384. Kolk, H. H. J., van Grunsven, M. F., & Keyser, A. (1985). On parallelism between production and comprehension in agrammatism. In M. L. Kean (Ed.), Agrammatism (pp. 165–206). New York, NY: Academic Press. Latash, M. L. & Anson, J. G. (1996). What are ‘normal movements’ in atypical populations? Behavioral and Brain Sciences, 19, 55–106. Levelt, W. J. M. (1983). Monitoring and self-repair in speech. Cognition, 14, 41–104. Levelt, W. J. M. (1989). Speaking: From Intention to Articulation. Cambridge, MA: The MIT Press. Mackay, D. G. (1987). The Organization of Perception and Action: A Theory for Language and Other Cognitive Skills. New York, NY: Springer. Menn, L. & Obler, L. (1990). Agrammatic Aphasia: A Cross Language Narrative Source Book. Amsterdam: John Benjamins. Miceli, G., Mazzuchi, A., Menn, L., & Goodglass, H. (1983). Contrasting cases of Italian agrammatic aphasia without comprehension disorder. Brain and Language, 35, 24–65. Norman, D. A. & Shallice, T. (1986). Attention to action: Willed and automatic control of behaviour. In R. J. Davidson, G. E. Schwarts, & D. Shapiro (Eds.), Consciousness and SelfRegulation: Advances in Research and Theory, Vol. 4 (pp. 1–18). New York, NY: Plenum Press. Oomen, C. C. E., Postma, A., & Kolk, H. H. J. (2001). Prearticulatory and postarticulatory selfmonitoring in Broca’s aphasia. Cortex, 37, 627–641. Phaf, R. H., Van der Heijden, A. H. C., & Hudson, P. T. W. (1990). SLAM; A connectionist model for attention in visual selection tasks. Cognitive Psychology, 22, 273–341. Postma, A. & Kolk, H. H. J. (1990). Speech errors, disfluencies and self-repairs in stutterers under two accuracy conditions. Journal of Fluency Disorders, 15, 291–303.
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.28 (1891-1974)
Herman Kolk
Postma, A. & Kolk, H. H. J. (1993). The covert-repair hypothesis: Pre-articulatory repair processes in normal and stuttered disfluencies. Journal of Speech and Hearing Research, 367, 472–487. Postma, A., Kolk, H. H. J., & Povel, D. (1990). On the relation among speech errors, disfluencies and self-repairs. Language and Speech, 33, 19–29. Rochon, E., Saffran, E. M., Berndt, R. S., & Schwartz, M. F. (2000). Quantitative analysis of aphasic sentence production: Further development and new data. Brain and Language, 72, 193–218. Roelofs, A. (2003). Goal referenced selection of verbal action: Modeling attentional control in the Stroop task. Psychological Review, 110, 88–125. Saffran, E. & Martin, N. (1990). Effects of Syntactic Priming on Sentence Production in an Agrammatic Aphasic. Paper presented at the Academy of Aphasia, Baltimore. Shallice, T. & Butterworth, B. (1977). Short-term memory impairment and spontaneous speech. Neuropsychologia, 15, 729–735. Schlenck, K. J., Huber, W., & Wilmes, K. (1987). ‘Prepairs’ and ‘repairs’: Different monitoring functions in aphasic language production. Brain and Language, 30, 226–244. Vigliocco, G., Butterworth, B., & Garrett, M. F. (1996). Subject-verb agreement in Spanish and English: Differences in the role of conceptual constraints. Cognition, 61, 261–298. Vigliocco, G., Butterworth, B., & Semenza, C. (1995). Constructing subject-verb agreement in speech: The role of semantic and morphological factors. Journal of Memory and Language, 34, 186–215. Vigliocco, G. & Hartsuiker, R. J. (2002). The interplay of meaning, sound and syntax in sentence production. Psychological Bulletin, 128, 442–472. Vigliocco, G., Hartsuiker, R. J., Jarema, G., & Kolk, H. H. J. (1996). One or more labels on the bottles? Notional concord in Dutch and French. Language and Cognitive Processes, 11, 407–442.
Appendix Adaptation in other disorders: Dyslexia, stuttering, and movement disorders Adaptation (or compensation) is a very general process, which stems from the inherent flexibility of the human cognitive system. From this perspective, it is unlikely that adaptation is limited to agrammatic sentence production. I can suggest three domains were an adaptation approach appears fruitful. The first is developmental dyslexia. Children suffering from this disorder are somehow hampered in the development of normal reading skills. Two types of developmental dyslexia are commonly distinguished. The first type – sometimes referred to as developmental surface dyslexia – reads slowly and typically sounds out parts of the words during reading (e.g., business is read as: b-bus-sy-bussy-ness). The second type – sometimes referred to as phonological developmental dyslexia – reads at a normal rate but makes many word substitutions (e.g., cigar is read as sugar or healthy is read as alive). The two types are commonly assumed to suffer from a different underlying deficit. Cast in the terminology of the well-known dual-route model, the two deficits could be characterized as a deficiency of the lexical and the nonlexical routes respectively (e.g., Castles & Coltheart 1993). The lexical route refers to a procedure by which the correct pronunciation of the word is determined by retrieving this information from
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.29 (1974-2067)
An analysis of agrammatic aphasia
the lexicon. The nonlexical route refers to a procedure of computing the correct pronunciation by application of the grapheme-phoneme correspondence rules. A deficiency of the lexical route would lead to a greater role of nonlexical reading, leading to slow reading with sounding-out behaviors. A deficiency of the nonlexical route on the other hand would increase the involvement of lexical reading. This would lead to reading at a normal rate but with many word-substitution errors. In this analysis, the shift to one or the other route is purely a function of the deficit. Hendriks and Kolk (1997) wanted to explore to what extent the shift is under control of the dyslexic reader. Even if for a particular child one route is less efficient than the other, in particular circumstances the child may nevertheless shift to the less efficient route. One obvious possible reason has to do with speed and accuracy. The lexical route is faster but less accurate, while the nonlexical route is slower but more accurate. Different everyday situations may require different emphasis on speed and accuracy. A group of Dutchspeaking dyslexic children were asked to read aloud word lists and texts as fast or as accurately as possible. When required to read fast, the children showed more word substitutions and fewer sounding-out behaviors. The opposite happened when they were required to read accurately. In other words, they behaved more like phonological dyslexics in the speed condition and more like surface dyslexics in the accuracy condition. By way of anecdote, one child, while participating in the “accurate” condition, was reading laboriously and accurately. He then glanced at his watch and suddenly started to read fast and sloppily. When asked afterward why there was this sudden change, he replied that he had no time to talk because there was a swimming lesson that afternoon, and he had to get the bus. This outcome suggests that readers can control strategically which of the two routes is emphasized. Recently, Kello and Plaut (2003) came to the same conclusion on the basis of error patterns of normal readers in a tempo-naming task. Hendriks and Kolk suggest that selective access is obtained by means of eye movements, whereas Kello and Plaut propose “input gain” as the responsible mechanism. As to dyslexia, the results show that it is not only the underlying deficit that determines the dyslexia symptoms, but also the way reading style is adapted to that deficit. Developmental stuttering is a disorder of speech production. Core symptoms of stuttering are: (a) blockings or tense pauses just before a word is uttered, (b) sound repetitions (p- p- pplace) and prolongations (lllllips), and (c) part-word repetitions (cu- cu- cu- cups). Stuttering can be connected to speech monitoring, in particular to the notion of covert repair (Kolk 1991; Postma & Kolk 1993; Kolk & Postma 1997). Just as it has been assumed that aphasics covertly repair errors of sentence planning, one can assume that stutterers covertly repair errors of word planning. With word planning, I refer to the construction of a phonetic plan for a word, which will serve as input to the articulators. As indicated above, with covert repair, I mean repair that takes place before actual articulation, while overt repair manifests itself as an error, followed by a stop, followed by an error correction. When planning errors occur relatively late in the phonetic plan, covert repair would lead to part-word repetitions. When errors are located at an early, but still noninitial segment, sound repetitions or prolongations would ensue. Finally, covert repair of an initial segment would lead to a restart of prearticulatory positioning. No sound would be produced by continuous repositioning, but there would be a build-up of muscle tension and we would observe tense pausing or blocking. One line of evidence comes from studies in which required accuracy was manipulated in normal speakers (Postma, Kolk, & Povel 1990) and in stutterers (Postma & Kolk 1990). Participants were asked to rapidly repeat normal sentences as well as tongue twisters. As a result, speech errors, self-repairs, and disfluencies occurred. As disfluencies we counted blockings, sound, and part-word repetitions and prolongations. Instructions emphasized either accuracy, while speed had to be maintained, or just speed, which could go at the expense of accuracy. Speed had to be maintained in both conditions, in order
JB[v.20020404] Prn:20/07/2006; 13:33
F: LA9309.tex / p.30 (2067-2089)
Herman Kolk
to obtain a sufficient number of speech errors and repairs. Tongue twisters elicited many more speech errors than normal sentences. They also elicited many more overt self-repairs, the increase being proportional to the increase in errors. This makes sense because each speech error represents an occasion for a self-repair. When we compared high- and low-accuracy conditions, however, error and self-repair rates diverged. In the low-accuracy condition, the error rate increased, but the self-repair rate was unaffected. To account for this error-repair dissociation, we argued that the accuracy instruction had two effects. First, it would lead speakers to more closely monitor their speech plan, causing the number of self-repairs to increase. Second, it would focus the attention of the speaker also on the process of speech programming itself, causing the number of programming errors and self-repairs to decrease. These two tendencies would cancel each other out and the net result would be no effect of accuracy manipulation on self-repair. Now what would be the effect of accuracy manipulation on disfluencies? If stuttering is some kind of (perseverative) speech error (Mackay 1987), one would expect disfluencies to behave like speech errors, that is, to increase in the high-accuracy condition. If on the other hand stuttering results from speech monitoring, disfluencies should behave like overt self-repairs and should show no effect of accuracy manipulation. The latter result was obtained, both in normal speakers and in stutterers, confirming the covert repair hypothesis. Stuttering then may also be a symptom of adaptation rather than a symptom of impairment. For further discussion see Kolk and Postma (1997). For recent computational evidence, see Hartsuiker, Kolk, and Lickley (2005). As discussed above, motor control is characterized by redundancy: there are usually many ways for the motor system to reach a particular goal. Latash and Anson (1996) have argued that these degrees of freedom could be exploited for adaptive purposes by persons with movement disorders. Such disorders could bring about reorganization within the central nervous system, inducing changes in the priorities within the central nervous system. The latter would lead to the use of new, quite unusual coordinations. As examples of such reorganization, they describe changes in movement control in different populations such as patients with limb amputation, Parkinson’s disease, dystonia, and Down syndrome. In each case it is argued how these changes – such as bradykinesia in Parkinson’s disease – could be seen as resulting from adaptive processes. I conclude that these findings and considerations indicate that in disorders other than agrammatism there is also a process of adaptation in which the frequency of specific behaviors from the normal repertoire is increased.
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.1 (48-114)
chapter
Nonsententials and agrammatism Patricia Siple Kolk (Chap. 9, this volume) presents a theory of language production deficits associated with Broca’s aphasia, or agrammatism, that accounts for both the major features of agrammatism and the high degree of variability found between and within agrammatics. Kolk attributes within-individual variability to a telegraphic grammatical strategy, available to nondisordered speakers, but overused by agrammatics because their disorder slows processes involved in the production of full sentences. This chapter extends Kolk’s theory by demonstrating that the syntax of nonsententials described by Progovac (Chap. 2, this volume) provides a syntactic analysis of agrammatic telegraphic utterances. As predicted by the nonsententials analysis, these utterances lack Tense features and structural nominative Case, including, for English-speaking agrammatics, default accusative case pronouns in subject position in nonsentential constructions.
.
Introduction
Since the mid-19th century, specific brain regions have been associated with characteristic language functions. During the 1860s, Paul Broca presented papers describing stroke patients with damage to anterior cortical areas of the left hemisphere of the brain that resulted in loss of the ability to produce fluent speech (Translations in Eling 1994). This region, now called Broca’s area, was initially posited to house the motor component of speech production because of the closeness of the region to the motor cortex. This view dominated thinking for a century (e.g., Geschwind 1970) until developments in linguistics, psycholinguistics, and neurolinguistics led to a reconceptualization of Broca’s area as the principal locus of the syntactic component of language (e.g., Goodglass 1968; Zurif & Caramazza 1976). Currently, the brain region described by Broca and the tissue immediately surrounding it are associated with a specific set of production deficits related to syntax. Taken together, these deficits are classified as agrammatism, and individuals with brain damage producing these deficits are called agrammatic aphasics or, simply, agrammatics. As research has provided greater specification of the syntactic deficits of agrammatics, identification of Broca’s area as the syntax module of grammar has given way to more explicit accounts of these deficits
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.2 (114-163)
Patricia Siple
(see Avrutin 2001; Goodglass 1993; Grodzinsky 1990, 2000; Menn & Obler 1990a; Zurif 1990 for reviews). The goal of this chapter is to provide a framework for a theoretical account of syntactic deficits in agrammatic production that integrates psycholinguistic processing with a knowledge/competence component based in Minimalism. Following an expanded description of the nature of production deficits associated with agrammatism, a sample of theoretical accounts of agrammatism will be described. One of these accounts, based in psycholinguistics, focuses on cognitive processes underlying speech production. Another account, based in syntactic theory, attributes agrammatic deficits to inaccessibility or loss of some of the agrammatic’s language knowledge. One serious deficiency of these approaches is an inability to account for the high degree of variability found in production. As was pointed out by Berndt in her preface to Menn and Obler (1990a), these accounts avoid the problem of variability by “working with the idealized case” of agrammatism. In response to this deficiency in current accounts, Kolk (Chap. 9, this volume) has described a more comprehensive theory of agrammatic language production that accounts for variability in a principled way. Kolk argues that agrammatism results from a reduction in processing resources available for language production. To compensate for this deficit, Kolk proposes that agrammatics overuse a separate, syntactically simpler, grammatical strategy for producing “telegraphic” utterances in addition to attempting full grammatical productions. What Kolk’s approach does not provide, however, is a theoretical characterization of what counts as a syntactically simpler strategy. I will argue that the set of phrasal and small clause nonsentential structures proposed by Progovac (Chap. 2, this volume) can play an important role in expanding Kolk’s theory by providing an account of the structures produced by the “telegraphic” component and specifying the basis for the syntactic simplification that Kolk argued results from the “telegraphic” strategy. Identifying a nonsentential grammatical component with the “telegraphic” strategy in Kolk’s theory places the account within a broader framework that includes language acquisition and several aspects of nondisordered speakers’ production.
. Syntactic deficits associated with agrammatic production Agrammatism is most typically associated with disfluent, “telegraphic” speech, and it is these characteristics that have traditionally been used to diagnose the disorder (Goodglass & Kaplan 1972). Agrammatism is characterized by utterances that take longer to produce and are shorter in length than are utterances produced in similar situations by nondisordered speakers. Agrammatics also sometimes have wordfinding difficulty. Across a wide variety of languages, the major syntactic features that characterize agrammatic speech as compared with the speech of nondisordered individuals include:
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.3 (163-227)
Nonsententials and agrammatism
– – – –
Reduction in the variety of syntactic structures used Increased use of infinitives for finite verb forms Omission of function words relative to content words Omission or substitution of inflectional morphology
Thus, utterances produced by agrammatics typically lack functional categories, including complementizers, determiners, tense, and certain agreement markers. In addition, some agrammatics omit more verbs in their utterances than they do nouns, except in languages with a complex nominal inflection system (Menn & Obler 1990b). Agrammatics also tend to use the canonical word order of their language, and agrammatic speakers of free word order languages tend to use a “favorite” word order (Menn & Obler). Syntactic structures produced by agrammatics are more limited in variety and simpler in structure than those produced by nondisordered speakers. Complementizer Phrases are typically absent in spontaneous speech, and tasks specifically designed to elicit wh-questions and subordinate clauses introduced by a complementizer have resulted in little success (Myerson & Goodglass 1972; Menn & Obler 1990b; Friedmann 2001, 2002). Grammatical morphemes that are discourse controlled and are positioned at the beginning or end of a clause (e.g., conjunctions) are used often; and negation is typically indicated by a negative morpheme placed at the beginning or end of the phrase or clause being negated. Because of the rarity of utterances containing CPs and clauses containing internal negation, these structures have received little attention in the literature on agrammatics. Instead, the focus of much of the work, especially since the 1970s, has investigated inflectional morphology and related aspects of grammar that are crucial to sentencehood. In current syntactic theory, these elements are dominated by the Tense node. Tense and agreement deficits are frequent in agrammatic speech. Agrammatics typically produce an abundance of nonfinite forms in situations requiring finiteness; omit auxiliaries, copulas, and have as a main verb; and either substitute or omit tense and agreement affixes. Tense and agreement deficits have been reported for agrammatic speakers of several languages, including Japanese, English, Italian, Hebrew, Dutch, and German (Panse & Shimoyana 1955; Goodglass & Geschwind 1976; Miceli, Mazzucchi, Menn, & Goodglass 1983; Grodzinsky 1984; and Kolk & Heeschen 1992; see Menn & Obler 1990a for data from 14 different languages). Substitution of agreement markers is much more frequent than omission (Menn & Obler 1990b). Omission occurs only when the resulting uninflected form exists as a word in the language. When the uninflected stem is not a word in the language, substitutions for the grammatically correct form are produced. Whether the production of uninflected words in these contexts represents true “omission” of affixes or “substitution” of zero-morpheme forms is the subject of debate, and is important for some accounts of agrammatic production (see Grodzinsky 1990; Menn & Obler 1990b; Jarema 1998, for a discussion of this issue). Other inflections related to verbs are also affected in agrammatism. Modals and mood indicators are almost exclusively absent in the cross-linguistic data from
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.4 (227-284)
Patricia Siple
agrammatic speakers of 14 different languages collected for Menn and Obler (1990a). The most often omitted free morpheme categories in the Menn and Obler survey were auxiliaries, copulas, and forms of have used as a main verb (Menn & Obler 1990b). After verb-related free morphemes, the next most omitted categories of free morphemes (Menn & Obler 1990b) were pre/postpositions and determiners (pronouns and articles). Avrutin (2001) reports that subject DPs are omitted more frequently than object DPs when subject DPs are required by the language and that omission of subjects frequently co-occurs with nonfinite verbs. The speech of English-speaking agrammatics also includes utterances in which the speaker’s name or accusative case me is substituted for the nominative subject pronoun I. As important as understanding the salient features of agrammatic production is the recognition of the high degree of variability, both within the individual and across individuals, that is reported in the literature. Variability is the norm rather than the exception. Individual agrammatics rarely demonstrate a particular aspect of agrammatism all of the time, and different agrammatics demonstrate different patterns of agrammatic deficits (see discussions by Goodglass 1993; Kolk 1998, and Chap. 9, this volume). In fact, production data are often reported as percentages of omissions or substitutions of particular language features relative to the number that would have occurred as the output of the sentential grammar of the language. A study by Goodglass, Christiansen, and Gallagher (1993) illustrates both within- and across-individual variability. They report, for example, that, on average, seven agrammatic aphasics omitted determiners (articles) 50% of the time and auxiliaries and copulas 50% of the time in free narrative productions. Determiner omission varied, however, from 8% to 95% for different agrammatics, and auxiliary and copula omission varied from 30% to 88%. Similar results are reported by Rochon, Saffran, Berndt, and Schwartz (2000), in a comparison of 29 English-speaking agrammatics with 12 nondisordered speakers, and by Kolk (Chap. 9, this volume, citing Hofstede 1992), in a study of 22 Dutch-speaking agrammatics. These reports, showing different degrees of omission of functional categories for different agrammatics, demonstrate variability across individuals. Individual agrammatics also demonstrate within-individual variability in that few, if any, of those described in these studies demonstrate 100% omission of any of the language features reported. Data from these studies are consistent with Kolk’s proposal that agrammatic language production reflects two grammatical strategies, attempts at fully grammatical sentences and productions that result from a syntactically simpler strategy. They cannot, however, be used to characterize the nature of the structures resulting from these different strategies. It cannot, for example, be determined from these reports whether the pattern of omission of one feature co-occurs with that of another without examining each utterance of the corpus individually. Accounts of agrammatic production rarely address issues of variability, even when proposed mechanisms entail specific expectations regarding the nature and prevalence of some forms of variability. Instead, accounts of agrammatic language deficits focus on the most salient (most reliably observed) aspects of agrammatic speech or comprehension:
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.5 (284-359)
Nonsententials and agrammatism
– –
Omission or lack of use of function words relative to content words Omission or lack of use of inflectional morphemes on verbs and nouns
In addition to these core features, accounts sometimes include other aspects of agrammatic production, including word order patterns and greater prevalence or use of nouns than verbs.
. Theoretical accounts of agrammatic production In the 1970s, psycholinguistic theories of agrammatic language production replaced approaches that focused on motor aspects of speech production. A psycholinguistic theory typically provides a description of grammatical knowledge that constrains the speaker’s utterances and specifies underlying processing operations that underlie utterance planning. Early psycholinguistic theories proposed that agrammatic production deficits resulted from the loss or unavailability of syntactic knowledge. In these theories, processing systems were proposed to account for agrammatic production in the absence of syntactic knowledge. Caplan’s (1985) theory, summarized below, takes this kind of approach. Other approaches have posited little or no loss of grammatical knowledge, or competence. Instead, production deficits are attributed to an inability to realize fully constructed sentence representations or sentence representations with selective underspecification. Friedmann and Grodzinsky (1997) have taken this approach, assigning disordered language production to pruning of the syntactic tree. Their approach will also be described below and contrasted with the one taken in this chapter. Whether the focus of the approach is underlying processing mechanisms or language structure, an account of the interplay between processing mechanisms and knowledge related to language structure is necessary to adequately describe the resulting disordered production. Agrammatic language production is arguably better characterized by its variability than by its core features (Kolk 1998). Rather than focusing on an account of the core features of agrammatism, Kolk (e.g., 1995, 1998, and Chap. 9, this volume) explores the interplay between syntactic representation and processing operations in an approach he calls adaptation theory. In this theory, Kolk provides an account of the variability, as well as the core features, of agrammatic production. Kolk does not, however, provide a theoretical characterization of the syntactic properties of the core features that would be compatible with his approach. The approach taken in this chapter will expand on Kolk’s theory to provide such a characterization.
. Absence of hierarchical structure (Caplan) In an early psycholinguistic account of agrammatism, Caplan (1985) attributed core deficits to damage, or loss of access, to syntactic knowledge, resulting in disordered
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.6 (359-403)
Patricia Siple
utterances. In Caplan’s account of agrammatic production, lexical semantic information is available to the agrammatic speaker for utterance planning, but the mechanism responsible for constructing phrasal categories is not. Lexical items are retrieved from the “content” word lexicon, along with lexical semantic information and lexical syntactic category. These items are ordered for production by mapping specific semantic information onto lexical category sequences. For example, “theme of verb” is mapped onto V + N. This process results in linear orderings of major lexical categories. Because phrase-building mechanisms are unavailable, function words and inflectional items are not produced. Caplan meant this account to be conservative – that is, to provide an account of the most salient features of agrammatic production. As such, he understood that it cannot account for the high degree of within-individual and between-individual variability demonstrated by many agrammatics. Accounts that propose a complete loss of hierarchical structuring abilities are too rigid to provide adequate accounts of both kinds of variability. Function words do appear in agrammatic speech, for example, and they occur in phrasal segments (e.g., Determiner Phrases, Prepositional Phrases). Because no role is given to syntax in Caplan’s theory, it is not necessary to make reference to syntactic knowledge except to say that it is unavailable. It is clear, however, that at least some aspects of syntactic structure are available during utterance planning by agrammatics, at least some of the time. Given this, it becomes necessary to specify the nature of that knowledge and the way(s) it influences processing.
. Tree pruning in production Caplan attempted to account for agrammatic production data with descriptions of underlying processes operating on semantic knowledge. Friedmann and Grodzinsky have taken the opposite approach, providing an account of agrammatic production in terms of deficits related to the syntactic structural representations of utterances/sentences (Grodzinsky 1984, 1990, 2000; Friedmann & Grodzinsky 1997). Tense and agreement deficits, the most common forms of impairment described in reports of agrammatic production, have, for the most part, been treated together as a single class. Recently, however, examination of several cases of agrammatism led Friedmann and Grodzinsky (1997) to consider relative sparing of some types of grammatical elements and impairment of others. In their data, a Hebrew-speaking agrammatic demonstrated impaired tense inflection, but near-normal agreement inflection in sentence repetition and completion tests as well as in spontaneous speech. Attempts to elicit wh-questions were unsuccessful and repetition of wh-questions was poor. Examination of other cases reported in the literature indicated three patterns of syntactic deficits in agrammatic production: cases like those of Friedmann and Grodzinsky except with both tense and agreement deficits, cases with the same pattern as in their data, and cases like theirs but with relative sparing of both tense and agreement. Friedmann and Grodzinsky proposed a syntactic account of these data, the tree-pruning hypothesis. They argue that when a particular node of the syntactic tree is impaired (when it is underspecified), feature checking cannot occur and
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.7 (403-450)
Nonsententials and agrammatism
this may produce grammatical errors. They argue further that agrammatics are unable to construct nodes in the syntactic tree above the impaired node. That is, the syntactic tree is “pruned” in the lowest node displaying a production deficit. They noted that the split-inflection framework of Pollock (1989) provided an account of (and was supported by) their Hebrew data. In this framework, inflection does not occupy a single Inflection node; rather, tense and agreement are represented separately with the Tense node dominating the Agreement node and Negation falling between them. According to the tree-pruning hypothesis, then, impairment of both tense and agreement indicates pruning in the Agreement node; impairment of tense but sparing of agreement indicates pruning in the Tense node; and sparing of both tense and agreement but impairment of complementizer-related structures indicates pruning in the Complementizer node.1 The tree-pruning hypothesis represents an advance in syntactic accounts of agrammatic production in that it has a built-in mechanism for handling betweenindividual variability. Severity of impairment is related to the locus of pruning; and for a particular locus of pruning, certain patterns of deficits are expected and others are not. Specific predictions depend, however, on how the “pruning” metaphor is interpreted. Taken somewhat literally, individual trees would be constructed and then pruned prior to or during production. The pruning hypothesis is similar to a proposal described by Rizzi (1993/1994) to account for early child language acquisition – a proposal that has been called the truncation hypothesis. Although it is theoretically possible that full trees are created and then truncated for production, it seems unlikely. First, effort is given to building parts of a tree that are then unused. More importantly, for many constructions, several elements below the locus of pruning would contain only bundles of features (following movement of elements), leaving very few or no lexical items available for production. A more likely interpretation is that the tree is pruned prior to syntactic construction. Under this argument, agrammatics have lost the ability to construct nodes of the tree above a particular node because of the unavailability of language knowledge or the processes needed to use it in tree construction. This interpretation would appear to be too inflexible since conditions for selecting the pruning locus are not related to difficulty of constructing a particular tree but, instead, are more likely an inherent characteristic of the agrammatic. This interpretation would predict an all-or-none-like pattern of specific forms of production deficits. Thus, this interpretation would have difficulty accounting for the great within-individual variability found in agrammatic production. A third interpretation of tree pruning is consistent with a proposal put forward by Hagiwara (1995). Hagiwara, working within the Minimalist framework (e.g., Chomsky 1995), argues that the agrammatic begins to construct the syntactic tree from the bottom up, using the Merge operation. Borrowing an argument from Caplan and Hildebrandt (1988) that the amount of computational space available to the agrammatic for language processing is reduced, Hagiwara proposes that tree construction breaks down because of this limitation, truncating the tree at a node lower than the Complementizer node. This interpretation holds promise, but would need to be specified in greater detail than is found in Hagiwara (1995) or
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.8 (450-517)
Patricia Siple
in Friedmann and Grodzinsky (Friedmann & Grodzinsky 1997; Friedmann 2001) in order to account for the variability found in agrammatic production. Tree-pruning approaches assume that the utterances produced by agrammatics are ungrammatical, no matter how “pruning” is interpreted, since they stop short of building a complete tree. In contrast, the approach I explore below, consistent with Kolk’s analysis, proposes that much of agrammatic production can in fact be characterized as normal nonsententials.
. Processing in the temporal-window and adaptation theory (Kolk) Most accounts of agrammatic production are unable to account for the high degree of within- and/or between-individual variability found in the data. Kolk (1998), in a review of contemporary approaches to agrammatism, has called for an integration of these two types of approaches, arguing that neither representational nor processing approaches alone can provide an adequate account. Kolk (e.g., Chap. 9, this volume) has, for a number of years, taken this approach, integrating processing and representation in adaptation theory. In Kolk’s adaptation theory, agrammatic production is constrained by a reduction in the size of the temporal window available for construction of the syntactic tree (e.g., Kolk 1995, 1998, and Chap. 9, this volume). The amount of reduction, resulting from a slowing of retrieval of syntactic knowledge and/or abnormally fast decay of the syntactic construction, varies for individual agrammatics as a function of severity. Construction of a syntactic tree requires adequate processing resources to maintain synchrony of construction of the parts of the tree. Reduction of the temporal window for processing causes desynchronization in the construction of complex trees, with the likelihood of desynchronization increasing with syntactic complexity. Specific production deficits result from the interaction of syntactic complexity and the agrammatic’s specific temporal processing deficit. Individual differences in the amount of reduction of the temporal window (the processing deficit) account for between-individual variability in agrammatic production according to adaptation theory. Kolk also notes that agrammatics demonstrate a great deal of within-individual variability, especially as a result of differences in the methods used to elicit language production. Elicitation in unstructured free conversation demonstrates the greatest degree of deficit. In structured elicitation tasks, the greater the task structure, the fewer deficits in agrammatic speech (Hofstede & Kolk 1994). Kolk attributes this withinindividual variation to a process of adaptation. He argues that both nondisordered and agrammatic speakers possess two different strategies for producing utterances. The goal of one system is to produce full grammatical utterances whereas the goal of the other system is to reduce processing time through syntactic simplification that results in telegraphic output. Kolk calls this telegraphic speech ellipsis, but the term ellipsis is defined empirically, or operationally, rather than theoretically. Telegraphic forms are those forms found in nondisordered casual conversation consisting of nonfinite utterances and utterances without verbs. Kolk argues that agrammatics overuse
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.9 (517-569)
Nonsententials and agrammatism
the telegraphic strategy in order to compensate for processing difficulties encountered when attempting to produce full grammatical forms. In several studies of Dutch and German speakers, Kolk and his colleagues have demonstrated that telegraphic utterances in agrammatic free conversation are similar in structure to telegraphic utterances found in casual conversation of nondisordered adults (e.g., Kolk & Heeschen 1990, 1992; Kolk, Chap. 9, this volume); in telegraphic utterances produced by public officials talking with foreign immigrants (Hofstede 1992, cited in Kolk 1995); and in the early utterances of children (Kolk 2001, and this volume). Moreover, when nondisordered speakers were asked to limit their responses to two to three words in a picture description task, their utterances resembled those of agrammatics (de Roo, Kolk, & Hofstede 2003). For nondisordered speakers, use of the telegraphic strategy is reserved for a limited number of special situations involving casual conversation in which the telegraphic utterances can be understood from context. Agrammatics, however, turn to this telegraphic strategy much more frequently because of their often severe processing limitations. Kolk’s approach to agrammatic production is comprehensive in that it accounts for a wider range of features of agrammatic language impairment than do purely process models (e.g., Caplan 1985) or models based on syntactic representation (e.g., Grodzinsky 2000). In Kolk’s account, core features of agrammatic production reflect overuse of a separate language system that produces telegraphic utterances. Other features of syntactic impairment occur as a result of a processing deficit that restricts the agrammatic’s ability to produce full sentences. Both between-individual and within-individual variability are accounted for in a principled way. Between-individual variability is related to the amount of processing reduction imposed by severity of impairment. Within-individual variability is related to variability in processing requirements of the language task and the specific communication setting. Because of these features of Kolk’s approach, it will be taken as a starting point for the framework described below. The proposed framework will provide a formal description of syntactic representation, based in Minimalism, for both the telegraphic strategy and the full grammar strategy in Kolk’s account. Kolk describes the output of the telegraphic strategy empirically by examining output in situations where it is proposed to operate. Labeling telegraphic utterances as instances of ellipsis, however, can be taken to implicate formally described processes of ellipsis that do not result in syntactic simplification. In the following section, a formal description of syntactic representation for the telegraphic strategy is presented that is both syntactically simpler than full-sentence representations and can account for the general features of telegraphic language production. Commitment to a theory of language knowledge increases the predictive power of the approach.
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.10 (569-612)
Patricia Siple
. Extending Kolk: Syntactic structure and adaptation theory There is growing evidence that a nonsentential grammar, based in Minimalism, describes early child language development (Radford 1990; Roeper 1999). Radford describes the use of a nonsentential grammar as an early stage of language acquisition. Stage theories often include the constraint that later stages replace earlier ones, but this may not be the case for nonsentential grammar. Instead, it is possible that the nonsentential grammar remains available after a full grammar system has developed, and is used by adults in specific contexts (Roeper; Barton & Progovac 2005; Progovac, Chap. 2, this volume). This nonsentential grammar can provide a theoretical basis for the component of grammar that Kolk proposed to account for “telegraphic” utterances produced by agrammatics. Both the nonsentential component and the full grammar component would have been available to agrammatic aphasics prior to the event resulting in damage to Broca’s area. Thus, it should be possible to characterize syntactic productions of agrammatics in terms of the extent to which agrammatic utterances reflect each of these two components of grammar. At the same time, agrammatics serve as a test case for the existence of two distinct grammatical components. The two grammar components, a nonsentential component and a full sentential grammar component, provide the language knowledge or representational component of a model for understanding syntactic deficits in agrammatics. In contrast to most previous representational approaches (including some interpretations of the tree-pruning approach), grammatical knowledge, consisting of both a nonsentential and a sentential component, is assumed to be intact in the model proposed here. A disordered process component that produces a reduction in the processing resources available for language production is posited to underlie agrammatic production deficits. When fully sentential utterances are attempted by agrammatics, the processing limitation imposed by a reduction in available processing resources often results in disordered productions. Syntactic structures derived from the nonsentential component of the grammar (described below and in Progovac, Chap. 2, this volume) are less complex than fully grammatical structures. They are, therefore, less influenced by a reduction in processing resources since they require fewer resources for production than do attempts at fully grammatical utterances. Kolk (Chap. 9, this volume) has argued that agrammatics compensate for deficient processing resources by electing to produce syntactically less complex utterances in a wider range of contexts than do nondisordered individuals. The greater the processing deficit, the more likely the election of these types of constructions, with more severely impaired agrammatics producing these simpler constructions most, if not all, of the time. Kolk sometimes calls these syntactically less complex utterances “ellipsis.” If an elliptical analysis is used to account for these utterances, however, the resulting syntactic structures would be more complex than their full sentential counterparts because they involve additional operations such as movement and deletion (see Merchant, Chap. 3, this volume, and Barton, Chap. 1, this volume). In contrast to an elliptical analysis, I argue that the “elliptical” utterances described by Kolk are well-formed
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.11 (612-685)
Nonsententials and agrammatism
adult nonsententials described by the analysis proposed by Progovac (Chap. 2, this volume). These nonsententials, produced by the nonsentential component of grammar, are indeed syntactically less complex structures. Following Kolk’s argument, agrammatics produce nonsententials (use their nonsentential component of grammar) in a wider variety of contexts than do nondisordered speakers, with the proportion of well-formed adult nonsententials increasing with the severity of their disorder.
. Nonsentential grammar Progovac (Chap. 2, this volume) has argued that telegraphic utterances like those produced by agrammatic aphasics are systematic, rule-governed utterances that are the products of a nonsentential component of grammar that is governed by UG in the same way that a full sentential grammar is. In adults, this nonsentential component coexists with full grammar capability and is used instead of full sentential grammar under certain conditions. The core claim of the argument presented by Progovac is that the Tense node is absent in telegraphic utterances, resulting in less complex, phrasal and root small clausal structures. In these structures, time is given pragmatically, by context, or by the use of temporal adverbs. Typically, nonsentential utterances are set in the “here and now.” The predicate and any argument(s) are generated within the Predicate Phrase, such as Verb Phrase, Adjective Phrase, and Prepositional Phrase. NP arguments in nonsentential constructions may, but need not, have DP structure. NP arguments in subject position are argued not to receive structural nominative case, due to the lack of Tense, which in the theory of Minimalism is responsible for assigning nominative case. The lack of structural case on the subject in turn licenses the use of default case pronouns (normally accusative in English, but see the Epilogue of this volume for variation in this respect), and the possibility to omit articles, which in Minimalism are also required by the case-assignment mechanism. The analysis presented by Progovac provides well-formedness criteria for nonsentential utterances. Under this analysis, well-formed nonsentential constructions should include the following features: –
–
–
Systematic absence of Tense and Agreement on verbs, including Absence of finite forms Omission of copula Omission of auxiliaries Absence of structural nominative case on subject pronouns The use of default Case, which is typically accusative in English (but see the Epilogue of this volume for possible variation) Possible omission of determiners Some omission of pronoun subjects (when the language requires them) Correlation of tense and case features indicated above
Utterances demonstrating these characteristics should be seen in abundance in agrammatic production.
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.12 (685-738)
Patricia Siple
. Nonsentential utterances in agrammatic production The features of nonsentential constructions listed above have been commonly reported as features of agrammatic production across a wide variety of languages (e.g., Menn & Obler 1990b; Goodglass et al. 1993; Rochon et al. 2000). Omissions of tense and agreement are frequent in agrammatic speech. Nonfinite forms are common; auxiliaries are omitted, as are copulas and have as a main verb; and tense and agreement affixes are either substituted or omitted. Examples of nonsentential utterances produced by English-speaking agrammatics that demonstrate the occurrence of these features in both phrase and root small clause constructions are given in (1) through (17).2 Some of these utterances are taken from agrammatics’ attempts to describe what is going on in the Cookie Thief picture taken from the Boston Diagnostic Aphasia Test (Goodglass & Kaplan 1972). VP: (1) Fall down (Cookie Thief: A boy is standing on a stool that is tipping over) (2) Come over (Cookie Thief: Water is spilling over the side of the sink) (3) Pass out (for And then I passed out) (4) Mother fill tub (Cookie Thief: Mother is filling the tub/sink with water) (5) Baby cry (for The baby cries) (6) Him shoos3 (for He says shhh!) NP/DP: (7) Me . . . one guy (beginning of a response to a question about what the speaker did in the army) (8) Speech (to explain why he is at the hospital) (9) Cookie jar (Cookie Thief picture description) (10) Four years ago (to indicate Time for succeeding utterances)
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.13 (738-789)
Nonsententials and agrammatism
(11) Lower Falls . . . Maine . . . Paper (responding to a question about former employment in a paper mill in Lower Falls, Maine) (12) Nice day (for It’s a nice day) (13) Low pressure (for My blood pressure is low) (14) Red Riding Hood pretty girl (for Red Riding Hood was a pretty girl) AP: (15) Weather overcast (when asked about the weather) PP: (16) Various of work (for Various kinds of work) (17) Little Red Hood outside home (for Little Red Riding Hood runs outside the house) A crucial component of the analysis of nonsententials provided by Progovac is that lexical items in nonsententials differ from those in full sentential productions in that they are selected with default features for Tense and Case rather than receiving Case and Tense features through feature checking in TP. Progovac (Chap. 2, this volume) argues that the default case is accusative in English. Thus, subject pronouns in English nonsententials are marked with accusative case instead of nominative case. The use of accusative case pronouns in subject position is evident in agrammatic production. Goodglass (1968) described an agrammatic who always used me in subject position. This became a focus of his therapy and, after several reminders by his therapist, he produced the utterance in (18). (18) I. . ., I. . ., everybody tell me “I”. . .but me forget! This utterance includes the small clause nonsentential in (19), with the accusative case pronoun, me, in subject position, within a coordinate construction:
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.14 (789-843)
Patricia Siple VP
(19) me
V’ V forget
As Progovac has argued, structures with accusative case pronouns in subject position provide some of the strongest evidence for the existence of a separate nonsentential component of grammar because they cannot be accounted for by processes of ellipsis. Accusative case subject pronouns are often found in agrammatic production as can be seen in (6) and (7) above and in (25) below. A future study is necessary, however, to determine if there is variation in this respect, and if nominative case consistently correlates with the use of finiteness. Although there are many instances of the use of default accusative case in subject position in nonsententials, some agrammatics have found ways to avoid the use of subject pronouns altogether. In many cases they are simply omitted; but in some cases, agrammatics have developed other strategies for avoiding the use of subject pronouns. Goodglass (1976) reports the utterance in (20) in which the speaker’s name is substituted for the nominative pronoun I, and another case in which an agrammatic typically used the formulaic expression this guy instead of he in subject position, as in (21). (20) Dad and Peter Haney (referring to himself by his full name) and Dad . . . hospital (to indicate that he and his father had come back to the hospital) (21) This guy work here (attempting to repeat He works here in a sentence repetition task) Absence of Tense also results in nonsententials with nonfinite verbal forms, as in (1) through (6), or absence of the verb, as in (7) through (17). The utterance in (3) is a phrasal nonsentential with a nonfinite verb in a situation where a past-tense verb is called for to express the proposition and then I passed out: (22)
VP V’ V
P
pass
out
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.15 (843-903)
Nonsententials and agrammatism
Utterance (15), an adjective root small clause, illustrates omission of the verb – in this case the copula, which correlates with the omission of the article in the subject position: (23)
AP Weather
A’ A overcast
Agreement affixes are also omitted in nonsententials and in agrammatic speech. Goodglass (1976) reports that a severely impaired agrammatic repeated baby several times before producing the following utterance. (24) Baby cry (for The baby cries) In this utterance, the article the is omitted from the NP in addition to omission of the third-person agreement marker -s on the verb, which illustrates the correlation between tenselessness and caselessness as discussed above. Free narratives of agrammatics often include strings of nonsententials that, taken together, express complex meaning relationships. In the sample in (25) through (31), taken from Goodglass (1976), a 22-year-old man is talking about events during his late teens. (25) Me . . . uh fifteen (26) Uh, oh, I guess six month. . . (27) My mother pass away (for Six months after I became 15 my mother died.) (28) An’uh . . . an’en . . . uh . . . ah..seventeen (29) Seventeen (30) Go . . . uh High School (for When I was 17, I was going to High School.) (31) An un . . . Christmas . . . well, uh, I uh . . . Pittsburgh (for and I went to Pittsburgh for Christmas) This excerpt demonstrates the presence of several features of nonsententials and the correlation of tense features and case features in individual utterances in free narration. In utterance (25), the accusative pronoun me occurs in subject position and the verb/copula is missing. Utterance (26) is used to indicate time adverbially. Formulaic phrases like I guess often occur in agrammatic speech. Utterances (27) and (30) provide additional examples of nonfinite verb forms in situations that would normally
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.16 (903-982)
Patricia Siple
involve tense. In (30), the pronoun subject is also omitted. In (28) and (29), both the subject and copula are omitted. Finally, both the verb and prepositions are omitted in (31). This extended example also illustrates the avoidance of embedding, and the use of parataxis instead (for further discussion of this phenomenon in nonsententials in general, see the Epilogue of this volume). As discussed above, Progovac (Chap. 2, this volume) argues that another consequence of the use of the nonsentential grammar is the possible omission of articles in Noun Phrases. Agrammatics often omit articles, as in the omission of the in (4), (5), (15), and (17) and the omission of a in (12) and (14); but in some cases they do not. Goodglass, Fodor, and Schulhoff (1967) report the following utterance from an English-speaking agrammatic. (32) Write a letter (attempting to repeat I am writing a letter in a sentence repetition task) In this utterance, the nominative pronoun subject, the auxiliary, and the progressive affix -ing are all omitted, but the article a is included in the utterance. In utterance (4), the article the is omitted from both the subject NP and the object NP, and the auxiliary and the progressive affix -ing are omitted from the VP. In nonsententials, negation is realized by placing no either before or after a phrase or small root clause. Examples of negation by an English-speaking agrammatic are given in (33) and (34), taken from Myerson and Goodglass (1972).4 (33) No New York (34) Money no Agrammatics’ use of negation with nonsententials shows striking similarities with that in pidgins and in first and second language acquisition (see the relevant chapters in this volume). The examples given above demonstrate the presence of nonsententials in agrammatic production that conform to the syntactic analysis of nonsententials proposed by Progovac (Chap. 2, this volume). Nonsentential structures like these are found in large number in productions of English-speaking agrammatics. Kolk (Chap. 9, this volume) presents a similar set of nonsentential utterances produced by Dutch-speaking agrammatics. These, too, appear to correspond to the analysis presented by Progovac, providing cross-linguistic support for the analysis. An important aspect of the nonsentential analysis is that the individual features of nonsententials should co-occur in utterances. Several examples of co-occurrence are described above for English-speaking agrammatics. Recent studies of Dutch- and German-speaking agrammatics report similar types of co-occurrences of syntactic features predicted by a nonsentential analysis. In a case study of a Dutch-speaking agrammatic, de Roo (2001) reports that only 8% of utterances elicited in situations requiring spontaneous speech contained a finite verb. Of the other 92%, 35% included an infinitive or past participle verb and 57% did not contain a verb at all. The data
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.17 (982-1023)
Nonsententials and agrammatism
indicated a strong relation between nonfiniteness and subject omission. Subjects were omitted in nonfinite utterances 96% of the time, but were present in finite utterances 91% of the time. Infinitive forms sometimes included negatives or time adverbials, and objects were present more often than subjects. In another, multi-case study, Rugendijk, van Zonneveld, and Bastiaanse (1999) examined case assignment in 6 Dutch-speaking and 6 German-speaking moderately to severely impaired agrammatic aphasics. In both Dutch and German, determiners and pronouns are marked for case (nominative, accusative, dative). In sentences with no finite case assigner, subjects are marked with default nominative case. In agrammatic utterances without a finite verb, accusative and dative case pronouns and determiners were generally omitted. Pronouns that were produced without a finite verb were all in nominative (default) case. Dutch-speaking agrammatics tended to omit determiners in nonfinite utterances, whereas Germanspeaking agrammatics either omitted the determiner or used nominative (default) case in nonfinite utterances. Both patterns are consistent with a nonsentential analysis. These cross-linguistic data add further support for the presence of a nonsentential component of grammar that is used more often and in a wider range of contexts by agrammatics.
. Full-grammar versus nonsentential utterances Except for the most severely impaired, agrammatic aphasics sometimes attempt fullgrammar constructions in addition to their nonsentential constructions. The data described above from de Roo (2001) display a clear distinction between these two types of constructions. Subjects were omitted in nonfinite constructions, but were present in finite constructions. Similarly, in the study by Rugendijk et al. (1999), accusative and dative pronouns and determiners were omitted in utterances without a finite verb, but were present in utterances with finite verbs, providing evidence for the operation of two grammatical components. Other features of full-grammar constructions set them apart from nonsentential utterances. Attempts at full-grammar utterances include syntactic elements not predicted by nonsentential grammar and are often characterized by attempts at repair. An example of an agrammatic utterance that includes projections above TP (cited by Grodzinsky 1990) is given in (35). (35) I think I’ll wear um dress An example of an attempted repair (taken from Goodglass, Gleason, Bernholtz, & Hyde 1972) is given in (36b). The agrammatic is responding to an elicitation situation in which nondisordered speakers typically produce (36a). (36) a.
He wanted him to clean the gun (nondisordered speaker) b. To do again, he wanted do it again (agrammatic speaker)
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.18 (1023-1088)
Patricia Siple
Repairs like these lead to improved grammatical productions about half the time according to Goodglass et al. Other evidence supporting the distinction between fullgrammar and nonsentential utterances was described above in the discussion of Kolk’s adaptation theory, including decreases in nonsentential utterances with increases in the amount of syntactic structure provided in different elicitation tasks.
. Summary of the nonsententials approach The account of agrammatic language production taken in this chapter begins with Kolk’s adaptation theory as a starting point. Kolk’s theory has the advantage of being able to account both for the major features of agrammatic production and for the high degree of variability in production between and within individual agrammatics. Departing from other accounts that view agrammatic utterances as ill formed, Kolk (e.g., Chap. 9, this volume) has argued that many agrammatic utterances result from overuse of a grammatical strategy available to and used by nondisordered, as well as agrammatic, speakers, and are thus well formed. Kolk describes utterances produced using this strategy as elliptical or telegraphic forms found in nondisordered casual conversation consisting of nonfinite utterances and utterances without verbs. Kolk does not, however, provide a theoretical, syntactic analysis of these telegraphic forms. This chapter provides a theory-driven account of these telegraphic utterances using a nonsentential analysis described by Progovac (Chap. 2, this volume). Working within a Minimalist framework, Progovac (Chap. 2, this volume) argues that structures like those alluded to by Kolk are the products of a nonsentential component of grammar. Structures produced by the nonsentential component of grammar are not full sentences in that the Tense node is absent. Following Minimalist principles, nonsententials are characterized by the absence of syntactic phenomena that rely on Tense. According to the analysis presented by Progovac, well-formed nonsententials will lack finiteness and certain case-related properties. In English, this includes use of accusative default case pronouns in subject position. Importantly, tense features and case features of nonsententials should be correlated, or found together, in individual structures. I have demonstrated above that these types of structures are prevalent in agrammatic production. Tenselessness is a core feature of agrammatic production. Agrammatics use nonfinite verb forms in situations that would normally involve tense, omit verbs altogether, and omit copulas, auxiliaries, and agreement affixes. Caserelated features of nonsententials observed in agrammatic production include use of accusative case pronouns in subject position by English-speaking agrammatics, omission of pronoun subjects, use of strategies to avoid subject pronouns, and omission of articles in some NP constructions. All of these tense and case features are predicted by the nonsententials analysis described by Progovac. Examples of the tense and case features characteristic of nonsententials have been widely reported in studies of agrammatism. There is also evidence that these features co-occur in individual utterances. In fact, Myerson and Goodglass (1972: 42) report an analysis of a small corpus of data from a severely impaired English-speaking agram-
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.19 (1088-1135)
Nonsententials and agrammatism
matic and conclude that “he does not reach the sentence level but confines himself to the phrase level.” From the description of the grammar of this agrammatic, the entire corpus of data appears to consist of well-formed nonsentential utterances. In less severely impaired agrammatics, well-formed nonsententials that include both tense and case features described by Progovac are found in greater number and in a wider range of contexts than in nondisordered speakers. Nonsentential utterances are distinguished from attempts at fully grammatical utterances in that attempts at fully grammatical utterances: (a) typically include the co-occurrence of finite verbs and case-related features that are assigned by tense, (b) include projections above TP, and (c) are characterized by attempts at repair. These attempts at fully formed sentences also differ from nonsententials produced by agrammatics in that they are generally ill-formed utterances. Kolk has suggested that agrammatic production includes nonsententials in greater number and in more contexts because processing limitations make full-grammar productions difficult or impossible. Data from agrammatics support the proposal that grammatical knowledge consists of two components, a syntactically simpler nonsentential component and a full-grammar component. The nonsentential analysis described by Progovac (Chap. 2, this volume) provides a theory-driven account of the nonsentential utterances produced by agrammatics. Specific details of this approach, however, remain to be investigated in future research.
. Conclusion A review of syntactic deficits in language production by agrammatic aphasics indicates that the central feature of agrammatic production is impairment of tense. Other deficits also occur in agrammatic production, and these are primarily related to omission (and sometimes substitution) of functional categories. Both processing and representational accounts of agrammatic production have been offered, but neither of these types of approaches alone are able to account for the nature and variability of agrammatic language production. For example, theoretical approaches that describe agrammatic deficits as tree pruning (e.g., Grodzinsky 2000) or as impaired access to, or processing of, functional categories (e.g., Caplan 1985) account for some features of agrammatic production, but they cannot account for the large amount of withinindividual variability in production. All of these approaches ascribe ill-formedness to agrammatic utterances. Unlike other approaches, Kolk’s adaptation theory (e.g., 1995, 1998, and Chap. 9, this volume) includes both processing and representational components. Syntactic deficits and within-individual variability in agrammatic production are attributed to reduced processing resources. According to adaptation theory, agrammatics adapt to their reduction of processing resources for production by overusing telegraphic constructions available to all adult speakers. It is therefore important in Kolk’s account that these telegraphic constructions be grammatically less complex than attempts at
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.20 (1135-1190)
Patricia Siple
fully grammatical constructions. Kolk defines telegraphic constructions empirically and demonstrates the similarity between telegraphic utterances produced by agrammatic and nondisordered speakers. Kolk argues that within-individual variability results from the degree to which telegraphic constructions are used in different contexts. He does not, however, offer a theoretical account of the syntactic properties of these telegraphic utterances that would be consistent with his analysis. In the account proposed here, Kolk’s adaptation theory is taken as a starting point. Within the proposed account, grammatical knowledge, represented within the Minimalist Program, is taken to be intact, and syntactic deficits are attributed to a reduction in processing resources required for realizing syntactic constructions. Similar to adaptation theory, both agrammatic and nondisordered individuals have two grammatical components available for use in production. One of these is a nonsentential component that guides construction of well-formed nonsentential utterances and the other component is the full sentential grammar of their language. A nonsentential grammar proposed by Progovac (Chap. 2, this volume) provides a good account of nonsentential utterances produced by agrammatics. This nonsentential grammar captures key features of agrammatic production related to tense and case and makes specific predictions about the co-occurrence of these tense and case features. Nonsentential grammar has the added advantage of describing simpler syntactic structures for these utterances (an assumption of Kolk’s adaptation theory) whereas other accounts of these kinds of structures, such as elliptical analyses (Barton, Chap. 1, this volume; Merchant, Chap. 3, this volume), describe syntactic structures that are more complex than their full sentential counterparts. Agrammatics use nonsentential grammar to a much greater extent than nondisordered individuals in order to compensate for their reduction in processing resources. The syntactic analysis of nonsententials provided by Progovac (Chap. 2, this volume) provides a formal description that can be used to identify and classify nonsentential, telegraphic productions and dissociate them in principled ways from attempts at full sentential productions. The strength of the nonsentential analysis described by Progovac is that it not only makes specific, theory-derived predictions about the occurrence of features and co-occurrence of sets of features that account for a large subset of agrammatic productions in a principled, systematic way; but it is also consistent with Kolk’s theory, which provides a mechanism for both between- and within-individual variability. The features of this extension of Kolk’s adaptation theory make it possible to provide a better account of the full range of agrammatic productions and, in addition, place this account of agrammatism within a broader, more general account of nonsentential constructions. At the same time, the ability of Progovac’s analysis of nonsententials to account for a subset of agrammatic productions representing core features of the disorder underscores the strength and generalizability of this analysis.
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.21 (1190-1277)
Nonsententials and agrammatism
Notes . Ouhalla (1993) has presented a related proposal that limits agrammatic production to VP constructions. This could be viewed as pruning above the VP node. In Ouhalla’s account, functional categories are posited to form a separate component of UG, a UG lexicon. Ouhalla argues that this component is impaired in agrammatism, preventing access to the categories in the UG lexicon. As it is presented, this account does not provide the flexibility needed to account for the wide variety of constructions found in agrammatic production. It can account for neither the between-individual nor the within-individual variability found in agrammatic data. . Examples (1), (2), (4) through (6), and (10) are taken from Myerson and Goodglass (1972); (3), (12)–(14), and (16)–(17) are from Menn (1990); (7) through (9) are from Goodglass and Kaplan (1972); (11) is from Goodglass (1968); (15) is from Geschwind (1970). . It is unclear in this example whether the verb is inflected or the speaker is having difficulty pronouncing shhh or shush because of his disfluency. . No glosses are given for these utterances.
References Avrutin, S. (2001). Linguistics and agrammatism. GLOT International, 5(3), 1–11. Barton, E. & Progovac, L. (2005). Nonsententials in Minimalism. In R. Elugardo & R. J. Stainton (Eds.), Ellipsis and Non-Sentential Speech (pp. 71–93). Dordrecht: Kluwer. Caplan, D. (1985). Syntactic and semantic structures in agrammatism. In M.-L. Kean (Ed.), Agrammatism (pp. 125–152). New York, NY: Academic Press. Caplan, D. & Hildebrandt, N. (1988). Disorders of Syntactic Comprehension. Cambridge, MA: The MIT Press. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: The MIT Press. de Roo, E. (2001). Root nonfinite and finite utterances in child language and agrammatic speech. Brain and Language, 77, 398–406. de Roo, E., Kolk, H., & Hofstede, B. (2003). Structural properties of syntactically reduced speech: A comparison of normal speakers and Broca’s aphasics. Brain and Language, 86, 99–115. Eling, P. (Ed.). (1994). Reader in the History of Aphasia: From (Franz) Gall to (Norman) Geschwind. Amsterdam: John Benjamins. Friedmann, N. (2001). Agrammatism and the psychological reality of the syntactic tree. Journal of Psycholinguistic Research, 30, 1–24. Friedmann, N. (2002). Question production in agrammatism: The tree pruning hypothesis. Brain and Language, 80, 160–187. Friedmann, N. & Grodzinsky, Y. (1997). Tense and agreement in agrammatic production: Pruning the syntactic tree. Brain and Language, 56, 397–425. Geschwind, N. (1970). The organization of language and the brain. Science, 170, 940–944. Goodglass, H. (1968). Studies on the grammar of aphasia. In S. Rosenberg & J. H. Koplin (Eds.), Developments in Applied Psycholinguistics Research (pp. 177–208). New York, NY: Macmillan. Goodglass, H. (1976). Agrammatism. In H. Whitaker & H. A. Whitaker (Eds.), Studies in Neurolinguistics, Volume 1 (pp. 237–260). New York, NY: Academic Press. Goodglass, H. (1993). Understanding Aphasia. San Diego, CA: Academic Press.
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.22 (1277-1376)
Patricia Siple
Goodglass, H., Christiansen, J. A., & Gallagher, R. (1993). Comparison of morphology and syntax in free narrative and structured tests: Fluent vs. nonfluent aphasics. Cortex, 29, 377–407. Goodglass, H., Fodor, I. G., & Schulhoff, C. (1967). Prosodic factors in grammar: Evidence from aphasia. Journal of Speech and Hearing Research, 10, 5–20. Goodglass, H. & Geschwind, N. (1976). Language disorders (aphasia). In E. C. Carterette & M. P. Friedman (Eds.), Handbook of Perception, Volume VII: Language and Speech (pp. 389–428). New York, NY: Academic Press. Goodglass, H., Gleason, J. B., Bernholtz, N. A., & Hyde, M. R. (1972). Some linguistic structures in the speech of a Broca’s aphasic. Cortex, 8, 191–212. Goodglass, H. & Kaplan, E. (1972). The Assessment of Aphasia and Related Disorders. Philadelphia, PA: Lea & Febiger. Grodzinsky, Y. (1984). The syntactic characterization of agrammatism. Cognition, 16, 99–120. Grodzinsky, Y. (1990). Theoretical Perspectives on Language Deficits. Cambridge, MA: The MIT Press. Grodzinsky, Y. (2000). The neurology of syntax: Language use without Broca’s area. Behavioral and Brain Sciences, 23, 1–71. Hagiwara, H. (1995). The breakdown of functional categories and the economy of derivation. Brain and Language, 50, 92–116. Hofstede, B. T. M. (1992). Agrammatic Speech in Broca’s Aphasia: Strategic Choice for the Elliptical Register. Ph.D. dissertation, University of Nijmegen. Hofstede, B. T. M. & Kolk, H. H. J. (1994). The effects of task variation on the production of grammatical morphology in Broca’s aphasia: A multiple case study. Brain and Language, 46, 278–328. Jarema, G. (1998). The breakdown of morphology in aphasia. In B. Stemmer & H. A. Whitaker (Eds.), Handbook of Neurolinguistics (pp. 221–234). San Diego, CA: Academic Press. Kolk, H. (1995). A time-based approach to agrammatic production. Brain and Language, 50, 282–303. Kolk, H. (1998). Disorders of syntax in aphasia: Linguistic-descriptive and processing approaches. In B. Stemmer & H. A. Whitaker (Eds.), Handbook of Neurolinguistics (pp. 249–260). San Diego, CA: Academic Press. Kolk, H. (2001). Does agrammatic speech constitute a regression to child language? A threeway comparison between agrammatic, child, and normal ellipsis. Brain and Language, 77, 340–350. Kolk, H. & Heeschen, C. (1990). Adaptation symptoms and impairment symptoms in Broca’s aphasia. Aphasiology, 4, 221–231. Kolk, H. & Heeschen, C. (1992). Agrammatism, paragrammatism and the management of language. Language and Cognitive Processes, 7, 89–129. Menn, L. (1990). Agrammatism in English: Two case studies. In L. Menn & L. Obler (Eds.), Agrammatic Aphasia: A Cross-Language Narrative Sourcebook (pp. 1369–1389). Amsterdam: John Benjamins. Menn, L. & Obler, L. (1990a). Agrammatic Aphasia: A Cross-Language Narrative Sourcebook. Amsterdam: John Benjamins. Menn, L. & Obler, L. (1990b). Cross-language data and theories of agrammatism. In L. Menn & L. Obler (Eds.), Agrammatic Aphasia: A Cross-Language Narrative Sourcebook (pp. 1369– 1389). Amsterdam: John Benjamins. Miceli, G., Mazzucchi, A., Menn, L., & Goodglass, H. (1983). Contrasting cases of Italian agrammatic aphasia without comprehension disorder. Brain and Language, 19, 65–97.
JB[v.20020404] Prn:25/07/2006; 12:11
F: LA9310.tex / p.23 (1376-1429)
Nonsententials and agrammatism
Myerson, R. & Goodglass, H. (1972). Transformational grammars of three agrammatic patients. Language and Speech, 15, 40–50. Ouhalla, J. (1993). Functional categories, agrammatism and language acquisition. Linguistische Berichte, 143, 3–36. Panse, F. & Shimoyana, T. (1955). Zur auswirkung aphasischer störungen im Japanischen. Archive für Psychiatrie und Zeitschrift für Neurologie, 193, 131–138. [English translation in H. Goodglass & S. Blumstein (Eds.), Psycholinguistics and Aphasia (pp. 171–182). Baltimore, MD: Johns Hopkins University Press.] Pollock, J.-Y. (1989). Verb movement, universal grammar, and the structure of IP. Linguistic Inquiry, 20, 365–424. Radford, A. (1990). Syntactic Theory and the Acquisition of English Syntax. Oxford: Blackwell. Rizzi, L. (1993/1994). Some notes on linguistic theory and language development: The case of root infinitives. Language Acquisition, 3, 371–393. Rochon, E., Saffran, E. M., Berndt, R. S., & Schwartz, M. F. (2000). Quantitative analysis of aphasic sentence production: Further developments and new data. Brain and Language, 72, 193–218. Roeper, T. (1999). Universal bilingualism. Bilingualism: Language and Cognition, 2(3), 169–186. Rugendijk, E., van Zonneveld, R., & Bastiaanse, R. (1999). Case assignment in agrammatism. Journal of Speech, Language, and Hearing Research, 42, 962–971. Zurif, E. B. (1990). Language and the brain. In D. N. Osherson & H. Lasnik (Eds.), An Invitation to Cognitive Science, Volume 1 (pp. 177–198). Cambridge, MA: The MIT Press. Zurif, E. B. & Caramazza, A. (1976). Psycholinguistic structures in aphasia: Studies in syntax and semantics. In H. Whitaker & H. A. Whitaker (Eds.), Studies in Neurolinguistics, Volume 1 (pp. 261–292). New York, NY: Academic Press.
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.1 (48-122)
chapter
Reduced syntax in (prototypical) pidgins Donald Winford This chapter is a pretheoretical attempt to account for the characteristics of prototypical pidgin syntax. I argue that pidgins are the result of processes of grammar construction similar to those that operate in early second language acquisition. The earliest stages of both are characterized by minimal structures that often lack verbs and functional projections. These later develop into prototypical pidgins and stable forms of early interlanguage (IL), which have regular word order and syntactic conventions. Such grammars are best understood as manifestations of universal processes, not of syntactic reduction, but of grammar building, or restructuring. In some respects, the grammar that produces early IL and pidgin structures shares characteristics with the grammar of the telegraphic speech produced by adults already competent in a language.
Second language acquisition [and pidginization] research has been limited in the past by a tendency to focus on product rather than process, form rather than function, single rather than multiple levels of linguistic analysis, and isolated utterances rather than conversational context. (Extra & van Hout 1996: 99)
.
Introduction
Pidgins have traditionally been defined as languages with a minimal and highly reduced lexicon and grammar, lacking the range of grammatical devices and expressive power associated with other natural languages. Bloomfield (1933: 474) defined a pidgin as “a variety whose grammar and vocabulary are very much reduced. . . The resultant language must be native to no one.” The key elements of this definition, reduction and nonnative status, are generally accepted as the chief defining characteristics of prototypical pidgins (Thomason 1995). This is the sense in which I use the term pidgin throughout this chapter, unless otherwise specified. Such pidgins arose in limited contexts of trade and similar marginal contact between speakers of different languages. They are the main focus of my chapter, as opposed to other languages like Tok Pisin, Bislama, Sango, Yimas Pidgin, and others that have also been referred to as “pidgins.”1 Typical examples of prototypical pidgins include:
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.2 (122-195)
Donald Winford
– – – – – – –
–
Russenorsk (used in trade between Russians and Norwegians in the 19th century) Early South Pacific Pidgin English (used between English whaling crews and South Pacific Islanders in the 19th century) Eskimo Pidgin (used between the Inuit and various European traders in the 19th century) Hawai’i Pidgin English (used on Hawai’ian plantations in the 19th century) Chinese Pidgin English (used along the coast of China in the 18th and 19th centuries) Ndjuka-Trio Pidgin (used between the Ndjuka and the indigenous Trio community in Suriname from the 18th century to the present) Delaware Pidgin and Mobilian Pidgin (used among Native Americans and Europeans during the 17th and 18th centuries, and from at least the 16th to the mid-20th century, respectively) Chinook Pidgin (used among Native Americans and Europeans especially during the 18th and 19th centuries, and surviving until the mid-20th century)2
The following texts illustrate the main features of such pidgins. The following Russenorsk text was recorded and annotated by Brun (1878), and reproduced by Peterson (1980: 102). (1) Nor: Kjøp i sei-ka, tres-ka, tiksa og baldus-ka? buy you pollock cod haddock and halibut ‘Are you buying pollock, cod, haddock, and halibut?’ Rus: Da, da. Moja kop-om altsamm-a, davaj på skip kom. yes, yes. I buy-V all please prep ship come ‘Yes, yes, I’ll buy all of it. Please come on board.’ Nor: Spasiba! Har i mokka, har i groppa? thank-you have you flour have you grain ‘Thank you! Do you have flour? Do you have grain?’ Rus: Da, da! Davaj på skip kom, brat, på tsjaj drikke yes yes please prep ship come brother prep tea drink ‘Yes, yes. Please come on board, brother. Drink some tea’ Nor: Blagodaru pokorna! Kak tvoja betal-om for sei-ka? thank-you humbly what you pay-V for pollock ‘I humbly thank you! What are you paying for pollock?’ Rus: Pet pudof sei-ka for odin pud moki five pood pollock for one pound flour ‘Five poods of pollock for one pood of flour’ (1 pood = 36 lbs) The following is a pidgin text from the Marquesas, South Pacific. (Source: Herman Melville 1846, Chap. 33, quoted by Clark (1983: 14).)
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.3 (195-264)
Reduced syntax in (prototypical) pidgins
(2) Why you no like to stay? Plenty moee-moee (‘sleep’) – plenty ki-ki (‘eat’) – plenty whi–henee (‘young girls’) – Oh, very good place Typee! Suppose you no like this bay, why you come? The following is an early Melanesian Pidgin text (1859) from Tanna, New Hebrides (source: Clark 1983: 19). (3) You see . . . no good missionary stop Tanna. Suppose missionary stop here, by and by he speak, “Very good, all Tanna man make a work.” You see that no good: Tanna man he no too much like work. By–and–bye missionary speak, “No good woman make a work; very good, all man he only get one woman.” You see Tanna man no like that: he speak, “Very good plenty woman; very good woman make all work.” Tanna man no savé (‘know’) work . . . he too much lazy; he too much gentleman! As these texts demonstrate, prototypical pidgins entirely lack inflectional morphology as well as any complex (especially bound) derivational morphology of the sort found in their source languages.3 They also lack categories such as tense, aspect, person, and number. Notions like time reference are expressed through use of adverbials, as in the following examples from Russenorsk: (4) Stari gammel, snart på kjæka slipom. old old soon on church sleep ‘I’m old, I shall die soon.’ (Source: H. Stanges, in Jahr 1996: 119) Modal verbs like ville (‘want’) and skulle (‘should’) can convey futurity, but they are not tense auxiliaries. (5) Moja vil spraek på principal. ‘I want to speak with the captain.’
(Broch & Jahr 1984: 45)
In fact, not just categories like tense and aspect, but practically all functional categories are either missing or very rare in pidgins. They lack copulas as in (6) and articles as in (5) and (6), employ only one or two prepositions at most (Russenorsk has only one, på/po from Norwegian and Russian respectively), have no complementizers or subordinators (Russenork has one apparently subordinating conjunction kak, from Russian kak, which seems to be essentially a question word (see (7))). (6) Russman bra mann. Russian-man good man ‘The Russian is a good man.’ (7) Moja smotrom kak ju pisat. I see that(how?) you write ‘I see how you write.’
(Broch & Jahr 1984: 37)
Pidgins, in short, have a very basic syntax, lacking any kinds of embedding or subordination, with very limited exceptions (see the possible exception of kak clauses above).
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.4 (264-332)
Donald Winford
Clauses are combined only by paratactic means, either through juxtaposition or via coordinating conjunctions. Some pidgins have variable word order. For instance, the basic word order of Russenorsk is SVO, though it has a tendency to use SOV order in sentences with adverbials. Compare (8) and (9). (8) Moja kopom fiska. ‘I buy fish.’ (9) Moja tvoja på vater kasstom. I you on water throw ‘I’ll throw you into the water.’
(Jahr 1996: 115–116)
We will see similar examples of variation in word order below. Also worth mentioning here is the fact that pronominal subjects in pidgin utterances can take various forms, including nominative, accusative, and other oblique forms taken from their lexifiers. Thus, Russenorsk employs moja and tvoja, derived from possessive forms in Russian, as first- and second-person pronouns respectively. Most English-lexicon pidgins seem to prefer accusative forms of first-person pronouns, but nominative forms of others, such as he. One of the motivations for selection of pronominal forms appears to be their salience in the lexifier language. For instance, pidgins based on American Indian languages usually employ emphatic or other free forms of the lexifier pronouns in all cases, instead of the bound forms used by their lexifiers. Compare the following examples from the Tsimshian variety of Chinook Pidgin, and from Chinook, which has preverbal subject pronominals and postverbal full-noun subjects (cited in Thomason 1983: 847). (10) CJ:
Ixt man yaka kuli kupa lamatáy one man 3sg go to mountain ‘One man went to the mountain’
(11) Chi: Wixt áiuu i + qísqis again he-went-on m.sg + bluejay ‘Again the bluejay went on’ It is fairly certain that, whatever the forms pidgin pronouns take, these forms generally remain invariant and rarely, if at all, manifest case distinctions.
. Structural characteristics of pidgins The characteristic structural features of pidgins, as described by Bickerton (1981) and Drechsel (1996: 1225) among others, can be summarized as follows: Morpho-syntax: –
Absence of morphological apparatus such as affixation and inflection; hence no morphological expression of categories like number, person, and agreement
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.5 (332-404)
Reduced syntax in (prototypical) pidgins
– – – – –
Absence of other functional categories such as tense and aspect, with limited expression of deontic modality (e.g., desire) Minimal inventory of function morphemes such as articles, quantifiers, prepositions, conjunctions, and complementizers Restricted number of question words and pronouns. Most pidgins have only three pronouns: first, second, and third person, undifferentiated for gender or number Use of oblique (usually accusative) pronominal forms as subjects Use of one universal negative marker
Syntax: – –
–
Analytic structures, with word order as the primary means of determining grammatical functions such as “subject,” “object,” and so on A reduced number of sentence patterns, due to lack of rules for changing word order to create derived structures, for example, “movement” rules for topicalization, passivization, inversion in questions, and so on A lack of derivational depth, as reflected in minimal, if any, mechanisms for subordination or embedding (e.g., of relative or complement clauses)4
Such characteristics appear to be universal across pidgins, regardless of the typology of their source languages. Thus, pidgins based on highly inflectional or even polysynthetic sources such as Native American languages display the same kind of basic syntax found in pidgins derived from more analytic languages such as English. Examples like the following, from Eskimo Pidgin, are particularly instructive (source: van der Voort 1997: 376): (12) Ila kaktuña he hungry ‘He’s hungry’ Compare this with the Eskimo equivalent: (13) Kaak-tok hungry(V)-he/she/it ‘He is hungry’ Several points stand out in the pidgin example. First, as van der Voort demonstrates, the pronoun ila, like others, derives from the emphatic form of the Eskimo pronoun. Moreover, the form kaktuña actually derives from the (inflected) Eskimo form kaaktuña, literally ‘hungry(V)-I’ (‘I’m hungry’). Hence, in Eskimo, sentence (12) would be anomalous, equivalent to saying He I-am hungry. This is a good illustration that apparent inflectional morphology from a lexifier becomes “frozen” in the pidgin, playing no role in the syntax. Drechsel (1981) and others discuss similar kinds of frozen morphology in other pidgins based on Native American languages. In general, it is quite striking that, despite considerable variation in lexicon, phonology, and often word order, pidgins of all language backgrounds display a re-
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.6 (404-469)
Donald Winford
markable similarity in their syntax. Thus, referring to Eskimo, Delaware, Chinook, and Mobilian Pidgins, Drechsel (1981: 98) notes: In spite of differences and variations in word order, the four pidgins shared, with respect to their overall form, more similarities with each other than with the related target or base languages, most of which were of the synthetic or even polysynthetic type.
In this chapter, I argue that these structural characteristics of pidgins must be explained in terms of the processes by which they were formed, which are essentially the same as those involved in early second language acquisition (see Section 4). Attempts to explain pidgin grammar as a reduction of either a target language grammar, or of a learner’s L1 grammar, fail to account for many properties of pidgins. While pidgins do derive certain structural properties, not to mention their phonology and lexicon, from their source languages, they also manifest properties that have no source in either. Some scholars have attributed such properties to the influence of linguistic universals that, unfortunately, are not clearly defined. I argue that it is a combination of influences from the source languages, constrained by universal principles, that best explains pidgin structure. This combination of influences is also characteristic of early second language acquisition. Pidgin grammar, like early interlanguage (IL) grammar, is made up of minimal rules of predication, yielding structures quite similar to those of other nonsentential speech (see Progovac, Chap. 2, this volume). The similarities suggest that speakers who produce such small utterances share an understanding of what constitutes a minimal predicative structure.
. Processes of pidginization Hymes (1971: 70) points out that pidginization is “a complex process, comprising the concurrence of several component processes,” among them simplification, reduction of “inner form,” and admixture. Terms like “simplification” and “reduction” are somewhat problematic, since they can be used to refer to both the processes by which pidgins came into being, and the results of such processes. Pidgins are clearly reduced and simpler by comparison with the L1s of their creators. But the actual mechanisms involved in creating them are not so easy to pin down. Pidgin origins have been explained in term of three broad theories: the “foreigner talk” theory, the relexification of speakers’ L1s, and universal linguistic processes of “simplification.”
. The role of “foreigner talk” Foreigner talk theories maintain, in effect, that the main input to pidgin formation is the simplified and reduced forms of the dominant language that speakers of this language use to communicate with speakers of other languages. It is assumed that the
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.7 (469-533)
Reduced syntax in (prototypical) pidgins
“foreigner talk” produced in this way provided the major lexifier input to pidgin formation. In this respect, the learners of pidgins are said to share something in common with immigrant learners in natural SLA settings, who often rely on simplified input from target language (TL) speakers as a basis for constructing their version of the TL. This position has been supported by the argument that, with very few exceptions (e.g., Russenorsk), pidgins derive the bulk of their vocabulary from one language, which some refer to as the lexifier language. Pidgins have therefore typically been viewed as reduced and simplified versions of their respective lexifier languages. It seems reasonable to assume that foreigner talk did play a role in pidgin formation. This is supported by the fact that the processes of reduction found in foreigner talk are very much like those found in pidgins; hence the two share many structural characteristics. Hinnenkamp (1982) and Harding (1984) discuss the kinds of foreigner talk used by German and English native speakers respectively in talking to foreign immigrants. As Hinnenkamp (1982: 157) shows, such varieties display many of the kinds of reduction and simplification associated with pidgins. Mühlhäusler (1986: 104) provides the following example of German foreigner talk as created by the writer Karl May in his work Das Kafferngrab (originally published in 1879): (14) Mynheer rett Quimbo. Mynheer helf arm Quimbo. Quimbo will Myneer save Quimbo Minheer help poor Quimbo. Quimbo want nicht gut schmeck Strauss, oh, oh, Mynheer, aber Mynheer nicht not good taste ostrich oh oh Mynheer but Mynheer not treff Quimbo, den Quimbo bin sonst tot. Quimbo lass liegen hit Quimbo for Quimbo be otherwise dead Quimbo let lie Sau? Oh, oh, Mynheer, Quimbo ess viel schoen Sau. pig oh oh Mynheer Quimbo eat much beautiful pig Mühlhäusler (1986: 104) notes the following characteristics of this foreigner talk variety, among others: – – – – –
Consistent use of verb stem, instead of inflected verb forms Absence of articles and other determiners Uninflected attributive adjectives The use of viel (‘much’) instead of sehr (‘very’) Logical order is expressed by sequential order
For English foreigner talk, which he elicited from Australian adults, Mühlhäusler (1986: 101–102) notes the following characteristics: – – – – – –
Avoidance of embedded constructions, and use of parataxis instead Reinforcement of the message by gestures Avoidance of do-support in negative clauses Use of no as a universal negator Avoidance of tensed verb forms Addition of -um to verb forms
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.8 (533-595)
Donald Winford
– –
Lexical replacement: talk (‘to say’) or see (‘to look’) Selection of me rather than the I form of the pronoun
The similarities with the features of prototypical pidgins are striking. Still, there are some caveats regarding the role that foreigner talk might have played in pidgin formation. The first is that, as Mühlhäusler (1986: 101) points out, attempts to elicit foreigner talk from native speakers of English have shown that the input they provide is highly variable, since individual speakers differ markedly in the application of the various simplificatory devices. This implies that the input to pidgin formation was itself highly variable, and provided more than one model that learners could imitate. This conflicts with any assumption that a pidgin is based on a single reduced form of the TL. The second caveat is that it denies or at best ignores any role that learners of the putative TL might have played in this process. Clearly, pidgin learners could not simplify the TL in the same way as its native speakers, since they had no command of it. It seems obvious that both native speakers of the TL and the speakers of the other languages involved in the contact must have played a role in creating the new contact language.
. The role of L1 influence The view of pidgins as merely reduced versions of the TL has rightly been challenged, for example, by Bickerton and Givón (1976: 2). They criticize earlier approaches for making “erroneous assumptions about what the process was acting on,” and for confusing the product with the process. Yet, Bickerton and Givón’s own view of the process of pidginization is itself subject to question. They argue that pidgins arise when speakers in contact gradually relexify their own grammar, “slotting newly acquired vocabulary into surface structures characteristic of their own language” (3). This simply reverses the older view, by treating pidgins as reduced versions of the learners’ L1s. Thus Bickerton and Odo (1976: 304) argue that the pidgin learner “will begin by making what appear to him to be the minimal necessary (or possible) adaptations to his own phonology, grammar and lexicon, and will extend these as soon as he is able to do so.” For Bickerton and Odo, it is “this replacement process, rather than any kind of ‘simplification’ [that] appears to be the dominant factor in early pidginization.” It is true that learners’ L1s do influence early learner varieties to a significant extent, in both pidginization and early L2 acquisition. Examples of this include the variable word order found in the Hawai’ian Pidgin English varieties spoken by groups speaking different L1s, as described by Bickerton and Odo. They note that there is a strong tendency to use SOV order among more “conservative” Japanese (JPE) and Korean pidgin (KPE) speakers, though they also use some SVO order, as in the following examples from Bickerton and Odo (1976) (page numbers are given after each example): (15) JPE: Ada vakyum klina no kaen yuz? other vacuum cleaner no can use ‘Can’t you use the other vacuum cleaner?’ (134)
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.9 (595-668)
Reduced syntax in (prototypical) pidgins
(16) JPE: Da pua piipl awl potato iit. the poor people all potato eat ‘The poor all eat potatoes.’ (135) (17) KPE: Aeswai koria kim neim wan moa taim mi meri. that’s-why Korea Kim name one more time me marry ‘That’s why I married a second time, a Korean called Kim.’ (163) (18) KPE: Mi onli chachi-chachi go palei. me only church-church go pray ‘I only went to church to pray.’ (163) On the other hand, “conservative” Filipino pidgin (FPE) speakers show a tendency to use VSO order, characteristic of their L1s, Ilocano and Tagalog (examples again are from Bickerton & Odo, with page numbers). (19) FPE: Wok haad dis piipl ‘These people work hard’ (173) (20) FPE: Mo plaeni da ilokano en da tagalog ‘Ilocanos were more numerous than Tagalogs’ (173) In addition to word order, L1 influence in these speakers extends to lexical borrowing, as well as various strategies for expressing time (Bickerton & Odo 1976: 137–141). By contrast, Japanese and other speakers who have fully acquired HPE follow SVO order and rely less on borrowing from their native languages. It is therefore premature to claim that the pidgin varieties used by early pidgin learners, who are more subject to L1 influence, are representative of HPE in general. The historical evidence in fact suggests that Japanese and other immigrants who came to Hawai’i after 1900 or so were in fact targeting an already established pidgin, which showed little L1 influence, and had developed as an independent system of communication in its own right. Note also that, even the more conservative pidgin speakers employ a variety of strategies, including pronoun copying (Bickerton & Odo: 249–251), simple forms of conjunction (Bickerton & Odo: 216–222), and paratactic relativization strategies (Bickerton & Odo: 141–145), which have no model in either English or their native languages. Clearly, more than L1 influence was involved in the creation of the varieties of HPE.
. The role of language universals The view that pidgin formation is regulated by universal principles in fact goes back as far as Coelho (1880), and was further developed by Naro (1971), Kay and Sankoff (1974), and Bickerton (1981). Different scholars have proposed a variety of such principles to explain how pidgins were created. Macedo (1986) addressed the issue of universals in terms of the theory of markedness, and the core versus periphery theory advanced within the Principles and Parameters framework. According to this view, pidgins have a core grammar consisting of
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.10 (668-714)
Donald Winford
structures that are “maximally general and unmarked,” reflecting the unmarked parameter settings of Universal Grammar (Macedo: 73). He further argues that “pidgin development is governed by universal principles which must meet at least one fundamental condition: maximal learnability” (66). Elaboration of pidgin grammar, from this perspective, involves the development of more “peripheral” rules and structures that are marked. However, Macedo illustrates his claims mostly with data from Creoles rather than (prototypical) pidgins, so it is difficult to identify what he calls the core grammar found in pidgins. Moreover, as he himself admits, Principles and Parameters theory failed to distinguish between core grammar and periphery in any principled way. Bresnan (2000) also attempts to explain universals of pidginization by appealing to the notion of markedness. She argues that pidgin formation involves simplification, which she defines as “the elimination of marked features” (150). For her, simplification by native speakers and imperfect learning by others contribute to pidgin formation “through a process of mutual accommodation, eliminating marked features.” This raises, first, the question of how marked structures can be distinguished from universally unmarked structures in the adult grammar. To answer this, Bresnan employs the framework of optimality theory to demonstrate how the reduced nature of pidgin pronominal systems (particularly, their use of forms characterized by iconicity and lack of allotaxy) can be explained. More problematic is the question of how it is possible for speakers to target a marked structure for elimination. Bresnan argues that her theory of markedness and simplification (at least of pronominal systems) explains this in terms of adults’ ability “to access universal markedness properties of all languages, starting only from knowledge of their own vernaculars” (165). How far this “explains” anything is still not clear, nor is it clear how the approach might be extended to the kinds of “reduction” found in syntax. Such approaches have a somewhat post hoc character to them. The actual processes involved in these highly reduced varieties, that is, the nature of, and the constraints on, the kinds of “reduction” that take place, have not figured prominently in the search for universals. In the following section, I argue that reconciliation of the various views on pidgin formation is possible if we view pidgin formation as essentially a process of second language acquisition in which three factors played a role, TL input, L1 influence, and universals of learning.
. Pidgins and second language acquisition The similarities in structure between pidgins and early SLA have long been noted, for example, by scholars such as Schumann (1978), Andersen (1981), Huebner (1983), and others. Schumann’s study of the early L2 English of Alberto, a Spanish native speaker, was among the earliest comprehensive attempts to describe similarities between pidgins and early IL. The study revealed that Alberto’s early IL shared many characteristics with pidgins, including the following:
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.11 (714-780)
Reduced syntax in (prototypical) pidgins
– – – – – –
Lack of possessive inflection Lack of verbal inflection Absence of pronouns A single no negator Lack of auxiliaries Lack of movement in questions
Schumann also mentions other characteristics of pidgins that could be found in early IL, though he doesn’t examine these in Alberto’s IL. They include: – – –
Use of word order to “replace” inflectional morphology Lack of articles and a copula Use of topic-comment word order
On the basis of his comparison, Schumann (1978) concludes that Alberto’s English appears to be pidginized. He discusses various social and psychological factors that seem to constrain SLA, but he has little to say about the processes that yield these reduced L2 varieties. Like other researchers (e.g., Hall 1966), he uses the term pidginization to refer to the differences between the lexifier/TL and the pidginized variety, that is, the outcome.
. The basic variety More recent studies of the early L2 varieties used by immigrant learners in Europe confirm the similarities between pidgins and early ILs. Of particular interest are the studies done by Klein, Perdue, and their associates, who studied the acquisition of second languages by 40 adult immigrants in Europe over the period 1981 to 1988, recording their L2 production for 30 months in each case (Perdue 1993a, 1993b). The subjects were of various language backgrounds, including Arabic, Finnish, Italian, Punjabi, Spanish, and Turkish. The TLs were various European languages, including Dutch, English, French, German, and Swedish. The study was designed to allow for comparison of acquisition of the same TL by different immigrant groups, as well as different TLs by the same group. Interestingly, the investigation revealed that all learners developed a Basic Variety of each TL that was surprisingly uniform in its general structure. About a third of the learners got stuck at this stage of acquisition, though some continued to add more vocabulary items to their system. The Basic Variety (BV) was characterized by a small but expanding lexicon, drawn mostly from the TL, but with some L1 items.5 It consisted mostly of nouns and verbs and a smaller inventory of adjectives and adverbs. It also contained a small number of function words such as quantifiers, a few determiners, a single negative marker, and a few prepositions. The pronominal system distinguished only between the speaker, the hearer, and a third person. This BV completely lacked inflectional morphology and, by extension, grammatical categories such as tense and aspect, case, gender and num-
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.12 (780-857)
Donald Winford
ber, agreement, and so on. In syntax, it lacked complementizers and subordination in general. The following section presents examples of such features of the BV.
.. Pidginlike characteristics of the BV The following example sentences illustrate several of the pidginlike characteristics of the BV: (21) Lack of TMA marking (L2 English of Santo (L1 = Italian), Perdue & Klein 1992: 267): After, come back the brigade fire ‘Then, the fire brigade arrived.’ (22) Lack of copula (L2 English of Ricardo (L1 = Spanish), Butterworth & Hatch 1978: 236): He champion. (23) Use of a single negator (usually no) (L2 English of Andrea (L1 = Italian), Perdue & Klein 1992: 264): Researcher: Do you watch television much? Andrea: I no see every day. (24) Frequent lack of articles (L2 German of “V,” (L1 = Italian), Klein & Perdue 1997: 330): Mädchen nehme brot girl take bread (25) Paractactic linkage (L2 Dutch of Mohamed (L1 = Arabic), Extra & van Hout 1996: 102): Ik ga met die jongen familie (= neef ) naar die naar die I go with that boy family (= nephew) to those to those twee broers morokkans Ik ga naar hem dan alles en dan two brothers Moroccan I go to him then everything and then wij alles gaan naar tilburg we everything go to Tilburg Examples like these confirm the observation that there are marked similarities in structure between early IL and pidgins. The question arising here is whether these similarities can be accounted for in terms of similar processes and principles of grammar construction.
.. Sources of the BV grammar Klein and Perdue (1997: 308) reject the view that such basic learner varieties are “imperfect imitations of . . . the target language,” arguing instead that “they are systems in their own right, error-free by definition.” Klein and Perdue (330) also argue that the L1 plays only a partial role in shaping the grammar of the BV. They acknowledge that there is a certain degree of L1 influence on the syntax of early IL, particularly with re-
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.13 (857-899)
Reduced syntax in (prototypical) pidgins
spect to word order transfer. But they argue that this kind of transfer only plays a role in cases where the basic constraints that regulate utterance organization in the learner variety are in conflict (see discussion below). For them, the early IL is governed by its own rules, which are not parasitic on those of the input languages. Like pidgins, early IL varieties are not merely reduced versions of the TL, nor relexified versions of the learners’ source languages, but possess word order and other syntactic conventions of their own. Klein and Perdue seem to adopt a “universals” explanation for the structure of the BV, suggesting that “the organizational constraints of the BV belong to the core attributes of the human language capacity” (301). They also note that pidgins have properties similar to those of the BV, though they have not been systematically investigated with respect to organizational principles of the type associated with the BV (Klein & Perdue: 340). Studies such as these provide ample justification for treating pidgins as L2 varieties that arose in much the same way as other cases of early SLA in natural settings. It therefore seems reasonable to assume that the processes involved in the creation of early IL, pidgins, and foreigner talk are quite similar, and subject to the same universal principles and constraints. The question then is: how do such conventions arise? To better understand the processes of pidgin and early IL formation, we must first distinguish the stages of their development, and then examine the overall structure of their grammars at each stage, to uncover the principles that govern their organization.
. Stages of development in pidginization and early IL Most of the generalizations made about pidgin and early IL syntax are based on the relatively stable varieties that emerge after a certain period of acquisition. With the exception of studies like those of Klein, Perdue, and their associates, very little attention has been paid to the earliest stages of learning and development of L2 systems. With regard to pidgins, the picture is worse. Apart from studies such as those of Clark (1983) and Bickerton and Odo (1976), studies of the early stages of pidginization are hard to come by. Even the pidgin varieties studied by Bickerton and Odo do not represent the earliest stages of pidginization in Hawai’i, though they argue that those stages can be reconstructed through observation of the contemporary situation (303). Despite these shortcomings, a comparison of pidginization with the earliest stages of SLA can still be revealing. It is well known that, during the acquisition process, all learners go through a series of learner varieties (Klein & Perdue 1997: 307). Pienemann (1999: 5) suggests that there is a certain sequence by which procedural skills develop in the learner, and proposes that there is a hierarchy of processability that constrains that sequence of acquisition. According to Pienemann (1999: 83), learners go through five major stages in the development of their L2 processor. Stage 1 is characterized by complete absence of any language-specific procedure. While new words are added to the emerging IL, learners have no access to syntactic information about the L2 lexical items; hence their (L2) language-production process is blocked. As a result, “all that can happen at
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.14 (899-967)
Donald Winford
this stage is the mapping of conceptual structures onto individual words and phrases” (Pienemann: 83). In Stage 2, lexical items are assigned to grammatical categories, and some lexical morphological markers can be produced. However, grammatical information from the L2 is still blocked at this stage. Still, L2 learners “have a set of well-defined semantic roles which they will attempt to map onto L2 forms. Since this mapping is not possible using L2 procedures, the only solution is to use simplified procedures” (Pienemann: 83). One such procedure is the use of a fixed serial word order, such as NVN. The third stage of acquisition involves the development of phrasal procedures, though phenomena such as agreement and nonserial word order are still blocked. However, the learner uses general cognitive principles in concert with strict word order to convey more complex messages. One manifestation of this is the use of “pragmatic” word order options that “allow the learner to imitate a range of L2 syntactic phenomena without full access to L2 procedures” (Pienemann 1999: 85). Later stages of acquisition involve the development of S-procedure rules that allow phrases to be assembled into sentences (Stage 4) and the emergence of the subordinate clause procedure, allowing for complementation and other subordination phenomena (Stage 5).
. Stage 1 pidgins and ILs The first stage of Pienemann’s (1999) model seems quite relevant to the earliest stages of pidginization and IL formation. At this point, the lexicon consists of a relatively small repertoire of nounlike words and verblike words, some adjectives and adverbs, a three-way pronominal system, and a few quantifiers and function words (e.g., a negator and a few modals). Stage 1 IL is characterized by very minimal syntactic structures, many of which lack either arguments or predicates. Examples include the following: (26) L2 English of Santo (L1 = Italian) (Perdue & Klein 1992: 264): I lift ‘I take the lift’ (27) Then back Morocco ‘Then he went back to Morocco’
(Extra & van Hout 1996: 99)
Very similar phenomena are found in cases of early pidginization, as in the following examples from Hawai’i Pidgin English (Bickerton & Odo 1976: 163, 145, 230): (28) KPE: Aena tu macha churen, samawl churen, haus mani pei. and too much children small children house money pay beri haad taim, no moa moni . . . pua very hard time, no more money poor ‘And I [had] too many children, small children, I [had] to pay the rent. [It was a] very hard time, I had no money, [we were] poor’
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.15 (967-1033)
Reduced syntax in (prototypical) pidgins
(29) KPE: Baimbai mi oni stoa, rait? by-and-by me only store right ‘Soon I [had] only [the] store, right?’ (30) FPE: Sam taim as handrid piptin dala ei dei sometimes us hundred fifteen dollar a day ‘Sometimes we [made] a hundred and fifteen dollars a day’ Bickerton (1981) suggests that such structures represent an early pregrammatical or “macaronic” stage, which would apparently correspond to Stage 1 in Pienemann’s (1999) scheme. Bickerton and Odo (1976: 129) argue that these kinds of pidginization “never gave rise to a stable and homogeneous artefact which subsequent speakers acquire.” These kinds of output are consistent with what some call the “jargon” stage of pidginization. Mühlhäusler (1986: 136, 143) provides other examples: (31) No mani, no kom Depending on the context, this utterance can be interpreted in various ways, to mean, for example, ‘I had no money, so I didn’t come,’ ‘If I have no money, I won’t come,’ and so on. (32) Me, tamaree . . . plenty kanaka Martair. ‘When I was a boy, there were many people at Ma’atea.’ In short, in Stage 1 pidgins and ILs, rules of predication are ill developed, and there is very heavy reliance on context, including previous utterances as well as factors outside the utterance, for interpretation. In this stage, as Roberts (1998: 17) points out, “lexical items such as nouns and verbs do not project onto phrasal constituents, and therefore do not govern and subcategorize complements.” This corresponds closely to Pienemann’s (1999) account of the earliest stage of SLA. As van der Voort (1997: 388) notes, such early stage pidginized varieties or “jargons” do not have the status of separate languages, though they may evolve into (Stage 2) pidgins. Unlike jargons, “pidgins are not grammatically ‘impoverished’ or ‘reduced’ versions of the input languages, but more independent entities” (van der Voort: 388). Bickerton & Odo (1976: 129) seem to suggest that all of the structures produced by their older Japanese and Filipino Pidgin speakers represent a pregrammatical stage of pidginization. In fact, most of the data they present, as we saw earlier, show evidence of regular rules of predication, consistent with Stage 2 pidginization. There is also a great deal of consistency across speakers, suggesting that their pidginization strategies were not in fact completely individual or idiosyncratic. Much of the consistency, in fact, appears to be due to similar appeals to knowledge of Japanese or Filipino. Given the right circumstances, such varieties could easily have become targets of learning in its own right. As Stage 1 pidgins and ILs develop into Stage 2 varieties, they can also include very minimal predicative structures, but interpretation of these is highly dependent on context. Extra and van Hout (1996: 99) shed some light on the nature of this early-stage
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.16 (1033-1099)
Donald Winford
IL. They distinguish between “referential functions” and “linguistic devices,” and note that the former can be expressed in earliest IL without any overt linguistic marking. They provide examples like the following (Extra & van Hout: 99): (33) a.
He down ‘He went down’ b. They go away ‘They went away’
As Extra and van Hout point out, there is no overt expression of motion in (33a), or of time in (33a) and (33b). Learners may also express referential functions by using L2 linguistic devices in ways not found in the L2. Thus we find (Extra & van Hout 1996: 99): (34) a.
He at home ‘He went home’ b. He at home ‘She went home’
Such utterances can cause misunderstandings about direction (34a) and person (34b). These can only be resolved by appealing to the context of the utterance. Extra and van Hout point out that, since the learner has no command of the L2, but is constructing his own expressive devices, such zero-markings can not be termed “deletions” (99).
. Stage 2 pidgins and ILs: The syntax By contrast with the earliest stage of pidginization, Stage 2 pidgins (synonymous with what I earlier referred to as “prototypical pidgins”) and the relatively stable early IL referred to as the Basic Variety, are characterized by a clear though rudimentary grammatical organization, in other words, regular though simple rules of predication. In this respect, they are reminiscent of Pienemann’s (1999) second stage of language acquisition/production. As noted, by this stage grammatical categories have emerged, along with basic syntactic procedures. Moreover, recall Pienemann’s observation that learners at this stage resort to certain pragmatic strategies to compensate for their lack of access to full L2 procedures. We can now examine these aspects of Stage 2 pidgin and IL varieties. Bickerton (1981: 38) suggests that HPE has the following phrase structure rules: (35) S
→
NP V (NP) NP (NP) V V NP
These rules appear to account for pidgin varieties that manifest SV order, as well as varieties used by Japanese/Korean and Filipino speakers, which may manifest OV and VS order respectively. In addition, we can assume that V in these cases refers to predicates of all types, including verbal, nominal, adjectival, and prepositional phrase predicates.
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.17 (1099-1180)
Reduced syntax in (prototypical) pidgins
Putting aside the word order differences, which I explained as due primarily to influence from speakers’ L1s, these syntactic patterns display remarkable similarity to those described by Klein and Perdue (1997: 314) for what they call the Basic Variety. They found that their informants used three basic utterance patterns (with some variation), as follows: (36) PH1a PH1b PH1c PH2 PH3
NP1 – V NP1 – V – NP2 NP1 – V – NP2 – NP3 NP1 – copula – NP2/AdjP/PP V/Cop – NP2
Pattern (PH1c) occurs with verba dicendi and other ditransitive verbs such as give. Hence there are basically just two (PH1) patterns. With regard to variation in word order, Perdue and Klein (1992: 262) argue that this is conditioned, in part, by certain semantic and pragmatic principles (see below).
. Pidgin and IL structures as small clauses The basic syntactic structures outlined here manifest very similar characteristics to those of small clauses as discussed in this volume. These include various kinds of basic predication, yielding VP, AP, PP, and DP structures as described by Progovac (Chap. 2, this volume). Moreover, like small clauses, their general properties include: – – – – –
Absence of a tense node Lack of inflection, including tense and agreement A tendency for subjects to be marked by accusative or other oblique cases, rather than nominative case6 Lack of copulas Possible omission of articles
Finally, like other small clauses, Stage 2 pidgin and IL utterances still express predicateargument relationships, despite their lack of Tense and, often, articles. A brief comment on some of these properties in pidgins and early IL is appropriate. In the first place, some features such as lack of copulas and omission of articles may well be due to L1 influence. Lack of articles in the varieties of Hawai’ian pidgin English spoken by Chinese and Japanese speakers, for instance, may have been due to the absence of articles in the respective L1s. The same may apply to the lack of copulas. In addition, while there is a tendency for articles to be absent in pidgins and early ILs, this is not true of determiners in general, if we take these to include demonstratives, numerals, and possessives. For instance, Bickerton and Odo (1976: 146) show that, while definite and indefinite articles are absent 91% and 82% of the time respectively in their most “conservative” speakers of Hawai’i Pidgin English, there is a very high frequency of demonstratives (dis, dat, etc.) and possessive adjectives (yu ‘your,’ mi ‘my,’ etc.). It’s not clear what this implies for the status of DP in this and other Stage 2 pidgins/ILs.
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.18 (1180-1236)
Donald Winford
. The role of pragmatics in the interpretation of pidgin/IL utterances As Progovac (Chap. 2, this volume) argues, small clauses, despite missing many elements of full clauses, can still express truth-evaluable assertions, which are recoverable by appeal to pragmatic factors. One example of this, shared by small utterances in general, is the use of time adverbials or contextual factors to provide time specifications in the absence of Tense. The role that pragmatics plays in the interpretation of early ILs and pidgins is also widely recognized. In fact, Perdue and Klein (1992) argue that syntactic patterns by themselves are inadequate to understand the production of early learners, since word order by itself does not determine the semantic roles of NPs or other constituents. They claim that the phrasal patterns are only “one type of principle at work in the way these learners structure their utterances; they interact with others of a more semantic or pragmatic type” (262).
.. Pragmatics and word order For instance, although word order in the basic variety often reflects that of the TL, Perdue and Klein (1992) note that the position of an NP does not always determine its role as subject or object. Instead, the issue is which argument takes which position. This, they argue, is determined by either a semantic principle (S), or a pragmatic principle (P), described as follows: (S) Controller first (P) Focus last Principle S affects the way learners introduce referents and maintain reference and places an NP whose referent controls the event, first. This is exemplified in subjectpredicate structures like the following (Perdue & Klein: 263): (37) a. The police drive the car. b. Chaplin have the bread. In these cases both the phrasal and semantic constraints interact to produce structures of the type in (37). Principle P dictates that constituents that are in focus appear in utterance-final position, while topics appear in initial position. This principle is exemplified in structures like the following (Perdue & Klein 1992: 268, 267): (38) a.
Have the young girl ‘There was this young girl’ b. After come back the brigade fire ‘Then the fire brigade arrived’
The topic/focus principle is well illustrated in answers to information questions like the following: (39) Who stole the bread?
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.19 (1236-1303)
Reduced syntax in (prototypical) pidgins
Here, the topic is stole the bread and the agent of the action is in focus. This means that the response to a question like (39) will be something like: (40) Stole the bread the girl topic focus In his study of the acquisition of English by Ge, a native speaker of Hmong, Huebner (1983) also argues that it is the interaction between syntactic patterns and pragmatic organization that best explains the structure of Ge’s early IL. Pragmatic principles seem to lie behind several aspects of Ge’s IL grammar. With regard to word order, Huebner (60) notes that Ge expresses case relations (strictly speaking, theta roles) primarily through word order, employing the basic order: (41) Agent/Experiencer – Verb – Object This is the same order used in Hmong. However, as Huebner demonstrates, Ge’s word order is determined primarily by pragmatic principles, following a pattern of presupposed information followed by asserted information. This is in fact also characteristic of Hmong as a topic-prominent language; hence the role of L1 influence cannot be ignored here. In some respects, Ge’s IL seems to follow the Pragmatic Constraint (P) “Focus last,” suggested by Perdue and Klein (1992). Huebner (1983) describes this in terms of the topic/comment distinction. For instance, in a question like: (42) Who came to Honolulu? The VP is presupposed, or “given” information. In such cases, Ge’s response typically has the following structure: (43) Keim to Honolulu, isa fai familii presupposed/topic asserted/focus ‘Five families came to Honolulu’ (Note that isa here functions as a focus marker.) The topic-prominent nature of Ge’s early IL also explains a variety of its other structural characteristics. According to Huebner (70–84), these include (among others): (44) Word order that places topics in initial position, as in: Holii, bat ai sii, ai no tok. ‘As for Hawley, I see (him). I don’t talk (to him).’ (45) Lack of “dummy” subjects like it, there: Aen haeva beibii ‘And there were babies’ (46) Verb-final sentences: Hos, ai reis ‘As for horses, I raised (them)’
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.20 (1303-1361)
Donald Winford
Huebner (1983) points out that the topics in Ge’s utterances are invariably accompanied by a pause and rising intonation, making the distinction between topic and comment quite clear. According to Huebner, these topic-comment structures cannot be derived by a set of rules applied to “base structures,” because “there is no evidence that Ge has any rules to apply.” Rather, the topic-comment structures themselves are basic, and constitute the vast majority of Ge’s utterances (69% as opposed to 31% subject-predicate structures (Huebner: 86)).
.. Pragmatics and missing elements As Progovac (Chap. 2, this volume) notes, many nonsentential utterances that lack arguments or other elements can still express predicate-argument relationships because the missing elements are recoverable from the discourse context. Klein and Perdue (1997: 316) explain the recovery process by appealing to the distinction between “given” and “new” information. Phenomena such as zero anaphora in early IL can be explained in terms of this principle, in that referents that are maintained from a preceding utterance may be referred to by “Ø” under certain circumstances. This is illustrated by the following: (47) L2 English of Santo (L1 = Italian) (Perdue & Klein 1992: 266): a. Have the one girl b. And take in the window in the shop the bread c. And go away. Huebner (1983) also argues that pragmatic principles account for the frequency of zero anaphora in Ge’s IL. One regular pattern involves coreferential constituent deletion, controlled by the topic, rather than the subject. (48) Q. You built a house? Ge: Yae, biwd da haws ‘Yes (I) built the house’ Similarly, in response to information questions like: (49) What happened to John? John is presupposed information. In these cases, such information need not be repeated if it can be retrieved from the context. Hence, the following answer is quite possible: (50) A snake bit – The principle Ge follows here seems to correspond to Klein and Perdue’s (1997: 316) pragmatic constraint P, having to do with the “given-new” distinction. The studies by Klein and Perdue (1997) and Huebner (1983) demonstrate that contextual factors – the pragmatics of conversational exchange – clearly play a major role in determining the syntactic strategies employed by early learners. The nature of
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.21 (1361-1427)
Reduced syntax in (prototypical) pidgins
the communicative task the learner has to perform (e.g., narrative, response to questioning, describing, etc.) also appears to play a role in this (Klein & Perdue: 309). Presumably pragmatic factors play a role in pidgins as well, and in certain kinds of nonsentential speech produced by native speakers of a language.
.. Pidgin syntax and pragmatics The absence of adequate textual data is a serious disadvantage in attempts to generalize about pidgin syntax. In particular, we lack data from various genres, such as casual conversational exchange, narratives, description, and so on, which have proven quite illuminating in studies of SLA. However, some sources have provided at least some of these kinds of data. Roberts (1998: 16–17) discusses cases of zero anaphora in HPE, as illustrated in the following excerpt from a conversation between a deputy sheriff (A) and Mr. Sumida, a Japanese immigrant (B) in 1904 court testimony. (51) A: How much you pay kela [‘that’] sake, how much money makana [‘give’]? B: Me Honolulu buy. Me letter Honolulu hapai [‘bring’]. Me hapai, steamer come, me hapai camp, me charge two dollar, me pay steamer. A: Who boss Honolulu? B: Suga. A: You got license? B: Wahine [‘wife’] hapai license. Me hapai one tub of sake kela Miazaki house Camp Four. No makana dala this time. Pay day makana dala, me hapai Honolulu boss. As Roberts notes, in cases like these, “Null arguments must be interpreted pragmatically and cannot be predicted by regular syntactic processes” (17). Bickerton and Odo (1976: 240, 136) provide further examples. Example (52) is from a conversation between an interviewer and male Filipino speaker: (52) Int.: MF: Int.: MF: Int.: MF:
Yeah, tell me love story. No kan, ae? Why no kan? No kan! (emphatic) How come? Sheim, dakain (‘I’m kinda embarrassed’ Lit. ‘shame, kind of ’)
Example (53) is from a conversation between an interviewer and a male Japanese speaker: (53) MJ:
Ai laik dat potato, ye swiit poteto, bat no, difrent kain swiit poteto. Riil ono (‘tasty’), insaid paat prpl, no, its miin Okinawa wrd, no aes tumaikru, e, da poteto neim, yo. Awl kain get. Hawai’i, ai no sii – hawai’i, onli ai sii, a, red wan, ai sii.
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.22 (1427-1488)
Donald Winford
‘I like that potato, yeah sweet potato, but no, [there are] different kinds of sweet potato. [It’s] real tasty, [the] inside is purple, no, there’s an Okinawa word, tumaikru. That’s the name of the potato. There are all kinds. I don’t see [it] [in] Hawai’i. [In] Hawai’i I only see, ah, [the] red one, I see.’ One might note, incidentally, that pro-drop, which occurs very frequently in the pidgin output of Japanese speakers, may be another result of influence from the speaker’s native language. Clearly, pidgin structures, like early IL utterances, need to be explained as much in pragmatic as in syntactic terms. As Klein and Perdue (1997: 311) point out: In-depth contextual interpretation is therefore necessary in order reliably to establish regular form-function correspondences. Once an interpretation has been established, the surest way of missing learner-language regularities is to imagine a “corresponding” utterance in another language – the target language or the source language – then attribute its organization back to the learner’s utterance.
Unfortunately, most of the research on pidgins as well as early IL has suffered limitations because of the tendency to focus on isolated utterances rather than the conversational context. Lack of the latter can lead to serious problems for the analyst, not to mention the interlocutors.
. Conclusion By emphasizing the developmental stages through which pidgins and early ILs pass, I hope to have shown that these varieties cannot be treated as reductions of any of their source languages, but as independent creations that draw in part on L1 and TL input, and are guided by universal constraints on predication. This idea of the autonomy of pidgins and early ILs as systems in their own right has been expressed repeatedly, as discussed earlier in this chapter. In addition, the structural similarities we have seen among different varieties exhibiting minimal types of predication, from foreigner talk, to pidgins and early IL, to the simplified production of mature speakers, seem to support the idea that the use of nonsentential speech represents a part of every human’s language capacity, and is thus cross-linguistically available (Progovac, Chap. 2, this volume). As we have seen, earlier scholars such as Macedo (1986) and Bresnan (2000) also express the view that the uniformity in structure across pidgins is due to speakers’ ability to identify the universally least marked properties of languages, based on their knowledge of the essential or core properties of their own L1s. The approach to nonsentential speech in this volume offers yet another perspective on this issue, by appealing to some of the assumptions of the Minimalist Program (see Progovac). Moreover, this approach recognizes that syntactic principles explain only part of the picture. It must also be assumed that the producers of small utterances of all types also
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.23 (1488-1545)
Reduced syntax in (prototypical) pidgins
rely heavily on pragmatic knowledge, which may itself be universal in nature. Klein and Perdue (1997: 339) support this view with regard to the early stages of second language acquisition, and suggest that, in addition to universal syntactic principles, the semantic and pragmatic constraints they propose to explain the production of early learners are also part of Universal Grammar. Much still has to be done to demonstrate that this is in fact the case, and the production of nonsentential speech by pidgin and other speakers may have an important role to play in further exploration of this possibility.
Notes . As is well known, the languages that have been referred to as “pidgins” constitute a motley assortment, which are characterized by significant diversity of structure and varying degrees of reduction in their syntax. I reserve the term prototypical pidgin only for that subset of such languages that is characterized by minimal syntax of the sort described in this chapter. Languages such as Yimas Pidgin and the so-called Melanesian Pidgins do not fit this designation. . It is probably more appropriate to reserve the label prototypical for earlier Chinook Jargon, since it later developed into a more complex lingua franca when it became a language of everyday communication in the Grande Ronde reservation (Zenk 1988). . Pidgins, of course, do have strategies of word formation, but these typically do not involve bound morphology. Some scholars (e.g., Bakker 2003) claim that certain pidgins preserve at least relics of inflectional morphology, either inherited from their lexifier languages or borrowed from other languages. Such cases are rare, involving mostly more complex pidgins or simplified languages, and very few prototypical pidgins. In the latter cases, at any rate, the so-called inflectional morphology appears to be “frozen” and lacking either productivity or any significant role in the syntax. . In addition to the example of a possible embedding strategy from Russenorsk (above), pidgins such as Eskimo Pidgin and Mobilian Pidgin also seem to have developed minimal strategies of subordination (Drechsel 1981: 98). . Not all researchers agree that there is a well-defined Basic Variety that is uniform across all early learners in the early stages of second language acquisition. For further discussion, see Jordens (1996). . See the Epilogue to this volume, where the coeditors discuss variability in case marking of subjects in “telegraphic speech,” and argue that it’s more easily explained by a nonsentential analysis than by a full sentential analysis that implies subsequent deletion.
References Andersen, R. (1981). Two perspectives on pidginization as second language acquisition. In R. Andersen (Ed.), New Dimensions in Second Language Acquisition Research (pp. 165–195). Rowley, MA: Newbury House. Bakker, P. (2003). Pidgin inflectional morphology and its implications for creole morphology. In G. Booij & J. van Marle (Eds.), Yearbook of Morphology 2002 (pp. 3–33). Dordrecht: Kluwer.
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.24 (1545-1653)
Donald Winford
Bickerton, D. (1981). Roots of Language. Ann Arbor, MI: Karoma. Bickerton, D. & Givón, T. (1976). Pidginization and Syntactic Change: From SXV to SVX. Unpublished manuscript. Bickerton, D. & Odo, C. (1976). Change and Variation in Hawai’ian English, Vol. 1: General Phonology and Pidgin Syntax. Honolulu, HI: Social Sciences and Linguistics Institute, University of Hawai’i. Bloomfield, L. (1933). Language. New York, NY: Holt. Bresnan, J. (2000). Pidgin genesis and optimality theory. In J. Siegel (Ed.), Processes of Language Contact: Studies From Australia and the South Pacific (pp. 145–173). Montréal: Fides. Broch, I. & Jahr, E. H. (1984). Russenorsk: A new look at the Russo-Norwegian pidgin in northern Norway. In P. S. Ureland & I. Clarkson (Eds.), Scandinavian Language Contacts (pp. 21–65). Cambridge: CUP. Brun, A. W. S. (1878). Skildringer fra den norske Kyst. Norg og Syd, 11. Trondhjem. Butterworth, G. & Hatch, E. (1978). A Spanish-speaking adolescent’s acquisition of English syntax. In E. Hatch (Ed.), Second Language Acquisition: A Book of Readings (pp. 231–245). Rowley, MA: Newbury House. Clark, R. (1983). Social contexts of early South Pacific pidgins. In E. Woolford & W. Washabaugh (Eds.), The Social Contexts of Creolization (pp. 10–27). Ann Arbor, MI: Karoma. Coelho, F. A. (1880). Os dialectos românicos ou neolatinos na África, Asia, e América. Bolletim da Sociedade de Geografica de Lisboa, 2, 56–167. (Republished 1967 in J. MoraisBarbosa (Ed.), Estudos Linguísticos Crioulos. Lisbon: Academia Internacional de Cultura Portuguesa.) Drechsel, E. J. (1981). A preliminary sociolinguistic comparison of four indigenous pidgin languages of North America (with notes towards a sociolinguistic typology in American Indian Linguistics). Anthropological Linguistics, 23(3), 93–112. Drechsel, E. J. (1996). Native American contact languages of the contiguous United States. In S. A. Wurm, P. Mühlhäusler, & D. T. Tryon (Eds.), Atlas of Intercultural Communication in the Pacific, Asia and the Americas, Vol. 2 (pp. 1213–1239). Berlin: de Gruyter. Extra, G. & van Hout, R. (1996). Second language acquisition by adult immigrants: A multiple case study of Turkish and Moroccan learners of Dutch. In P. Jordens & J. Lalleman (Eds.), Investigating Second Language Acquisition (pp. 87–113). Berlin: de Gruyter. Hall, R. A., Jr. (1966). Pidgin and Creole Languages. Ithaca, NY: Cornell University Press. Harding, E. (1984). Foreigner talk: A conversational-analysis approach. In M. Sebba & L. Todd (Eds.), Papers From the York Creole Conference, September 24–27 1983, York Papers in Linguistics (pp. 141–152). Heslington: University of York. Hinnenkamp, V. (1982). Foreigner Talk and Tarzanisch. Hamburg: Helmut Buske. Huebner, T. (1983). A Longitudinal Analysis of the Acquisition of English. Ann Arbor, MI: Karoma. Hymes, D. (Ed.). (1971). Pidginization and Creolization of Languages. Cambridge: CUP. Jahr, E. H. (1996). On the pidgin status of Russenorsk. In E. H. Jahr & I. Broch (Eds.), Language Contact in the Arctic: Northern Pidgins and Contact Languages (pp. 107–122). Berlin: de Gruyter. Jordens, P. (1996). Input and instruction in second language acquisition. In P. Jordens & J. Lalleman (Eds.), Investigating Second Language Acquisition (pp. 407–449). Berlin: de Gruyter. Kay, P. & Sankoff, G. (1974). A language-universals approach to pidgins and creoles. In D. DeCamp & I. F. Hancock (Eds.), Pidgins and Creoles: Current Trends and Prospects (pp. 61–72). Washington, DC: Georgetown University Press.
JB[v.20020404] Prn:25/07/2006; 12:03
F: LA9311.tex / p.25 (1653-1730)
Reduced syntax in (prototypical) pidgins
Klein, W. & Perdue, C. (1997). The Basic Variety (or: Couldn’t natural languages be much simpler?). Second Language Research, 13, 301–347. Macedo, D. P. (1986). The role of core grammar in pidgin development. Language Learning, 36, 65–75. Melville, H. (1846). Narrative of a Four Months’ Residence Among the Natives of a Valley of the Marquesas Islands: Or a Peep at Polynesian Life. London: John Murray. (titled Typee in later editions) Mühlhäusler, P. (1986). Pidgin and Creole Linguistics. Oxford: Blackwell. Naro, A. J. (1971). The origin of West African Pidgin. In CLS 9: Papers from the 9th Regional Meeting, 442–449. Chicago, IL: Chicago Linguistic Society. (Reprinted 1974 in Sociolinguistics Newsletter, 5, 8–11.) Perdue, C. (Ed.). (1993a). Adult Language Acquisition: Cross-Linguistic Perspectives. Volume 1, Field Methods. Cambridge: CUP. Perdue, C. (Ed.). (1993b). Adult Language Acquisition: Cross-Linguistic Perspectives. Volume 2, The Results. Cambridge: CUP. Perdue, C. & Klein, W. (1992). Why does the production of some learners not grammaticalize? Studies in Second Language Acquisition, 14, 259–272. Peterson, R. E. (1980). Russenorsk: A little known aspect of Russo- Norwegian relations. Studies in Language, 4, 249–256. Pienemann, M. (1999). Language Processing and Second Language Development: Processability Theory. Amsterdam: John Benjamins. Roberts, S. J. (1998). The role of diffusion in the genesis of Hawai’ian Creole. Language, 74, 1–39. Schumann, J. H. (1978). The Pidginization Process: A Model for Second Language Acquisition. Rowley, MA: Newbury House. Thomason, S. G. (1983). Chinook Jargon in areal and historical context. Language, 59, 820–870. Thomason, S. G. (1995). Language mixture: Ordinary processes, extraordinary results. In C. Silva-Corvalán (Ed.), Spanish in Four Continents: Studies in Language Contact and Bilingualism (pp. 15–33). Washington, DC: Georgetown University Press. van der Voort, H. (1997). New light on Eskimo Pidgins. In A. Spears & D. Winford (Eds.), The Structure and Status of Pidgins and Creoles (pp. 373–393). Amsterdam: John Benjamins. Zenk, H. B. (1988). Chinook Jargon in the speech economy of Grande Ronde Reservation, Oregon: An ethnography-of-speaking approach to an historical case of creolization in process. International Journal of the Sociology of Language, 71, 107–124.
JB[v.20020404] Prn:25/07/2006; 12:00
F: LA9312.tex / p.1 (48-108)
chapter
Copula variation in Guyanese Creole and AAVE Implications for nonsentential grammar* Walter Edwards This chapter explores copula absence in Guyanese Creoles (GC) and African American Vernacular English (AAVE), and considers implications for the underlying grammar of small clauses. GC and AAVE data show that the copula is absent in all three varieties in nonpast utterances across syntactic contexts. This observation has important consequences: first, GC and AAVE are similar codes in that they share a key linguistic feature; and second, these varieties provide an example of the existence of grammars that are fully intact yet lack the syntactic Tense node. Such grammars allow for the nonexpression of the semantically empty copula in the present tense, and exemplify what Progovac (Chap. 2, this volume) calls small clauses.
.
Introduction
Over the last decade or so Creole and African American Vernacular English (AAVE) scholars have been discussing copula verbs in these varieties both to support their positions on the linguistic provenance of AAVE and to correctly characterize the syntax and semantics for these verbs. This chapter begins with a brief examination of copula behavior in Rural Guyanese Creole (RGC) and Urban Guyanese Creole (UGC) and proceeds to consider what these syntactic behaviors imply for the underlying grammar of nonsentential small clauses, particularly in nonpast contexts.
. The copula in Rural Guyanese Creole RGC has three copula forms: an optional equative copula a, an optional locative copula de, and an obligatory past-tense copula bin.1 The strings in (1) through (3) exemplify these forms respectively: (1) Evlin (a) wan tiicha. ‘Evelyn is a teacher.’
JB[v.20020404] Prn:25/07/2006; 12:00
F: LA9312.tex / p.2 (108-180)
Walter Edwards
(2) Tairon (de) in di haus. ‘Tyrone is in the house.’ (3) Evlin bin wan tiicha. ‘Evelyn was a teacher.’ The selection of the optional copula in (1) and (2) seems to depend on the sense of permanence the speaker is encoding. With the copula the sentence assumes a slightly more temporary interpretation than without the copula. However, RGC does not allow a copula before predicative adjectives in the present tense as (4) through (7) reveal. Thus, it is clear that this variety is copulaless in this environment. (4) *Di bai a tal. ‘The boy is tall.’ (5) *Di bai de tal. ‘The boy is tall.’ (6) *Di bai a na leezi. ‘The boy isn’t lazy.’ (7) *Mi moda de na leezi. ‘My mother isn’t lazy.’ Since tall is a permanent (individual-stage) characteristic and lazy can be so interpreted, it is not surprising that (4) through (7) are not receptive to the copula. Several researchers have commented on the nonexistence of copulas before predicative adjectives in English-based creoles, including Bailey (1966) for Jamaican Creole. Arends, Muysken, and Smith (1995: 323–324) commented on the nonexistence of present-tense copulas in Guyanais, and the optionality of copulative sa in Principense. Some researchers (including Winford 1993: 155–210; Rickford 1999: 227–228) have explained copula absence before adjectives as being a consequence of the fact that predicate adjectives are really verbs. However, these accounts seem to leave unexplained why the copula is frequently absent in other environments in creoles and is mandatory before these same predicate adjectives in past-tense environments in RGC and in all environments in other languages, including English. Given that the present-tense copula has been shown repeatedly to be nonexistent or optional in conservative (a.k.a. basilect) creoles like RGC, an interesting question emerges about the status of this copula in less conservative varieties of creole, and in AAVE, which is considered by many scholars to be an evolved creole. In this chapter I present evidence that UGC, a variety that is spoken in Georgetown, the capital city of Guyana, is also a copulaless variety in some environments, and that the same analysis is accurate for AAVE, as that variety is spoken in Detroit. My thesis in this chapter is that the absence of the copula in present tense has important consequences for the grammars of varieties of Guyanese Creole (GC) and AAVE in two respects. The first is that GC varieties and AAVE are similar codes in that they share a key linguistic feature; and second, it provides an additional example, in common with all the studies in this volume, of the existence of gram-
JB[v.20020404] Prn:25/07/2006; 12:00
F: LA9312.tex / p.3 (180-236)
Copula variation in Guyanese Creole and AAVE
mars that are fully intact without full-sentence projections. Grammars of this latter sort would allow for the nonexpression of the semantically empty copula verb in the present tense, in all or in selected environments if, as is the case in RGC, the present tense is the default tense. The copulaless present-tense sentence would exemplify what Progovac (Chap. 2, this volume) calls small clauses.
. Urban Guyanese Creole and the present-tense copula In Edwards (1980) I discussed inter alia the patterning of the copula in a short sample from a typical UGC speaker and concluded that, “In the overwhelming number of cases the zero copula construction occurs in the non-past tense and before predicate adjectives” (301). I also found that the most restrictive environment for copula absence was ___NP, while the most permissive was ___Loc, followed by ___Adj. However, the sample used in the 1980 study only allowed a single opportunity for the copula to have occurred in the ___Loc context, making the ___Loc result dubious. In the 1980 article I urged linguists who are studying Black English Vernacular (BEV) “to examine the structure of urban varieties of Creole English in the Caribbean for possible insights into the structure and origin of BEV” (304). In the present chapter I would like to use the data and conclusions from my more recent work on UGC and AAVE to support my claim that UGC and AAVE are varieties that allow the formation of acceptable sentences without a present-tense copula as a full verb.
. Copula variation in Urban Guyanese Creole The speech data used in this part of the chapter were obtained from working-class Guyanese who were long-time residents of Georgetown, the capital city of Guyana, at the time they were interviewed. Five speakers, M32, M24, M35, M 27, and F45 in Table 12.1 below, were interviewed by me in 1974 as part of the database for Edwards (1975). One speaker, F28, was interviewed in 1979 by a former student of mine. This latter sample was analyzed and discussed in Edwards (1980). In Table 12.1, I also include an idealized characterization of urban Guyanese speech published in the Daily Argosy in 1930. In the count for copula tokens, I excluded sentences with existential is [Iz], and negatives. I excluded is because in UGC it is categorically surfaces as is [Iz], for example. (8) Because is laboring and laboring is manpower. (M24) ‘Because it is laboring, and laboring is manpower.’ (9) Then from there is manpower again. (M24) ‘Because from there it is manpower again.’
JB[v.20020404] Prn:25/07/2006; 12:00
F: LA9312.tex / p.4 (236-395)
Walter Edwards
Table 12.1 Copula variation in urban Guyanese speech: Following environment Speaker F28 M32 M24 M35 M27 F45 Totals %
___Adj + – 1 5 1 4 0 0 11 24
14 2 4 5 8 8 35 76
0
7
___Loc + – 0 2 0 0 0 0 3 33
1 0 0 1 3 3 6 66
___NP + – 8 4 12 8 1 1 36 97
0 0 0 1 0 0 1 3
___V+ing + – 1 3 1 1 0 0 6 13
Uncle Stapie from Sunday Argosy 1930 0 1 7 0 0
___Gun +
–
7 5 8 14 3 3 39 87
0 0 0 0 0 0 0 0
3 1 0 3 0 0 13 100
9
0
0
(10) You see, is kind of difficult. (M24) ‘You see, it is kind of difficult.’ I excluded negatives because the present-tense copula almost never occurs explicitly before negative markers in UGC. Thus, He is not eating is rare in this variety; instead He not eating is typical. Table 12.1 displays the patterns I discuss below. The ages of informants range from 28 to 45. There are four males (M) and two females (F). Note that the ___Gun environment refers to the copula preceding the UGC future marker gun.2 Table 12.2 presents some examples of the utterances from the speech data that illustrate copula presence and absence in UGC. The UGC material is disappointing in terms of the number of tokens identified in the ___Loc and ___Gun environments. In both of these contexts, the copula is almost always absent but the sparse data and the smallness of the sample size discourage any firm conclusions about the copula in these contexts. In UGC the ___Gun and ___Loc environments are problematic for the present study. As a native speaker of UGC, I don’t recall ever hearing a sentence like *we is gun/gunna play in which the formative is precedes gun, the future marker. Typically, gun is used alone. However, in RGC the copula a sometimes collocates with gun in expressions such as Awi a gu/gun play now (‘We are going to/will play now’).3 Thus, for UGC, it seems artificial to include ___Gun as a context for is/are variation in UGC. In the ___Loc environment the formative de is frequently used as the locative copula in the past and nonpast tense, thus complicating the count for is/are variation. However, the samples used for this study had very few tokens of the copula in locative contexts. An example of de usage in UGC is given at the end of column 1 in Table 12.2. The UGC data is revealing in the ___Adj, ___NP, and ___V+ing environments. The copula is most frequently absent in the ___V+ing environment, followed by ___Adj. In the ___Adj position, there is a strong tendency for the copula to be absent. In fact, in the speech of the only informant with more than 10 ___Adj tokens
JB[v.20020404] Prn:25/07/2006; 12:00
F: LA9312.tex / p.5 (395-411)
Copula variation in Guyanese Creole and AAVE
Table 12.2 Copula presence and absence in UGC (regularized orthography) Column I Copula Presence
Column II Copula Absence
The only thing does kill me is the vomit But I is what? I is a fair person Pressure is not a good thing That is what I know That is a sweet, sweet piece a joke That is using That is what I telling you That is the kind of labouring I does do So the most important thing you need is strength So, that is why That is the only difficulty Bartica is kind of hard There’s no comparison between Bartica and Georgetown That is the popular band Ordinary dance is Saturday night I believe they are doing they own thing This man is a rich man The first one is “abide with me” That’s why the food gat to be African That person is the conductor Everybody else is chorus I’m on leave from my job George de in the yard
When I say I sick, I sick I pregnant She coming and stretching she hand I wrong I talking I know the girl sick She sick and been admitted. . . Yu gyafing and thing When I doing night life, I serious They talking about a type of clothes Everybody doing they own thing I knocking, you know I fighting We fighting I still fighting Their way of life different Country people more nice People dem all different When dem talking to black people. . . We gun sing till the conductor signal She gun pass it to another female The conductor gun say “nancy” The chorus gun say “tori” They in de (‘They’re in there’) Milton out the house already I on the playground Everybody at home
(F28) the copula was present only once. The data also show that the copula is nearly always absent before ___V+ing and nearly always present in the ___NP environment. Although Table 12.1 does not give information about the influence of the preceding environment on the presence or absence of the present-tense copula, the illustrative examples in Table 12.2, column 1, reveal that the copula is often present when the subject is a nonpersonal pronoun such as that; or when the subject is an NP. In fact, the copula is almost categorically present in these contexts. Table 12.2, column 2 shows that the copula is frequently absent when the subject is a personal pronoun, including I.4 Table 12.3 includes an analysis of two tales, “Theo” and “Love and Stone Wall,” by the fictional raconteur Uncle Stapie, whose column, “Uncle Stapie Pon The People,” appeared regularly in the Sunday Argosy, a now defunct Guyanese newspaper, in the 1930s. Uncle Stapie’s popularity depended on the authenticity of his vernacular lan-
JB[v.20020404] Prn:25/07/2006; 12:00
F: LA9312.tex / p.6 (411-490)
Walter Edwards
Table 12.3 Copula variation in Uncle Stapie stories Copula Presence
Copula Absence
You is boss for the neighbourhood Some a dem so simple Dat is all dat concern you A pon de young people side Another story dat a hear is a ole confidence trick Dinner time come de rowin (quarrelling) wo
guage and the wittiness of his stories. The stories analyzed for the information on the chart appeared in the May 4, 1930 edition of the paper, on page 4. The language of the stories exhibit copula usage patterns similar to the speech samples: the present-tense copula is uniformly absent before adjectives and V+ing constructions, and present after NP subjects. The examples in Table 12.3 are taken from the stories.
. Guyanese Creole and the nonsentential analysis The data and analyses so far have revealed that in both RGC and UGC the copula is optional in the present tense and in RGC the present-tense copula is categorically absent in the ___Adj environment. These facts seem to support the proposal that some grammars include small clauses and that such entities do not have Tense nodes (see particularly Progovac, Chap. 2; Barton, Chap. 1; and Paesani, Chap. 6, in this volume). In such grammars the present assumes the role of default time and needs no grammatical signal. A small clause reading of the GC data provides a coherent explanation of the absence of the present-tense copula in all environments. One could reasonably assume that GC speakers, in common with English, French, and Slavic speakers, possess two grammars simultaneously. One grammar would have full-sentence projections that insert present-tense copulas in all environments, except the ___Adj environment in RGC and the ___Gun environment in UGC; while the other, the small clause grammar, lacks a Tense node and, thus, a tensed copula to insert. Time would then be interpreted as the present time, which, in this model, is the default time. Past and future time would be lexicalized by conventionalized markers such as bin and go/gun respectively, except when narrative contexts authorize the omission of these markers. If one explains the data in this way, then one can further speculate that the small clause grammar is more dominant in the linguistic epistemology of GC speakers than in English, French, or Slavic speakers, since utterances with small clause characteristics seem to be more regular and systematic in GC than in those languages. The pronominal forms in the data in Tables 12.1 and 12.2 could possibly be interpreted as providing additional support for the nonsentential analysis. The small clause analysis proposes that there is a default case that small utterances exhibit and that typically is the accusative but in some varieties could be nominative (see Progovac, Chap. 2, and Work, Chap. 8, this volume, respectively). RGC data are split between the choice of nominative and accusative forms as seen in Table 12.4, where mi and dem are accusative forms but awi, hi, and shi are nominative.5 The data also show that UGC
JB[v.20020404] Prn:25/07/2006; 12:00
F: LA9312.tex / p.7 (490-733)
Copula variation in Guyanese Creole and AAVE
Table 12.4 Personal pronouns in RGC and UGC Person
Singular
Case
Plural
Case
RGC
1st 2nd 3rd
mi yu hi/shi
Nom, Acc Nom, Acc Nom, Acc
awi alyu dem
Nom, Acc Nom, Acc Nom, Acc
UGC
1st
I me you he/she
Nom Acc Nom Nom, Acc
we
Nom
alyou dem/they dem
Nom Nom Acc
2nd
speakers select the nominative more frequently than the accusative for personal pronoun forms. An indication of this tendency in UGC is the choice of they rather than their to show possession in the string I believe they are doing they own thing (Table 12.2).6 Although this evidence is inconclusive, the pronoun data could be interpreted as showing a tendency to a default case of accusative for RGC and nominative for UGC in personal pronoun choice. This reading would be consistent with my proposal (Edwards 1980) that RGC and UGC derive from different linguistic inputs, with the latter absorbing more English features than the former.
. Copula variation in AAVE The AAVE material was extracted from transcriptions of interviews with six workingclass Detroiters (ages 18–51) recorded in 1988. Following practices in the recent literature, I looked at the environments preceding and following the copula, and distinguished between the auxiliary and main-verb functions of the copula.7 In the preceding environment, I was interested in the subject type: personal pronouns versus nouns. In the following environment, I looked at adjectives, locatives, NPs, V+ing, and gonna. I excluded negative sentences and, following Rickford (1999), excluded sentences in which the subject is the pronoun I or it. Tables 12.5, 12.6, and 12.7 display the results. The data in Tables 12.5, 12.6, and 12.7 indicate that in Detroit AAVE the copula is most frequently absent after personal pronouns, before V+ing, before adjectives, and before locatives (in that order); and most often present before a predicate nominal and when the subject is a nonpersonal pronoun. Even though the AAVE sample used for this chapter is rather small, the results are similar to the finding in Rickford (1999) and Rickford and Rickford (2000). In the latter study, Rickford and Rickford pointed out that:
JB[v.20020404] Prn:25/07/2006; 12:00
F: LA9312.tex / p.8 (733-758)
Walter Edwards
Table 12.5 Copula variation in AAVE: Preceding environment
Informant
Total possible
315 310 303 168 64 55 %
13 13 11 8 9 47
Preceding Environment Personal pronoun Pronoun + – + –
+
–
1 0 0 1 3 9 25
5 4 0 – 2 6 77.2
1 0 2 – 0 2 22.7
4 5 7 6 2 18 75
2 4 1 1 2 11 91.3
0 0 1 0 0 1 8.7
NP
Table 12.6 Copula variation in AAVE: Following environment
Informant
Total possible
315, 18M 310, M21 303, M35 168, F24 64, F44 55, F51 %
12 12 11 8 9 44
___Adj + –
Following Environment ___NP ___Loc + – + –
2 1 0 2 1 7 43.3
1 6 1 0 2 10 64.5
1 1 3 3 0 9 56.7
0 4 0 3 0 4 35.5
___V+ing + –
2
1
2
3
0
1
0
6
3 55.5
2 44.5
4 3 34.6
2 6 65.4
Note. Only one token of copula before gonna appeared in the sample. The fact that only one token was found in this normally productive environment is disappointing, but was clearly the result of the interview prompts, which encouraged speakers to discuss current social issues and past experiences.
AAVE speakers . . . delete both is and are more often when these copulas come after a pronoun . . . than when they come after a noun . . . and they delete is and are least often before a noun . . . more often before an adjective . . . and most often before going to or its reduced form gonna and gon. (116)
Some scholars, including Walker and Meechan (1999) and Walker (2000), have suggested that copula absence in AAVE is really copula deletion, which results from a phonological not syntactic process. They begin with the observation first made by Labov (1972) that AAVE speakers can delete the present-tense copula in precisely those environments in which the Standard English (SE) speakers can contract the copula; and that AAVE copula deletion is impossible in those SE positions where the full copula is mandatory. However, in a presentation at the 2001 meeting of the Linguistic Society of America, Bender (2001) looked thoroughly at the idea of copula deletion as a phonological product and disputed it. The following are the AAVE examples she found that show that the contracted copula and the zero copula are not interchangeable:
JB[v.20020404] Prn:25/07/2006; 12:00
F: LA9312.tex / p.9 (758-778)
Copula variation in Guyanese Creole and AAVE
Table 12.7 Copula absence and copula presence in AAVE Copula Absence: Examples What happening to our young people You perfetratin [perpetrating] Knowing what you saying When you at home they don’t know what they talking about They allowin these people to keep these old raggedy cars He startin on it already You goin school just to see can you do it The older people still going to bars Young people going to discos They going to another neighbourhood I know she having a good time I know she having a great time They right back on the street doing the same thing all over again A lot of senior citizen around here now When they right down the street Copula Presence: Examples Where they’re fixing them cars We’re getting up there the young girls He’s building up and he’s tearing down I think he’s doing a nice job He’s building up and he’s tearing down Person I can see that’s doing something for theyselves People that’s on the street The only kind of game they play that’s good is Pac Man People that’s gonna cause trouble Young drug boys are coming to their house I’m glad that you all, you know, are doing this The only reason they’re there Most of them is off That’s sad for our young people This is a disgrace The young mens that’s around here. . .
(11) How old you think his baby is? (12) How old you think his baby? (Deletion is possible) (13) *How old you think his baby’s? (Contraction is not allowed. One might argue that contraction here is not allowed to prevent baby’s from being confused with babies.) However, note the following sentences: (14) The baby’s fine. (15) The babies fine.
JB[v.20020404] Prn:25/07/2006; 12:00
F: LA9312.tex / p.10 (778-848)
Walter Edwards
(16) How old you think his babies? (17) Tha’s the man they say is in love. (18) Tha’s the man they say in love. (Deletion is allowed) (19) *Tha’s the man they say’s in love. (Contraction is not allowed.) Bender (2001) concludes that copular deletion cannot be understood as a product of contraction, but must be understood as an allomorph of the full copula. Significantly for this study, she has shown that since the zero and contracted copulas follow different rules, prosody cannot account fully for both. The copular forms must be governed by rules that fall beyond phonology. This evidence points back toward the original question of the syntactic provenance of the zero copula. The story told by the information in Tables 12.5, 12.6, and 12.7 mimics quite closely the story told by the information in Tables 12.1 and 12.2 that characterized copula behavior in the speech of the Guyanese UGC informants, supporting the claim that UGC and AAVE exhibit similar patterns of copular behavior in the present tense, that is, that the copula is frequently absent before adjectives and locatives. As expected, SE patterns show up numerically more frequently in the AAVE data than in the UGC data, since the former variety, more than the latter, continues to encounter influence from the colloquial standard vernacular. Even though both samples are too small to be representative, and the number of tokens are so few that general patterns cannot be claimed, it seems clear that in both varieties the copula is frequently absent in ___V+ing and ___Adj environments.8 Whatever the grammatical or phonological motivations for these linguistic orientations, it seems reasonable, if not inescapable, that UGC and AAVE speakers use common principles to organize their copula use in the nonpast tense. It seems to me, therefore, that these additional data support the suggestion I made cursorily in Edwards (1980); namely, that Urban Caribbean Creoles might provide evidence of the linguistic relatedness of AAVE and Caribbean language varieties. Taken together with the evidence in the literature (Alleyne 1980; Bickerton 1975, 1980; Rickford 1987) that many conservative Englishbased creoles, including RGC, do not have present-tense copulas before predicate adjectives, the data appear to establish a firm historical link between creoles and AAVE with regard to the nonpast copula construction.
. AAVE, Guyanese Creole, and the nonsentential analysis The data and analyses just presented show that, like RGC and UGC, AAVE speakers in Detroit very rarely use the present-tense copula before adjectives, locatives, and V+ing. The data from my Detroit sample are similar to data presented and analyzed in Baugh (1980) and reproduced in Table 12.8. On that chart Baugh showed that the most favorable environment for the absence of the copula in the present tense was the ___Adj environment with a perfect score of 1.000. Focusing only on the forms and environ-
JB[v.20020404] Prn:25/07/2006; 12:00
F: LA9312.tex / p.11 (848-916)
Copula variation in Guyanese Creole and AAVE
Table 12.8 Environmental influences on copular deletion (adapted from Baugh 1980: 42– 51) -C__ -V__ __C__VNP__ Pro__ __gon(na) __V+ing __ NP __ Loc __ Adj __ DP
0.061 0.000 0.322 0.000 0.000 0.714 0.601 0.402 0.000 0.682 1.000 0.741
ments used in my analysis, the table shows that the VARBRUL program Baugh used predicted that based on his AAVE data there was the strongest possibility (1.000) that the copula would be absent before predicate adjectives. There was also a better than 60% chance that the copula auxiliary would be absent before gonna, the future marker in AAVE. Recall that I pointed out above that in UGC the copula auxiliary is never used before gun, the future marker. Thus, it seems reasonable to conclude that a nonsentential reading of the AAVE data is as feasible as it was for the GC data. I would propose that AAVE speakers also have a complex linguistic competence that permits them to access their small clause grammar that lacks present-tense copulas.
. Discussion The data and analyses present a case for classifying RGC, UGC, and AAVE as varieties that exhibit central characteristics of small clause status within the nonsentential analysis. What distinguishes these two varieties from SE is the fact that their nonsentential constructions are not responses to full-sentence prompts, but are initial, base-generated, well-formed expressions. There is additional independent support for the absence of Tense in creoles like Guyanese Creole in noncopula constructions. Bickerton (1975) proposed that in the Guyanese Creole Basilect (the most conservative variety of GC, i.e., RGC in this chapter) there is no tense marker. Time is indicated by the speaker and inferred by the listener based on whether the verb is stative or nonstative. Thus, in (20) the tense/time is nonpast whereas in (21) the tense/time is past. The crucial distinction is thus a semantic one based on the character of the verb. (20) Mi gat wan ball. ‘I have a ball.’
JB[v.20020404] Prn:25/07/2006; 12:00
F: LA9312.tex / p.12 (916-989)
Walter Edwards
(21) Mi iit wan mango. ‘I ate a mango.’ Bickerton proposes a [+ Anterior] marker bin for the basilect with the meaning of [+ past] for statives and [+ past before past] for nonstatives. Thus (22) and (23) have simple-past and pluperfect meanings, respectively. (22) Mi bin gat a ball. ‘I had a ball.’ (23) Mi bin iit the mango. ‘I had eaten the mango.’ UGC preserves this [Anterior] system with di as the past marker, as (24) through (27) show. (24) A gat a ball. ‘I have a ball.’ (25) A eat the mango. ‘I ate the mango.’ (26) A di gat a ball. ‘I had a ball.’ (27) A di eat the mango. ‘I had eaten.’ Note, parenthetically, that Gibson (1982) takes issue with Bickerton’s [+/– Anterior] system for GC mainly because bin is not a unique anterior marker, performing as it does other functions in the GC verbal system. Gibson’s position is stated this way: I have posited a primary tense opposition on [+/–NONPRESENT] for Guyanese Creole. The fact that the PRESENT is not formally marked, its use from present, past, nonpast and future meanings, and its use for relative time referencing and sequencing, indicate that the PRESENT is the unmarked member of the PRESENTNONPRESENT opposition. (246)
Whether Bickerton’s or Gibson’s analysis is correct, it seems clear that present tense, even in noncopular constructions, poses a theoretical problem for the sentential analysis of GC, a problem that does not apparently exist for the nonsentential analysis. As a sociolinguist I am particularly interested in the implications of the nonsentential analysis for creoles and for AAVE since it brings into play the importance of dealing with variation within a coherent theoretical framework. Since creoles are typically spoken in societies with non-creole, politically more dominant languages, the creole speaker typically has a linguistic competence that embraces features of both the non-creole and creole languages. If the assumptions and descriptive details of the nonsentential analysis are correct then the typical GC and AAVE speaker’s linguistic competence includes a grammar that generates well-formed sentences within a
JB[v.20020404] Prn:25/07/2006; 12:00
F: LA9312.tex / p.13 (989-1044)
Copula variation in Guyanese Creole and AAVE
TP framework as well as a grammar that generates well-formed sentences outside of this framework. The speaker must also possess as part of his conceptual framework a mechanism for switching between grammars. Such a mechanism might resemble the markedness metric developed by Myers-Scotton (1991) or such code-switching models as those proposed by Gardner-Chloros (1991). Thus, the nonsentential analysis is likely to prompt a rich theoretical rapprochement between scholars working on universal grammars and their formal characterizations and those working to find principles that explain and theorize language variation.
Notes * I acknowledge with gratitude the invaluable help of my research assistant Leslie Ash in preparing this chapter. She transcribed the interviews with AAVE speakers, conducted related research, and helped organize this chapter. The AAVE speakers used in this paper were interviewed for sociolinguistic work in Detroit that led to Edwards (1992). . The past-tense copula is often not expressed in narrative contexts, especially when past time is clear from the preceding narrative. . The UGC future marker occurs in such expressions as Dem gun go home (‘They will go home’). Note, too, that in Table 12.1 and elsewhere in this chapter the auxiliary copula and the full-verb copula are discussed together as if they pattern exactly alike in the copulaabsence/copula-presence variation. I do not consider that a settled matter, but will make that assumption in this chapter, following scholars, including Rickford (1999) and Baugh (1980), who have done so for AAVE. The reader will also notice that in discussing UGC I have chosen, for ease of presentation, to use regularized English orthography instead of the phonemic orthography I used in discussing RGC. . See Edwards (1984) for a discussion of the differences between RGC and UGC. . In this regard UGC differs from AAVE since in the latter variety the present-tense copula is almost always present after the first-person singular pronoun. . Note that in conservative Jamaican Creole the personal pronouns are mainly accusative: mi, him, yu, her, unu, we, alyu, and dem (see Bailey 1966). . In fact neither RGC nor UGC has separate forms for the genitive in the second or third persons. . In the data for RGC and UGC the preceding environments were not analyzed. This fact lessens the reliability of the comparison between the two data sets. . The figures for ___Adj and ___Loc in the AAVE samples in this study show a lower incidence of copula absence than that found in other studies. This is probably a consequence of the fact that only six speakers were used. Additional samples are being analyzed for copula variation.
References Alleyne, M. (1980). Comparative Afro-American. Ann Arbor, MI: Karoma Press.
JB[v.20020404] Prn:25/07/2006; 12:00
F: LA9312.tex / p.14 (1044-1135)
Walter Edwards
Arends, J., Muysken, P., & Smith, N. (1995). Pidgins and Creoles: An Introduction. Amsterdam: John Benjamins. Bailey, B. (1966). Jamaican Creole Syntax: A Transformational Approach. London: CUP. Baugh, J. (1980). A Reexamination of the Black English Copula. In W. Labov (Ed.), Locating Language in Space and Time (pp. 83–106). New York, NY: Academic Press. Bender, E. M. (2001). AAVE Copula Absence Is Not Phonological Deletion. Paper presented at the Linguistic Society of America, Washington, DC. Bickerton, D. (1975). Dynamics of a Creole System. Cambridge: CUP. Bickerton, D. (1980). Decreolization and the creole continuum. In A. Valdman & A. Highfield (Eds.), Theoretical Orientations in Creole Studies (pp. 109–128). New York, NY: Academic Press. Edwards, W. F. (1975). Sociolinguistic Behavior in Rural and Urban Circumstances in Guyana. Ph.D. dissertation, University of York. Edwards, W. F. (1980). Varieties of English in Guyana: Some comparisons with BEV. Linguistics, 18, 289–309. Edwards, W. F. (1984). A community-based approach to the provenance of Urban Guyanese Creole. Special Edition of York Papers in Linguistics, 2, 83–94. Edwards, W. F. (1992). Sociolinguistic behavior in a Detroit inner-city Black neighborhood. Language in Society, 21, 93–115. Gardner-Chloros, P. (1991). Language Selection and Switching in Strasbourg. Oxford: OUP. Gibson, K. (1982). Tense and Aspect in Guyanese Creole: A Syntactic, Semantic and Pragmatic Analysis. Ph.D. dissertation, University of York. Labov, W. (1972). Language in the Inner City: Studies in the Black English Vernacular. Philadelphia, PA: University of Pennsylvania Press. Myers-Scotton, C. (1991). Motivation and Structures in Code-Switching: Evidence from Africa. Oxford: OUP. Rickford, J. (1987). Dimensions of a Creole Continuum: History, Texts and Linguistic Analysis of Guyanese Creole. Stanford, CA: Stanford University Press. Rickford, J. (1999). African American Vernacular English. Malden, MA: Blackwell. Rickford, J. & Rickford, R. (2000). Spoken Soul: The Story of Black English. New York, NY: Wiley. Walker, J. A. (2000). Rephrasing the copula: Contraction and zero in early African American English. In S. Poplack (Ed.), The English History of African American English (pp. 35–72). Malden, MA: Blackwell. Walker, J. A. & Meechan, M. E. (1999). The decreolization of Canadian English: Copula contraction and prosody. In J. Jensen & G. Van Herk (Eds.), Actes du Congrès annuel de l’Association canadienne de linguistique 1998/Proceedings of the 1998 Annual Conference of the Canadian Linguistics Association (pp. 431–441). Ottawa: University of Ottawa, Cahiers Linguistiques d’Ottowa. Winford, D. (1993). Predication in Caribbean English Creoles. Amsterdam: John Benjamins.
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.1 (47-119)
Epilogue Wherefrom and whereto?* Ljiljana Progovac, Kate Paesani, Eugenia Casielles, and Ellen Barton This Epilogue concludes this volume on the syntax of nonsententials by bringing together themes that connect the chapters. Section 1 discusses variability within and across individuals as well as within and across contexts and registers of language, including adult speech, child speech, nonnative speech, aphasic speech, and pidgins and creoles. It is suggested that the continuity of grammar may lie not in sentential grammars for all, but in nonsentential grammars for all speakers. Against the backdrop of the nonsentential analysis, Section 2 explores some connections between language acquisition and language history. Section 3 provides a general comparison of sentential and nonsentential analyses, looking in particular at connectivity effects, while Section 4 addresses some remaining issues and implications for future research.
.
Nonsententials, individual variation, and multiple grammars
As several chapters in this volume have shown, adults often produce nonsentential utterances in casual conversation and in a variety of special contexts and registers. These include different types of (isolated) phrases that can appear in short answers to questions (1), labels (2), telegrams (3), and out of the blue utterances (4). (1) A: Who prepared dinner? B: Mark. (2) Airport. (3) In trouble. (4) From Spain. (while holding a letter) Furthermore, and perhaps even more interestingly, adult nonsentential utterances also include telegraphic subject-predicate structures that turn up not only in telegrams (5), but also in headlines (6), note taking (7), the so-called Mad Magazine register (8), medical diagnoses (9), and possibly many other contexts. (5) Car dead.
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.2 (119-195)
Ljiljana Progovac et al.
(6) Failed vasectomy claim rejected. (7) Experiment performed twice. (8) Him stop smoking?! (9) Lump in throat. These utterances are characterized by the absence of finite verbs, nominative case, determiners, and so on, and therefore seem to depart from regular adult speech. Although data produced by unimpaired adult native speakers is usually not characterized as variable, the contrast between the nonsentential adult utterances in (1) through (9) and their sentential counterparts seems to point to the existence of variable productions in adult speech. Notice that what is special about this type of adult variation, as opposed to the well-known cases of sociolinguistic variation, is that this variation not only occurs across individuals who might speak different varieties and therefore have different grammars, but also crucially within individuals, who are assumed to have one grammar. There are different ways of handling this type of individual variability. One way would be to set these adult data aside, outside the scope of syntactic analysis. Such a view would seem to avoid the problem rather than to try to account for it. In addition, such a view would also have to explain why adult nonsentential utterances would be beyond the scope of syntactic theory when syntactic theory has a great deal to say about child, nonnative, and aphasic nonsentential and sentential productions (see below). If we resist the temptation to put aside the not-so-regular data, a second, more interesting possibility arises, which is to explore how syntactic theory could account for nonsentential and sentential adult utterances. This can be done in at least two ways, both of which have been pursued in this volume to different degrees. The first is not to accept variability and assume that what looks like isolated phrases and telegraphic speech is just elliptical adult speech. The second is to embrace variability and expect syntactic theory to account for it. Let us briefly consider each of these possibilities. The first view would maintain that grammar only generates full sentences (elliptical or nonelliptical) and that adult speech is not variable. Pursuing this view for adult data makes sense since adults often produce elliptical sentences such as VP-ellipsis, sluicing, and so on. Since there are well-known analyses of ellipsis, which are needed for independent reasons, extending these analyses to account for adult nonsentential data is a worthwhile enterprise, which can give fruitful results (see Merchant 2004, and Chap. 3, this volume).1 If on the right track, this would mean that adult nonsentential utterances (or at least those for which we can recover what seems to be missing) are only superficially similar to other nonsentential data (such as those found in child, aphasic, and nonnative speech) in that they involve the generation of full sentences that later undergo elliptical processes. That is, from this perspective, adult nonsentential utterances would not lack any sentential syntactic structure but only its pronunciation. If right, adults always produce sentential syntax and any possibility of a unified account of nonsentential utterances in terms of isolated phrases and small clause predicate-subject structures has to be abandoned.
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.3 (195-270)
Epilogue
The second view, which several chapters in this volume have explored, entails challenging syntactic theory to account for the data without converting it into the sentential type easily handled by syntactic theory. From this perspective, not all adult utterances are homogeneous; adults produce fully sentential, elliptical, and nonsentential utterances and syntactic theory needs to account for that. A contrast between different possibilities for answering a question can serve as an illustration (see Casielles, Chap. 5, this volume). The question in (10) can receive a sentential answer as in (11), an elliptical answer as in (12), or a phrasal answer as in (13): (10) Who prepared dinner? (11) Mark prepared dinner. (12) Mark did. (13) Mark. Some evidence that short answers such as that in (13) might be phrasal as opposed to sentential comes from the fact that in some languages where nouns are case-marked, these short answers can show no case whatsoever, as in the following examples from Korean (see, e.g., Morgan 1973): (14) A: Nu-ka ku chaek-ul sa-ass-ni? who this book-acc bought ‘Who bought the book?’ B: Yongsu. Yongsu-no.case Interestingly, even in English we can see a difference between the case of a subject in an elliptical answer as in (15) and the case that shows up in a phrasal answer as in (16) or in tenseless subject-predicate structures of the type in (17) and (18).2 (15) A: Who prepared dinner? B: I did. (16) A: Who prepared dinner? B: Me. (17) Me preparing dinner. (written under a picture) (18) Me prepare dinner?! If this view is on the right track, then adults produce a variety of structures, not all of which are full sentences. These would include isolated phrases, subject-predicate small clauses (without tense or nominative case), elliptical sentences, and full sentences. From this perspective, individual variation would also be a feature of adult speech and perhaps, more generally, a pervasive feature of human language.3 What is interesting about this possibility is that it allows us to compare this type of individual variation with other well-known cases such as those found in children, aphasics, and nonnative learners. In the fields of neurolinguistics and language acqui-
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.4 (270-323)
Ljiljana Progovac et al.
sition, variability is a particularly well-known feature that has been at the center of theoretical debates. Any account of root infinitives and null subjects in first language (L1) acquisition needs to explain the fact that children sometimes use inflection and overt subjects and sometimes they do not. Variability across individuals is even greater in cases of second language (L2) acquisition where some L2 learners achieve nativelike competence, whereas many others do not come even close to the target grammar (see Work, Chap. 8, this volume). The same issue comes up in neurolinguistics. Any theory of agrammatism has to deal with variable productions within and across individuals (see Kolk, Chap. 9, and Siple, Chap. 10, this volume, for details). A syntactic analysis of the nature of structures lacking functional categories such as Tense, Agreement, and Case has to account at the same time for the fact that sometimes these categories are manifested and sometimes they are not. Now, how could syntactic theory account for adult variability? In this case we cannot appeal to any of the issues relevant for aphasic, child, or nonnative variability such as limited processing capabilities, maturation issues, or limited access to Universal Grammar (UG). Interestingly, Roeper (1999) has suggested that every speaker, including adults, has a set of mini-grammars for different domains and that it is this universal bilingualism that can also account for optionality in the acquisition of an initial grammar. In the spirit of Roeper’s “Universal Bilingualism,” many contributions to this volume have explored the possibility that adult speakers may have access to multiple grammars or, more specifically, that adult grammar may allow for multiple syntactic outputs. Furthermore, this volume has pointed to another interesting set of data, which is also characterized by a high degree of variability and which often shows some of the features mentioned above such as the absence of tense and case: the case of pidgin and creole languages (see Edwards, Chap. 12, and Winford, Chap. 11, this volume). This set of data is important because if a unified analysis of nonsententials proves to be plausible, then a proposal about the syntax of these structures would have to take into account not only factors such as limited processing capabilities (for aphasics), maturation issues (for children), or lack of access to UG and possible transfer from L1 (for L2 learners), but it would also have to consider other equally important factors such as the role of language contact and the processes of language creation and language change.4 Future research will have to examine in detail the precise nature of what look like phrasal, small clause, elliptical, and full-sentence utterances in each particular case. However, if the grammar of an individual adult speaker allows for multiple syntactic outputs, then some of the problems that have been pointed out in the literature on aphasia and L1 and L2 acquisition will have to be reconsidered. For example, producing sentences with and without tense and with and without nominative case might not be as deviant as linguists thought. Needless to say, we would still have to explain why a sentential grammar seems unavailable for some agrammatics in some situations and available in others (see Kolk, Chap. 9, this volume) or how children acquire a sentential grammar and what might be the role of nonsentential utterances in the acquisition
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.5 (323-385)
Epilogue
process. Yet, we would not need to come up with child-specific or aphasia-specific constraints to explain variability in terms of expression of Tense and Case; nor would we need to explain how children get rid of the particular early grammar that allows for telegraphic subject-predicate expressions. Maybe they do not. The possibility of producing phrasal utterances and root small clauses may stay with us and surface in situations in which economy of expression is crucial, such as telegrams and headlines; in situations that require expressive force and additional meaning, such as humor or poetry; or when damage to certain areas of the brain interferes with the usual access to the mechanisms that are required to build whole sentences. A similar view has been defended recently by Herschensohn (2000). She points out that if variability of category projection is allowed, in the spirit of Safir (1993), Grimshaw (1994), and Chomsky’s (1995) Minimalist Program, the continuity/discontinuity debate in L1 acquisition and its counterpart in L2 is obviated. From this perspective, some of the perceived differences between so-called grammar and so-called usage, noted recently by Newmeyer (2003), may arise from looking at adult data in isolation and from assuming that only one (sentential) grammar option is available to each individual. Once we allow for multiple grammar options and consider adult native data in the context of other natural language, the picture that emerges is that of a variety of different syntactic possibilities. This would explain the different syntactic features present in the data that the same speaker can produce depending on the context (e.g., a formal written document, an e-mail, a lecture, a conversation in a bar, a family discussion, a diary entry, a telegram, a personal note). Although further research is needed to carefully carry out a cross-linguistic analysis of the syntax of different registers, it seems to be the case that English speakers can produce not only CP roots (with cased DPs and tense) or VP roots (root small clauses and root infinitives with caseless or default-cased NPs and no tense), but also structures that seem to be in the middle, such as tensed, subjectless clauses (as in diary entries and note taking, e.g., Shall now stop writing for a day) or tensed clauses with empty objects (as in recipes, e.g., Place in bowl with mushrooms) (see Paesani, Chap. 6, this volume). Recently, there have been some proposals outside mainstream Minimalist views such as Culicover and Jackendoff (2005), who have defended a nonelliptical analysis of nonsentential utterances in the context of what they refer to as a simpler syntax (see also Ginzburg & Sag 2000 for an HPSG view). Some of the research in this volume suggests that a nonsentential approach can be explored also in a Minimalist framework. Although there are many details that need to be investigated in regard to syntax/semantics/pragmatics interface issues, we hope the research presented in this volume will stimulate further investigation into the kind of possible syntactic outputs and their interpretation, not only in adult nonsentential utterances but also in those produced by children, aphasics, nonnative speakers, and speakers of pidgin and creole languages. A more detailed analysis of further data might show that nonsentential utterances are a crucial core feature of natural languages and their development, rather than a peripheral accident in language use.
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.6 (385-452)
Ljiljana Progovac et al.
. Why nonsententials: A historical perspective . Tracking (root) small clauses through acquisition and history Potts and Roeper (Chap. 7, this volume) point out that root small clauses, once used abundantly (in fact, exclusively, at an early, two-word stage) by young children, fall out of use in adult speech.5 One of the goals of Potts and Roeper’s chapter is to explain why and how such small clauses fall out of use, or change their interpretation in adulthood. In this section, we consider their explanation, as well as the one pursued in Progovac (Chap. 2, this volume), taking into account both language acquisition and history. Potts and Roeper (Chap. 7, this volume) observe that something happens to small clauses through the acquisition process: while children seem to be able to use utterances such as Me worry to make a statement, adults do not do so, at least not in root contexts. Such small clauses are still used in embedded contexts, however, with perception verbs, for instance (You never see [me worry]). But are such clauses completely barred from root contexts? According to Potts and Roeper, they are. This is because they analyze examples such as (19) below as a special type of expressive clause, which, moreover, may not even be a small clause, but rather some kind of loose coordination of two incompatible ideas/phrases: Me and then Worry. (19) Me worry?! (Never!) Their semantic analysis in terms of coordination rather intriguingly finds support in German, where in fact an overt conjunction can be used between the two elements (albeit optionally, as pointed out by Nicola Work, personal communication). (20) Ich (und) Angst haben?! I (and) fear have-inf ‘Me afraid?!’ According to Potts and Roeper, children do not use such expressive clauses because they still do not use coordination to combine syntactic constituents. Thus, for Potts and Roeper, there is no connection between the children’s small clause Me worry, which typically expresses a statement, and the adult structure Me worry?, which cannot express a statement, and which may not be a small clause. On the other hand, for Progovac (Chap. 2, this volume), root small clauses (e.g., Me worry) are continuously available through childhood and adulthood, and are structurally the same type, even though such clauses in adult speech differ in interpretation. The argument is that once finite clauses (e.g., I worry/I am worried), which specialize for indicative, truth-evaluable statements/assertions, emerge, nonfinite root clauses (e.g., Me worry) begin to specialize for nonindicative/irrealis/incredulity interpretations (unless embedded in some pragmatic frame, as discussed in Section 4.1). Progovac’s analysis, however, has nothing to say about why German would use conjunctions to form such incredulity clauses.
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.7 (452-507)
Epilogue
Progovac relates this acquisition path to the similar path of development of Indicative through history. It has been claimed in the Indo-Europeanist (IE) literature that once Indicative (tense-marked) forms emerged in Pre-Indo-European (Pre-IE), the Injunctive form, which had prior to that been used for everything, eventually narrowed its meaning to basically everything but Indicative, that is, to a range of other (irrealis) moods, such as Imperative, Exclamative, Optative, and Desiderative. Progovac suggests that this same range of moods also underlies the irrealis interpretations in adult root small clauses. Moreover, some Indo-Europeanists, including Kiparsky (1968), argue that Injunctive-type forms have been preserved in frozen expressions in a variety of languages and circumstances, such as English Long Live the King or God Forbid, and Progovac’s suggestion is that the bare form of the verb in nonsententials (e.g., Him worry?!) in English might also be analyzable in this fashion, as an underspecified mood/tense form, comparable to Injunctive. Thus according to Progovac’s analysis root small clauses (e.g., Me worry) are continuously available from childhood to adulthood. What changes in adulthood is only their meaning, which is affected by the emergence of the Indicative, a more specialized form. This change in meaning arguably parallels the historical development of the Indicative from the underspecified Injunctive. This is in fact consistent with Potts and Roeper’s (Chap. 7, this volume) general claim, that the most important milestone in the reinterpretation/reanalysis of children’s small clauses is the acquisition of finiteness, that is, IP/TP, and that “increased syntactic sophistication correlates with decreased semantic ambiguity.” In this respect, children’s small clauses such as Me worry can be seen as vastly underdetermined with respect to mood, freely allowing the expression of the indicative mood, while adult small clauses are less ambiguous in this sense, excluding the indicative option under normal circumstances. Very roughly speaking, this process can also be likened to the change in meaning of the word cow after the borrowing of beef.6 Before beef was available, the word cow was used to express the whole range of meanings, including table meat, just as it is still possible to do with words such as chicken or duck. But once beef was introduced as a more specific term for table meat, precisely this element of meaning was taken away from the meaning of cow, and cow was no longer normally used in this sense. However, arguably, the same word cow continues to be available, even though its meaning has narrowed down to yield part of its semantic territory to a more specific form. There are of course other ways to look at this difference between child and adult clauses such as Me worry. One is to argue that such utterances in child language are not root small clauses, but rather full finite clauses that have undergone some kind of phonological adjustments. So, Me worry is really I worry, and Me dancing is really I am dancing. This kind of approach would have nothing to explain in terms of the syntactic transition from Me dancing to I am dancing, or in terms of the change in interpretation of Me worry through acquisition, because for that hypothesis there is no change (see, e.g., Guasti (2002) and references therein). This is considered to be the main virtue of the continuity hypothesis in language acquisition – that no explanation for change is necessary.
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.8 (507-566)
Ljiljana Progovac et al.
It should be pointed out, however, that the nonsentential analysis offers a novel way of looking at the transition between children’s small clauses and both adult small clauses and finite clauses, which can moreover embrace the surface differences between Me worry and I worry. Given the nonsentential analysis, continuity of form can still be maintained (even if not continuity of meaning), but in a completely different fashion. For this approach, the continuity is not in the fact that both adults and children only project full finite clauses; rather, the continuity is in the fact that both children and adults use nonfinite clauses/phrases at the root.
. Where do root small clauses and other nonsententials come from? Syntactic research has been very successful in identifying various types of clauses, as well as in proposing insightful analyses for their structure: finite, infinitive, small clause. However, syntacticians rarely ask themselves why it is that human languages avail themselves of all these different clause types. If predicate-argument relationships can be expressed by any of these clauses, why do we have a multitude of options? Why not just one? This section speculates on a possible answer to these questions, which may also help explain why nonsentential speech occurs at all. It is uncontroversial that there are three basic possibilities for embedded clauses, on a continuum from most economical to least economical: small clause (21), infinitive clause (22), and finite clause (23): (21) We consider [nonsententials intriguing]. (22) We consider [nonsententials to be intriguing]. (23) We consider/think [that nonsententials are intriguing].
(AP) (TP?) (CP)
From the synchronic point of view, it is clear why small clauses and infinitives are often used in embedded contexts – they are syntactically more economical than their finite counterparts. However, the real question is the following: why would all three of these options be available to a human language? Let’s assume for a moment, as many syntacticians actually do, that a root clause must be a finite clause. That would mean that a language has to have finite clauses. If so, why not just have every clause be a full finite clause, and have only the option in (23) for embedding? Even though in each particular case the structure of the embedding would be more complex than in the case of small clauses or infinitives, the grammar overall would be much simpler that way, since it would not include a multitude of categories, some of which are rather unclear in status. The differences in meaning among (21) through (23) are not very clear-cut, and the uncertain nature (and analysis) of these meaning differences suggests that they might not have been the driving force behind the development of the multiplicity of embedding options, but more likely its consequence. The situation can again be likened to lexical borrowings. As pointed out in the previous subsection, when English borrowed the word beef from French, it was not because it did not already have a way
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.9 (566-624)
Epilogue
of expressing the notion of cow meat, or because expressing the concept of meat by a separate lexical item was particularly important (cf. chicken, duck, etc.). But once the two words cow and beef were made available, their meanings specialized and adjusted with respect to each other. But if there is no clear semantic factor responsible for the existence of the three clausal types illustrated above, what can be the reason for their existence? It is conceivable that small clauses (possibly also infinitives) are historical relics of ancient root sentential formation processes. Let us speculate for a moment that at one point in the history of language (for concreteness, let us say Pre-IE, also discussed in the previous subsection), the only way to express predication may have been by using a small clause, the smallest structure that can bring together the predicate and its argument(s). This claim is speculative, and it would be difficult to prove or disprove it, but this is also the case with the alternative claim, commonly assumed by syntacticians, that humans always talked, and always talk, in finite sentences. As a reviewer points out, it would be problematic to make this argument for Pre-IE if it is a requirement of UG and of the interfaces that adult root clauses in unmarked registers be tensed. The discussion here addresses the issue under the opposite premise: the premise that it is not a requirement of UG that every root clause be tensed, and that the so-called marked registers are just as naturally accommodated by UG as the unmarked ones, the claim for which several chapters in this volume provide empirical and theoretical arguments. The historical scenario presented here should thus be seen as a possible alternative to the more commonly held view. Keeping the speculative nature of this scenario in mind, consider how the category of Tense might have developed/grammaticalized through history. It has been proposed that the English infinitive particle to was originally a preposition, which was grammaticalized as a tense particle. According to Fischer, van Kemenade, Koopman, and van der Wurff (2000: 96), this grammaticalization in English was only partial, in comparison to German or Dutch counterparts, leaving to still as a category difficult to define. Hyde (2000) analyzes English infinitives as PPs even synchronically, proposing that the italicized element in (24) is a PP small clause. If so, then examples from the headline register, such as (25), may also be analyzed as root small clauses. (For a discussion of these types of examples, as well as of the placement of negation in such infinitive clauses, see Paesani, Chap. 6, this volume.) (24) They expected John to visit China. (25) President to visit China. It is tempting to speculate on the reasons why the preposition to would (partly) grammaticalize as a tense particle, and moreover as a tense particle with a sense of future time. It has been noted repeatedly that embedded infinitive clauses such as (24) refer to a future event with respect to the matrix clause (see, e.g., van Gelderen 1993; Los 2005: 17; Stowell 1982). This future orientation is even clearer with root infinitive clauses such as (25). The question now is why (25) is future oriented, but the small clause in (26), with the preposition in, pertains instead to the present moment.
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.10 (624-695)
Ljiljana Progovac et al.
(26) Class in session. This may be because the preposition in conveys the sense of something already being in a location, and thus the sense of present time; on the other hand, to seems to convey the sense of change of location, and thus a new location in the future. Prepositions can thus have intrinsic temporal values, and it may be that grammaticalization processes exploited these properties on the road toward developing Tense, starting with typical small clauses, and proceeding through intermediate stages such as English to-infinitive clauses.7 Under this (hypothetical) scenario, languages eventually also developed fullfledged finite clauses, whose defining characteristics in English are the use of structural nominative case in the subject position, and finite verb forms (see previous section for the emergence of finiteness and the Indicative in Pre-IE, as discussed in Kiparsky 1968). But, if this is a plausible scenario for the emergence of finiteness, the issue is the following: once finite clauses emerged, did infinitives and small clauses fall out of use? The obvious answer would be “no,” given that both infinitives and small clauses are uncontroversially still used, at least in embedded contexts, across various languages. It is in that sense that the emergence of the finite forms under this scenario would have to be seen as a process of addition, rather than as a process of substitution, again, much like a lexical borrowing such as beef did not replace cow, but was added to the vocabulary. This same process is arguably at work with root small clauses such as Me worry, on their path of development through acquisition and history (see previous subsection).8 If this scenario is on the right track, then it would explain not only why we have several options when it comes to embedding, but also why we sometimes still use small clauses or infinitives in root contexts. We use them regularly in embedded contexts because they are more economical structures, whose temporal properties can be easily determined through their association with the matrix clause. We sometimes use them in root contexts because our ability to communicate using tenseless root objects is still available to us, even though restricted to a great extent, through the emergence of more specialized finite forms. While the scenario for the historical development of clause structure as outlined above is undoubtedly speculative, it is theoretically and empirically plausible. It is often claimed that if human grammars did not, from the very beginning, exhibit the whole array of functional categories that they have today (the well-studied languages, that is), then it would be impossible to explain how they ever acquired such functional categories (similar claims have also been made regarding language acquisition; see previous subsection for some discussion). What we have presented in this section is a possible scenario of how the functional category of Tense could have developed through grammaticalization, and how the stages of these transitions might be felt in the synchronic stage as well, in various forms of nonsentential speech.9 The remaining sections of this Epilogue return to the synchronic aspects of nonsententials.
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.11 (695-745)
Epilogue
. Elliptical versus nonsentential approaches As pointed out in the Introduction, this volume is primarily designed to explore an alternative, nonelliptical approach to the data that we refer to here as nonsententials. Nevertheless, since both Merchant’s (2004, and Chap. 3, this volume) elliptical analysis and the nonsentential analysis deal with question-answer pairs (QAPs), a direct comparison now becomes possible in this area at least, especially given that both analyses are couched within the same basic theoretical framework of Minimalism. The nonsentential analysis as explored in various chapters of this volume extends to a much wider set of data than QAPs, but these data will not be discussed in this section in any detail since there is no well-formulated elliptical analysis of such data. Elliptical accounts of short answers, in general, have some clear advantages over nonelliptical accounts. First, they do not need to invoke special mechanisms for interpretation, other than the ones already in place for full sentences. Thus, an elliptical approach need not concern itself with the difficult question of how the pragmatic context interacts with the syntactic output, the issue also addressed in Section 4.1. Next, the empirical advantages come from so-called connectivity effects, that is, restrictions that seem to apply in the same manner to full sentences and to nonsententials, such as island effects, Case, and preposition stranding (see Barton, Chap. 1, and Merchant, Chap. 3, this volume). On the other hand, the main theoretical advantage of a nonsentential approach is in its economy of syntactic derivation, which becomes especially appealing when one considers, for example, short answers produced by young children or by impaired aphasics. Casielles (Chap. 5, this volume) points out that children produce such short answers (e.g., Me) quite effortlessly from early on, before they in fact produce full sentences (e.g., I ate the cake) or elliptical constructions (e.g., I did), and before they are believed to have acquired structures involving A-bar movement, such as wh-movement. Both A-bar movement and ellipsis would need to be in place for Merchant’s elliptical analysis of QAPs to carry over to the L1 acquisition data as well. On the empirical side, a nonelliptical approach finds support in patterns that occur in nonsentential data, but that cannot be derived from sentential sources in any straightforward fashion, such as the correlation between the use of the default case and the use of nonfinite forms (see Progovac, Chap. 2, this volume). In this section we look more closely at some of these factors that are relevant for distinguishing between the two approaches, in particular, island effects, Case, and preposition stranding, and suggest future prospects for each approach.
. Island effects In deriving QAPs such as (27), Merchant (Chap. 3, this volume) posits full-sentence sources that reduce to fragments, not by sporadic deletions of several unconnected elements that do not form a constituent (28), as was the case in some previous elliptical approaches (see Barton, Chap. 1, this volume), but rather by the movement of the
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.12 (745-811)
Ljiljana Progovac et al.
fragment to the left-peripheral (pre-TP) position, and a subsequent sluicing operation that deletes the remnant TP (29). (27) A: Who did she see? B: John. (28) She saw John. (29) John [TP she saw t]. Merchant’s innovative approach to analyzing short answers is rather elegant, since it utilizes independently motivated syntactic operations.10 Since Merchant’s analysis is a movement analysis, a very important testing ground is island effects (i.e., the postulated syntactic obstacles for movement). Indeed, as noted initially in Morgan (1973), small answers do seem to show island effects: (30) Coordination: A: Did [John and Bill] leave this morning? B: ?No, Harry. (cf. No, John and Harry left this morning.) (31) Subject Island: A: Did [John’s seeing Martha] upset the President? B: ?No, Thelma. (cf. No, John’s seeing Thelma upset the President.) In contrast, it is pointed out in Merchant that, with the intonation rise/accent on the italicized constituent, the short answer in (32) below is grammatical since it does not involve an island. In other words, while the bracketed constituents in (30A) and (31A) constitute islands for movement, the bracketed clause in (32A) below does not – hence the contrast. (32) A: Did Abby think [Ben wrote the letter]? B: No, Charlie. (cf. No, Abby thought Charlie wrote the letter.) However, there are examples demonstrating that short answers do not necessarily obey island effects. Casielles (Chap. 5, this volume) points out that part of the problem with examples such as (30) and (31) is that these yes/no questions do not unambiguously mark the intended contrast: they are yes/no questions, which primarily put focus on truth versus falsity, by their very nature. That is why Merchant needed to accent relevant constituents in questions such as (32A) – to make it possible to answer such questions with the short answer in (32B). But, Casielles points out, if such accent is placed on constituents inside islands, they improve just as much in QAPs: (33) A: Did [John’s seeing Martha] upset the President? B: No, Thelma. (cf. No, John’s seeing Thelma upset the President.)
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.13 (811-875)
Epilogue
Thus, there seems to be no real contrast between examples with islands like (30) and (31) and those without islands like (32): both seem unacceptable with unmarked intonation, and both improve with relevant intonation marking. According to Casielles, rather than island effects, these examples show a focus structure mismatch, which can be partly rescued by prosodically marking narrow focus on the element being questioned. Another way to provide the relevant focus is by creating an echo question, which can also occur within an island. As pointed out by Stainton (Chap. 4, this volume) the derivation of the short answer in the QAP in (34) within Merchant’s move-then-sluice analysis would not be possible given the ungrammaticality of the movement itself (35). (34) A: The Pope likes beer and what? B: Tomato juice. (cf. The Pope likes beer and tomato juice.) (35) *Tomato juice [TP the Pope likes beer and t]. Notice that both echo questions (36) and prosodically marked questions (37) can affect parts of words, suggesting that neither can be derived by syntactic movement: (36) A: The name is Peter De-what? B: Peter De-Gaul. (37) A: Is the name Peter De-Witt? B: No, De-Gaul. It follows that the island effects discussed above do not provide conclusive evidence for movement in short answers. There is another complicating factor in using island effects to argue for sentential derivations. Such an argument can be made conclusively only if there is a good analysis of islands in the first place and, moreover, an analysis that relies on a sentential derivation. There is, however, no good analysis of islands in sentential terms, and certainly no unified analysis: most accounts resort to stipulating which syntactic nodes, and/or which combination of nodes, and/or nodes in which syntactic positions, constitute barriers/obstacles to movement (see, e.g., some classic accounts given in Chomsky 1986; Huang 1982; Lasnik & Saito 1984). Moreover, some of these obstacles are weak and some are strong, and some wh-phrases are not subject to the same island constraints to which other wh-phrases are. To get a glimpse into the extent of the problems in capturing islandhood in purely syntactic terms, in spite of the abundance of attempts, the reader is referred to a recent state-of-the-art article on islands (Szabolcsi & den Dikken 2003: 232), which also quotes a recent interview with Chomsky, in which he says that “there is no really principled account of many island conditions” (for the complete interview, see Belletti & Rizzi 2000). Szabolcsi and den Dikken (2003) conclude that a substantial portion of the island effects is semantic in nature. Of note is also another line of research on islandhood, advocated by Kroch (1989) and Cinque (1990), among others, suggesting that island
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.14 (875-928)
Ljiljana Progovac et al.
effects also have a pragmatic component, having to do with, for example, whether a wh-phrase is D(iscourse)-linked or not, that is, whether it is topic-like, and hence independently licensed in the left periphery of the clause. If nonsentential answers are available, then it seems that a comparison between sentential and nonsentential answers to questions may help tease apart purely sentential/syntactic aspects of islands, from the semantic/pragmatic ones.
. Case cross-linguistically The proponents of the nonsentential analysis in this volume rely heavily on default case (see in particular Progovac, Chap. 2, Section 4.3). The crux of the proposal is that structural nominative case does not get checked in nonsententials; instead nonsententials operate with lexical items with the unmarked/unspecified/default features for Case and Tense. The default case in (modern) English is arguably accusative, and that explains why, for example, subjects of root small clauses (Him worry?!) or isolated NPs (Me!) appear in the accusative case. This unexpected case is not just an idiosyncrasy of English: for example, Irish also uses accusative case as default, and such case is available both with NP/DP answers to questions, and in root small clauses (see Progovac, Section 5, for references and examples). However, accusative as default case seems to be a rather rare phenomenon crosslinguistically, and in fact even English used to have nominative as default. This is not surprising given that nominative was the default case in Proto-Indo-European, as well as in early IE, according to Kiparsky (1968) and references cited therein. A variety of other languages, such as German, Serbian, and Spanish, use nominative as their default case (see in this volume Potts & Roeper, Chap. 7, and Work, Chap. 8, for German; Progovac, Chap. 2, for Serbian; Casielles, Chap. 5, for Spanish). A careful reader of the volume might have also noticed that accusative is not typically or necessarily the form of subject pronouns in other special situations that use root small clauses. Thus, in L2 acquisition, Work cites examples such as She no old from the Basic Variety, and concludes that nominative represents the default in these cases, possibly by transfer from the learners’ native languages. Winford (Chap. 11) offers pidgin examples such as He at home, where nominative seems to be used as the default case. Also, the examples from Urban Guyanese Creole (UGC) from Edwards (Chap. 12, Table 12.2), illustrate the use of nominative pronoun forms (e.g., I sick). Some preliminary independent evidence that UGC might have nominative as a default is suggested by its use also in possessive/genitive contexts, as in I believe they are doing they own thing. Even though the nonsentential analysis explored in Progovac (Chap. 2, this volume) is not specifically designed for QAPs, it extends straightforwardly to these data as well, as argued in Casielles (Chap. 5, this volume). This analysis predicts that NP/DP answers corresponding to subjects in English can only be in the default (accusative) case, unless accompanied by a tense element. Thus, as illustrated in Section 1, an answer to a question (e.g., Who prepared dinner?) can have a nominative subject and a
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.15 (928-996)
Epilogue
finite verb (e.g., I did), or it can have neither (e.g., Me) (see also Barton, Chap. 1, this volume, and references therein). On the other hand, Merchant’s elliptical analysis seems better equipped to explain why it is that in many languages with richer case systems, such as German and Serbian, the NP/DP answers to questions have to be marked with the same case as they would be in full-sentential counterparts (see Merchant, Chap. 3, this volume, and references cited therein). Such connectivity effects are illustrated below for Serbian short and long answers: (38) A: Ko-ga je Ana posetila? who-m did Ana visit B: Ver-u./ *Vera. Vera-acc Vera-nom C: Ana je posetila Veru./*Ana je posetila Vera. ‘Ana visited Vera.’ (39) A: Ko dolazi? who comes B: Vera./ *Veru. Vera-nom/Vera-acc C: Vera dolazi./*Veru dolazi. ‘Vera is coming.’ Such data should constitute a strong argument for an elliptical analysis given that Case is one of those structural phenomena that needs to be checked/assigned in relation to another sentential element. (At least that is what one would expect if these languages worked exactly like English in terms of their case-checking properties.) If, according to the nonsentential analysis, the answers to questions in (38B) and (39B) are not elliptical, but rather base-generated NP/DPs, then the challenge for this analysis is to determine what checks the features of Case in these NPs. We return to this question after we take a closer look at Merchant’s solution for English accusative answers corresponding to subjects. Notice that Merchant’s (2004, and Chap. 3, this volume) move-then-sluice analysis does not extend to QAPs such as (40), given that there is no corresponding sentence such as (41) with a gap/trace in the subject position. Instead, Merchant proposes to derive (40) differently from typical short answers (see (29)), that is, from the underlying left-dislocated structure (42), which may not involve any movement, but which still involves TP sluicing.11 (40) A: Who prepared dinner? B: Me. (41) *I/Me, [TP t prepared dinner]. (42) Me, [TP I prepared dinner.]
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.16 (996-1061)
Ljiljana Progovac et al.
The nonsentential analysis seems to fare better here, since it does not need a stipulation for accusative case in such answers – this is simply the case used in all positions in which structural case is not checked/assigned, including NPs/DPs occurring in syntactic isolation (as in labels or titles), in predicative positions, or in root small clauses (see Progovac, Chap. 2, this volume, Section 4.3, for the variety of uses for default case). At the same time, the nonsentential analysis needs to explain why such default accusative answers are not available in the other languages discussed by Merchant. However, it is not entirely clear that Merchant’s analysis extends seamlessly to these other, better-behaved languages either. His fix for English subject answers in fact has the potential to undermine the argument coming from Case connectivity effects. Notice that the case of the dislocated DP need not coincide with the case of the DP inside the sentence with which it is associated, as is obvious from the English example (42). While this mismatch helps in the case of English QAPs, it is problematic for languages that show connectivity. For example, in German, where the default case is nominative, the dislocated NP surfaces in the nominative case even when associated with the object position (Fromkin et al. 2000): (43) Der Johan, ich habe ihn gesehen. the-nom Johan I have him-acc seen ‘As for Johan, I have seen him.’ However, as Merchant points out, the answers to questions corresponding to direct objects are accusative in German, not nominative, suggesting that left dislocation cannot provide a basis for short answers in German. Still, Merchant’s Case connectivity effects constitute a challenge for a nonsentential analysis. Recall that for Merchant, English is quirky. However, many other languages, particularly those with richer case morphology, behave as predicted by his analysis. In these languages, answers corresponding to subjects will have nominative case. But here is a problem: the theory of structural case checking, on which Merchant’s analysis rests, is designed primarily for English (and languages with similar case systems), but it is exactly in English that it does not work. Merchant’s typological survey of QAPs across a variety of languages can only provide support for an elliptical analysis if case checking works exactly the same way across all of these languages as it does in English. However, even a cursory look at case patterns of one such language, Serbian, shows that the case system is profoundly different, including the nominative-tense relationship, and the verb-accusative relationship, and that in fact an analysis of such patterns within Minimalism does not require an elliptical analysis of short answers. Other languages may be different from both English and Serbian, and an in-depth look at each is necessary to understand these connectivity effects completely. The Minimalist framework allows for more than one option when it comes to grammatical features. If features are Uninterpretable (purely formal/grammatical), such as Case in English is taken to be, then they must be structurally checked and eliminated before they reach interfaces. On the other hand, if features are Interpretable, then they need not be checked, and they certainly should not be eliminated since
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.17 (1061-1120)
Epilogue
they will need to be interpreted. One example of an Interpretable feature would be the gender feature on a pronoun (he vs. she). So, it is possible to make an argument that Case in Serbian, at least non-nominative case, involves Interpretable features (see below). On the other hand, nominative cannot be analyzed as involving Interpretable features, given that it can range over a variety of thematic roles, including agents, experiencers, and themes, which do not seem to have anything in common when it comes to interpretation. But if nominative in Serbian is structural case, which moreover has Uninterpretable features that have to be checked in TP, then one would indeed expect not to be able to have nominative answers to questions without a full-sentence projection, that is, without a TP. As pointed out above, nominative is also the default case in Serbian. Thus, nominative answers to questions can be analyzed as involving default nominative case, rather than structural case, regardless of what the ultimate analysis of nominative case in Serbian full sentences turns out to be (note that, unlike in English, nominative case in Serbian need not be adjacent to tense, nor does it need to surface at all in tensed clauses, given that Serbian is a pro-drop language). If so, then nominative NP/DP answers will not force a full-sentential projection in Serbian. This approach predicts the difference between English and Serbian in this respect: NP/DP answers to questions pertaining to subjects are accusative in English because default case in English is indistinguishable from accusative; such answers to questions are nominative in Serbian, because default case is nominative in Serbian (also in German, as discussed above), as can be established on independent grounds. Thus, neither English nor Serbian (nor German) is quirky under this analysis. We return now to the possibility of analyzing other cases in Serbian as involving Interpretable features. Dative objects are typically associated with the theta-role of the goal/recipient, and it is conceivable that dative has Interpretable features, rather than Uninterpretable formal features. Serbian also has distinct morphological marking for cases associated with instruments (instrumental) and location (locative), which are traditionally considered “inherent” cases, that is, cases directly related to a thematic role, and thus not subject to structural case checking/elimination. In the context of this system, the accusative case in Serbian is much more meaningful/informative than it is in English, and it may be considered to have Interpretable features as well. An accusative case knows that it is not an agent, not a goal/recipient, not an instrument, not a locative – it is typically in fact a theme/patient (except for the cases where it is governed by a preposition, to which we return below). In the light of the fact that accusative also need not occur in a fixed syntactic position in Serbian (the way accusative objects in English have to be adjacent to their verbs), it may indeed be that accusative has Interpretable features that are not in need of structural checking and elimination. In addition to its association with the theme/patient theta-role, accusative in Serbian also has a quantificational feature (arguably, that of universal quantification), in that it is interpreted as affecting the whole object, which is most obvious in its interaction with aspect (see Progovac 2002, 2005). This is evident from the following contrast in Serbian, which shows that one and the same verb can govern either accusative or
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.18 (1120-1175)
Ljiljana Progovac et al.
genitive case, the former expressing whole-object affectedness, and the latter exhibiting only existential quantification: (44) a.
Dodaj vodu. add water-acc ‘Add (all) the water.’ b. Dodaj vode. add water-gen ‘Add (some) water.’
Notice, however, that this unmarked, typical interpretation of accusative is overridden when accusative is governed by a preposition: (45) Seo je na stolicu. sat-m.sg is on chair-acc ‘He sat on a chair.’ (46) Ljuti se na Milicu. angers self on Milica-acc ‘He is angry at Milica.’ Here, it seems that the relationship between the preposition and its DP complement is purely formal, arbitrary. The accusative here is interpreted neither as affecting the whole object, nor as being a theme/patient argument necessarily. This may be exactly the reason why prepositions have to be adjacent to their complements in Serbian, providing tangible evidence for a formal checking relationship between the two. Prepositions in Serbian impose fixed word order, that is, adjacency, on their complements. As discussed in the following section, this may also be the reason why short answers to questions cannot consist of just complements to prepositions. But when it comes to case not bound by prepositions, if accusative (as well as other oblique cases) in Serbian can be analyzed as involving Interpretable features, then it can occur in syntactic isolation without causing a crash. However, if nominative is a default case in Serbian, one may still wonder why nominative cannot serve as an answer corresponding to a nonsubject: (47) A: Koga je Ana posetila? who-acc is Ana visited? ‘Who did Ana visit?’ B: Vera! In fact, nominative responses such as the one in (47) are not unacceptable – they simply do not serve as direct answers to questions. Such nominative responses would be interpreted as avoiding the answer, possibly by virtue of not being maximally informative (in the sense of Grice 1975). In other words, answering in the nominative, when a more informative, accusative response is available, signals that Speaker B is avoiding answering the question. His response is then interpreted as a comment that has something to do with the question. Notice that when it comes to English accusative
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.19 (1175-1236)
Epilogue
answers pertaining to subjects, accusative cannot be interpreted as avoiding a nominative answer, given that a nominative answer is unavailable in the first place, for syntactic reasons. Needless to say, this issue deserves further attention (for some discussion regarding Grice’s Cooperative Principle, as well as avoidance strategies in English, see Barton, Chap. 1, this volume). This discussion barely scratches the surface of the complexities of Case and its interpretation in Serbian, as well as in English. Even so, the discussion should suffice to show that Case considerations can be profoundly different from language to language, and that each language should be studied in its own right, before one can conclude if short answers to questions require a sentential analysis or not. More generally, this section has shown that the theory in fact gives options when it comes to Case, and that certain options allow case-marked nonsententials. Nonetheless, connectivity effects will continue to provide fruitful testing ground not only for elliptical versus nonsentential analyses, but also for the theoretical analysis of Case in general.
. Preposition stranding Merchant (2004, and Chap. 3, this volume) provides yet another connectivity argument for an elliptical analysis: the cross-linguistic correlation between the ability to strand prepositions in movement operations and to have answers without prepositions, as illustrated below for two language types, English (48), which allows preposition stranding, and Serbian (49), which does not (see also Casielles, Chap. 5, this volume, for some discussion regarding Spanish): (48) A: To whom did John speak?/Who did John speak to? B: To Emma./Emma. ˇ (49) A: Na ˇcega je Stefan seo? / *Cega je Stefan seo na? on what did Stefan sit B: Na stolicu. / *Stolicu. on chair-acc / chair-acc This pattern follows nicely from Mechant’s move-then-sluice analysis: only a constituent that can independently move to the left periphery can serve as an answer to a question. What might be a way to explain such connectivity effects under a nonsentential approach? Descriptively speaking, languages that do not allow preposition stranding, such as Serbian, also do not allow complements of prepositions to be separated from their prepositions in any other way; in other words, there is an adjacency requirement. This generalization would equally affect sentential and nonsentential derivations. What is lacking so far, for both an elliptical and a nonsentential analysis, is the theoretical reason for this correlation. For Merchant’s analysis, the question is why English can strand its prepositions in wh-questions, but Serbian cannot. For a nonsentential analysis, the question is why there would be an adjacency requirement on
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.20 (1236-1300)
Ljiljana Progovac et al.
prepositions in Serbian, but not in English. This is in fact the same question, only posed from two different perspectives. Having an answer to this question is crucial for distinguishing between the two analyses. If the answer turns out to affect only a subsentential part, such as the PP in question, then a nonsentential analysis is still possible; on the other hand, if the answer necessarily invokes the full-sentence projection, TP, then the nonsentential analysis cannot be right. It was pointed out in the previous section that in Serbian, direct complements of the verb (not embedded in PPs), including those marked with accusative case, can be analyzed as carrying Interpretable case features, which do not get checked and eliminated. In contrast, the case on the complement of a preposition in Serbian (accusative or other) is purely formal, determined in an arbitrary fashion by the preposition itself, and thus must be checked and eliminated. Accusative case in Serbian arguably has two meaningful (Interpretable) properties: first, it is typically associated with the theme/patient thematic role (in opposition to dative, instrumental, and locative cases) and, second, it involves (universal) quantification, which results in a whole-object affectedness interpretation (in contrast to genitive case, which involves existential interpretation). In this kind of case system, if a bare accusative NP/DP stolicu (‘chair’) were to be used as an answer to a question such as On what did Stefan sit?, it would wrongly contribute semantic features that are incompatible with the intended meaning, namely, it would wrongly imply that the chair is a theme/patient argument that is completely/thoroughly affected by the action. The use of the governing preposition may then be required in short answers precisely to override this default interpretation of accusative case. But, crucially, nothing above PP need project in this scenario. The reason why English accusative answers are grammatical even without governing prepositions may have to do with one basic difference between English and Serbian: whereas Serbian accusative case, when used in isolation, necessarily carries Interpretable features, English accusative never does, and it can be the spell-out of default case (see previous section for details).12 Needless to say, even if this idea is on the right track, it would require a much more detailed analysis, not only for Serbian and English, but also for the other languages that show connectivity effects. Nonetheless, the fact that there are matches between sentential and short answers does not in itself constitute an argument for a sentential derivation, in the sense that correlation does not necessarily imply causation. A third factor, and moreover a subsentential one, may be causing the patterns in both. We see this as an important future challenge for an elliptical approach: to provide evidence that the connectivity effects in general cannot reduce to subsentential factors. At the same time, Merchant’s analysis, as well as previous elliptical approaches, has identified an important challenge for a nonsentential analysis: to determine that there is in fact a common third factor, and moreover a subsentential factor, that causes the patterns of connectivity across sentential and nonsentential speech. We believe that the resolution of this issue, as well as of other connectivity issues, will significantly further our understanding of not only the syntax of nonsententials, but also full-sentence grammars.
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.21 (1300-1355)
Epilogue
. Some remaining issues and future prospects . The role of (pragmatic) context As pointed out in Section 3, a nonsentential analysis needs to explain how syntactically well-formed, but semantically incomplete, nonsentential utterances utilize context (for some general discussion, see Barton, Chap. 1, and Progovac, Chap. 2, this volume). Root small clauses are often used in the sense of labels, as on classroom doors or photos, suggesting that the pragmatic context of each situation contributes to the interpretation of the small clause: (50) a. Class in session. b. Me in Rome. c. Me dancing. Small clauses can also be used against the background of a linguistic context, as would be the case for example with answers to questions (51) (see Casielles, Chap. 5, this volume), or with embedded small clauses (52): (51) A: What scares him? B: Me dancing. (52) You never saw [me dancing]. Notice that infinitive clauses can likewise either be anchored in the linguistic context as embedded clauses (53), or they can be pragmatically anchored in root contexts, as is arguably the case with the headline register (54), as discussed in Paesani (Chap. 6, this volume). (53) They expected [John to visit China]. (54) President to visit China. One important spatio-temporal frame for anchoring a small clause is that of the immediate present, or the “here and now”’ (e.g., Class in session). Children’s overwhelming reliance on the here and now in their early speech may partly explain the preponderance of nonsentential data, before, but even after, they have acquired Tense. The use of nonsententials in pidgin languages might also be attributed to limited communicative needs, mostly having to do with the here and now (see Winford, Chap. 11, this volume). But even full-fledged languages in fact make use of this pragmatic commodity. Hebrew, for example, regularly omits the present-tense copula with what seem to be stage-level predicates, that is, predicates that are anchored in the present moment. Rothstein (1995) analyzes such Hebrew sentences as root small clauses (for details, as well as for root small clauses in Irish and Russian, see Progovac, Chap. 2, this volume). In addition, Edwards (Chap. 12, this volume) presents arguments in favor of a root small clause analysis of copulaless data in African American Vernacular English, as well as in Guyanese Creoles.
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.22 (1355-1416)
Ljiljana Progovac et al.
Moreover, Potts and Roeper (Chap. 7, this volume) argue that expressive small clauses of the type You fool/You idiot also express “a momentary attitude linked to a situation.” In fact, they say that the most prominent feature of expressive content is its strong tie to the utterance situation. This is in contrast to what would be sentential counterparts in (55), which constitute more permanent characterizations. (55) You are a fool/an idiot. For Potts and Roeper, this semantic difference precludes the possibility for an elliptical analysis of such expressive clauses (this is also mentioned in Merchant, Chap. 3, this volume). The point of contact between the output of syntactic theory and the input of pragmatic context needs to be studied in much more depth and detail than this volume has attempted. Some promising recent research in this respect is Avrutin (1999, 2004), who makes an elaborate argument to the effect that functional categories, including Tense, are not the only way to anchor a sentence temporally: Discourse context can provide such an anchor as well. This line of research, especially needed for nonsententials, promises to develop a modular account of meaning, that is, one that takes into account, in a precise and systematic fashion, not only the elements of structure, but also the elements of context.
. Word order in nonsentential utterances The data presented in the chapters of this volume point to word order as another important topic for future research, given that word order even in nonsententials may vary across languages and contexts. Although word order in root small clauses normally mimics that of embedded small clauses in that the subject precedes the predicate, intriguingly, we have also found the opposite ordering in adult informal conversation, as in (56) and (57), and in L2 acquisition, as in (58). (56) Nice lady, your mom.13 (57) Silly me/Silly Chris. (58) Schön diese Haus. ‘Beautiful that house.’ This pattern is also attested in other languages, as the following Serbian example shows: (59) Pametna žena, tvoja komšinica.14 ‘Smart woman, your neighbor.’ It is well known that in languages with so-called free word order like Spanish, where word order variation can express a different packaging of information, both orders (subject-predicate and predicate-subject) are possible in regular, embedded small clauses:
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.23 (1416-1486)
Epilogue
(60) Considero a Juan inteligente. consider-1sg part Juan intelligent ‘I consider Juan intelligent.’ (61) Considero inteligente a Juan. This different word order correlates with a different interpretation. Whereas in (60) the predicate is the focal element, in (61) the focal element would be the subject of the small clause, namely, Juan. So, (60) would be a good answer to a question such as What do you think of John?, while (61) would answer a question such as Who do you consider intelligent? Thus, in these languages, unless a particular element has been focalized either prosodically or by a specific focalizing structure such as a cleft or a pseudocleft, the order of elements in an utterance will reflect the information order topic-comment. It will place toward the beginning of the clause those elements considered to be topical, and in sentence-final position those that are the most informative or focal. As pointed out by Winford (Chap. 11, this volume), pidgins tend to follow this topic-comment information order as well. (62) A: Who came to Honolulu? B: Keim to Honolulu, isa fai familii. came to Honolulu five families ‘FIVE FAMILIES came to Honolulu.’ What is interesting about the utterances in (56) through (58), however, is that they seem to follow the opposite order, focus-background, in which the most informative element seems to be the predicate: nice lady, silly, or beautiful. As is well known, at least since the pioneer work of Weil (1844/1879) and Mathesius (1961/1975), in addition to the order topic-comment, speakers can start their utterances with the most important element and produce focus-background structures. Weil refers to this order as the pathetic order and says, “When the imagination is vividly impressed, or when the sensibilities of the soul are stirred, the speaker enters the matter of his discourse at the goal” (45). Similarly to Weil, Mathesius points out that this is the type of arrangement used in agitation: “First the speaker impatiently states the new element of the intended statement and only afterwards adds the known elements from which he actually starts” (83–84).15 Syntactic structures with this focus-background order include prosodically prominent preverbal subjects, as in (63), so-called Focus Preposing structures, where a nonsubject is prosodically focalized and placed in sentence-initial position, as in (64), as well as constructions involving the right-dislocation of nonfocal elements, as in (65): (63) JOHN ate the cake. (64) THIRTY POUNDS it weighs. (65) She is a nice lady, your mom. From this perspective, (56) (NICE LADY, your mom) might be the nonsentential counterpart of the right-dislocated structure in (65).
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.24 (1486-1549)
Ljiljana Progovac et al.
Although these examples deserve much more attention than we can provide here, they could be observing a focus-background information structure. If so, a more detailed analysis of constructions such as these might in fact shed some light on the nature of right-dislocation.
. Parataxis and embedding The chapters in this volume seem to converge on the conclusion that nonsententials do not embed. For instance, Potts and Roeper (Chap. 7, this volume) argue that expressive small clauses in adult native speech, such as Me worry or You idiot, cannot embed. They in fact develop a semantic analysis for expressive small clauses that predicts this salient property. They call such unembeddable clauses, exemplified in (66), “syntactic isolates.” (66) *I consider you fool. The tendency to avoid embedding is also evident outside of the adult native context. Kolk (Chap. 9, this volume) and Siple (Chap. 10, this volume) note that the speech of agrammatic aphasiacs includes very little subordination. Likewise, Winford (Chap. 11, this volume) points out that embedding is virtually nonexistent in pidgins and that instead parataxis (i.e., juxtaposition, loose joining) is used. The following examples from Korean Pidgin English (67) and Hawaiian Pidgin English (68) illustrate: (67) Aena tu macha churen, samawl churen, haus mani pei. beri and too much children small children house money pay very haad taim, no moa moni . . . pua. hard time no more money poor ‘And I [had] too many children, small children, I [had] to pay the rent. [It was a] very hard time, I had no money, [we were] poor.’ (68) No mani, no kom. ‘No money, no come.’ Work (Chap. 8, this volume) also reports that subordination does not occur in early varieties of L2 acquisition, and Potts and Roeper (Chap. 7, this volume) make the same claim for L1 acquisition (see also Stainton, Chap. 4, this volume, Section 3.1). A reviewer suggests that the claim that there is no subordination in these varieties is problematic, given that subordination is an instance of recursion, an important characteristic of human language. The reviewer further suggests that a grammar that disallows subordination is a more complex grammar because it would have to have a special stipulation that blocks subordination, given that UG allows it. However, as was pointed out in Section 2 regarding Tense, this is not the only way to look at the phenomenon of embedding in nonsententials. Given that languages typically employ complementizers/subordinators in embedded clauses, it is plausible that these special vocabulary items, or the functional projections with which they are associated (e.g.,
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.25 (1549-1622)
Epilogue
CP), are required for subordination. If so, then even if UG allows subordination in principle, subordination will not occur unless the language (or the speaker) has specialized words or morphemes that can build the relevant functional projection, and thus facilitate subordination. In certain speakers, such as agrammatic aphasics or children acquiring their L1, there may be a discrepancy between production and comprehension that might explain variability with respect to embedding. For instance, although complementizers/subordinators might be understood as signaling embedding in comprehension, unless they become part of a speaker’s active vocabulary or grammar, a speaker might not be able to use one of these function words to build the projection of CP in production. If such variation is indeed linked to the ability to understand and use specific words, the difference between production and comprehension might be related to the difference between active and passive vocabulary. Alternatively, vocabulary size and complexity, or the abstractness of the function words necessary to produce embedded structures, may cause difficulties in production (see, e.g., Bates & Goodman 1997; Bird, Franklin, & Howard 2002). Needless to say, this issue needs further research. Although embedding is rare or unattested in some nonsentential contexts, it seems to occur in others, though in a restricted form. As Paesani (Chap. 6, this volume) points out, embedded structures are attested in the diary (69), note-taking (70), and telegraphese registers (71). (69) Wonder if ever make it so? (70) What did was take “sha” and. . . (71) Regret cannot attend. Embedded structures are also allowed in headlines, as illustrated in (72). (72) China suggests U.S. examine its own human rights record. (Detroit Free Press, March 4, 2005) The headline register, however, often uses paratactic structures. These structures include juxtaposed phrases (73), and the use of colons to report additional information about a person or event (74) or to report direct or indirect speech (75). In these cases, parataxis appears to be a strategy used to avoid embedding. (73) Forces take Uzbek town, capture rebel.
(Detroit Free Press, May 20, 2005)
(74) U.S.-Mideast relations: Peace and protests. (Detroit Free Press, May 23, 2005) (75) Woman: Jackson plotted smear.
(Detroit Free Press, March 4, 2005)
In cases such as (72), the embedded structure may simply be an instance of the sentential grammar being chosen over the nonsentential one, suggesting that embedding is in fact not a feature of the nonsentential grammar. If this is the case, the question arises as to how to analyze examples such as (76), where the embedded clause has an overt subject, but the matrix clause lacks both a
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.26 (1622-1685)
Ljiljana Progovac et al.
subject and a verb. Notice that (77) and (78), in which the embedded small clause exhibits the absence of tense but not the absence of a subject, are not acceptable. (76) Very glad [you are here]. (77) *Very glad [you here]. (78) *I am very glad [you here]. Small clauses do occur in embedded contexts, but only in well-defined positions, where the subject of the small clause can receive so-called Exceptional Case Marking (ECM) from the preceding verb, as in He wants [you here]. In this respect, root small clauses are different from embedded small clauses in that the latter check structural Case, whereas the former do not (see Progovac, Chap. 2, this volume, Section 2). Even though the adjective glad, not being transitive, cannot check (ECM) Case on you in (77) and (78), the question is why the subject of the small clause you cannot receive default accusative case in such instances, and thus render the examples grammatical. As pointed out above, it may well be that any kind of embedding, even of a small clause, requires some functional projection, or some structural relationship, such as case checking, to facilitate the embedding (see also Progovac, fn. 10). If so, it would follow that small clauses characterized by default case cannot embed. Notice that contrasts like the one in (76) through (78) would be just as problematic if one were to adopt an elliptical analysis. If the deletion of the copula is in principle possible, then it is not clear what would exclude it in (77) or (78). On the other hand, one would expect structures of the type in (76) to be possible, where the embedded clause is sentential, but the matrix clause is either sentential or nonsentential. The data presented in Paesani (Chap. 6, this volume), in which subjects are deleted from the embedded clause, appear problematic for this prediction, however. This phenomenon is particularly prevalent in the diary register. (79) Think will cross that last bit out.
(Fielding 1996: 25)
Haegeman (2002) and Haegeman and Ihsane (1999) point out that there is a correlation between the absence of the matrix and embedded subjects in such embedded contexts. According to their corpus of null subjects in the diary register, structures such as (79), in which both subjects are null, are the most common, followed by structures in which the matrix subject is null and the embedded subject is overt (80). Cases in which the matrix subject is overt but the embedded subject is null are unattested (81). (80) Think I will cross that last bit out. (81) *I think will cross that last bit out. These data suggest that there is some kind of correlation or dependency between the absence of a lower subject and that of a higher one. A similar correlation exists with respect to the absence of determiners in subject position in main and embedded clauses, as illustrated in the headline data in (82) through (84) (from Stowell 1991):
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.27 (1685-1767)
Epilogue
(82) Bronx man claims policeman shot a bystander. (83) Bronx man claims a policeman shot a bystander. (84) *A Bronx man claims policeman shot a bystander. It is unclear at this point how either the nonsentential analysis or an elliptical analysis might account for the data in (79) through (84), and it is a question we leave open for future research. While there is no doubt that the use of embedding is a very useful tool in human languages, alternatives to embedding are nonetheless a prevalent and important feature. Indeed, even deep wisdoms can be expressed by mere nonsententials and parataxis:16 (85) No pain, no gain. (86) Easy come, easy go. (87) Nothing ventured, nothing gained.
Notes * In conceptualizing the Epilogue, we were largely inspired by its counterpart in DeGraff (1999). . See also Newmeyer (2003), where it is suggested that speakers mentally represent fullsentential structures even when they produce nonsentential utterances. . This is analyzed in Progovac (Chap. 2, this volume) as default case (see also Section 3.2). . For a recent multidisciplinary study of variation and the universal properties of language, see for example Jenkins (2004). . See the contributions in DeGraff (1999). . The reader is also referred to Radford (1990) for an elaborate discussion of L1 acquisition of English, where typical two-or-more-word utterances by young children are analyzed as small clauses (see also Lebeaux 1988). For a good overview of alternative views on language acquisition, which assume full functional structure in children’s early speech, see Guasti (2002) and references therein. . A host of loan words was borrowed from French after the Norman Conquest and among them are words for table meat: beef, mutton, veal, and pork (see, e.g., Pyles & Algeo 1993: 295). . A reviewer points out that in Quebec French and certain 17th-century dialects of French, for example, the preposition après (‘after’) (e.g., Il viendra après diner ‘He will come after dinner’) is also used in verbal periphrastic constructions, not to express future, but to express the progressive (e.g., Il est après manger ‘He’s eating’). We do not have much to say about this particular grammaticalization process, which seems unrelated to the intrinsic meaning of the preposition. Nonetheless, prepositions are often reanalyzed into functional categories by grammaticalization processes, including those operative in creolization (see, e.g., the contributions in DeGraff 1999). . It is worth pointing out that it does not follow from our discussion here that the brains of humans who would speak in small clauses, say Pre-IE humans, would have to be fundamentally
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.28 (1767-1831)
Ljiljana Progovac et al.
different from the brains of humans who speak in tensed sentences, just as it does not follow from current syntactic theory that the brains of speakers of languages with grammaticalized aspect (e.g., Slavic), or of languages with an elaborate system of focus marking (e.g., Hungarian), are fundamentally different from the brains of speakers of languages without such categories or projections. As is in fact commonly assumed by syntacticians, the variation among even presentday languages is determined by the nature and number of functional projections (see, e.g., Borer 1984), and there is no reason to believe that such variation could not have existed historically (across time) as well. In this respect, Kiparsky (1968) makes a convincing argument that Pre-IE did not have grammaticalized tense, and that it expressed time optionally with, for example, temporal adverbials (for further details, see Progovac, Chap. 2, this volume; also Section 2.1). . The discussion in this section may also have implications for the evolutionary development of these categories, but since little is understood or agreed on in the studies of language evolution, we only refer the reader to the most recent views on the topic, including Jackendoff (2002), Bickerton (in press), and Pinker and Jackendoff (2005), as well as references cited therein. . A reviewer points out that the accusative answers corresponding to subjects can be derived from clefts by deleting the crossed-out material: (i)
It was me who ate the pie.
Such derivations from clefts may reintroduce unconstrained elliptical mechanisms, precisely the ones that Merchant’s move-then-sluice analysis seeks to eliminate (see Barton, Chap. 1, this volume). For recent approaches to clefts and pseudoclefts, as well as for semantic reasons why not to derive short answers from them, the reader is referred to Casielles (Chap. 5, this volume). . Incidentally, if left-dislocation can serve as a basis for short answers to questions, then the question arises as to why all short answers cannot be so derived, including answers corresponding to objects. Notice that in that case one may not expect to find evidence for movement (e.g., island effects; see previous section), given that left-dislocation is typically analyzed as not derived by movement. . Presumably, it would also follow from a combination of these two factors (that bare accusative is Interpretable in Serbian, and that prepositions check Uninterpretable features on their complements, including accusative) that prepositions cannot be stranded. . Notice that the article is missing from the first NP in (56), suggesting that the nonsentential grammar option is involved. Also, if the subject were a pronoun it would appear in the default accusative case: Nice lady, her (see Progovac, Chap. 2, this volume, for article omission and default case as a characteristic of nonsentential grammars). . Interestingly, the regular order subject-predicate in both English and Serbian can only have the incredulity reading of Mad Magazine sentences as in (i): (i)
Your Mom a nice lady?!
. As mentioned in Casielles (Chap. 5, this volume) the order focus-background has also been found in early child language (see Bates 1976; Bowerman 1973; Leonard & Schwartz 1978; Sinclair de Zwart 1973, for early studies). As Bates (1976) notices for Italian, sometimes the child blurts out the new information first and then the rest in decreasing order of interest. . Notice the striking similarity between the example in (85), and the pidgin example in (68).
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.29 (1831-1935)
Epilogue
References Avrutin, S. (1999). Development of the Syntax-Discourse Interface. Dordrecht: Kluwer. Avrutin, S. (2004). Optionality in child and aphasic speech. Lingue e Linguaggio, 1, 67–97. Bates, E. (1976). Language and Context: The Acquisition of Pragmatics. New York, NY: Academic Press. Bates, E. & Goodman, J. C. (1997). On the inseparability of grammar and the lexicon: Evidence from acquisition, aphasia and real-time processing. Language and Cognitive Processes, 12, 507–584. Belletti, A. & Rizzi, L. (2000). An Interview on Minimalism, With Noam Chomsky. Conducted at the University of Siena, November 8–9, 1999; revised March 16, 2000; available at: http://www.media.unisi.it/ciscl/pubblicazioni.htm Bickerton, D. (in press). Language evolution: A brief guide for linguists. Lingua. (Online publication 2005, doi:10.1016/j.lingua.2005.02.006) Bird, H., Franklin, S., & Howard, D. (2002). ‘Little words’ – not really: Function and content words in normal and aphasic speech. Journal of Neurolinguistics, 15, 209–237. Borer, H. (1984). Parametric Syntax: Case Studies in Semitic and Romance Languages. Dordrecht: Foris. Bowerman, M. (1973). Early Syntactic Development: A Cross-Linguistic Study With Special Reference to Finnish. Cambridge: CUP. Chomsky, N. (1986). Barriers. Cambridge, MA: The MIT Press. Chomsky, N. (1995). The Minimalist Program. Cambridge, MA: The MIT Press. Cinque, G. (1990). Types of A’-Dependencies. Cambridge, MA: The MIT Press. Culicover, P. & Jackendoff, R. (2005). Simpler Syntax. New York, NY: OUP. DeGraff, M. (Ed.). (1999). Language Creation and Language Change: Creolization, Diachrony, and Development. Cambridge, MA: The MIT Press. Fielding, H. (1996). Bridget Jones’s Diary. New York, NY: Picador. Fischer, O., van Kemenade, A., Koopman, W., & van der Wurff, W. (2000). The Syntax of Early English [Cambridge Syntax Guides]. Cambridge: CUP. Fromkin, V. A. et al. (Eds.). (2000). Linguistics: An Introduction to Linguistic Theory. Oxford: Blackwell. Ginzburg, J. & Sag, I. (2000). Interrogative Investigations: The Form, Meaning, and Use of English Interrogatives. Stanford, CA: CSLI. Grice, H. P. (1975). Logic and conversation. In P. Cole & J. L. Morgan (Eds.), Syntax and Semantics: Speech Acts, Volume 3 (pp. 41–58). New York, NY: Academic Press. Grimshaw, J. (1994). Minimal projection and clause structure. In B. Lust, M. Suñer, & J. Whitman (Eds.), Heads, Projections and Learnability: Syntactic Theory and First Language Acquisition: Cross-Linguistic Perspectives, Volume 1 (pp. 75–83). Hillsdale, NJ: Lawrence Erlbaum. Guasti, M. T. (2002). Language Acquisition: The Growth of Grammar. Cambridge, MA: The MIT Press. Haegeman, L. (2002). Non-overt subject pronouns in written English. In S. Scholz, M. Klages, E. Hantson, & U. Römer (Eds.), Language, Context and Cognition: Papers in Honour of Wolf Dietrich Bald’s 60th Birthday (pp. 135–149). Munich: Langenscheidt, Longman. Haegeman, L. & Ihsane, T. (1999). Subject ellipsis in embedded clauses in English. English Language and Linguistics, 3, 117–145.
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.30 (1935-2050)
Ljiljana Progovac et al.
Herschensohn, J. (2000). The Second Time Around: Minimalism and L2 Acquisition. Amsterdam: John Benjamins. Huang, J. (1982). Logical Relations in Chinese and the Theory of Grammar. Ph.D. dissertation, Massachusetts Institute of Technology. Hyde, B. (2000). The structures of the to-infinitive. Lingua, 110, 27–58. Jackendoff, R. (2002). Foundations of Language: Brain, Meaning, Grammar, Evolution. Oxford: OUP. Jenkins, L. (Ed.). (2004). Variation and Universals in Biolinguistics. Amsterdam: Elsevier. Kiparsky, P. (1968). Tense and mood in Indo-European syntax. Foundations of Language, 4, 30– 57. Kroch, A. (1989). Amount Quantification, Referentiality, and Long Wh-Movement. Manuscript, University of Pennsylvania. Lasnik, H. & Saito, M. (1984). On the nature of proper government. Linguistic Inquiry, 15, 235– 289. Lebeaux, D. (1988). Language Acquisition and the Form of Grammar. Ph.D. dissertation, University of Massachusetts, Amherst. Leonard, L. B. & Schwartz, R. G. (1978). Focus characteristics of single-word utterances after syntax. Journal of Child Language, 5, 151–158. Los, B. (2005). The Rise of the To-Infinitive. Oxford: OUP. Mathesius, V. (1975). A Functional Analysis of Present Day English on a General Linguistic Basis. Transl. by L. Dušková, ed. by J. Vachek. Prague: Academia. (Originally published 1961 in Czech under the title Obsahový rozbor souˇcasné angliˇctiny na základˇe obecnˇe lingvistickém.) Merchant, J. (2004). Fragments and ellipsis. Linguistics and Philosophy, 27, 661–738. Morgan, J. (1973). Sentence fragments and the notion ‘Sentence.’ In B. Kachru, R. Lees, Y. Malkiel, A. Pietrangeli, & S. Saporta (Eds.), Issues in Linguistics: Papers in Honor of Henry and Renee Kahane (pp. 719–751). Urbana, IL: University of Illinois Press. Newmeyer, F. (2003). Grammar is grammar and usage is usage. Language, 79, 682–707. Pinker, S. & Jackendoff, R. (2005). The faculty of language: What’s special about it? Cognition, 95, 201–236. Progovac, L. (2002). Aspect Phrases, perfective prefixes and congruent Prepositional Phrases in Serbian. In J. Toman (Ed.), Formal Approaches to Slavic Linguistics 10: The Second Ann Arbor Meeting 2001 (pp. 201–221). Ann Arbor, MI: Michigan Slavic Publications. Progovac, L. (2005). A Syntax of Serbian: Clausal Architecture. Bloomington, IN: Slavica. Pyles, T. & Algeo, J. (1993). The Origins and Development of the English Language. New York, NY: Harcourt Brace Jovanovich. Radford, A. (1990). Syntactic Theory and the Acquisition of English Syntax. Oxford: Blackwell. Roeper, T. (1999). Universal bilingualism. Bilingualism: Language and Cognition, 2(3), 169–186. Rothstein, S. (1995). Small clauses and copular constructions. In A. Cardinaletti & M. T. Guasti (Eds.), Syntax and Semantics 28: Small Clauses (pp. 27–48). San Diego, CA: Academic Press. Safir, K. (1993). Perception, selection and structural economy. Natural Language Semantics, 2, 47–70. Sinclair de Zwart, H. (1973). Language acquisition and cognitive development. In T. Moore (Ed.), Cognitive Development and the Acquisition of Language (pp. 9–25). New York, NY: Academic Press. Stowell, T. (1982). The tense of infinitives. Linguistic Inquiry, 13, 561–570. Stowell, T. (1991). Empty Heads in Abbreviated English. Paper presented at the GLOW conference, Leiden.
JB[v.20020404] Prn:20/07/2006; 15:07
F: LA93EP.tex / p.31 (2050-2066)
Epilogue
Szabolcsi, A. & den Dikken, M. (2003). Islands. In L. Cheng & R. Sybesma (Eds.), The Second Glot International State of the Article Book: The Latest in Linguistics [Studies in Generative Grammar 61] (pp. 213–240). Berlin: de Gruyter. van Gelderen, E. (1993). The Rise of Functional Categories. Amsterdam: John Benjamins. Weil, H. (1879). De l’ordre des mots dans les langues anciennes comparéees aux langues modernes. Paris: Joubert. (Translated as The Order of Words in the Ancient Languages compared with that of the Modern Languages, Boston: Ginn, 1887 by Charles W. Super. Revised edition, 1978, Amsterdam/Philadelphia: John Benjamins.)
JB[v.20020404] Prn:16/08/2006; 14:03
F: LA93INL.tex / p.1 (47-152)
Index of languages
A AAVE (see African American English Vernacular) African American English Vernacular (AAVE) (see also Black English Vernacular (BEV)) , , , –, –, Afrikaans , American Indian (see also Native American) , Arabic (see also Moroccan Arabic) , , , , Australian , B BEV (see Black English Vernacular) Bislama Black English Vernacular (BEV) (see also African American Vernacular English (AAVE)) Brazilian Portuguese (see also Portuguese) Bulgarian , C Canadian French (see also French; Quebec French) Catalan Caucasian Chinese , , , , , Chinese Pidgin English Chinook Pidgin , Tsimshian variety Czech , D Danish Delaware Pidgin
Deutsch (see German) Dutch , , , , , , , , –, , , , , , , , , , , , , , , , , , ,
E Early South Pacific Pidgin Eskimo , , , Eskimo Pidgin , ,
F Filipino Pidgin English (FPE) (see Hawai’i/Hawaiian Pidgin English) Finnish , , , FPE (see Filipino Pidgin English) French (see also Canadian French; Quebec French) , , , , , , –, , –, –, , –, –, , , , , , , , , , , ,
G Gbe , GC (see Guyanese Creole) German , , , , –, , , , , , , , , , , , , , , –, , , , , –, –, –, , , , , , , , , , , , , , , – Greek , –, , , Guyanais Guyanese Creole (GC) , , , , –, , –,
Rural Guyanese Creole (RGC) , –, , , , , Urban Guyanese Creole (UGC) , , , –, –, Guyanese Creole Basilect H Haitian Creole Hawai’i/Hawaiian Pidgin English (HPE) , , , –, , Filipino Pidgin English (FPE) Variety , Japanese Pidgin English (JPE) Variety , , Korean Pidgin English (KPE) Variety , , , , Hebrew , , , , , , , , , , Hmong HPE (see Hawai‘i/Hawaiian Pidgin English) Hungarian –, I Icelandic IE (see Indo-European) Ilocano Indo-European (IE) , –, , –, , – Pre-Indo-European (Pre-IE) , –, – Proto-Indo-European (Proto-IE) , Irish , –, , , , Old Irish , Italian , , , , , , , , , , , , , , ,
JB[v.20020404] Prn:16/08/2006; 14:03
F: LA93INL.tex / p.2 (152-242)
Index of languages
J Jamaican Creole , Japanese , , , , , , , , , , , –, , Japanese Pidgin English (JPE) (see Hawai’i/Hawaiian Pidgin English) K Korean , , , , , , , , , , , , , , , Korean Pidgin English (KPE) (see Hawai’i/Hawaiian Pidgin English) L Latin M Malagasy , , , Marquesas Melanesian Pidgin Mobilian Pidgin , Moroccan Arabic (see also Arabic) , N Native American (see also American Indian) , Ndjuka-Trio Pidgin Nieuan
Norwegian , O Old Irish (see also Irish) , P Polynesian Portuguese (see also Brazilian Portuguese) , Pre-Indo-European (Pre-IE) (see Indo-European) Principense Proto-Indo-European (Proto-IE) (see Indo-European) Punjabi , Q Quebec French (see also Canadian French; French) R RGC (see Rural Guyanese Creole) Rural Guyanese Creole (RGC) (see Guyanese Creole) Russernorsk Russian , , , , , , , , , , , , , , S Sango
Sanskrit , , Scandinavian Serbian , , , , , , , , , –, , Slavic (see also Czech; Russian; Serbian) , Spanish , , , , –, , –, , , , , , , , , , , –, , , Swedish , , –, , T Tagalog Tok Pisin Tongan Tsimshian (see also Chinook Pidgin) Turkish , , , U UGC (see Urban Guyanese Creole) Urban Caribbean Creole Urban Guyanese Creole (UGC) (see Guyanese Creole) Urdu Y Yiddish , Yimas ,
JB[v.20020404] Prn:18/08/2006; 12:50
F: LA93INN.tex / p.1 (47-185)
Index of names
A Abney, S. , Aboh, E. O. , Akmajian, A. , , , , Alexiadou, A. Algeo, J. Alleyne, M. Ameka, F. Anagnostopoulou, E. Andersen, R. Anderssen, J. Andruski, J. Anishchanka, A. Anson, J. G. , Arends, J. Ash, L. Authier, J.-M. Avrutin, S. , , , , , , –, , , , , B Baauw, S. Badecker, B. Bailey, B. , Baker, M. , Baker, N. Bakker, P. Ball, A. Bardovi-Harlig, K. Bar-Hillel, Y. Barton, E. , , , , , , –, –, , –, , , , , –, , –, , , , , , , , , , , , , , , , –, , , , , , , , , , , , Bastiaanse, R. , Bates, E. , , , Baugh, J. , , Bausch, R. Beck, M.-L. Becquet, K.
Belletti, A. Bender, E. M. , Berndt, R. , , , , Bernholtz, N. A. Bickerton, D. , , , –, , –, Bierwisch, M. , Bird, H. Blair, D. Blake, R. Bley-Vroman, R. Bloom, L. Bloom, P. Bloomfield, L. , Bobaljik, J. D. Bock, J. K. , Borer, H. , Boring, E. G. Boškovi´c, Ž. , , , Bouchard, D. , , Bowerman, M. , Brame, M. , Bresnan, J. , , Broca, P. –, , , , Broch, I. Brown, C. Brown, R. Brugmann, K. Brun, A. W. S. Bryant, D. Bub, D. Bühler, K. Burge, T. Burzio, L. Butterworth, G. , , , C Caplan, D. , –, , Caramazza, A. , , Cardinaletti, A. , , Carlson, G. ,
Carnie, A. Carston, R. , Casielles, E. , , –, , , , , , , , , , , , , , , , , , , , , , , , , –, , , Castles, A. Chafe, W. Chaudron, C. , Chertkow, H. Chien, Y.-C. Chierchia, G. , Chomsky, N. , , , , , –, , , –, , , , , , , , , , , Christiansen, J. A. Chung, S. , , Churchward, C. M. Cinque, G. Clark, R. , , Coelho, F. A. Coetzee, A. Cohen, J. Cohen, S. Collins, P. C. Coltheart, M. Comrie, B. Contreras, H. , , Corder, S. P. , Cruse, D. A. Culicover, P. , , , , –, , , –, Culy, C. , Curat, H. , D Davies, E. , De Cat, C. Déchaine, R.-M. DeGraff , M. , , , , , ,
JB[v.20020404] Prn:18/08/2006; 12:50
F: LA93INN.tex / p.2 (185-324)
Index of names Dell, G. S. den Dikken, M. , , de Roo, E. , , , DeVilliers, J. Dietrich, R. , Draak, M. Drechsel, E. J. –, Dunbar, K. E Eberhard, K. M. Edwards, W. F. , –, , , , , , , , , , , , , , , Eling, P. Ellis, R. , , Emonds, J. Engel, D. , , , , –, , , Erteschik-Shir, N. Eubank, L. , , Extra, G. , , – F Farah, M. J. Ferguson, C. A. , Fielding, H. Fischer, O. Fodor, I. G. Follett, W. Fowler, C. A. Fox, B. A. Frana, I. Franck, F. Franklin, S. Fraser, C. Friedmann, N. , , , , , , Fromkin, V. A. G Gagnon, D. A. Gallagher, R. Gardner-Chloros, P. Garrett, M. F. Gazdar, G. Geschwind, N. , , Giannakidou, A. Gibson, K. Ginzburg, J. , , , , –, , Givón, T.
Gleason, J. B. Goodglass, H. , , –, –, Goodman, J. C. Greenbaum, S. , Greenfield, P. M. Grice, H. P. , , , , Grimshaw, J. , , , Grodzinsky, Y. , , , , , , , , , , , , Groenendijk, J. Grohmann, K. Grondin, N. , Guasti, M. T. , , , , , , , Guéron, J. H Haarmann, H. J. , Haegeman, L. , , , –, , , –, , , Hagiwara, H. Hagoort, P. Haider, H. Hale, K. , Hall, R. A. Hamblin, C. L. Hankamer, J. , , , , , , , , , Harding, E. Harley, H. B. Hartsuiker, R. J. –, , , Hatch, E. Hausser, R. R. Havekort, M. Hawkins, R. Haznedar, B. Heeschen, C. , , , , Heim, I. , Hendriks, A. W. , , Herschensohn, J. , , , , Heycock, C. , , , Higginbotham, J. Higgins, F. R. , Hildebrandt, H. , Hinnenkamp, V. Hoekstra, T. Hofstede, B. T. , , , –, , ,
Hollebrandse, B. Howard, D. , Huang, J. , , Huber, W. Huddleston, R. Hudson, P. T. W. Huebner, T. , , Huinck, P. Hyde, B. , , , Hyde, M. R. Hymes, D. I Ihsane, T. , , , –, Indefrey, P. , Ionin, T. , , Isserlin, M. J Jackendoff, R. , , , , , –, , , –, , Jackson, R. Jahr, E. H. , Jakobson, R. , Janda, R. D. , , , Jarema, G. , Jenkins, L. Jespersen, O. , Johnson, M. , Jordens, P. , , Jumel, C. K Kaplan, D. , , Kaplan, E. , , , Kartunnen, L. Kawahara, S. , , , Kay, P. Kayne, R. Keenan, E. Kello, C. T. , Keyser, A. Keyser, S. J. , Kim, S. King, J. Kiparsky, P. , –, , , , , Kitagawa, Y. , , Klein, W. , , , , –, , , , , , , –, –
JB[v.20020404] Prn:18/08/2006; 12:50
F: LA93INN.tex / p.3 (324-445)
Index of names Koizumi, M. Kolk, H. , –, , , –, , , , , –, –, –, –, –, –, , –, , Kolk, P. Koopman, H. Koopman, W. Krashen, S. , Kratzer, A. , , , Krifka, M. Kroch, A. , L Labov, W. , Lado, R. Ladusaw, W. Lambrecht, K. , Lang, E. Larson, R. , Lasnik, H. , Lasser, I. , Latash, M. L. , Laury, R. Leaper, C. Lebeaux, D. , , Lechner, W. Lee, J. F. , Leech, G. , Leonard, L. B. , , Levelt, W. J. M. Levinson, S. Lickley, R. J. Liu, H. , , Lobeck, A. Lockhart, C. , Loebell, H. Longobardi, G. , , , , Los, B. , Ludlow, P. Lyons, J. M Macedo, D. P. , , Mackay, D. G. MacWhinney, B. Manzini, M. R. Martensen, H. , Martin, N. , , Massam, D. , Mathesius, V.
Matthews, P. May, K. Mazzuchi, A. McClelland, J. McCloskey, J. , , McConnell-Ginet, S. , McCready, E. , McLoughlin, P. Meechan, M. E. Meinunger, A. Meisel, J. M. Melville, H. Menn, L. , , , –, , Merchant, J. , , , , , , , –, , , , , , , , , –, –, , , –, –, , , , , , , , , , , , , –, , , , , , Miceli, G. , Miller, C. A. Minor, D. Montague, R. , Monville-Burston, M. Morgan, J. –, –, , , , , , , , , , , –, , , , , Moro, A. Mühlhäusler, P. , , Munn, A. Muysken, P. Myerson, R. , , , Myers-Scotton, C. N Napoli, D. , , Naro, A. J. Nathan, G. Newmeyer, F. , , Nicol, J. Norman, D. A. , O Obler, L. , , –, Odo, C. , , –, , Ono, T. , Oomen, C. C. E. , O’Shea, M. V. Ouhalla, J. , , ,
P Paesani, K. , , –, , , , , , , , , , , , , , , , , , , , P¯anini Panse, F. Parodi, T. , Partee, B. H. , , –, , , Paul, H. Paul, I. Perdue, C. , , –, , , , , , , –, – Pérez-Leroux, A. Pesetsky, D. , Peterson, R. E. Phaf, R. H. Pienemann, M. – Pierce, A. Pinker, S. Pirvulescu, M. Platzack, C. , , Plaut, D. C. , Pollock, J.-Y. , , , Postal, P. M. Postma, A. , , Potsdam, E. , Potts, C. , , , , , , , , , , , , , , , , , , , , , , Povel, D. Prévost, P. , , , , , , , , Prince, E. F. , Progovac, L. , –, , , , , , , , , , , , , , , , , , , , , , , , , , , , –, , , , –, , , , , , , , , , , , , –, , , , , , , , , , , –, , , , , , , , , , , , , , , , , – Pullum, G. K. , Pyles, T.
JB[v.20020404] Prn:18/08/2006; 12:50
F: LA93INN.tex / p.4 (445-581)
Index of names
Q Quirk, R. , R Radford, A. , , , , , , , , , , , , , , , , , , Raposo, E. , Ratliff, M. Reilly, J. Reinhart, T. Richards, J. , Rickford, J. R. , , , , Rickford, R. Ritter, E. Rizzi, L. , , , , , , Roberge, Y. , , Roberts, I. Roberts, S. J. , Rochon, E. , , , Roelofs, A. Roeper, T. , –, , , , , , , , , , , , , , , , , , , , , , Rooth, M. , Rosen, S. T. Rosengren, I. , , Ross, J. R. Rothstein, S. , , Rugendijk, E. Rutkowski, P. S Sadock, J. , , Saffran, E. M. , , , Safir, K. Sag, I. , , , , , , , , –, , Saito, M. Sankoff, G. Saroyan, W. Sauter, K. , Schachter, P. , Schegloff, E. A. Schein, B. Schlangen, D. , Schlenck, K. J. , Schmitt, C. Schulhoff, C. Schumann, J. H. ,
Schütze, C. , , , , , , Schwabe, K. Schwartz, B. D. , , , Schwartz, M. F. , , Schwartz, R. G. , , Schwarz, F. Segal, G. , Selinker, L. Semenza, C. Seuren, P. A. M. Shallice, T. , Shimoyana, T. Shopen, T. , Sinclair de Zwart, H. , Siple, P. , –, , , , , , , , , , , Smith, J. H. Smith, N. Speas, M. Spencer, A. Sperber, D. Sportiche, D. Sprouse, R. A. Stainton, R. , , , , , , , , , , , , , , , , , , , , , , , , Stanges, H. Stanley, J. , , , , Starke, M. Stokhof, M. Stowell, T. , , , , , , –, , , , , , , Svartvick, J. , Svenonius, P. Sweet, H. Szabolcsi, A. , T Tesak, J. – Thomason, S. G. , Thompson, S. A. Turvey, M. T. U Uriagereka, J. V Vainikka, A. , , , Vallduví, E.
van der Heijden, A. H. C. van der Voort, H. , van der Wurff, W. van Gelderen, E. , van Grunsven, M. F. , –, van Hout, R. , , – van Kemenade, A. van Riemsdijk, H. , , van Zonneveld, R. Verbuk, A. , Vigliocco, G. Vinet, M.-T. , –, Visser, J. W Walker, J. A. Ward, G. Wassenaar, M. Waugh, L. R. Wearing, C. Weil, H. Weinberg, A. Wexler, K. , –, , , , , White, L. –, , , , –, , Wilder, C. , Williams, E. Williams, K. Wilmes, K. Wilson, D. Winford, D. , –, , , , , , , , , , , , , Woolf, V. , Work, N. , –, , , , , , , –, , , , , , Y Yanofsky, N. , , , , , Young-Scholten, M. , , Z Zanuttini, R. Zenk, H. B. Zhang, S. , , , Zurif, E. B. , Zwarts, F. Zwicky, A. ,
JB[v.20020404] Prn:18/08/2006; 12:54
F: LA93INS.tex / p.1 (47-116)
Index of subjects
A A-bar movement (see Movement) Absence of categories (see also Function words) Absence of arguments (see also Null subjects; Object drop) Absence of articles (see Absence of determiners) Absence of auxiliaries (see also Absence of Tense/Finiteness) –, –, , , –, , –, , –, , , , , , , , , Absence of copula (see also Absence of Tense/Finiteness) , , –, , , , , , , , Absence of determiners , –, , –, , , , , , , –, , , , , , , , , , , , , , , Absence of subordination (see Subordination, Lack of _ ) Absence of Tense/Finiteness (see also Absence of auxiliaries; Absence of copula) –, –, , , , , , , , , , , , , , , , , , Absence of verb (see also Absence of copula; Gapping) , , , , , Absolute ablative Abstract case (see Case)
Access to UG (see also Universal Grammar (UG)) , Accusative Case (see also Case) , , –, , , , –, , , –, , , , , , , , , , , , , , , , , , , , , , , , –, , , , , , , , , –, , Default accusative (see also Default Case) , , , –, , , , , , , , , , Acquisition , –, , , , , , , , , , , , , , –, , , , , –, , , , , , , , , , , , , , , , , –, , , , , , First Language Acquisition (L1 acquisition) , , , , , –, , , , , , , –, , , , , , , , , , , Natural order , , Second Language Acquisition (SLA; L2 acquisition) –, , –, –, –, , , –, , , , , , , Stages of _ , , , , –, , –, , , , –, Early stages , , , One-word stage
Pregrammatical stage Two-word stage Adaptation in disorders (see also Compensation) , , , –, , –, , , , , – Corrective adaptation , –, Preventive adaptation , , , – Adaptation Theory , , , , , – Adjacency in word order (see also Word order) , , Adjacency pairs Adjunction (see also Parataxis) , , , , , Adverb(ial)s , , , , , , , , , , –, Sentential Temporal/Time , , , , , , , , , Agent (see Thematic Roles) Agrammatism (see also Broca’s aphasia; Impairment; Decay; Delay; Paragrammatism) , –, , , , , , , –, , , , –, , –, , , , , , , , , Agrammatic production –, , , –, , – Symptoms of _ , –, –, , Morphological symptoms , , ,
JB[v.20020404] Prn:18/08/2006; 12:54
F: LA93INS.tex / p.2 (116-187)
Index of subjects Rate symptoms , , Syntactic symptoms , , , , , , –, , Agreement (see also Agreement Phrase) , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , Gender-number agreement Subject-predicate agreement , Subject-verb agreement , , Agreement Phrase (AgrP) , , , AgrP (see Agreement Phrase) Alzheimer’s disease Ambiguity , , Anaphora (see also Binding; Reflexives) , , , Bound forms Bound Variable Anaphora Deep anaphora Surface anaphora Zero anaphora , Antecedents (see also Binding) , , , , –, –, , , –, , , , , , , , , , Linguistic antecedents , , , , –, , , –, , , , , Anticonnectivity (see also Connectivity; Mismatches) , , , , , , Antihonorific (see also Honorific) , Aphasia (see also Agrammatism; Impairment) , –, , , , , , , , , , , –, , , , , , , , , , Broca’s aphasia (see also Agrammatism) –, , , , Wernicke’s aphasia Appositive
Arguments (see also Predication; Thematic Roles; Visibility of arguments) , , , , –, –, –, , , , , , , , , , , , –, , , , , , , , , , , , ,, Articles (see also Determiners; Absence of determiners) , , , , , , –, , , , , –, , , , , , , , , Aspect (see also Aspect Phrase) , , , , , , , , , , , , , , , , Perfect , , Perfective , Progressive , , , Aspect Phrase (AspP) , , AspP (see Aspect Phrase) Assertion (see also Declarative; Propositions; Statements; Tense and Assertion) , , , , , , , , –, , , , Aufgabe Autonomy of syntax , Auxiliary (AUX) (see also Absence of auxiliaries) , , , , , , , , , , , , , , , , , , , , , , , , , , B Bare argument ellipsis (see Ellipsis) Base-generation , , , , , , –, , , , , , , , , –, , , , , –, , , , , , , , Basic syntax (see also Minimal (default) grammar; Small structures; Telegraphic speech) , , , , Basic Variety (BV) , , , , –, , , –, –, –, , Basilect , , Basilect Creole ,
Between – patient variability in agrammatism (see Variability) Bilingualism (see also Universal Bilingualism) , , , Binding (see also Anaphora; Antecedents; Reflexives) , , , , , , , , , , Binding Principles Principle A Principle B Binding Theory , , , Boolean coordinator (see also Coordination) Borrowings , , , , Boston Diagnostic Aphasia Test Bound Variable Anaphora (see Anaphora) Bradykinesia Brain (see also Agrammatism; Broca’s area; Left hemisphere) , , , , , , , , Brain damage , , Brain dysfunction Brain imaging – Broca’s aphasia (see Aphasia) Broca’s area (see also Brain) , , BV (see Basic Variety) C CA (see Contrastive Analysis) Case (see also Default Case) –, , , –, –, –, –, –, –, , –, , –, –, –, , –, , –, , , –, , , , –, , , –, –, –, , –, , –, , –, , , , , , , , –, , , , , –, , –, –, –, – Abstract , , ,
JB[v.20020404] Prn:18/08/2006; 12:54
F: LA93INS.tex / p.3 (187-250)
Index of subjects
Accusative (see Accusative) Case-checking (see also Feature-checking) , , , –, , , –, Case features , , , , , , , , , , , –, Caseless (see Default Case) Case Theory – Dative , , Exceptional Case Marking (ECM) , , Genitive , , , , Inherent Instrumental , Locative , , , , Nominative (see Nominative) Oblique , , , , Structural , –, , , , , , –, –, , , , , , , , , , –, Vocative , , , Casual conversation (see also Conversation) , , , , , , , , , , , , , , , , , , , , C-command C-I (see Conceptual-Intentional) Clefts (see also Pseudoclefts) , , , , , , Clitics (see also Weak Pronouns) , , , , , , Code-switching Cognition/cognitive , , , , , , , Communication , , , , , , , , , , , , , , , , , , , , , , Communicative demands Communicative function Compensation in disorders (see also Adaptation) , , , , , , ,
Complementation (see Subordination) Complementizer (C) (see also Subordinator) , , , , , , , , , , , , , , , , , , Complementizer Phrase (CP) , –, , , , , –, , , , Conceptual-Intentional (C-I) (see Interfaces) Conceptual Structure Conjunction (see also Coordination) , , , , , , , , Conjunction Reduction Connectivity effects (see also Anticonnectivity) , , , , , , , –, , , , , , –, , , , , , , Binding , , , , , , , , , , Case , , , , –, –, , , – Islands , , , , , , , , , , –, Preposition stranding , –, , – Proper Constituents , Subcategorization , – Contact (see Language contact) Content words , Context (see also Discourse context; Pragmatic context) , , –, –, , , , , –, , , , , , , –, , –, , , , , , –, –, , , , , , , , , , , , , , , , , , , , , , , , , , , Context of the utterance , , Linguistic context , , –, , , ,
Nonlinguistic/Extralinguistic context –, , –, , , Continuity (see also Discontinuity) , , , , , Continuity Hypothesis – Contraction , Contrastive Analysis (CA) , Contrastive emphasis (see Information structure; also Emphasis) Controller First (see Information structure) Conversation (see also Casual conversation) , , , , , , , , , , , , , , , , , Conversation analysis , Conversational context (see also Context) , , , Conversational implicature (see also Grice) , Cooperative Principle (see also Conversation; Grice) Coordination (see also Conjunction) , , , , , , , , Copula , , , , , , , , , –, , –, , –, , , , , , , –, , , , , , , , , , , , , –, , , Copula omission (see Absence of categories) Core (vs. Periphery) , , , , , Covert repair (see Repair in agrammatism) CP (see Complementizer Phrase) Creole languages , , , , , , , –, , –, , , ,
JB[v.20020404] Prn:18/08/2006; 12:54
F: LA93INS.tex / p.4 (250-326)
Index of subjects
D Dative (see Case) Decay in agrammatism (see also Agrammatism) –, , Declarative (see also Assertion; Propositions; Statements) , , Deep anaphora (see Anaphora) Default accusative (see Accusative) Default Case (see also Case; Default Accusative Case; Default Nominative Case) , , , , , , , , , , , –, , , , –, , , , , , , , , , , , , , , , , , –, , , , , , , , , , , , , , –, , – Default features (see Features) Default grammar (see Minimal (Default) Grammar) Default nominative (see Nominative) Default tense (see Tense; see also Absence of Tense/Finiteness) Default time (see Time) Deixis (see also Demonstratives; Ostension; Pointing) , , , , Delay in Agrammatism (see also Agrammatism) , , , Deletion (see also Ellipsis) , , , , –, , , , , , , , , , , –, , , , , , , , , , , , , , , , , , , –, Nonconstituent deletion , , , , , , Recoverability of _ , –, , , , Stripping Unconstrained deletion , , , Demonstratives (see also Deixis; Ostension; Pointing) Deontic modality
Desiderative mood (see also Mood; Optative mood) , , Determiners (see also Absence of Determiners; Determiner Phrase (DP)) , , , , , , , , , , , , , , , , , , , , , , , , , , , , , Definite , , , Expletive , , , Referential –, , , , , , Substantive – Determiner Phrase (DP) , , , –, , , , , , , , , , , , , , –, Developmental stages (see Acquisition) Diachronic development , , , , , , , , , , , –, – Diary register (see Special registers) Direct interpretation (see Base-generation) Disapprobation clauses (see also Expressive small clauses) –, –, , , Self-disapprobation –, , , Discontinuity (see also Continuity) Discourse , , , , –, –, , , , , , , , , , , , –, , , –, , , , –, , , , , , , , , , Discourse context (see also Context; Pragmatic context) , –, , , , , , –, , , , , , , , Discourse initial utterances , , –, , , , –, –, –
Disfluencies in agrammatism (see also Nonfluencies) , , Dislocation , , , , , Left dislocation , , , – Right dislocation , , Disorder (see Impairment; also Aphasia; Agrammatism) Dissociation in agrammatism , , , , Dominant language , Down Syndrome DP (see Determiner Phrase) Dyslexia , Dystonia E EA (see Error Analysis) Early stages (see Acquisition) Echo questions , ECM (see Exceptional Case Marking) Economy in syntax , , , , , , , , , , , , , , , , Einstellung Ellipsis (see also Deletion) –, –, , , –, , , –, , , , –, –, , , –, , –, –, , –, –, –, –, –, , , –, , , , , , , , , –, –, –, , , , –, , , , , , , – Bare argument ellipsis E feature , , –, General Ellipsis , , Limited Ellipsis , , , , , , , , , , , Stripping VP ellipsis , , , –, , Embedded small clauses (see also Small clauses) , , Embedding (see Subordination) Emotive force (see Expressives)
JB[v.20020404] Prn:18/08/2006; 12:54
F: LA93INS.tex / p.5 (326-401)
Index of subjects
Emphasis (see also Contrastive emphasis) , , Emphatic pronouns (see Pronouns) Empty categories (see also Null subjects; Object drop) , , , Empty object (see Object drop) Enrichment , Epithets , , Error Analysis (EA) ESC (see Expressive Small Clauses) Eventive predicates (see also Stage-level predicates; Stative predicates) , Evolution , Exceptional Case Marking (ECM) (see Case) Exclamatives (see also Mood) , –, Existential quantification (see Quantification) Experiencer (see Thematic Roles) Expletives , , , , , Expletive determiners (see Determiners) Expletive pronouns (see Pronouns) Expressives , , , –, –, , , , , , , Expressive feature Expressive force , Expressive modifier – Expressive small clauses (ESC) (see also Disapprobation clauses; Root small clauses) , , , –, –, , External negation (see Negation) Extra-linguistic context (see Context)
F Features Default (see also Unmarked features; Unspecified features; Valueless features) , ,
Feature-checking , , –, –, , , , , –, Formal (see also Uninterpretable features; Grammatical features) , , , , , , , , – Grammatical (see also Formal features; Uninterpretable features) , , , Interpretable –, , Strong (see also Weak features) , Uninterpretable (see also Formal features; Grammatical features) , , , –, Unmarked (see also Markedness) , , , , , , , Unspecified (see also Underspecification) , , , Valueless (see also Default features; Unmarked features; Unspecified features) Weak (see also Strong features) Finiteness (see also Absence of Tense/Finiteness; Nonfinite forms) , , , , , , , , , , , , , , , , , , , , , , , , –, , –, First Language Acquisition (L-1 acquisition) (see Acquisition) Fixed word order (see Word order) Focus (see also Information structure) , , , , , , –, , –, –, –, –, , –, , , , , , , –, Focus-background (see Information structure) Focus last (see Information structure) Focus Movement/Preposing (see Movement)
Narrow focus –, , , Wide focus FocusP (FP) , , , , , , , , Foreigner talk , –, , Formal features (see Features) Form-function correspondence Form-meaning mapping (see also Syntax-semantics mapping) Fossilized forms , , , , , FP (see FocusP) FP (see Functional Projection) Fragments (see also Short answers) , , –, , –, , , , , , , , , Free-choice items , Free word order (see Word order) Fronting-then-deletion (see Move and Sluice) Frozen expressions (see Fossilized forms) Function words (see also Content words) , , , Function word omission (see also Absence of categories) , , Functional categories (see also Grammatical categories) , , , , Functional projection (FP) , , –, , , , , , Functional structure , , , , , , Functional equivalence Multiple-route plasticity Redundancy , Functional reorganization , Future tense (see Tense)
G Gapping (see also Verbless forms; Absence of categories)
JB[v.20020404] Prn:18/08/2006; 12:54
F: LA93INS.tex / p.6 (401-482)
Index of subjects
Gender agreement (see Agreement) General Ellipsis (see Ellipsis) Genitive (see Case) Given-new (see Information structure) Goal (see Thematic roles) Grammatical categories (see also Functional categories) , Grammatical features (see Features) Grammatical morphology (see also Inflectional morphology; Verbal morphology) , , , , , , , Grammaticalization , , Grice (see also Conversation; Cooperative Principle) , , , H Headlines (see Special registers) Here and now (see Pragmatic context; Time) Historical development (see Diachronic development) Historical Present , , Honorifics (see also Antihonorifics) , –, , HPSG , Hypercorrection (see also Prescriptive rules) I IL (see Interlanguage) Impairment (see also Aphasia; Agrammatism) , , , , , , –, –, , , Imperative mood (see also Mood) , , , –, , , –, , , Implicit arguments (see also Absence of arguments) Incorporation Incredulitive mood (see also Mood) , , , , , , ,
Indicative mood (see also Mood; Nonindicative mood) , , , , –, , , , , Individual-level predicates (see also Stage-level predicates) , , Infinitival (see also Non-finite forms) , , , , , , , , , , , , , –, , –, , , , , , , , , , , , , , , , –, To-infinitive –, INFL (see also Inflection) , , , , Inflection (see also INFL) , , , , , , , , , , , , , –, , , , , , , , , Inflection omission (see Absence of Tense/Finiteness) Inflectional morphology (see also Grammatical categories; Verbal morphology) , , , , , , , , , , Informal speech (see Casual conversation) Information structure (see also Focus; Topic) , , and Word order Controller First Focus-background , , , , , Focus Last , Topic-comment , , , Topic-focus Contrastive emphasis (see also Emphasis) , , Given-new information Inherent case (see Case) Initial state in acquisition (see also Acquisition) Injunctive mood (see also Mood) , , , Inner loop , Inner speech Instrument (see Thematic Roles)
Instrumental (see Case) Interactive situation Interfaces –, , , , , , , , , , Conceptual-Intentional (C-I) (see also Logical Form) – Sensory-Motor (S-M) Syntax-semantics , , , Interlanguage (IL) , , , , –, , –, , , , –, –, , Internal negation (see Negation) Interpretable features (see Features) Interrogative mood (see also Mood) –, , Intonation (see also Prosody) , , , , , , , , , , , , , Intonation questions , Inversion , , Irrealis mood (see also Mood; Nonindicative mood; Tenselessness and Irrealis mood) , , –, , Islands –, , , –, , , , , , –, Island effects (see Connectivity) Isolates (see also Unembeddable forms) , J Jargon , L L1 acquisition (see First Language Acquisition) L2 acquisition (see Second Language Acquisition) Labels (see Special registers) Language Acquisition Device (LAD) Language change (see also Diachronic development) Language contact , , Language use (see also Usage) , ,
JB[v.20020404] Prn:18/08/2006; 12:54
F: LA93INS.tex / p.7 (482-556)
Index of subjects Learnability , Learner language (see also Second Language Acquisition) –, –, , , –, Learner variety (see Learner language) Left dislocation (see Dislocation) Left hemisphere (see also Brain) Left periphery , , , , , , , , , Lexical categories (see Content words) Lexifier , , , , LF (see Logical Form) Light verbs Limited Ellipsis (see Ellipsis) Lingua franca Linguistic context (see Context) Loan words (see Borrowings) Locative (see Case) Logical Form (LF) (see also Interfaces) , , , , M Mad Magazine utterances , , , , Markedness (see also Periphery; Unmarked features) , , , , , , , , , , , , , , , , , , , , , Matching effects (see Connectivity effects) Maturation Meaning and Truth (see also Truth) Memory , , , Short-term memory , Mental set , Merge in syntax , , , , , , , , Minimal (default) grammar (see also Basic syntax; Reduced syntax; Small structures; Telegraphic speech) , , , , , Minimalism , , , –, , –, , , , , , , –, , , , –, , , , , ,
Minimal Structure Principle Minimal structures (see also Small structures) , Minimal Tree (see also Tree-pruning) Mismatches (see also Anticonnectivity) , , , , , , Missing copula (see Absence of Copula) Missing object (see Object drop) Missing subject (see Null subject) Mixed analysis of nonsententials , , , , Modality Modals , , , , , , , , , Momentary aspect (see also Here and Now) , Monitoring of speech , , , Montagovian type Mood (see also Desiderative; Exclamative; Imperative; Incredulitive; Indicative; Injunctive; Interrogative; Irrealis; Nonindicative; Optative) –, , , , Morphological symptoms (see Agrammatism) Motor control –, , , , , Move and Sluice analysis (see also Movement) , , , , , , –, , , , , Movement (see also Passive; Transformations) , , , , , , , , , , –, , , , –, –, , , , , , –, –, , , , , , , , , , , , –, , , A-bar movement , , , , , Focus Movement/Preposing , Topicalization , Tough Movement
Verb Raising Wh-movement , Yiddish-movement Movement disorder (see also Motor control) Multiple-route plasticity (see Functional Equivalence) N Narrative , , , , , , , , Narrow focus (see Focus) Natural Order in acquisition (see Acquisition) Negation , , , , , , –, , , , , , , , , , , , –, , , , , , , , , , External negation , Internal negation , , Phrasal negation , , , , , Sentential negation , , , , , Negative polarity (see also Polarity) , , , NegP , , Neurolinguistics , , Nominative –, , , , , , , –, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , –, , , – Default nominative (see also Default Case) –, , – Nonconstituent deletion (see Deletion) Nonfinite forms (see also Absence of Tense/Finiteness; Infinitivals; Participles) , , , , , , , , , , , , , , , –, , , , , , , , , , , ,
JB[v.20020404] Prn:18/08/2006; 12:54
F: LA93INS.tex / p.8 (556-632)
Index of subjects
Nonfluencies in agrammatism (see also Disfluencies) , Rate symptoms (see Agrammatism) Nonindicative mood (see also Indicative mood; Irrealis mood; Mood) , , , –, , Nonlinguistic context (see Context) Nonpast (see Tense) Nonpropositional (see Propositions) Nonsentence Nonsentential speech act , , Note-taking (see Special Registers) Null arguments (see Absence of arguments) Null objects (see Object drop) Null subjects , , , , , , , , , , , , Numerals O Object drop , , –, , Oblique case (see Case) Omission of categories (arguments, auxiliaries, copula, determiners, finiteness, objects, subjects, subordination, Tense, verbs) (see Absence of categories) Optative mood (see also Desiderative mood; Mood) , , , , Optimality Theory Ostension (see also Pointing) , , , Outer loop , OV (see Word Order) Overt repair (see Repair in agrammatism) P Paragrammatism (see also Agrammatism) Parameters of variation , , ,
Parameter setting Parataxis (see also Adjunction; Subordination, Lack of _ ) , , , , , , Parkinson’s disease Participles (see also Nonfinite forms) , , , , , , , , , , , , Passive (see also Movement) , , , , , , , , Past time (see Time) Patient (see Thematic roles) Perfect (see Aspect) Perfective (see Aspect) Periphery (vs. Core; see also Markedness) , , , Person agreement (see Agreement) Phase Phrasal negation (see Negation) Picture description task , , , , Pidgin languages , , , , , , , , , –, –, –, , , , , , Pidginization , , , , , , – Prototypical Pidgins , , , , , , Planning in agrammatism , , , , , Pointing (see also Ostension) , – Polarity (see also Negation, Negative Polarity) , , , , , Polysemy Polysynthetic languages , , Possessives , , , PP (see Prepositional Phrase) Pragmatic context (see also Context; Discourse Context; Here and Now) , , –, , –, , , , –, , , , – Predication (see also Arguments; Thematic-roles) , , , , , –, –, , , , , , , , , , –, –,
–, , , –, –, , , , , –, , , , –, , , , Predicate-argument (see Word order) Predicate-nominal (see Word order) Pregrammatical stage (see Acquisition, Stages of _) Prepositional Phrase (PP) (see also Prepositions) , , , , , , , , , , , , , –, , , , Prepositions (see also Prepositional Phrase) , , , , , , , , , , , , –, –, Preposition stranding (see Connectivity) Prescriptive rules (see also Hypercorrection) , , Present Tense (see Tense) Present time (see Time) Presupposition , , , , , Preventive adaptation (see Adaptation in disorders) Priming , , Self-priming Syntactic priming , Principle A (see Binding) Principle B (see Binding) Principles and Parameters framework , , Processing , , , , , , –, , , , –, , , Processing limitations Pro-drop (see Null subjects) Production , –, , , , –, , –, , , , , , Progressive (see Aspect) Pronouns , , , , , , –, , , , , , , , , , , , , , , , , –, , , , –, , , , ,
JB[v.20020404] Prn:18/08/2006; 12:54
F: LA93INS.tex / p.9 (632-713)
Index of subjects , , , , , , Bound Clitic (see also Weak pronouns) , , , , , , Emphatic (see also Emphasis) , Expletive Resumptive , , , Strong (see also Tonic pronouns) Tonic (see also Strong pronouns) , Weak (see also Clitic pronouns) Proper constituents (see Connectivity) Propositions (see also Assertion; Declarative; Statements) , , , , , , –, , –, , –, –, , –, , –, –, , Nonpropositional Propositional function , , Subpropositional Prosody (see also Intonation) Prototypical Pidgins (see Pidgins) Proverbs Pruning of the tree (see Tree-pruning) Pseudoclefts (see also Clefts) , , ,
Q QAPs (see Question-Answer Pairs) Quantification , , , , , , , Existential , , , , Universal , , , , , , –, , , , , , , , , , Quantifier Float Question-Answer Pairs (QAPs) –, , , , , , –, ,
, , , , , , , , , , , , R Rate symptom (see Agrammatism; Nonfluencies) Recipes (see Special registers) Recipient (see Thematic roles) Recoverability (see also Deletion, Recoverability of _) , Recursion (see also Subordination) , Reduced syntax (see Basic syntax; Reduction) Reduction (see also Reduced syntax; Simplification) , , , , , , –, , , , , , , Redundancy in agrammatism (see Functional Equivalence) Referential determiners (see Determiners) Reflexives (see also Binding) , Relative clauses , , Relevance Theory Relics (see Fossilized forms) Repair in agrammatism , , , , , , , , Covert repair , , , , Overt repair , Self-repair , Representational approaches to aphasia (vs. Processing approaches) , , Responsive form Resumptive pronouns (see Pronouns) Right dislocation (see Dislocation) Root clauses , , , Root infinitives , , , , , , , , , , –, Root phrases , , , Root Principle Root small clauses (see also Expressive small clauses) , , , , –, , , –, –, , , , , –, , –, , ,
, , , , , , , , –, , , –, S Scrambling (see Word order) Second Language Acquisition (SLA) (see Acquisition) Self-disapprobation (see Disapprobation clauses) Self-priming (see Priming) Self-repair (see Repair in agrammatism) Semantic types , , , , , Sensory-motor (SM) (see Interfaces) Sentencehood (see also Tense and Sentencehood) , , , , Sentence-peripheral , Sentential adverbs (see Adverb(ial)s) Sentential negation (see Negation) Short answers (see also Fragments; Question-answer pairs (QAPs)) , –, , –, –, , Shortest Derivation Requirement Short-term memory (see Memory) Simplification (see also Reduction) , , , , , –, Simulation , Simulation model Synchron SLA (see Second Language Acquisition) Sluicing (see also Deletion; Move and Sluice) , , , , , , –, , , –, , , , –, , , , –, , , S-M (see Sensory-motor) Small answers (see Short answers) Small clauses (see also Embedded small clauses; Expressive small clauses; Root small clauses) , , , ,
JB[v.20020404] Prn:18/08/2006; 12:54
F: LA93INS.tex / p.10 (713-785)
Index of subjects –, , , , –, , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , Small grammar (see Minimal (default) grammar) Small structures (see Fragments; Minimal (default) grammar; Basic syntax) , , , , , , , , Small utterances (see Small structures) Social factors , , , SOV (see Word order) Spatio-temporal reference (see also Time) , , Special registers , , , , , , , , Diary , , , –, –, –, , , Headlines , , , , , , –, –, –, , , , , , , Labels , , , , , , , Note-taking , , , , , , , , Recipes , , , , –, – Telegrams/telegraphese (see also Telegraphic speech) , , –, , , –, , –, –, , , Titles , , , Speech Acts , , , , , , , , Nonsentential speech acts , , , , , , Speech monitoring , , Split-INFL hypothesis , Stage-level predicates (see also Eventive predicates; Individual-level predicates) , , , Statements (see also Assertion; Declarative; Propositions) , , ,
Stative predicates (see also Eventive predicates) , , Stories (see Narratives) Stripping (see Ellipsis) Strong features (see Features) Strong pronouns (see Pronouns) Stroop task Structural case (see Case) Stuttering (see also Impairment) – Subcategorization , , , Subject drop (see Null subjects) Subject omission (see Null subjects) Subject-predicate (see Word order) Subject-predicate agreement (see Agreement) Subject-verb agreement (see Agreement) Subjunctive Subordination , , , , , , , , , , , , , , , , , , , , , , , , , , , , – Lack of _ , , , , , , , , , – Subordinators (see also Complementizers) , – Subpropositional (see Propositions) Subsentential , –, , Substitution in agrammatism , , , Surface anaphora (see Anaphora) SV (see Word Order) SVO (see Word Order) Symptoms of Agrammatism (see Agrammatism) Synchron (see Simulation) Syntactic Buffer Hypothesis , Syntactic symptoms in agrammatism (see Agrammatism) Syntax-semantics interface (see Interfaces)
Syntax-semantics mapping (see also Form-meaning mapping) , T Target Language (TL) , , –, , , , –, –, , Teacher Talk , , – Telegrams (see Special registers) Telegraphese (see Special registers) Telegraphic speech (see also Fragments; Minimal (default) grammar; Basic syntax) , –, , , , , , , , , –, –, , , , , Temporal reference (see Time) Temporal/Time adverbs (see Adverb(ial)s) Temporal-window Hypothesis , – Tense (see also Tenselessness; Tense Phrase; Time) Default (see also Absence of Tense; Tenselessness; Zero Tense) , , , , Future (see also Future time) , , , , , , , , Nonpast , , , , – Past (see also Past time) , , , , , , , , Present (see also Historical Present; Present time) , , , , , , , , –, , , Tenselessness (see also Absence of Tense/Finiteness; Nonfinite forms) , , , , , –, , , , , , , , , , Tenselessness and Irrealis mood , , –, – Tense and Assertion (see also Assertion; Tenselessness and Irrealis mood) –
JB[v.20020404] Prn:18/08/2006; 12:54
F: LA93INS.tex / p.11 (785-870)
Index of subjects
Tense and Sentencehood , –, Tense and Time , –, , –, – Tense Phrase (TP) , , , –, , , , , , , , –, , , , , , , –, , , , , , , , , , , , , , , , –, –, , , Thematic (Theta) Roles (see also Theta Criterion) , , , , , , , , Agent , , , , Experiencer Goal Instrument Locative , , , , Patient , , Recipient , Theme , , , , , Theta Criterion Time (see also Tense; Tense and Time) , –, –, , –, , , , , , , , , , –, , , Default (see also Default tense) , , , , –, , , , , Future (see also Future tense) , , , , Here and now (see also Present time; Present tense) , , , , , , , , , , , , , Past (see also Past Tense) , , , –, Present time (see also Present Tense) –, , , , Temporal/Time adverbs (see Adverb(ial)s) Timeless Titles (see Special registers) TL (see Target Language) To-infinitive (see Infinitivals) Tonic pronouns (see Pronouns)
Topic (see also Information structure) , , , , , , , , , , , , –, Topicalization (see Movement) Topic-comment (see Word order) Topic drop Topic-focus (see Word order) Topic-prominent languages , Tough Movement (see Movement) TP (see Tense Phrase) Transfer in acquisition –, , , , Transformations (see also Movement) , Transparency Assumption , Tree-pruning (see also Truncation) , , , –, , , Truncation (see also Tree-pruning) , , , , , Truth , Meaning and Truth Truth-conditions , Truth-evaluability , , , , , , Two-word stage (see Acquisition) Type-lifting
U UG (see Universal Grammar) Unconstrained deletion (see Deletion) Underspecification (see also Features, Unspecified) , , , , , Unembeddable forms (see also Isolates; Subordination, Lack of _) , , , , , , Uninterpretable features (see Features) Universal Bilingualism (see also Bilingualism) , , Universal Grammar (UG) (see also Access to UG; Universals)
, , , , , , , , , , , Universal quantifiers (see Quantification) Universals (see also Universal Grammar) , , , Unmarked features (see Features) Usage (vs. Grammar) V Valueless features (see Features) VARBRUL Variability – Variability in acquisition , , – Variability in agrammatism , –, , , –, –, , , , , Between-patient –, , Within-patient , , , –, , , –, –, Variable word order (see Word order) Verb final (see Word order) Verb Raising (see Movement) Verb-second (see Word order) Verbal morphology (see also Grammatical categories; Inflectional morphology) Verbless forms (see Absence of Categories; also Gapping) Visibility of arguments –, Vocative (see Case) VP ellipsis (see Ellipsis) VS (see Word order) VSO (see Word order) W Weak features (see Features) Weak pronouns (see Pronouns) WEAVER Wernicke’s aphasia (see Aphasia) Wh-movement (see Movement) Within-patient variability in agrammatism (see Variability)
JB[v.20020404] Prn:18/08/2006; 12:54
F: LA93INS.tex / p.12 (870-909)
Index of subjects Word Order , , , , , , , –, , , , , , –, , , and Information structure (see Information structure) Canonical , Fixed vs. Free word order (see also Scrambling; Variable word order) , , , , , Predicate-argument , , , Predicate-nominal , Predicate-second Predicate-subject ,
Scrambling (see also Fixed vs. Free word order; Variable word order) Subject-predicate , , , , , , , , Types: OV SOV , , SV SVO , , , VS , , , , , VSO , Variable word order (see also Fixed vs. Free word order; Scrambling) , Verb final
Verb-second , Würzburg School
X X-bar Theory , , Xmax Generalization ,
Y Yiddish-movement (see Movement)
Z Zero anaphora (see Anaphora) Zero tense (see Tenselessness)
Linguistik Aktuell/Linguistics Today A complete list of titles in this series can be found on the publishers’ website, www.benjamins.com 100 Schwabe, Kerstin and Susanne Winkler (eds.): On Information Structure, Meaning and Form. Generalizations across languages. Expected December 2006 99 Martínez-Gil, Fernando and Sonia Colina (eds.):
Optimality-Theoretic Studies in Spanish Phonology. Expected December 2006 98 Pires, Acrisio: The Minimalist Syntax of Defective Domains. Gerunds and infinitives. xiv, 179 pp. + index. Expected Novermber 2006 97 Hartmann, Jutta M. and László Molnárfi (eds.): Comparative Studies in Germanic Syntax. From A(frikaans) to Z(urich German). Expected December 2006 96 Solstad, Torgrim and Benjamin Lyngfelt (eds.): Demoting the Agent. Passive, middle and other voice phenomena. Expected December 2006 95 Vogeleer, Svetlana and Liliane Tasmowski (eds.): Non-definiteness and Plurality. vi, 353 pp. + index. Expected Novermber 2006 94 Arche, María J.: Individuals in Time. Tense, aspect and the individual/stage distinction. 2006. xiv, 281 pp. 93 Progovac, Ljiljana, Kate Paesani, Eugenia Casielles and Ellen Barton (eds.): The Syntax of Nonsententials. Multidisciplinary perspectives. 2006042931, 2006. ix, 372 pp. 92 Boeckx, Cedric (ed.): Agreement Systems. 2006. ix, 346 pp. 91 Boeckx, Cedric (ed.): Minimalist Essays. 2006. xvi, 399 pp. 90 Dalmi, Gréte: The Role of Agreement in Non-Finite Predication. 2005. xvi, 222 pp. 89 Velde, John R. te: Deriving Coordinate Symmetries. A phase-based approach integrating Select, Merge, Copy and Match. 2006. x, 385 pp. 88 Mohr, Sabine: Clausal Architecture and Subject Positions. Impersonal constructions in the Germanic languages. 2005. viii, 207 pp. 87 Julien, Marit: Nominal Phrases from a Scandinavian Perspective. 2005. xvi, 348 pp. 86 Costa, João and Maria Cristina Figueiredo Silva (eds.): Studies on Agreement. 2006. vi, 285 pp. 85 Mikkelsen, Line: Copular Clauses. Specification, predication and equation. 2005. viii, 210 pp. 84 Pafel, Jürgen: Quantifier Scope in German. 2006. xvi, 312 pp. 83 Schweikert, Walter: The Order of Prepositional Phrases in the Structure of the Clause. 2005. xii, 338 pp. 82 Quinn, Heidi: The Distribution of Pronoun Case Forms in English. 2005. xii, 409 pp. 81 Fuss, Eric: The Rise of Agreement. A formal approach to the syntax and grammaticalization of verbal inflection. 2005. xii, 336 pp. 80 Burkhardt, Petra: The Syntax–Discourse Interface. Representing and interpreting dependency. 2005. xii, 259 pp. 79 Schmid, Tanja: Infinitival Syntax. Infinitivus Pro Participio as a repair strategy. 2005. xiv, 251 pp. 78 Dikken, Marcel den and Christina M. Tortora (eds.): The Function of Function Words and Functional Categories. 2005. vii, 292 pp. 77 Öztürk, Balkız: Case, Referentiality and Phrase Structure. 2005. x, 268 pp. 76 Stavrou, Melita and Arhonto Terzi (eds.): Advances in Greek Generative Syntax. In honor of Dimitra Theophanopoulou-Kontou. 2005. viii, 366 pp. 75 Di Sciullo, Anna Maria (ed.): UG and External Systems. Language, brain and computation. 2005. xviii, 398 pp. 74 Heggie, Lorie and Francisco Ordóñez (eds.): Clitic and Affix Combinations. Theoretical perspectives. 2005. viii, 390 pp. 73 Carnie, Andrew, Heidi Harley and Sheila Ann Dooley (eds.): Verb First. On the syntax of verbinitial languages. 2005. xiv, 434 pp. 72 Fuss, Eric and Carola Trips (eds.): Diachronic Clues to Synchronic Grammar. 2004. viii, 228 pp. 71 Gelderen, Elly van: Grammaticalization as Economy. 2004. xvi, 320 pp. 70 Austin, Jennifer R., Stefan Engelberg and Gisa Rauh (eds.): Adverbials. The interplay between meaning, context, and syntactic structure. 2004. x, 346 pp. 69 Kiss, Katalin É. and Henk van Riemsdijk (eds.): Verb Clusters. A study of Hungarian, German and Dutch. 2004. vi, 514 pp.
68 Breul, Carsten: Focus Structure in Generative Grammar. An integrated syntactic, semantic and intonational approach. 2004. x, 432 pp. 67 Mišeska Tomić, Olga (ed.): Balkan Syntax and Semantics. 2004. xvi, 499 pp. 66 Grohmann, Kleanthes K.: Prolific Domains. On the Anti-Locality of movement dependencies. 2003. xvi, 372 pp. 65 Manninen, Satu Helena: Small Phrase Layers. A study of Finnish Manner Adverbials. 2003. xii, 275 pp. 64 Boeckx, Cedric and Kleanthes K. Grohmann (eds.): Multiple Wh-Fronting. 2003. x, 292 pp. 63 Boeckx, Cedric: Islands and Chains. Resumption as stranding. 2003. xii, 224 pp. 62 Carnie, Andrew, Heidi Harley and MaryAnn Willie (eds.): Formal Approaches to Function in Grammar. In honor of Eloise Jelinek. 2003. xii, 378 pp. 61 Schwabe, Kerstin and Susanne Winkler (eds.): The Interfaces. Deriving and interpreting omitted structures. 2003. vi, 403 pp. 60 Trips, Carola: From OV to VO in Early Middle English. 2002. xiv, 359 pp. 59 Dehé, Nicole: Particle Verbs in English. Syntax, information structure and intonation. 2002. xii, 305 pp. 58 Di Sciullo, Anna Maria (ed.): Asymmetry in Grammar. Volume 2: Morphology, phonology, acquisition. 2003. vi, 309 pp. 57 Di Sciullo, Anna Maria (ed.): Asymmetry in Grammar. Volume 1: Syntax and semantics. 2003. vi, 405 pp. 56 Coene, Martine and Yves D’hulst (eds.): From NP to DP. Volume 2: The expression of possession in noun phrases. 2003. x, 295 pp. 55 Coene, Martine and Yves D’hulst (eds.): From NP to DP. Volume 1: The syntax and semantics of noun phrases. 2003. vi, 362 pp. 54 Baptista, Marlyse: The Syntax of Cape Verdean Creole. The Sotavento varieties. 2003. xxii, 294 pp. (incl. CD-rom). 53 Zwart, C. Jan-Wouter and Werner Abraham (eds.): Studies in Comparative Germanic Syntax. Proceedings from the 15th Workshop on Comparative Germanic Syntax (Groningen, May 26–27, 2000). 2002. xiv, 407 pp. 52 Simon, Horst J. and Heike Wiese (eds.): Pronouns – Grammar and Representation. 2002. xii, 294 pp. 51 Gerlach, Birgit: Clitics between Syntax and Lexicon. 2002. xii, 282 pp. 50 Steinbach, Markus: Middle Voice. A comparative study in the syntax-semantics interface of German. 2002. xii, 340 pp. 49 Alexiadou, Artemis (ed.): Theoretical Approaches to Universals. 2002. viii, 319 pp. 48 Alexiadou, Artemis, Elena Anagnostopoulou, Sjef Barbiers and Hans-Martin Gärtner (eds.): Dimensions of Movement. From features to remnants. 2002. vi, 345 pp. 47 Barbiers, Sjef, Frits Beukema and Wim van der Wurff (eds.): Modality and its Interaction with the Verbal System. 2002. x, 290 pp. 46 Panagiotidis, Phoevos: Pronouns, Clitics and Empty Nouns. ‘Pronominality’ and licensing in syntax. 2002. x, 214 pp. 45 Abraham, Werner and C. Jan-Wouter Zwart (eds.): Issues in Formal German(ic) Typology. 2002. xviii, 336 pp. 44 Taylan, Eser Erguvanlı (ed.): The Verb in Turkish. 2002. xviii, 267 pp. 43 Featherston, Sam: Empty Categories in Sentence Processing. 2001. xvi, 279 pp. 42 Alexiadou, Artemis: Functional Structure in Nominals. Nominalization and ergativity. 2001. x, 233 pp. 41 Zeller, Jochen: Particle Verbs and Local Domains. 2001. xii, 325 pp. 40 Hoeksema, Jack, Hotze Rullmann, Víctor Sánchez-Valencia and Ton van der Wouden (eds.): Perspectives on Negation and Polarity Items. 2001. xii, 368 pp. 39 Gelderen, Elly van: A History of English Reflexive Pronouns. Person, Self, and Interpretability. 2000. xiv, 279 pp. 38 Meinunger, André: Syntactic Aspects of Topic and Comment. 2000. xii, 247 pp. 37 Lutz, Uli, Gereon Müller and Arnim von Stechow (eds.): Wh-Scope Marking. 2000. vi, 483 pp. 36 Gerlach, Birgit and Janet Grijzenhout (eds.): Clitics in Phonology, Morphology and Syntax. 2001. xii, 441 pp. 35 Hróarsdóttir, Thorbjörg: Word Order Change in Icelandic. From OV to VO. 2001. xiv, 385 pp. 34 Reuland, Eric (ed.): Arguments and Case. Explaining Burzio’s Generalization. 2000. xii, 255 pp.